2016-03-01 03:57:09 whoops I ran into the apk name conflict problem again :&& 2016-03-01 04:03:36 oh huh since when does coreutils --enable-single-binary 2016-03-01 07:49:37 nmeum: aw! we pushed 4.1.18 kernel 2016-03-01 07:49:44 must have some kind of regression 2016-03-01 08:30:13 what's the simplest way to regexp a pkgver ? It might be n.x, or n.x.y, or n.x.y.z, or 0_git* 2016-03-01 08:31:47 nmeum: it was edge you upgraded? 2016-03-01 08:31:57 from v3.3 to edge? or just apk upgrade in edge? 2016-03-01 08:33:25 fcolista: regexp a pkgver? check if string is a valid version? 2016-03-01 08:33:45 ncopa, not really. I want to check if a $pkver + 1 is available. 2016-03-01 08:34:44 but pkgver format might be very different. 2016-03-01 08:35:01 And increasing the "last" digit would not suffice 2016-03-01 08:35:09 correct 2016-03-01 08:35:28 it's just a starting point, though 2016-03-01 08:36:01 you want check if there are any new upstream release? 2016-03-01 08:36:19 i'm doing a basic script that parses APKBUILD, check the pkgver and source, and tries to wget the $pkgver + 1 2016-03-01 08:36:20 right 2016-03-01 08:36:24 new upstream release. 2016-03-01 08:36:36 This is not the _only_ check that should do 2016-03-01 08:36:56 but i cannot know in advance what's the next version 2016-03-01 08:37:15 might be a new major 2016-03-01 08:37:26 or source might change 2016-03-01 08:37:49 some has version or part of the version in the url 2016-03-01 08:37:51 in the path 2016-03-01 08:37:52 but i think that a "upstream" field should be implemented in APKBUILD. 2016-03-01 08:38:31 likd source="http://download.gnome.org/pub/source/foo/$pkgver/foo-$pkgver.tar.gz" 2016-03-01 08:38:50 correct 2016-03-01 08:39:09 check what void linux does 2016-03-01 08:39:11 https://github.com/voidlinux/void-packages/blob/master/srcpkgs/linux/update 2016-03-01 08:39:41 https://github.com/voidlinux/void-packages/blob/master/srcpkgs/ImageMagick/update 2016-03-01 08:39:55 nice 2016-03-01 08:40:35 so it has a update field where's the regexp allowing to match the next ver 2016-03-01 08:40:53 so, do we have plans to build 64bit arm packages now that there's a 64bit RPi3? :) 2016-03-01 08:44:19 fabled: i just received my APU2C4 2016-03-01 08:44:28 ScrumpyJack: #5197 2016-03-01 08:45:03 clandmeter, cool 2016-03-01 08:45:21 3pc's ;) 2016-03-01 08:45:40 did you order it aswell? 2016-03-01 08:45:44 ScrumpyJack, doing aarch64 is lot of work. though supporting rpi3 in 32-bit mode should be doable fast. 2016-03-01 08:45:54 clandmeter, not yet. but i should probably get a few 2016-03-01 09:42:57 i can redescribe the problem again, i am currently locked out of my laptop because since the kernel update yesturday i cannot boot from encrypted devices anymore because the chifre is missing 2016-03-01 09:43:06 that is a huge problem 2016-03-01 09:43:10 any ideas what changed? 2016-03-01 09:49:18 leo-unglaub_: nmeum reported the same 2016-03-01 09:49:30 i wonder if this is related: https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/commit/crypto?h=linux-4.1.y&id=5d545a70cca8594f6120a9e3d70187920e62ce77 2016-03-01 09:50:43 hmmm 2016-03-01 09:50:47 good question 2016-03-01 09:50:54 i dont know really 2016-03-01 09:50:59 i am not sure about it 2016-03-01 09:51:20 it sounds like a kernel module is missing in initramfs 2016-03-01 09:51:32 a crypto kernel module 2016-03-01 09:51:35 but how can they change something like that in a bugfix release in a longstable? 2016-03-01 09:51:53 yup, thats a good question 2016-03-01 09:52:15 the kenel developers always have bad timing ... i was sitting in a meeting last night and all of the sudden i cannot decrypt and boot my laptop ... 2016-03-01 09:52:43 in your emergency shell, can you see if you have any crypt modules in /lib/modules/*/ 2016-03-01 09:53:05 it could also be modprobe or mkinitfs bug 2016-03-01 09:54:02 there are actually a lot of modules in that folder 2016-03-01 09:54:11 subfolder 4.1.18.0 2016-03-01 09:54:22 4.1.18.0? 2016-03-01 09:54:29 and then there are modules and there is also a folder kernel/crypto/* 2016-03-01 09:54:45 äh, 4.1.18-0-grsec 2016-03-01 09:54:50 i suppose /lib/modules/4.1.18-0-grsec/ 2016-03-01 09:55:36 yeah, i misstyped ... i am currently on a fucking unstable raspberry pi with debian on it an i have no idea on how to change the keyboard layout because it always crashes when i want to switch it +g* 2016-03-01 09:56:16 is the kernel/lib/libcrc32c.ko there? 2016-03-01 09:56:42 no 2016-03-01 09:56:56 i dont know if it needs to 2016-03-01 09:57:00 but a wild guess 2016-03-01 09:57:01 asn1_decoder and oid_registry 2016-03-01 09:57:16 the crypt modules needs libcrc32 2016-03-01 09:57:18 or similar 2016-03-01 09:57:31 and previously they had entry in modules.dp 2016-03-01 09:57:35 modules.dep 2016-03-01 09:57:40 but not anymore 2016-03-01 09:57:59 can you try load the kernel crypto modules manually? 2016-03-01 09:58:02 modprobe 2016-03-01 09:59:20 the only loaded modules are dm_crypt, cryptd 2016-03-01 10:04:46 i only find kernel/arch/x86/crypto/crc32.ko 2016-03-01 10:04:57 and modprobe crc32 also does not help 2016-03-01 10:09:51 what does dmesg say? 2016-03-01 10:09:57 damn webirc crashed again :( 2016-03-01 10:10:20 i think that the kernel module for aes-xts-plain64 may be there, but that there are a missing dep 2016-03-01 10:12:10 ncopa: i agree, it has to be there 2016-03-01 10:12:23 because no files where removed from the disk 2016-03-01 10:15:44 if you modprobe aes-something? 2016-03-01 10:15:54 i dont know what happened 2016-03-01 10:16:55 i modprobed everything i could find 2016-03-01 10:17:14 if you are at a loss to, then i really have a big problem 2016-03-01 10:17:15 heheheh 2016-03-01 10:17:17 ;) 2016-03-01 10:17:18 and no errors? 2016-03-01 10:17:23 no errors 2016-03-01 10:17:31 also nothing in dmesg 2016-03-01 10:23:11 lol, i am reading the kernel commit log on the bug 2016-03-01 10:23:24 signed off by *@oracle.com ... okay, no more questions *g* 2016-03-01 10:25:13 you find info in kernel commits? 2016-03-01 10:26:22 ncopa: cryptsetup 1.7.1 contains patches for the kernel bug 2016-03-01 10:26:28 release 2 days ago 2016-03-01 10:27:12 the cryptsetup people did a workaround because otherwise all old kernels would have been broken :/ 2016-03-01 10:27:45 #284 @ cryptsetup bugtracker 2016-03-01 10:30:33 http url? 2016-03-01 10:30:47 this sounds like a real mess 2016-03-01 10:31:11 what happened with the "dont break userspace"? 2016-03-01 10:31:20 i googled it on the smartphone, because everytime i open a new tab on this raspberry shit my webirc crashes 2016-03-01 10:31:23 and in a *stable* branch 2016-03-01 10:31:42 according to a git bisect they fucked up in 5 branches or so 2016-03-01 10:31:49 no idea what they wanted to do there 2016-03-01 10:31:56 but that crypto linux code is horrible 2016-03-01 10:32:03 compared to openbsd for example 2016-03-01 10:32:16 thats propobly why google replaced that part of linux with the openbsd counterpart 2016-03-01 10:32:27 i have really no idea what he wanted to do with that commit in there 2016-03-01 10:32:51 my biggest problem now is how can i get the new cryptsetup 1.7.1 on my laptop 2016-03-01 10:33:02 its kind of a chicken <-> egg problem *g* 2016-03-01 10:37:00 ncopa: could you please update the cryptsetup package to 1.7.1 so that it gets build and i can place it on my laptop to boot? 2016-03-01 10:45:45 leo-unglaub: done 2016-03-01 10:45:56 clandmeter: awesome, big thanks!! 2016-03-01 11:26:35 alright, lets dd that alpine iso on my stick and then boot my laptop 2016-03-01 11:27:10 what would you do? chroot into my /root and update or simply wget the package and then unpack it? 2016-03-01 11:28:13 i suppose you can: apk upgrade -U -a --root $mountpoint 2016-03-01 11:28:35 where $mountpoint is your mounted root 2016-03-01 11:28:51 you probably also need regenreate the initramfs 2016-03-01 11:29:39 hmm, indeed 2016-03-01 11:29:46 might its enough if you take your initramfs-grsec file, gunzip -c | cpio it to temp location 2016-03-01 11:29:47 for that a chroot would be better 2016-03-01 11:30:01 untar new cyptsetup-libs 2016-03-01 11:30:16 and put the new libs in tempdir 2016-03-01 11:30:33 and ffinally cpio $tempdir | gzip -c > initramfs-grsec 2016-03-01 11:32:46 i will give it a try 2016-03-01 11:33:55 sorry about the messup 2016-03-01 11:34:26 dont worry, it happens ... just the timing was bad *g* 2016-03-01 11:35:42 http://serverfault.com/questions/753918/cryptsetup-luks-check-that-kernel-supports-aes-xts-plain64-cipher 2016-03-01 11:37:07 fabled: the kernel regression: https://lkml.org/lkml/2016/2/19/748 2016-03-01 11:38:08 i am not sure it this is exactly my issue, because my key is not that long 2016-03-01 11:38:35 ncopa, new cryptsetup should fix it too 2016-03-01 11:38:44 apparently it is a bug in cryptsetup 2016-03-01 11:38:57 which happened to work on older kernels 2016-03-01 11:39:40 okay, the normal update is not enought 2016-03-01 11:39:45 i do have to regenerate it 2016-03-01 11:39:59 hmm, the cryptsetup people see that a little bit different *g* 2016-03-01 11:40:08 oh they fight if it is kernel or userspace bug... 2016-03-01 11:41:36 yes *g* 2016-03-01 11:41:45 if i was not affected by it i would have fun with that thread *g+ 2016-03-01 11:44:10 is there something in apk that allows me to retrigger all scripts? 2016-03-01 11:44:18 for the mkinitfs rebuild 2016-03-01 11:44:30 i normally do: apk fix mkinitfs 2016-03-01 11:44:37 which will reinstall it 2016-03-01 11:46:27 hmm, it fails 2016-03-01 11:47:10 it says that /tmp/mkinitfs.RANDOM/lib/modules/4.1.18-0-grsec: no such file or directory 2016-03-01 11:53:30 ah, wair 2016-03-01 11:53:32 wait 2016-03-01 11:54:03 it works 2016-03-01 11:54:13 i mounted /proc into a wrong folder 2016-03-01 11:54:31 brb, reboot 2016-03-01 11:57:46 yuhuuuuu, back on my desktop :) 2016-03-01 12:02:24 new cryptsetup solved it? 2016-03-01 12:05:13 ncopa: yes, it solves it 2016-03-01 12:05:28 we should update it in 3.3-stable then 2016-03-01 12:05:31 so from now on no one who installs the kernel and the cryptsetup will have this bug 2016-03-01 12:05:39 before we push the new kernel to there 2016-03-01 12:05:48 yeah, i would recommend that highly 2016-03-01 12:38:56 wah, this ssh bug is still presend even in the latest update 2016-03-01 12:38:57 crap 2016-03-01 12:44:02 ssh bug? 2016-03-01 12:47:52 ncopa: yeah, ssh alpine -> alpine always freezes after around 60 secounds ideling 2016-03-01 12:48:08 and then after two minutes of freeze i get a broken pipe 2016-03-01 12:48:31 sounds like a stateful firewall issue? 2016-03-01 12:48:54 i think its more of an ipv6 issue 2016-03-01 12:49:07 because it mostply happens since i use ipv6 2016-03-01 12:54:45 ncopa, any thought on alpine-fudge, seems nice as it would more dev pull to AL 2016-03-01 13:15:36 hmm.. alpine-fudge == alpine-sdk-advanced , that installs -- AFL, BFF, zzuf, LLVM (LibFuzzer)... etc + gdb 2016-03-01 13:25:22 is it now safe to upgrade a laptop that uses aes-xts-plain64? 2016-03-01 13:52:07 kunkku: if you are on edge yes :) 2016-03-01 13:53:54 kunkku: make sure you get cryptsetup-1.7.1 2016-03-01 14:03:31 how can i tell to shell to interpolate a variable inside a variable? 2016-03-01 14:03:33 e.g. 2016-03-01 14:03:36 souurce = "http://pypi.python.org/packages/source/${_pkgname:0:1}/$_pkgname/$_pkgname-$pkgver.tar.gz" 2016-03-01 14:03:45 wget "$source" does not work 2016-03-01 14:03:58 because there are two variables, $_pkgname and $_pkgver 2016-03-01 14:04:08 those are not "interpolated" 2016-03-01 14:04:22 ncopa: regarding the aes-xts-plain64 issue: yes it was on edge 2016-03-01 14:04:25 then wget tries actually to download "http://pypi.python.org/packages/source/${_pkgname:0:1}/$_pkgname/$_pkgname-$pkgver.tar.gz" 2016-03-01 14:04:25 makes no sense 2016-03-01 14:04:32 these should be interpolated already in $source 2016-03-01 14:04:42 check what this variable holds with echo or something 2016-03-01 14:04:51 variables exists 2016-03-01 14:04:57 single quotes? 2016-03-01 14:05:05 double quotes 2016-03-01 14:05:18 still makes no sense :( 2016-03-01 14:09:19 i need to export them 2016-03-01 14:17:10 hmm, segfaulting ... https://paste.debian.net/plain/410632 2016-03-01 14:19:49 "souurce" ! 2016-03-01 14:20:27 use the souurce, luuke 2016-03-01 14:21:47 vkris, it's a typo, but still not working 2016-03-01 14:22:28 for pkg name AA,, if I have a subpackages BB and BB-doc, how do I buil BB-doc ? 2016-03-01 14:22:58 define subpackages BB-doc in bb() {} ? 2016-03-01 14:26:40 or BB-doc$_BB_doc in main def ? 2016-03-01 14:33:07 ok doing the later i.e BB-doc$_BB_doc 2016-03-01 14:42:29 I mean BB-doc:_BB_doc, seems it worked 2016-03-01 14:45:18 ncopa: what would better add xvfb-run + xvfb-run.1 on a git repo and make a new pkg or patch xorg-server ? 2016-03-01 15:10:59 whaoo 2016-03-01 15:11:05 looks like openssl update broke curl 2016-03-01 15:11:15 they broke ABI compat 2016-03-01 15:11:49 argh!!!! 2016-03-01 15:11:53 /o\ 2016-03-01 15:13:08 :-/ 2016-03-01 15:23:50 so we can have some fudge now !! 2016-03-01 15:25:36 ncopa: hahaha 2016-03-01 15:25:49 it seams like people really cannot version there software anymore 2016-03-01 15:25:57 first the kernel, now openssl ... 2016-03-01 15:26:31 what a bunch of fucking javascript hipsters ... i bet they all have a fancy github page, but keeping abi compat is to much to ask .. 2016-03-01 15:32:02 gee !! did not know openssl devs did javascripting ;) 2016-03-01 15:32:38 i wonder if we should take the opportunity to switch to libressl 2016-03-01 15:32:50 yeah!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 2016-03-01 15:32:59 unfirtunally i dont have time right now 2016-03-01 15:33:03 ncopa: i love you right now in this moment :) 2016-03-01 15:39:28 switching to libressl may not be that bad of an idea :) 2016-03-01 15:41:10 ncopa: the ssh issue ... it seams to be alpine speciffic ... when i ssh to a debian box i can idle as long as i want, the connection is stable 2016-03-01 15:46:24 ncopa: I told you to switch to libressl before it became cool. 2016-03-01 15:46:38 (to tell you this.) 2016-03-01 15:46:45 (switching to libressl was always cool.) 2016-03-01 15:49:54 skarnet, yeah. there's new openssl major version out soon, so we need to be doing rebuild soon. i'm kinda hoping it's against libressl and not openssl 2016-03-01 15:51:45 how do i contact freenode tech staff? 2016-03-01 15:52:41 fabled: via IRC? 2016-03-01 15:52:42 :) 2016-03-01 15:52:52 well... #freenode requires voice to talk 2016-03-01 15:53:06 void linux has been doing libressl for a while 2016-03-01 16:06:58 to answer my self.. "/stats p" gives list of handles with 'staff' status 2016-03-01 16:16:16 see you later 2016-03-01 16:20:02 swichting to libressl would really be a good idea would probably require patching a few packages though 2016-03-01 16:22:05 hmmm, why http://git.alpinelinux.org/cgit/aports/commit/?id=ad476430906e49fc46d2fac75a1ffbbe8466ec70 ?! 2016-03-01 16:22:23 dne, it breaks ABI to change those config options 2016-03-01 16:22:30 we would need to recompile everything 2016-03-01 16:22:42 the cve fix still disables sslv2 even with those config options 2016-03-01 16:22:52 aha I see 2016-03-01 16:24:02 nmeum: void linux has already done it. we can probably find patches from them if needed 2016-03-01 16:24:09 but its a pretty big move 2016-03-01 16:24:15 big project 2016-03-01 16:24:20 gentoo also has some packages with a libressl use flag and libressl patches 2016-03-01 16:24:31 good to konw 2016-03-01 16:25:04 libressl flags sound like a bad idea 2016-03-01 16:25:34 skarnet: it's a gentoo useflag, we don't have useflags 2016-03-01 16:26:07 oh yeah, use flags. Meh. 2016-03-01 16:29:59 ncopa: http://tpaste.us/GB8q <- main/xorg-server: add xvfb-run 2016-03-01 16:30:06 pls have a look 2016-03-01 16:34:20 and I found a warning, so this may be useful too, 2016-03-01 16:34:31 http://tpaste.us/2NJ6 <- remove unused option --enable-glx-tls 2016-03-01 16:37:32 also move testing/php-suhosin to main pls ... 2016-03-01 16:41:29 now I can test firefox versions more easily, without full desktop install 2016-03-01 16:41:36 ;) 2016-03-01 18:20:25 Can someone here please tell me how to create the squashfs modloop for the rpi edge release? (4.1.18) 2016-03-01 23:53:54 Does http://git.alpinelinux.org/cgit/alpine-iso/tree/Makefile while making the MODLOOP only copy *.ko files? Is that why it is bigger when I make it just after doing 'apk add linux-rpi2' ? 2016-03-01 23:54:46 I just cannot seem to make a modloop for 4.1.18, I really need to so I can compile my own module with the linux-rpi2-dev which is only in edge 2016-03-02 00:12:55 ok so i've split apart my ghc port again, ready to probably send things off as patches, would anyone want to look things overe beforehand? 2016-03-02 00:14:00 and should I send the patches to a different list than alpine-devel? 2016-03-02 00:21:40 mitchty: usually the aports list: http://lists.alpinelinux.org/alpine-aports/summary.html 2016-03-02 00:22:50 andyshinn: cool will do, also quick question since this is actually 5 new apks, should I split them all out into their own patches for each apk? 2016-03-02 00:23:01 Is firmware compatible across kernel versions? 2016-03-02 00:23:22 If i use the new modules for 4.1.18 but the firmware from 4.1.15 will it work? 2016-03-02 00:24:27 yea, make a patch for each one so they can be reviewed / accepted independently 2016-03-02 00:25:20 ok will do but they'll need to get accepted in a certain order to work :) 2016-03-02 00:25:34 going to check that all my i's are dotted and t's crossed first 2016-03-02 00:26:07 well, if something is dependent on another, you should just make sure the depend list in APKBUILD is correct 2016-03-02 00:26:32 yep, thats done actually 2016-03-02 00:26:51 this is similar to the go setup in community is all 2016-03-02 00:27:06 with a subdependency for the bootstrap compiler 2016-03-02 00:28:26 i've also got the 8.0 rc building too so once this is there my next question is how to handle unstable type things that might be nice to test but shouldn't go anywhere near production code 2016-03-02 00:37:00 to note, http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#Send_a_patch points to alpine-devel 2016-03-02 05:18:11 pkgs: broken for everyone or just for me 2016-03-02 05:27:22 just you 2016-03-02 05:28:40 no longer for me, either 2016-03-02 08:52:25 fabled: i think i know whats wrong with varnish 2016-03-02 08:52:39 #4938 2016-03-02 08:53:43 out of stack 2016-03-02 08:53:56 thread is created here: https://github.com/varnish/Varnish-Cache/blob/master/bin/varnishd/waiter/cache_waiter_epoll.c#L195 2016-03-02 08:54:16 thread function is here: https://github.com/varnish/Varnish-Cache/blob/master/bin/varnishd/waiter/cache_waiter_epoll.c#L72 2016-03-02 08:54:27 struct epoll_event ev[NEEV] 2016-03-02 08:54:36 #define NEEV 8192 2016-03-02 08:54:55 the struct epoll_event is bigger than 10 bytes 2016-03-02 08:55:08 so it uses stack bigger than 80k 2016-03-02 08:55:35 it crashes on THR_SetName("cache-epoll"); 2016-03-02 09:01:15 sounds like that 2016-03-02 09:01:21 good catch 2016-03-02 09:01:40 i wonder how to fix it 2016-03-02 09:01:44 seems simple enough to adjust stack size 2016-03-02 09:01:48 we just set a hard stack size? 2016-03-02 09:01:56 yeah 2016-03-02 09:01:59 what should we set it to? 2016-03-02 09:02:02 256k? 2016-03-02 09:02:26 https://www.varnish-cache.org/lists/pipermail/varnish-misc/2009-November/017711.html 2016-03-02 09:02:41 looks like 256 is good for them 2016-03-02 09:02:47 but not sure if that was using epoll 2016-03-02 09:02:59 apparently they have different threads for different poll impl 2016-03-02 09:03:19 http://www.gossamer-threads.com/lists/varnish/misc/15274 2016-03-02 09:03:50 there's even patch 2016-03-02 09:04:03 that is cache_backend_poll.c 2016-03-02 09:04:10 i think we need fix cache_backend_epoll.c 2016-03-02 09:04:14 or all of them... 2016-03-02 09:04:26 ah 2016-03-02 09:04:28 yes 2016-03-02 09:04:32 probably all of them 2016-03-02 09:04:34 i also wonder if i should care about params->wthread_stacksize 2016-03-02 09:04:50 seems 256k is common 2016-03-02 09:04:56 i think yes 2016-03-02 09:05:04 if it's committed upstream 2016-03-02 09:12:07 https://github.com/varnish/Varnish-Cache/blob/master/bin/varnishd/mgt/mgt_pool.c#L194 2016-03-02 09:19:40 the poll waiter does not alloc big array on stack 2016-03-02 09:19:45 i think it shoudl work as is 2016-03-02 09:27:22 oh. how about just malloc the epoll stuff then? 2016-03-02 09:28:09 though it should still honor the stack size 2016-03-02 09:34:20 good morning 2016-03-02 09:43:42 i think we maybe just malloc the epoll stuf for now 2016-03-02 09:43:46 i think thats the simplest fix 2016-03-02 09:55:34 fabled: i suspect that they keep the array on the stack to avoid cpu caches misses 2016-03-02 09:56:22 it does not matter much. as long as it is an array they get grouped together 2016-03-02 09:56:36 it's so huge array that it's gonna trash the cache in any case 2016-03-02 09:58:16 probably yes 2016-03-02 09:59:05 ha 2016-03-02 09:59:14 child (32214) Started 2016-03-02 09:59:14 Child (32214) said Child starts 2016-03-02 09:59:19 and no more crashing 2016-03-02 10:05:39 cool! 2016-03-02 10:23:36 moving the default ssh port to a highport alone stops around 450 000 password guessing bots every day on my servers ... 2016-03-02 10:23:49 so i would highly recommend that to every server admon 2016-03-02 10:24:39 i use fw rule to only allow 3 attempts per 30 sec 2016-03-02 10:24:40 or so 2016-03-02 10:24:53 3 tcp connections per 30 sec on port 22 2016-03-02 10:32:06 ncopa: yeah, i had something similar, but its still a load difference if they just dont reach the port at all 2016-03-02 10:32:42 I use a script of mine that bans to a certain amount of failures 2016-03-02 10:34:09 I don't care one bit being scanned on the 22 2016-03-02 10:34:26 it doesn't load the bandwitdh not the OS 2016-03-02 10:36:51 coredumb: it depends on how much scanns you have 2016-03-02 10:37:06 if you have around 450 000 a day on one server you actually tend to notice it 2016-03-02 10:41:53 yeah I never get that much 2016-03-02 10:42:34 + other highport tend to get problematic from "firewalls you don't own" point of view 2016-03-02 10:53:37 why do you get 450000 password guessing attempts? what is on that host? something valuable? is password auth enabled on that? would you have that many if password auth would be disabled? 2016-03-02 10:59:07 ncopa: the problem with your fw rule is it becomes trivial to DoS you 2016-03-02 10:59:15 ncopa: you better whitelist a few IPs 2016-03-02 11:00:55 IP white/blacklisting isn't exactly secure, but it's an obvious low-hanging fruit for ssh connections. It will block the large majority of bots. 2016-03-02 11:02:38 bDEK2NLtrzOd: password auth is of course disabled. on all my servers 2016-03-02 11:03:23 i have no idea why i have so much bots scanning my servers, we do have a lot of customers so i guess the ip's are known all over the workd and bots tend to grap them 2016-03-02 11:48:43 i'll do new musl soonish - cherry pick the proper hosts parsing fix, and the BRE fixes as well 2016-03-02 11:51:18 "This release adds support for out-of-tree builds, search domains in resolv.conf" 2016-03-02 11:51:21 oh 2016-03-02 11:51:25 finally 2016-03-02 11:52:24 put ssh on port 80, don't serve unencrypted http to internet :&& 2016-03-02 11:52:47 fcolista, that one is in edge already. just some cherry picks for bre regressions 2016-03-02 11:53:12 oh when did that get implemented 2016-03-02 11:56:33 fabled: which one of apk and abuild depends on pkgconf again? 2016-03-02 11:58:18 I just spent some time making scripts to cross-compile pkgconf so it's available on the host. I'm an idiot: it's needed at build-time, so it should be built for the build platform. -_- 2016-03-02 11:59:21 ammunta: put ssh on port 80, get even more connection attempts than you had on port 22 :P 2016-03-02 12:00:09 okay true but can you imagine how confused it'll make them 2016-03-02 12:00:19 (not very, since they'll just move on but) 2016-03-02 12:00:23 afternoon folks 2016-03-02 12:01:50 to answer my own question: it's apk that depends on pkgconf 2016-03-02 12:02:30 abuild just depends on openssl - I built it with libressl, no patching needed 2016-03-02 12:04:11 fabled: how do you feel about enabling USB Gadget support from the rpi2 kernel build? 2016-03-02 12:04:29 ScrumpyJack, sounds ok 2016-03-02 12:04:32 s/from/for 2016-03-02 12:06:05 it would only work on the RPi zero and the A+ (the other models have a seprate usb hub) 2016-03-02 12:06:24 but it's good fun 2016-03-02 12:07:23 shall i stap it into a feature request or maybe send a patch for 2016-03-02 12:07:41 a patch for config-rpi2.armhf? 2016-03-02 12:08:07 but rpi zero uses the rpi.armhf kernel config, right? 2016-03-02 12:08:08 (sorry, i'm on a foreign windows box. copy paste issues) 2016-03-02 12:08:17 does it? 2016-03-02 12:08:34 i think so 2016-03-02 12:08:46 and the A+? 2016-03-02 12:08:46 or which chipset rpi zero is? 2016-03-02 12:09:04 lemme check 2016-03-02 12:09:19 A+ uses rpi kernel 2016-03-02 12:09:31 rpi2 is basically only for the rpi2 model b 2016-03-02 12:09:51 i'm pretty sure zero, compute module, a, a+, b, and b+ all have rpi kernel 2016-03-02 12:10:02 they are all single core with the original chipset 2016-03-02 12:10:10 only rpi2 has the newer quadcore chipset 2016-03-02 12:10:22 should also enable rpi3 support while at it 2016-03-02 12:11:18 wikipedia agrees with you: zero and A+ is BCM2835 2016-03-02 12:13:45 i'm in italy with a really crappy wifi connection at the hotel so probably shouldn't have opened my mouth about a patch 2016-03-02 12:15:37 it would be cool to have a seperate zero/A+ image that boots with the usb serial or ether gadget loaded and a dhcp + ssh server running so that you could just plug it into you host and boom, it's there! 2016-03-02 12:15:56 a bit like the beaglebone :) 2016-03-02 12:16:18 ok, *exactly* like the beaglebone ;) 2016-03-02 12:16:29 i wonder how many per-board images we want to do 2016-03-02 12:16:51 why, is there a list? 2016-03-02 12:17:03 probably as few as possible right? 2016-03-02 12:17:09 kinda my point :) 2016-03-02 12:17:24 but yeah, we should have a mechanism to allow build your own image easily 2016-03-02 12:17:43 yeah, a bit like the alpine-iso stuff. that would be neat 2016-03-02 12:18:55 maybe if we ask nicely, clandmeter will set up some community hosting service where people can "share" their isos and images and containers :) 2016-03-02 12:18:57 fabled: I'm working on something that can automatically bootstrap abuild and apk entirely from scratch, I'll make it available as soon as I can 2016-03-02 12:19:06 skarnet, cool 2016-03-02 12:19:38 once you have abuild and apk, can you delegate all the rest to apk packages or do you need more bootstrap? 2016-03-02 12:19:39 i think we would like to bootstrap one or two architectures in near future 2016-03-02 12:21:04 well I got x86_64 and arm working. It's currently 2-stages, the second stage involving a qemu boot, but right now I'm getting stuff out of native host compilation and into cross-compilation, I wanna see whether I can get abuild and apk going without any native compilation at all 2016-03-02 12:21:42 apk requires a patch btw, just replace the hardcoded /usr/lib/libfetch.a with -lfetch 2016-03-02 12:22:41 ok 2016-03-02 12:22:58 we should probably bundle libfetch with apk 2016-03-02 12:22:58 i had cross-compile working for bootstrapping new arch 2016-03-02 12:23:14 alpine arm was cross-built from alpine x86_64 2016-03-02 12:23:22 not with /usr/lib/libfetch.a, you didn't 2016-03-02 12:23:22 but it needs a bit of work to make it work again 2016-03-02 12:23:41 unless you installed an arm libfetch.a into your x86_64 /usr/lib :) 2016-03-02 12:24:28 ncopa: currently I'm bundling libfetch with my scripts - it's easier than tracking it across the net :P 2016-03-02 12:25:07 and i think apk is the only package using it 2016-03-02 12:25:26 but my question still stands: once you have abuild and apk, are you good? or do you need more stuff to bootstrap Alpine? 2016-03-02 12:25:44 yes. fakeroot 2016-03-02 12:25:49 oh jeez 2016-03-02 12:25:52 fuck fakeroot 2016-03-02 12:27:22 in theory, we could have a list of what file should have what owner/group and permission + uid/gid when greating the tarball 2016-03-02 12:27:26 creating* 2016-03-02 12:27:33 then we wouldnt need the fakeroot 2016-03-02 12:27:39 for my cross-build, the list of packages is: 2016-03-02 12:27:47 linux-headers musl libc-dev busybox busybox-initscripts binutils make pkgconf pkgconfig \ 2016-03-02 12:27:48 zlib gmp5 mpfr3 mpc1 isl cloog gcc \ 2016-03-02 12:27:48 openssl apk-tools \ 2016-03-02 12:27:48 openrc alpine-conf alpine-baselayout alpine-keys alpine-base build-base \ 2016-03-02 12:27:48 pax-utils patch sudo attr libcap fakeroot \ 2016-03-02 12:27:48 abuild mkinitfs $KERNEL_PKG $DEBUG_PKG 2016-03-02 12:28:12 what are isl and cloog? 2016-03-02 12:28:22 libs used by gcc 2016-03-02 12:28:28 optional though 2016-03-02 12:28:41 used for certain compiler optimizers 2016-03-02 12:29:10 the scripts are at http://dev.alpinelinux.org/~tteras/bootstrap/ 2016-03-02 12:29:44 where is fakeroot used? in abuild? 2016-03-02 12:29:50 yes 2016-03-02 12:30:03 make install part is done in fakeroot 2016-03-02 12:30:04 my cross-build list ordered 2016-03-02 12:32:16 you need to separate build dependencies and host dependencies 2016-03-02 12:32:38 because you need pkgconf on the build, for instance 2016-03-02 12:32:51 even if you don't on the host 2016-03-02 12:33:42 what I was thinking of was: have a separate "prod" boostrap and a "build" bootstrap 2016-03-02 12:33:58 on the host, you just need to bootstrap up to apk 2016-03-02 12:34:16 then you can install everything via apk 2016-03-02 12:34:33 on the build, you need to bootstrap up to abuild, which is more complex 2016-03-02 12:34:41 and involves a toolchain etc. 2016-03-02 12:36:11 but if I don't want to make a native build machine, I don't need a toolchain on my host, I just need an apk binary 2016-03-02 12:36:30 does it make sense ? 2016-03-02 12:37:21 yeah - we did separate the build/host/target dependencies for cross-compile 2016-03-02 12:38:36 then your dependency list is confusing, because it looks like a build dependency (toolchain, abuild) but it's not (lack of pkgconf) 2016-03-02 12:52:33 about abuild: its Makefile uses $(DESTDIR)/$(foobar) 2016-03-02 12:52:44 the convention is $(DESTDIR)$(foobar) 2016-03-02 12:53:09 not a big problem, but unaesthetic when you see double slashes 2016-03-02 12:53:29 $(foobar) is an absolute path anyway 2016-03-02 12:56:02 holy mother of 2016-03-02 12:56:34 ncopa: does abuild needing fakeroot mean it will not work on a machine that is only using static binaries? 2016-03-02 12:57:44 ncopa: resend xvfb patch links ? 2016-03-02 13:50:49 hopefully I did the ghc patch set right last night 2016-03-02 15:07:10 skarnet: unfortunally yes 2016-03-02 15:07:56 ncopa: then it's unusable for me -_- unless I build everything as root, with FAKEROOT="" 2016-03-02 15:08:38 other ideas how to set permissions of the files in tar archive? 2016-03-02 15:08:51 some of the files 2016-03-02 15:11:24 make everything root:root 0644 by default (0755 for dirs) and have a permission files along the tarball? it's ugly, but still less ugly than fakeroot. 2016-03-02 15:11:33 permissions file* 2016-03-02 15:12:12 could even be a text file read via shell "read": file, owner, group, mode 2016-03-02 15:12:37 i'd like a tar packager that could read that permission file 2016-03-02 15:12:43 yeah 2016-03-02 15:12:44 and stuff that data into the tar archive 2016-03-02 15:12:57 that would be awesome 2016-03-02 15:13:04 sould be simple 2016-03-02 15:13:27 we can code it, but then I'd advise cpio instead of tar... much simpler 2016-03-02 15:13:48 i dont know how easy it is to switch nowdays 2016-03-02 15:14:26 but cpio is probably not a bad idea 2016-03-02 15:14:43 the difference is we could hack cpio, we can't reasonably hack tar 2016-03-02 15:15:31 you could plan a major version number update for this :P 2016-03-02 15:16:06 but yeah, huge maintenance effort until the tar-based packages expire :/ 2016-03-02 15:28:51 then the permission file would surely be easier would it not ? 2016-03-02 15:31:52 ncopa wanted the archive to understand the permissions file - advantage: you can unpack in-place without race conditions 2016-03-02 15:32:25 but short-term, packing the permissions file into the archive and using it after unpacking works 2016-03-02 15:40:26 Python has a nice API for tar packages 2016-03-02 15:41:36 we could also write Lua bindings for libarchive 2016-03-02 15:55:40 this is apk I'm talking about, the basic base of the Alpine base, and it's already nontrivial to bootstrap 2016-03-02 15:55:55 please don't make it depend on anything else :P 2016-03-02 15:58:45 python can be cross compiled :-p 2016-03-02 15:59:03 maybe golang has some nice tar lib too? 2016-03-02 15:59:24 :) 2016-03-02 16:00:19 ACTION commits suicide 2016-03-02 16:10:22 please dont! 2016-03-02 16:10:24 ok ok ok 2016-03-02 16:10:31 no python, no lua and no golang 2016-03-02 16:10:41 lets do javascrupt!!! 2016-03-02 16:11:17 ACTION cries in his grave 2016-03-02 16:13:09 dont worry, nodejs cannot reap dead children... pid1 will take care of you. eventually 2016-03-02 16:14:43 even better idea. we could use phantomjs! 2016-03-02 16:15:10 lol 2016-03-02 16:15:11 then we could write the file list permissions in a boostrapped webpage with nice colors 2016-03-02 16:15:30 needs more blink tag 2016-03-02 16:16:26 that nodejs test that exhibited a flaw in bb init... that was fun 2016-03-02 16:23:41 golang does have an awesome tar package btw, but you would end up with a binary 5-10x the size of busybox 2016-03-02 16:25:56 make it 100x 2016-03-02 20:19:36 if you're going to use js, make sure to obfuscate it properly http://www.jsfuck.com 2016-03-02 23:51:18 how hard is lua to bootstrap anyway 2016-03-02 23:57:42 probably not hard, but it shouldn't be part of a package manager bootstrap... 2016-03-03 00:00:00 so wait, what are you doing? porting alpine to a new arch? rebuilding everything from scratch to see if you can? 2016-03-03 00:01:46 rebuilding everything from scratch to see if I can use apk as a package manager in a distro my customer wants to make. 2016-03-03 00:02:31 and hoping to use my experience to do Alpine things afterwards. 2016-03-03 00:06:16 tar isn't a great archive format tbh 2016-03-03 00:07:00 no it's not 2016-03-03 00:08:23 skarnet: what's the distro about ? apk + s6 init by default ? :) 2016-03-03 00:09:19 I'm not sure whether the standard tooling or the format is more horrifying 2016-03-03 00:09:23 it probably feeds into each other 2016-03-03 00:11:06 coredumb: an s6+s6-rc init, a busybox, a dnscache, a dropbear - everything you need to get a machine up and running and connected. Plus an apk binary. 2016-03-03 00:11:54 On the development one, it's more complicated: the same base, plus a toolchain, make, and everything you need up to abuild. 2016-03-03 00:12:06 wow where do you get a customer who appreciates this 2016-03-03 00:12:36 You don't get them like this. You mold them into this. 2016-03-03 00:13:11 heh 2016-03-03 00:13:22 where do you get a customer amenable to this sort of thing even :&& 2016-03-03 00:13:32 I wasn't under the impression that they're so easy to find 2016-03-03 00:14:21 I got lucky. 2016-03-03 00:14:45 I posted stuff on the busybox mailing-list, had some technical design discussions there, and he contacted me. 2016-03-03 00:15:26 He was looking for someone who'd look under the hood and design stuff from scratch, and he had money. I said "where do I sign?" 2016-03-03 00:16:02 skarnet: cool 2016-03-03 00:16:15 I'm trying to impose Alpine in the company 2016-03-03 00:16:25 that's enough of a shock already 2016-03-03 00:16:27 :P 2016-03-03 00:17:49 well impose... replace EL on strategic positions 2016-03-03 00:22:01 el = pnaelv el? 2016-03-03 00:23:21 and if so, ah, but from how many centuries ago 2016-03-03 00:23:44 EL = Enterprise Linux 2016-03-03 00:25:02 Enterprise Linux: how to combine the simplicity of enterprise software with the legendary user interfaces of Linux 2016-03-03 00:25:40 I mean, enterprise software uis aren't known for being spectacularly better than linux uis either 2016-03-03 00:25:54 oh come on 2016-03-03 00:26:10 RHEL is the complete opposite of simplicity 2016-03-03 00:26:25 That's... the joke 2016-03-03 00:26:40 doesn't make me smile anymore 2016-03-03 00:28:44 between their incredibly incompetent support/certified consultant and their extremely arrogant sales rep 2016-03-03 00:29:17 it's the perfect mix to open the golden corporate gates 2016-03-03 00:29:38 they even cost more than windows >_< 2016-03-03 00:29:45 wat 2016-03-03 00:30:12 yeah they're way too expensive 2016-03-03 00:31:04 6y ago we were paying more than windows licenses for half the server count 2016-03-03 00:31:06 lol 2016-03-03 00:32:00 their satellite costs 20k/year 2016-03-03 00:32:08 has anyone ever had luck getting bro working out of curiosity? 2016-03-03 00:32:33 bro? 2016-03-03 00:32:43 bro network 2016-03-03 00:32:52 bro network analysis software on alpine. 2016-03-03 00:33:03 Adran: I'm interested, let me know if you get some results :) 2016-03-03 00:33:04 it relies on glibc quite a bit, so curious if anyone had luck. 2016-03-03 00:33:13 time to sleep for me 2016-03-03 00:33:15 my thoughts were to get it working on alpine for a lightweight monitoring appliance 2016-03-03 00:33:16 gnite 2016-03-03 00:35:24 "lightweight monitoring appliance" "relies on glibc" "bro" 2016-03-03 00:35:35 yeah, I'm going to sleep too, this has finished me. 2016-03-03 00:36:16 skarnet: uhh, what is wrong with I said? 2016-03-03 00:37:00 you didn't say anything wrong, but I suspect the software itself is very, very wrong 2016-03-03 00:37:08 oh yes, thats what I'm afraid of. 2016-03-03 00:37:35 it appears to have glibcisms in it, which is why I was asking. but the ability to deploy it on a ova or similar the size of alpine would be much nicer then debian. :( *sighs* 2016-03-03 09:17:18 good morning 2016-03-03 09:27:28 !@#$% of a ^&*() mother!@#$% 2016-03-03 09:27:56 fabled: pkg-config (or pkgconf) is lying to you when you have a staging directory 2016-03-03 09:28:30 the .pc files know about the paths on the target 2016-03-03 09:28:59 but of course on the build machine, those paths are invalid - you need to prefix them with your DESTDIR 2016-03-03 09:29:37 how do you work around that? 2016-03-03 09:31:12 i don't remember that kind of problem. i used -sysroot for the toolchain. 2016-03-03 09:31:50 it's not about the toolchain, it's about application libs 2016-03-03 09:32:21 my example here is apk depends on openssl (libressl here) and zlib, and gets the include and lib paths via pkgconf 2016-03-03 09:32:38 I cross-built zlib and libressl, installed them into my staging rootfs 2016-03-03 09:33:10 but the libs must be accessible from the target, so the paths are /opt/zlib/include and /opt/zlib/dir 2016-03-03 09:33:24 /opt/zlib/lib* 2016-03-03 09:33:58 and on the build, pkgconf obviously says -I/opt/zlib/include 2016-03-03 09:34:15 but I want it to say -I/my/staging/dir/opt/zlib/include 2016-03-03 09:35:28 but that's what -sysroot does 2016-03-03 09:35:42 you do -sysroot /my/staging/dir 2016-03-03 09:36:12 uh 2016-03-03 09:36:50 that should only be valid for the toolchain itself 2016-03-03 09:40:09 are you telling me you're building a toolchain *specifically* for that build, that can only work if there's a valid sysroot in /my/staging/dir? 2016-03-03 09:47:30 gcc should take -sysroot 2016-03-03 09:48:19 actually, no need. The solution was in the man of pkgconf: the PKG_CONFIG_SYSROOT_DIR env variable 2016-03-03 09:48:34 http://dev.alpinelinux.org/~tteras/bootstrap/abuild-crossbuild-x86_64.conf 2016-03-03 09:49:03 yeah, you're using a very big hammer 2016-03-03 09:49:25 PKG_CONFIG_SYSROOT_DIR is the exact hammer size I want. :) 2016-03-03 09:59:50 kaniini 2016-03-03 10:01:00 ^: I suggest a clarification in pkgconf.1: PKG_CONFIG_SYSROOT_DIR does not prepend the paths defined on PKG_CONFIG_PATH with the given sysroot 2016-03-03 10:01:59 instead, it uses the verbatim PKG_CONFIG_PATH contents to find the .pc files, and prepends the paths obtained by querying those .pc files with the given sysroot 2016-03-03 10:02:06 (which is the right thing to do) 2016-03-03 10:43:10 clandmeter: help! 2016-03-03 10:43:18 i broke patchwork.a.o 2016-03-03 10:43:52 something probably changed in apache package 2016-03-03 10:44:40 hey ncopa :) 2016-03-03 10:44:55 hi 2016-03-03 10:50:18 clandmeter, http://bugs.alpinelinux.org/issues/5211 2016-03-03 10:50:21 see the last comment 2016-03-03 11:54:48 fabled: rpi3 is out 2016-03-03 11:55:31 and its back-compat with 1&2 2016-03-03 11:56:01 vkris, yes, there's a ticket about it too. it seems to require some kernel patches 2016-03-03 11:56:16 will probably pick them up on next kernel update 2016-03-03 11:57:54 I think they should try pro-strategy like http://freedom251.com/ and bring the cost more down 2016-03-03 11:59:02 promo-strategy* 2016-03-03 12:11:37 shouldn't we put a small note on website for funding rpi3 or donating ? 2016-03-03 12:14:26 ah The 900MHz 32-bit quad-core ARM Cortex-A7 CPU complex has been replaced by a "custom-hardened" 1.2GHz 64-bit quad-core ARM Cortex-A53 2016-03-03 12:17:04 seems the only kernel commit upstream is: https://github.com/raspberrypi/linux/commit/02ce9572cc77c65f49086bbc4281233bd3fa48b7 2016-03-03 12:21:50 after grsec announcement, all "hardened" keyword gives me same feeling ;) 2016-03-03 12:22:02 and on top of it "custom" 2016-03-03 12:31:35 what's this feeling ? 2016-03-03 12:32:28 proprietary, blobs, unavailability ... why ! are you getting something else ;) 2016-03-03 12:33:13 can they customize it, I thought arm licensed the arch 2016-03-03 12:34:50 well grsec is not proprietary though :) 2016-03-03 12:35:01 ARM on the other hand ... 2016-03-03 12:35:03 + extra-cost 2016-03-03 12:35:21 well, its mixed feelings ;) 2016-03-03 12:35:49 :) 2016-03-03 12:36:03 i've been having mixed feelings for ARM platforms for ages now 2016-03-03 12:36:19 and I still can't make myself buy one 2016-03-03 12:36:44 well I actually bought a couple of board but sent them back or resell them pretty quickly 2016-03-03 12:37:10 I planned rpi-zero, but its not available in india, otherwise to get pay $22 2016-03-03 12:37:34 was planning 4 rpi0 2016-03-03 12:43:22 and fun part is , after 1yrs they gonna sell it at $5 2016-03-03 12:45:47 and then nobody would buy 2016-03-03 12:48:31 thereby bringing IT (new) development to "0" 2016-03-03 12:49:08 no wonder we pioneered "0" 2016-03-03 13:27:03 ncopa: no time now :( 2016-03-03 13:27:16 business travel... 2016-03-03 13:29:50 ok 2016-03-03 13:29:53 sorry 2016-03-03 13:30:04 problems with libressl: 2016-03-03 13:30:19 - they break ABI every 6 months 2016-03-03 13:30:33 :) 2016-03-03 13:30:34 - they only support sec fixes for 1 year. we need 2(.5) 2016-03-03 13:30:41 arf 2016-03-03 13:31:10 alot of packages will need patching for the sslv3 removal: https://wiki.freebsd.org/OpenSSL/No-SSLv3 2016-03-03 13:31:28 about the ABI breakages: https://wiki.freebsd.org/LibreSSL/#History 2016-03-03 13:32:22 what is good though is that they version the shared libs properly 2016-03-03 13:34:02 i am not sure the provide all the engines we are interesetd in either 2016-03-03 13:34:06 like padlock 2016-03-03 13:37:06 that should be asked and feature-requested if missing 2016-03-03 13:41:02 ncopa: we're disabling sslv3 at arch now, and I think Debian also did this before, so patches are usually available over internets 2016-03-03 13:41:30 so thats fixable 2016-03-03 13:41:51 what worries me more is the support period 2016-03-03 13:41:54 sec fixes for 1 year – not really 2016-03-03 13:42:08 unless we do that on our own or suddenly bump that in stable branches w/ rebuild… 2016-03-03 13:42:26 (our own – I mean backporting) 2016-03-03 13:42:43 thats the cost we get 2016-03-03 13:43:15 ok i need to go 2016-03-03 13:43:17 see u 2016-03-03 15:22:52 i think we should harden alpines boot progess a little bit 2016-03-03 15:22:54 for example 2016-03-03 15:23:14 right now if you have an entry in fstab that cannot be mounted the entire boot progress fails 2016-03-03 15:23:39 in my opinion alpine should try to continue to boot as long as it can 2016-03-03 15:24:01 because if /home/foo/movies could not be mounted this is not a total boot blocker in my opinion 2016-03-03 15:26:23 use nofail? 2016-03-03 15:27:52 i did 2016-03-03 15:28:02 but it still fails 2016-03-03 15:28:17 i think the rc script expects a different return code 2016-03-03 15:40:33 leo-unglaub: local media 2016-03-03 15:41:21 sorry, local media? (I've had issues with that for network volumes, nofail still failing, needing both _netdev and nofail) 2016-03-03 15:46:44 barthalion: i double checked 2016-03-03 15:46:48 nofail still fails 2016-03-03 15:46:52 an internal drive 2016-03-03 15:46:55 my internal ssd2 2016-03-03 16:14:19 fabled: iirc, the build will fail if we dont specify fpu 2016-03-03 16:14:30 i remember i asked you which one to use. 2016-03-03 16:15:03 it succeeded now. looks like they fixed the mess upstream. 2016-03-03 16:15:51 ok good 2016-03-03 17:05:09 latest openssl breaks ABI :/ 2016-03-03 17:12:15 yep 2016-03-03 17:20:35 yep 2016-03-03 18:25:29 is there a problem with patchwork.alpinelinux.org at present? 2016-03-03 18:48:59 yes, i broke patchwork today 2016-03-03 18:49:01 sorry 2016-03-03 18:49:17 i didnt have time to investigate what happened to apache 2016-03-03 19:07:59 so, how does one become a pkg maintainer ? :D 2016-03-03 19:35:25 suqdiq: by contributing to aports 2016-03-03 20:40:31 suqdiq: do you remember an old ibm websphere exploit by the same name? that could decrypt the 3des encrypted db tables 2016-03-03 21:20:47 xming: It only breaks ABI in the default build configuration (if you re-enable the SSL2 support that stops it breaking the ABI, and doesn't undo the vulnerability fix) 2016-03-03 21:24:18 that's still unacceptable... a major release number wasn't made for dogs 2016-03-03 22:08:36 I wanted to flag package, but got 502 Bad Gateway 2016-03-04 02:47:30 it's just patchwork down, right, the ML is still working? 2016-03-04 13:57:23 monit update to v5.17 pls 2016-03-04 14:02:48 vkris, done 2016-03-04 14:03:02 thanks 2016-03-04 14:03:09 yw 2016-03-04 14:03:29 is pkg.a.o back online ? 2016-03-04 14:03:45 pw seems ok now 2016-03-04 14:06:39 pw has been fixed by clandmeter 2016-03-04 14:09:51 pkg.a.o sometimes gets accessible via proxies.. 2016-03-04 14:10:35 something possible not ok in its config (dns) 2016-03-04 14:15:04 does grsec build ok for rpi's ? 2016-03-04 14:15:17 was looking at, https://www.zephyrproject.org/news/linux-foundation-announces-project-build-real-time-operating-system-internet-things-devices 2016-03-04 14:21:08 since grsec patch is going through OOM state for AL (out of money) 2016-03-04 15:53:58 afaik we have an linux-grsec kernel for armhf 2016-03-04 17:28:15 any reason for chromium to be removed ? 2016-03-04 17:28:24 I haven't checked, did it got in community ? 2016-03-04 17:29:31 grsec seems 'proprietary' to me 2016-03-04 17:29:37 oh 2016-03-04 17:29:39 scrolled up 2016-03-04 17:35:35 ah, its 'mixed ...', all ingredients making it are not propotional 2016-03-04 22:45:18 hi, could please anyone take a look at this PR https://github.com/alpinelinux/aports/pull/34 ? 2016-03-05 11:03:15 does someone habe maybe docker with sabnzbd to share? 2016-03-05 13:02:48 hello :) 2016-03-05 19:32:51 oh openresty is an neat thing 2016-03-05 19:37:35 oh we have nginx logging to prefix/logs instead of hard-coding /var/log now 2016-03-05 19:37:38 and a symlink 2016-03-05 19:37:44 excellent 2016-03-05 19:39:03 I don't have to symlink /var/log/nginx/error.log -> /dev/stderr anymore, huh 2016-03-06 03:31:25 when using lbu to customize a livecd, it seems to be dropping the normal packages to do setup - like setup-timezone, automounting of the media apk directory and such, how do I reoslve this? 2016-03-06 18:54:59 good evening :) 2016-03-06 18:55:11 how is everybody doing here tonight? 2016-03-06 20:47:01 good 2016-03-06 23:05:35 quick question for you all, so I have apkbuilds for both in use package managers for ghc, neither technically needs ghc (though realistically they're less useful) but should they be named their names or haskell-...? 2016-03-06 23:06:24 do these packages require haskell to run? 2016-03-06 23:06:55 nope 2016-03-06 23:07:12 i built them to be independent of the ghc compiler that may be used 2016-03-06 23:07:17 https://raw.githubusercontent.com/mitchty/alpine-linux-ghc-bootstrap/master/cabal-7.10/APKBUILD example 2016-03-06 23:08:10 then i think it may not be prefixed with "haskell-" 2016-03-06 23:08:42 ok cool, was wondering what the rules were in general 2016-03-06 23:09:04 to be honest, I don’t know these rules, it’s just my humble opinion 2016-03-06 23:09:04 but to be honest i'm not planning on porting any other haskell compilers anytime soon 2016-03-06 23:10:38 ah gotcha, well would rather submit them to be "right...er" the first time so as not to waste review time 2016-03-06 23:58:36 what's the bootstrap bug? 2016-03-07 00:11:42 ammunta: it sends -static to shared libraries that get built 2016-03-07 00:13:06 i haven't drilled too far into fixing it, but it likely means patching a bunch of nonsense and this hack worked well enough 2016-03-07 00:19:34 https://github.com/emojicode/emojicode 2016-03-07 08:35:47 good morning 2016-03-07 08:37:52 morning 2016-03-07 08:54:36 http://dev.alpinelinux.org/archive/lua-aports/ does this, uh, not have documentation 2016-03-07 09:10:12 what are you guys using for virtual machines? 2016-03-07 09:13:57 by using you mean ? 2016-03-07 09:14:10 as hypervisor or ? 2016-03-07 09:15:30 you know, for testing an apline iso or doing a test installation 2016-03-07 09:15:33 something like tha 2016-03-07 09:16:00 vmm /w kvm 2016-03-07 09:18:49 leo-unglaub: qemu 2016-03-07 09:21:50 pkgs.alpinelinux.org -> 502 Bad Gateway 2016-03-07 09:24:13 ah, works again 2016-03-07 09:31:49 updata iptables to v1.6.x pls, cannot flag on pkgs.a.o 2016-03-07 09:33:44 vkris: cannot flag? 2016-03-07 09:34:39 ok now 2016-03-07 09:34:42 thanks 2016-03-07 09:34:53 iptables is already flagged 2016-03-07 09:37:50 seems that pkgs.a.o won't work on older browsers, like ancient FF 21 ;) 2016-03-07 09:38:13 so switched to FF42 2016-03-07 10:14:47 you know that something sucks if it has bash as a hard dependency ..... 2016-03-07 10:23:42 ncopa: are you sure that this is needed: http://git.alpinelinux.org/cgit/aports/tree/main/lxc/APKBUILD?id=19b0c63da322177b3942b5f1a790cd7d667d13e9#n82 ? 2016-03-07 10:26:25 leo-unglaub: http://tpaste.us/Ajvm 2016-03-07 10:28:06 clandmeter: i looked thru the alpine.in and i dont find a bash exclusive usage 2016-03-07 10:28:35 you only looked in the alpine template? 2016-03-07 10:29:07 yes, since its the only relevant one for me 2016-03-07 10:29:37 the only way would be the split them all up into subpkgs 2016-03-07 10:30:03 s/the/to 2016-03-07 10:30:23 or only install them when you need them :) 2016-03-07 10:31:17 patchwork has maybe stopped working - the openvswitch patch is on the mailing list - but not in patchwork 2016-03-07 10:32:09 clandmeter: hmm, that would be a good idea 2016-03-07 10:32:28 ACTION points into ncopa's direction regarding pw 2016-03-07 10:35:54 BitL0G1c: i applied it i think 2016-03-07 10:36:15 clandmeter, leo-unglaub i thought of shipping the alpine template with main package 2016-03-07 10:36:27 and the rest with the lxc-template 2016-03-07 10:36:32 i just noticed it because it also pulls in tar 2016-03-07 10:36:43 and this tar is now used instrad of the busybox tat 2016-03-07 10:36:54 but the gnu tar there is way slower 2016-03-07 10:37:02 so i noticed it 2016-03-07 10:37:17 thats interesting 2016-03-07 10:37:45 yeah, the gnutar is horrible .. 2016-03-07 10:37:51 well, all the templates needs diff depends, so adding bash doesnt magically make them work. 2016-03-07 10:38:20 so it makes sence to not support depends or split them all and do it correctly. 2016-03-07 10:41:49 ncopa: gnu tar is way slower for me when i extract the contents to another harddrive 2016-03-07 10:42:10 in the same directory there is no speed difference 2016-03-07 11:19:02 @ncopa, could you please take a look at https://github.com/alpinelinux/aports/pull/34 2016-03-07 11:19:03 ? 2016-03-07 11:19:46 looks like you are the upstream :) 2016-03-07 11:19:51 yeah 2016-03-07 11:20:34 pushed 2016-03-07 11:21:31 I have also https://github.com/jirutka/ssh-ldap-pubkey, but it requires Python; ssh-getkey-ldap is written especially for Alpine :) 2016-03-07 11:21:40 thanks! 2016-03-07 11:22:30 thats kinda cool 2016-03-07 11:22:42 so you can ssh login via ldap 2016-03-07 11:22:52 yeah 2016-03-07 12:17:09 could someone please bump hiawatha from 10.0 to 10.1? 2016-03-07 12:17:12 big thanks! 2016-03-07 12:18:45 ncopa, is there a note on site, to the effect that pkgs available in community/ are not packaged with same security delegence as that in main/ and any bug fixes/upgrades could be delayed ? 2016-03-07 12:19:16 i think not 2016-03-07 12:19:35 vkris: where do you thinkwe should write about it? 2016-03-07 12:19:58 faq+license 2016-03-07 12:21:05 in faq -> why are bugs in packages in community/ are not attended sooner ? 2016-03-07 12:22:04 its a wiki ;) 2016-03-07 12:24:35 its has to go somewhere, less main devs get responsible 2016-03-07 12:25:25 it AL is growing , already 4gb+ 2016-03-07 12:35:20 is comminuty really less security focused?? 2016-03-07 12:35:34 i actually dont have that feel when i look into it 2016-03-07 12:42:27 I don't think so, its only to absolve core devs from any unforseen responsibility 2016-03-07 12:42:51 we do have limited core devs ;) 2016-03-07 12:43:10 but unlimited community devs 2016-03-07 12:47:20 and any edgey (stalled) distro that rolls every 6months 2016-03-07 12:47:27 and an* 2016-03-07 12:47:48 stalled - kernel wise 2016-03-07 14:01:13 its not less security focused. its just has shorter support time 2016-03-07 14:06:51 ncopa: BitL0G1c helped me to debug the decrypting bug i had 2016-03-07 14:07:04 its like this: /etc/crypttab is basically obsolete 2016-03-07 14:07:29 now all things have to be put into vim /etc/conf.d/dmcrypt 2016-03-07 14:07:35 äh, without vim ;) 2016-03-07 14:09:30 hm 2016-03-07 14:10:27 I suppose it is this commit: http://git.alpinelinux.org/cgit/aports/commit/main/cryptsetup?id=f8ae4e7b0a1d9561f12be77def0337b975a52d5f 2016-03-07 14:11:17 i tend to prefer the debian style /etc/cryptsetup style over gentoo style /etc/conf.d/dmcrypt 2016-03-07 14:11:35 but i dont know... 2016-03-07 14:14:35 ncopa: yes, personally i also prefer the debian style 2016-03-07 14:14:48 because it seperates the config better and makes it mor clear 2016-03-07 14:15:01 but at least now i know why it did not work for me here 2016-03-07 14:23:17 i have to go 2016-03-07 14:23:20 see you later 2016-03-07 14:23:26 BitL0G1c: thanks for the help 2016-03-07 14:32:25 ;-) np 2016-03-07 14:58:07 iz: no that was before i cared about websphere :D but you're not the first to have mentioned that :D think i even found it on packetstorm before 2016-03-07 14:58:31 iz: but my name is not suck dick, it's Sucked IQ :D (or at least that's what i say when ppl from work ask me about it...ofc its suck dick) 2016-03-07 15:27:38 does anyone here have experience with apkovls, I'm trying to use them in conjunction with the livecd scripts, however it seems to replace the default livecd setup and capabilities rather then add files to the existing iso build. 2016-03-07 16:33:14 Hello, how do I build the edge alpine iso for the raspberry pi? (There is no release for it: http://dl-2.alpinelinux.org/alpine/edge/releases/armhf/) 2016-03-07 16:33:45 Also I can't build it from a standard install, building the iso seems to require a raspberry pi too 2016-03-07 16:41:54 Seems to be running on a fresh install on the rpi2 2016-03-07 16:43:30 Nope, ran out of disk space, suppose I'll have to do the persistant storage overlay hack 2016-03-07 19:10:28 @ncopa, I have question about writing abuild – how I can declare dependency on any version of openjdk-jre, i.e. currently openjdk7-jre or openjdk8-jre? 2016-03-07 20:51:11 Is there a technical reason the modloop is mounted directly from the boot drive? 2016-03-08 00:01:21 good evening 2016-03-08 00:01:32 still a lot of people here 2016-03-08 00:01:34 :) 2016-03-08 00:59:48 is someone with commit privileges around? 2016-03-08 01:06:21 kaniini: ping 2016-03-08 08:07:08 morning 2016-03-08 08:07:20 ncopa: i found a bug 2016-03-08 08:08:29 me too, likely! 2016-03-08 08:08:34 apk add mssql 2016-03-08 08:08:34 ERROR: unsatisfiable constraints: 2016-03-08 08:08:34 mssql (missing): 2016-03-08 08:08:35 required by: world[mssql] 2016-03-08 08:08:43 :D 2016-03-08 08:09:48 mssql ! 2016-03-08 09:27:41 it's a bug whic is coming back from the future (2017) 2016-03-08 09:46:48 good morning 2016-03-08 10:03:26 could someone please update pure-ftpd from 1.0.36 to 1.0.42 2016-03-08 10:03:29 thanks!!!! 2016-03-08 10:05:24 yes sir! 2016-03-08 10:07:53 leo-unglaub: its still in testing 2016-03-08 10:08:20 clandmeter: yes, i know 2016-03-08 10:08:33 but if you are already updating it, maybe you could be so kind and add --with-puredb to it 2016-03-08 10:08:45 thats another 5 euro 2016-03-08 10:08:51 that would allow pure-ftpd to use virtual users optionally without adding any dependencies 2016-03-08 10:09:05 ACTION gives clandmeter 5$ 2016-03-08 10:09:33 ACTION grabs it and buys a loly 2016-03-08 10:11:00 leo-unglaub: looking at gentoo, we are missing a lot of switches. none other are interesting? 2016-03-08 10:11:03 ACTION isn't sure what a loly is 2016-03-08 10:11:19 clandmeter: yes, there are a lot of missing switches 2016-03-08 10:11:38 but that one is the most important one because it allows you to use the core functionallity of pure-ftpd 2016-03-08 10:12:12 back in the days when i was the crux linux maintainer of it i had two more switches, but i think we can add them when someone needs them 2016-03-08 10:12:26 ACTION slaps skarnet with a dictionary 2016-03-08 10:12:40 ouch 2016-03-08 10:13:29 skarnet: loly -> something you can put in your mouth to have fun an pleasure 2016-03-08 10:13:48 ACTION knew about lollipos 2016-03-08 10:13:52 lollipops, even 2016-03-08 10:14:04 lol 2016-03-08 10:14:06 but loly... Google says something about shoes 2016-03-08 10:14:38 also pictures that are barely SFW 2016-03-08 10:14:58 i missed an L, stop hacking my keyboard! 2016-03-08 10:15:24 leo-unglaub: depending on your tastes, that could be a lot of things 2016-03-08 10:16:31 hmm seems lolly is dutch only... 2016-03-08 10:16:55 clandmeter: no, we germans say it to 2016-03-08 10:17:01 but we spell it lolli 2016-03-08 10:17:54 ACTION got hit with a heavy object because of a spelling mistake 2016-03-08 10:20:43 leo-unglaub: there it goes. 2016-03-08 10:25:15 hi, I'm hitting this issue using the openssl command, what should I do about it?: Error relocating /usr/bin/openssl: SRP_user_pwd_free: symbol not found 2016-03-08 10:25:56 byte[]: anything special about your system? 2016-03-08 10:26:03 the same thing worked a few weeks ago; I'm using gliderlabs/alpine:3.3 on docker 2016-03-08 10:26:27 ACTION sees docker and hides. 2016-03-08 10:26:49 clandmeter: big thx!! 2016-03-08 10:26:52 all it does is pull that image, do apk update and then apk add openssl ca-certificates before attempting to use openssl command 2016-03-08 10:27:10 did you apk upgrade? 2016-03-08 10:27:27 no will try 2016-03-08 10:28:17 clandmeter: did you add --without-unicode ? 2016-03-08 10:28:25 nope 2016-03-08 10:28:39 i think its a leftover from uclibc 2016-03-08 10:28:54 not sure though 2016-03-08 10:29:01 seems like a very old apkbuild 2016-03-08 10:29:14 which happens when it stays in testing. 2016-03-08 10:29:44 it needs love from a maintainer 2016-03-08 10:29:49 what does "with-unicode" even mean? if it's about UTF-8 support, that should indeed go away when building with musl 2016-03-08 10:29:51 ACTION looks in leo-unglaub direction 2016-03-08 10:30:15 clandmeter: well that worked, thanks! 2016-03-08 10:31:21 clandmeter: i would love to maintain the following packages: https://bitbucket.org/leo-unglaub/crux-ports-leo-unglaub/src 2016-03-08 10:31:37 my problem is that i cannot get the fucking lxc shit to wor with networks ... 2016-03-08 10:31:53 why not? 2016-03-08 10:31:57 i can even do that 2016-03-08 10:32:10 and i make spelling mistakes in a different language. 2016-03-08 10:32:31 i have no idea ... this entire container thing is not working with me ... 2016-03-08 10:32:39 it propobly feels the hate i have for it *g* 2016-03-08 10:32:54 be nice with lxc, else it will go complain to Docker 2016-03-08 10:32:58 how did you setup networking? 2016-03-08 10:33:04 bridged? 2016-03-08 10:33:48 i use the modprobe dummy method from the wiki 2016-03-08 10:33:52 http://wiki.alpinelinux.org/wiki/LXC 2016-03-08 10:34:11 you use your desktop for it? 2016-03-08 10:34:17 yes 2016-03-08 10:37:26 not sure i have used this dummy setup before. 2016-03-08 10:37:29 but looks sane 2016-03-08 10:39:02 sane or insane? 2016-03-08 10:39:43 i think i spelled it right this time 2016-03-08 10:40:42 clandmeter: mssql will most likely not be compiled against musl 2016-03-08 10:41:00 really?? 2016-03-08 10:41:24 ill have to complain to dalias 2016-03-08 10:41:26 mussql 2016-03-08 10:41:39 lol 2016-03-08 10:57:14 leo-unglaub: i removed the unicode switch, but what about the tls one? 2016-03-08 10:59:10 clandmeter: that is tricky 2016-03-08 10:59:39 i am personally for tls support, however, how good is tls support if it uses openssl as a backend without chifre control? 2016-03-08 10:59:51 you have to manually disable wreak chifren in pure-ftpd 2016-03-08 10:59:59 that is kind of bad design on there part 2016-03-08 11:00:06 i have no idea. 2016-03-08 11:00:11 its up to you 2016-03-08 11:00:14 let me see if i can get fucking lxc to work then i can play around with it 2016-03-08 11:00:26 because i would also prefer to change the file layout of it 2016-03-08 11:00:44 it should be in /etc/pure-ftpd and not in /etc/conf.d/pure-ftpd 2016-03-08 11:01:06 because you have in /etc/pure-ftpd the pure-ftpd.passwd pure-ftpd.pdb and the pure-ftpd.conf 2016-03-08 11:01:20 but i will send you a patch for it as soon as i have this lxc shit running 2016-03-08 11:01:32 ok, ill leave you at it. so you can send us your patch :) 2016-03-08 11:01:36 or if ncopa gives me rust on alpine i could writen and ftp server myself *f* 2016-03-08 11:01:38 hahahaha 2016-03-08 11:02:25 im sure he can with docker :p 2016-03-08 11:04:36 he can do everything with docker :) 2016-03-08 11:17:19 clandmeter: now that i thought about it a bit ... its actually interresting to see how bad ftp servers actually are if you keep in mind how simple they are 2016-03-08 11:17:51 leo-unglaub: with ftp servers as with virtually every piece of software in the world 2016-03-08 11:18:13 "this is a simple task, how the heck can the software manage to be so bad?" 2016-03-08 11:18:26 skarnet: well, writing a wifi driver or a network stack is very complicated ... so i would understand why they suck 2016-03-08 11:18:43 but if you look at ftp ... mkdir, ls, upload file, download file 2016-03-08 11:18:48 that is actually not that hard at all 2016-03-08 11:18:57 so it currently actually surprised me 2016-03-08 11:19:16 wifi drivers suck because wifi hardware sucks. Radio is hard. 2016-03-08 11:19:20 I mean anything that takes input is broken by default 2016-03-08 11:19:26 network stack isn't that complicated. 2016-03-08 11:19:51 My favourite example of something inherently complex is a graphical web browser. 2016-03-08 11:34:04 skarnet: well, thats way to much for one person 2016-03-08 11:34:16 but the simple stuff sucks to, thats so confusing 2016-03-08 11:34:31 smtp server ... i mean, really? smtp is not such a hard protocol 2016-03-08 11:36:44 smtp is very easy, but it's also insufficient, so a full-featured MTA is much more complex than just a SMTP client and server 2016-03-08 11:37:06 you need hooks for spam filtering, etc. etc. 2016-03-08 11:55:59 well, hooks are simple 2016-03-08 11:56:04 thats not the problem 2016-03-08 11:56:47 let's see, which protocols are nice and which protocols are ugly and which protocols are awful to work with 2016-03-08 12:01:47 you know 2016-03-08 12:01:59 somewhere out there, there are apparently people who run apache httpd 2016-03-08 12:02:11 somehow /nobody I know/ is any of them 2016-03-08 12:08:53 ammunta: we run it, but not because of choice... 2016-03-08 12:09:44 ammunta: that's a question of implementation, not a question of protocol 2016-03-08 12:18:35 skarnet: it's unrelated, it just reminded me 2016-03-08 12:24:12 damn you lxc ... 2016-03-08 12:24:16 i am getting mad here ... 2016-03-08 12:24:18 I didn't know that clandmeterwas Dutch 2016-03-08 12:24:38 xming: he is 2016-03-08 12:24:48 thats not a thing you would lie about *g* 2016-03-08 12:26:22 :D 2016-03-08 12:26:29 and that you were German 2016-03-08 12:27:05 :) 2016-03-08 12:28:10 esp8266 looks interesting to add wifi to rPi zero 2016-03-08 12:28:27 2 USD for a wifi module 2016-03-08 12:34:06 leo-unglaub: i hope you are not refering to Euro 2016, or ill be asking about your lxc setup every day... 2016-03-08 12:41:45 ;) 2016-03-08 12:43:40 please never mention lxc again 2016-03-08 12:43:55 i am serously thinking about just using my normal install to build the packages ... 2016-03-08 12:49:13 brb re boot 2016-03-08 12:52:43 re 2016-03-08 13:22:34 leo-unglaub1: how is your lxc setup? 2016-03-08 13:22:51 haha 2016-03-08 13:22:54 nice try 2016-03-08 13:29:32 leo-unglaub1: some are also using ovs 2016-03-08 13:34:52 clandmeter: ovs? 2016-03-08 13:35:08 openvswitch 2016-03-08 13:35:44 i never used it though. 2016-03-08 13:35:52 but i know some users on #alpine-linux do 2016-03-08 13:54:47 hmmm 2016-03-08 13:55:03 or i just ping clandmeter every time i have a patch *g* 2016-03-08 13:55:15 that way i dont have to deal with lxv 2016-03-08 13:55:17 lxc 2016-03-08 13:55:37 lxc-unglaub 2016-03-08 13:55:45 lol 2016-03-08 14:46:32 what's wrong with lxc? 2016-03-08 14:46:41 ? 2016-03-08 14:58:22 leo-unglaub1: docker run -it alpine:edge 2016-03-08 14:58:46 ncopa: ?? 2016-03-08 14:58:51 what would that do? 2016-03-08 14:59:05 so you don't need to deal with lxc ;) 2016-03-08 14:59:21 ncopa is sold to The Man now 2016-03-08 14:59:23 then i have a docker with networking and a build environment? 2016-03-08 14:59:49 you have a container with minimal alpine edge 2016-03-08 14:59:55 with networking 2016-03-08 15:00:01 with networking 2016-03-08 15:00:15 from there you can apk add alpine-sdk 2016-03-08 15:00:19 adduser leo 2016-03-08 15:00:35 etc 2016-03-08 15:02:25 here is a oneline script i use to test build a package in alpine 3.3: http://tpaste.us/GB8b 2016-03-08 15:02:42 it mounts my aports dir inside docker 2016-03-08 15:03:17 that way i can edit file from host system 2016-03-08 15:03:28 but build in alpine-3.3 environment 2016-03-08 15:25:06 damn, docker depends on iptables 2016-03-08 15:26:02 thats a problem, because i use nftables 2016-03-08 16:23:54 fabled: tvheadend wants to include ucontext.h and because you removed it, i temp added it to tvheadend. 2016-03-08 16:40:36 ncopa: clandmeter: how do I use lbu to add files to a livecd properly? e.g. if I use lbu there's no tz selection available, hostname isn't default, no keyboards. and I excluded those flies. 2016-03-08 16:40:59 clandmeter: does it only need some struct? or does it want to use ucontext? 2016-03-08 16:41:28 if it needs some struct then maybe we should have the mariadb test to try link 2016-03-08 16:42:05 Adran: i'm not following what you are trying to do 2016-03-08 16:42:18 if you boot livecd (iso image) which is read only 2016-03-08 16:42:23 then you cannot lbu to it 2016-03-08 16:42:35 and will need to have storage usb 2016-03-08 16:42:38 I'm trying to setup a custom livecd and you have the APKOVL= option, which once I set it, broke the ability to use the livecd to install a system 2016-03-08 16:43:15 because with that set, the system expects that you have already set things up 2016-03-08 16:43:20 Ah. :/ 2016-03-08 16:43:32 but its not that bad 2016-03-08 16:43:36 I just couldn't find any documentation on how to use this option better I'm afraid, hence asking. alright, so next question, how would I bundle data into the livecd but keep it as a standard installer? 2016-03-08 16:43:40 i thiink you can add the default services 2016-03-08 16:44:31 i suppose you would have to hack the makefile 2016-03-08 16:45:15 d'oh, I was hoping to have the files available on boot from the livecd, then be able to use lbu to copy it over to the fully installed system once finished. 2016-03-08 18:29:31 clandmeter, ok, yes, dalias noted that ucontext.h is probably needed. 2016-03-08 22:13:52 hu guys, I have a quick question about abuild 2016-03-09 04:46:39 ncopa: the xvfb patch pls 2016-03-09 07:39:50 noticed "privoxy:x:101:1001:Linux User,,,:/var/log/privoxy:/sbin/nologin" in /etc/passwd, wondering if its ok ? 2016-03-09 08:00:26 for privoxy, surely? 2016-03-09 08:01:41 1001 is not fine 2016-03-09 08:45:52 ... fhs demands that /srv exists, programs aren't supposed to care what's in there 2016-03-09 08:46:05 I'm. is there anything you can actually do usefully with that 2016-03-09 08:46:08 chdir into it 2016-03-09 08:49:37 would be put in b.a.o if its a bug 2016-03-09 08:49:56 hmmm.. smart algibot ! 2016-03-09 09:32:57 something in that reasoning bugs me 2016-03-09 09:33:13 doesn't it bug you? 2016-03-09 09:33:38 my dog barks b.a.o 2016-03-09 09:33:57 ok, algitbot is indeed smarter than I gave it credit for. :) 2016-03-09 11:25:36 ncopa, rnalrd : i'm going to push v4l-utils 1.10.0 2016-03-09 11:25:47 x86_64-edge:~/aports/main/v4l-utils$ check 2016-03-09 11:25:48 check checkapk 2016-03-09 11:25:48 x86_64-edge:~/aports/main/v4l-utils$ checkapk 2016-03-09 11:25:48 >>> No soname differences for v4l-utils. 2016-03-09 11:25:48 >>> No soname differences for v4l-utils-dev. 2016-03-09 11:25:48 >>> No soname differences for v4l-utils-doc. 2016-03-09 11:25:50 >>> No soname differences for qv4l2. 2016-03-09 11:25:52 >>> No soname differences for v4l-utils-dvbv5. 2016-03-09 11:25:54 >>> No soname differences for v4l-utils-libs. 2016-03-09 12:51:03 fcolista: we believe you when there is no soname change ;-) 2016-03-09 12:56:53 clandmeter, thx for the trust :) 2016-03-09 12:57:18 its quiet here 2016-03-09 12:57:34 just wanted to say something :) 2016-03-09 12:57:56 lol 2016-03-09 13:44:37 ncopa: what's your opinion on mitchtys GHC patches? %1682 %1683 and %1684 2016-03-09 14:48:42 i dont have any better suggetion how to solve ghc 2016-03-09 14:50:43 ncopa, nmeum : ghc can't use system llvm? 2016-03-09 14:53:01 fabled: ask mitchty if I recall correctly he needs this custom llvm setup to support arm in the future 2016-03-09 14:53:25 but mitchty is probably able to provide a more in-depth answer to this 2016-03-09 15:29:06 yep, long story short is you're stuck with one of 2 bad decsions 2016-03-09 15:29:39 either patch ghc to support the specific llvm you have, which means its using an llvm that may never be tested anywhere else 2016-03-09 15:30:24 or build the llvm version that ghc need and was validated against, given llvm doesn't have much forward guarantees i'm not sure what would be best 2016-03-09 15:30:52 and yes this is more for arm support which only uses llvm, x86_64 has an assembler backend too but thats more from legacy 2016-03-09 15:31:22 i patched ghc to use llvm 3.6, https://github.com/mitchty/alpine-linux-ghc-bootstrap/issues/8 2016-03-09 15:31:33 but now 8.0 is coming out and it uses 3.7.1 2016-03-09 15:31:47 so basically llvm upgrades could break ghc on arm's ability to build things 2016-03-09 15:32:21 https://ghc.haskell.org/trac/ghc/wiki/ImprovedLLVMBackend 2016-03-09 15:34:32 also for ghc you only need llc/opt not all of llvm 2016-03-09 15:45:11 hey everyone! 2016-03-09 15:46:06 fabled: while you were away mitchty answered your question http://paste42.de/10347/ 2016-03-09 15:48:27 Firefox 45.0 is here :) 2016-03-09 15:53:16 oh yeah sorry, didn't pay attention to who is online and have my client setup to not show joins/leaves to keep things tidy 2016-03-09 16:12:48 a third possible solution would be to have perhaps community/llvm3{5,6...etc} that could be used by other things 2016-03-09 16:13:10 but that was a bit of a bigger task than I was willing to bite off just for ghc 2016-03-09 19:48:03 hey :) 2016-03-10 02:32:13 Hi - just wanted to mention about the bind vulnerabilities and v.9.10.3-P4 2016-03-10 08:27:06 is by any chance musl by default balancing dns queries between different nameserver entries in resolv.conf ? 2016-03-10 08:40:09 coredumb: better ask in #musl 2016-03-10 08:42:22 coredumb: it queries all of them in parallel it seems: http://git.musl-libc.org/cgit/musl/tree/src/network/res_msend.c#n115 2016-03-10 08:43:19 dne: that's what I feared 2016-03-10 08:43:27 is there a way to disable the bhaviour ? 2016-03-10 08:46:17 better ask #musl I guess :) 2016-03-10 08:46:27 yup 2016-03-10 08:46:51 I would guess no 2016-03-10 08:49:58 coredumb: just curious. why do you want disable that feature? 2016-03-10 08:51:20 coredumb: it's documented here: http://wiki.musl-libc.org/wiki/Functional_differences_from_glibc#Name_Resolver_.2F_DNS 2016-03-10 08:53:37 dne: the faq needs to be updated 2016-03-10 08:54:08 it says: `musl's resolver does not support the "domain" and "search" keywords in resolv.conf. ` 2016-03-10 08:54:13 it does now 2016-03-10 08:55:12 right 2016-03-10 08:56:22 ncopa: cause my secondary DNS doesn't have some of my internal entries 2016-03-10 08:56:39 so it's always failing with host not found 2016-03-10 08:57:44 sounds like broken config 2016-03-10 08:59:02 nah sounds like totally acceptable config when you know that the secondary is only a failover to keep external queries working 2016-03-10 08:59:17 dne: thx for the link 2016-03-10 08:59:46 hey 2016-03-10 09:00:34 coredumb: there was a long discussion on musl mailing list 2016-03-10 09:00:46 apparently there are disagreements on it 2016-03-10 09:00:53 I bet so 2016-03-10 09:55:12 yeah, i think lxc finally works 2016-03-10 09:55:30 i now have networking in it over my eth0 and my host still is online via wlan0 2016-03-10 09:55:40 very strange config, but it seams to be online 2016-03-10 09:55:41 hehehe 2016-03-10 09:56:31 now i just need to map my ~/repos/alpine/aports folder into that container 2016-03-10 09:56:46 but i have a feeling simply using softlinks is a bad ide 2016-03-10 10:44:54 yeah. my build environment works ... kind of ... but it works ... 2016-03-10 10:45:25 i just realized how fucked up open source software has become if you need around 260 dependencies to build a simple file manager like thunar ... 2016-03-10 12:23:57 was thinking if we had alpine-utils pkgs ? 2016-03-10 12:24:49 one util is parallel from moreutils 2016-03-10 12:26:19 leo-unglaub: why strange config ? 2016-03-10 12:29:19 OpenJDK 64-Bit Server VM warning: INFO: os::commit_memory(0x00006ee795000000, 2555904, 1) failed; error='Operation not permitted' (errno=1) 2016-03-10 12:29:28 that is java on alpine 3.3 2016-03-10 12:30:20 openjdk8 2016-03-10 12:30:43 hm 2016-03-10 12:30:57 20 2016-03-10 12:32:13 happens in docker only 2016-03-10 12:55:53 1387688 Jun 15 2015 php-phalcon-2.0.3-r0.apk 2016-03-10 12:55:54 2130103 Mar 10 12:54 php-phalcon-2.0.3-r1.apk 2016-03-10 12:55:59 ncopa: ^^^ 2016-03-10 12:56:12 huge file size jump ! 2016-03-10 12:56:34 anything I need to change in abuild configs ? 2016-03-10 12:57:20 btw, update php-phalcon to v2.0.10 pls 2016-03-10 13:28:47 vkris: i cannot find php-phalcon-2.0.3-r1 in git log 2016-03-10 13:28:54 i have no idea where that comes from 2016-03-10 13:31:31 ncopa: I compared my current build for 2.0.3 with a very old build 2016-03-10 13:31:43 current release is 2.0.10 2016-03-10 13:32:12 aports:master |Natanael Copa| main/php-phalcon: upgrade to 2.0.10 | http://dup.pw/aports/eb3b0630 2016-03-10 13:32:19 i just pushed 2.0.10 2016-03-10 13:32:26 2243873 Mar 10 12:46 php-phalcon-2.0.10-r0.apk 2016-03-10 13:32:49 kinda double in size 2016-03-10 13:32:53 thanks 2016-03-10 13:33:07 checking whether HAVE_BUNDLED_PCRE is declared... no 2016-03-10 13:33:08 checking whether HAVE_JSON is declared... yes 2016-03-10 13:33:23 they probably bundle some 3rd party lib 2016-03-10 13:33:38 you'll have to ask php-phalcon devs 2016-03-10 13:33:43 ok 2016-03-10 13:41:10 did you get a chance to look at xvfb ? 2016-03-10 13:42:22 http://tpaste.us/GB8q <- main/xorg-server: add xvfb-run 2016-03-10 13:56:59 vkris: would it be an idea to ship it in a different package? 2016-03-10 13:57:43 vkris: who is maintaining this script? 2016-03-10 13:57:45 upstream? 2016-03-10 13:57:54 debian 2016-03-10 13:57:56 or do you want us to maintain it? 2016-03-10 13:58:47 what to you think about add a package named xvfb-run? 2016-03-10 13:59:42 thought of it, but later realized could be better in orig xvfb 2016-03-10 14:00:19 but if you say try making separate pkg 2016-03-10 14:00:51 debian pkgs them together 2016-03-10 14:02:31 anyone with some experience of the current alpine setup-* scripts have time to help midipix with new installer menu entries? 2016-03-10 14:07:11 it comes from http://http.debian.net/debian/pool/main/x/xorg-server/xorg-server_1.16.4-1.diff.gz 2016-03-10 14:35:53 ncopa: push testing/php-suhosin to main/ pls 2016-03-10 14:36:01 seem ok 2016-03-10 14:36:27 community i guess? 2016-03-10 14:38:17 ok 2016-03-10 14:38:43 how is php-suhosin wrt sec fixes and maintenance? 2016-03-10 14:38:50 would wait to see how active upstream is 2016-03-10 14:38:57 yes 2016-03-10 14:39:13 seems active and dev did email corresponding too 2016-03-10 14:39:32 and guided in tests 2016-03-10 15:23:53 https://lists.cypherpunks.ca/pipermail/otr-users/2016-March/002581.html :-/ 2016-03-10 15:33:52 ncopa: been working a lot with the current menu system, whats it switching to? 2016-03-10 15:39:50 Adran: midipix is working on something 2016-03-10 15:40:03 mmmmmmmmmm 2016-03-10 15:40:51 i think he needs some help with mapping what we currently have into his new system 2016-03-10 16:11:24 ncopa: would it be possible to update bind to -P4 2016-03-10 16:24:51 Hello, I have a couple of script ideas, one of which automates the build and running of a lxc at boot in a memory-install with config/scripts in the apkvol (~30line changes to init.d/lxc) and another that provides UnionFS/AUFS similar functionality with OverlayFS (~100line bash script). 2016-03-10 16:25:23 Where should I put these scripts I am working on for review with you guys? I was trying to use the wiki but the code tags break with scripts 2016-03-10 16:26:00 Just want a little feedback, the scripts may be usable on other dists but I want to focus them on Alpine and working with its memory install 2016-03-10 17:04:21 fabled: did you see my updated nldev patch? 2016-03-10 22:25:20 What is the process for a package to go from git's master branch (http://git.alpinelinux.org/cgit/aports/log/?h=master) to a -stable (let's say: http://git.alpinelinux.org/cgit/aports/log/?h=3.3-stable) 2016-03-10 22:26:06 process/time frame 2016-03-10 22:51:37 packages in stable only receive security updates and reliability fixes 2016-03-10 22:55:41 so with bind for example, -P4 came out (security related), and it's available in =master -- will it make it to -stable? 2016-03-10 22:55:54 (and if not -- what's the best way to utilize it under Docker) 2016-03-10 23:00:52 codex: actually the bind patches should have been backported, would you mind opening in issue in readmine for this? 2016-03-11 00:00:57 sure - can you point me in the right direction (semi-new to alpine) 2016-03-11 00:01:25 I recently migrated 100+ docker systems from Ubuntu -> Alpine and that's how I got into alpine 2016-03-11 00:02:07 Trying to catch up as quickly as I can on the methodologies and such 2016-03-11 03:36:11 Hello, is arm builder down? 2016-03-11 03:38:22 The reason I ask is a package was updated on Saturday, but not built. The x86 build happened same day though:http://git.alpinelinux.org/cgit/aports/commit/?id=fa39deeaf3eb163a4ee378ae67f933d24a96623a 2016-03-11 03:55:53 codex: packages in emaster will 2016-03-11 03:56:03 oops 2016-03-11 03:56:44 packages in git master will end up in Alpine 3.4 in a month or two, when it is released 2016-03-11 03:57:42 minor version bumps (security) should be backported immediately 2016-03-11 03:59:56 ubermacin: which builder? - edge, 3.3, 3.2, ... 2016-03-11 04:01:53 ACTION thinks arm edge is up, but the others might still be down 2016-03-11 04:02:12 ncopa ^^^ 2016-03-11 04:10:54 so what is the process for getting a package moved from testing over to main or community? 2016-03-11 04:23:08 Hey all, I was wondering about the build\development proccess of Alpline Linux. Specifically about the security part. 2016-03-11 04:23:08 I have seen the PIE and was wondering about the concept before trying to google so I would understand the intention of such a feature. 2016-03-11 05:04:23 tdtrask: testing 2016-03-11 07:32:54 arm builder for 3.3 is down yes 2016-03-11 07:33:02 i plan to drive to it today and fix it 2016-03-11 08:44:25 ncopa: I keep noticing this, http://i.imgur.com/zGNFpRF.jpg 2016-03-11 08:44:45 is bug in crond or ? 2016-03-11 09:41:19 I suggest adding http://xkcd.com/1654/ to algitbot on the "install" keyword 2016-03-11 09:50:00 hey there! 2016-03-11 11:23:04 any objections for pushing http://sprunge.us/KHaE ? 2016-03-11 11:23:39 the idea is to split c_rehash as standalone app to libcrypto1.0 so we keep the openssl package depndency out of some core dependency paths 2016-03-11 11:23:54 openssl binary is ~500kb and normally not required 2016-03-11 11:47:02 so now libcryto would have lib+bin 2016-03-11 11:47:16 in the pkg 2016-03-11 11:47:21 yes 2016-03-11 11:47:26 a small binary 2016-03-11 11:47:33 we have some libs like that already 2016-03-11 11:47:43 that need helper binaries to construct cache or similar 2016-03-11 11:47:48 yes I have noticed 2016-03-11 11:49:14 any freeing of unessessary deps sounds nice to me 2016-03-11 11:50:49 save as much space on ram based installs 2016-03-11 13:49:15 fabled: LGTM 2016-03-11 13:49:31 ncopa, the openssl change? 2016-03-11 13:50:01 yes 2016-03-11 13:52:45 build-3-3-armhf should be back online 2016-03-11 15:14:44 latex is still broken on alpine 2016-03-11 15:14:53 what a nightmare 2016-03-11 15:31:48 well it's not so much broken as never worked in the first place but we have trash in the repo 2016-03-11 15:41:50 uh? 2016-03-11 15:41:51 abuild-fetch: fcntl(F_SETLKW): Stale file handle 2016-03-11 15:41:54 what does it mean? 2016-03-11 16:16:17 nmeum: ACK for the nlmon and nldev stuff 2016-03-11 16:16:42 ncopa: you probably need to apply that patch yourself since I can't create snapshots 2016-03-11 16:16:51 or are snapshots automatically created by the builder? 2016-03-11 16:17:11 oh :-/ 2016-03-11 16:17:30 no 2016-03-11 16:17:33 its not 2016-03-11 17:18:02 ncopa: you could give me access to the archive if you don't want to apply those on your own either way it would be nice if those 2 patches could be merged ;) 2016-03-11 21:56:36 I'm starting to evaluate a replacement for RHEL as my virtualbox server platform. Alpine looks promising, but I am concerned that, over time, the community will become convinced that it, too, must move on to systemd. Systemd is one of the scariest new things to come along, if not THE scariest. It obscures and pretends, while not really providing anything better than openrc and similar tools. I am looking for longterm 2016-03-11 21:56:37 support that I don't have to worry about. 2016-03-11 21:57:23 I've been using systemd on my mageia for about a year now, and I can say with utter certainty that I am unimpressed, irritated, and constantly perplexed by it. 2016-03-11 21:58:26 My question is how committed is this community to assuring systemd never sees the light of day on alpine Linux, other than forcibly building/installing systemd on their own? 2016-03-11 21:58:48 I do not want to see or encounter this menace on any more of my systems. 2016-03-11 21:58:53 I hope that Alpine will NEVER adopt systemd. It’s totally against Alpine principles and it would be huge disappointment. 2016-03-11 21:59:06 THIS IS GOOD TO HEAR! 2016-03-11 21:59:24 However, this is question for @ncopa. I’m just a newbie Alpine user. 2016-03-11 22:00:06 thank you, I'll come back here in a bit; have to run. But I just found Alpine and wanted to ask. 2016-03-11 22:00:26 So far, everything I've read looks very nice. I'll be back in a few hours and I will look at people's responses. 2016-03-11 22:00:42 Thank you all for your work on Alpine. Keep it systemd-free always! 2016-03-11 22:01:33 Talking about OpenRC, there’s some initiative to replace OpenRC with someting even more lightweight, someday in future. 2016-03-11 22:09:50 bozonius: systemd will not support musl, alpine runs on musl, therefore alpine cannot use systemd 2016-03-11 22:10:10 ACTION does not see any of those things changing 2016-03-11 23:27:52 bozonius: as tdtrask said… even if we would like to use it, we can't 2016-03-11 23:28:29 and the will of majority of people here is another thing that would prevent such switch 2016-03-11 23:29:51 Alpine Linux – one of the last reasonable place in Linux ecosystem ^_^ 2016-03-11 23:52:47 bozonius: I don't think you have to fear systemd at all here. ncopa hates it almost as much as I do. 2016-03-11 23:52:59 (that means a lot) 2016-03-11 23:53:11 \(^_^)/ 2016-03-11 23:56:16 i didn't think openrc was that heavy 2016-03-11 23:56:18 skarnet: btw, during Christmas I’ve read most of information about s6 on http://skarnet.org/software/s6/… it’s indeed interesting, but I had some concerns… I’ll ask once I remember 2016-03-11 23:56:38 jirutka: sure 2016-03-11 23:57:03 Adran: OpenRC isn't so much "heavy" as "a giant hack" 2016-03-11 23:57:16 i didn't know that 2016-03-11 23:57:17 mmm 2016-03-11 23:57:31 I didn't either until I had to look in the code 2016-03-11 23:58:51 it’s like SysV init on steroids, but still very similar design; there are already better approches 2016-03-11 23:59:32 jirutka: the estimated time for the *beginning* of the work is towards the end of this year, something like Oct-Nov 2016-03-11 23:59:45 and there will be a whole lot of prep work 2016-03-12 00:00:39 I want to do things right, with no compromise, and this takes time 2016-03-12 00:01:33 one of things I’m miss in s6 are “soft” service dependencies like after/before in OpenRC 2016-03-12 00:02:16 (to what extent do we take the separation between mechanism and policy? what are our exact needs in terms of dependency management? what does apk provide? if we need something apk doesn't provide, that requires adding stuff to apk first.) 2016-03-12 00:03:18 jirutka: the equivalent of OpenRC is more s6-rc than s6. Is that what you're talking about? (Clarity is important.) 2016-03-12 00:03:18 I think that it’s very useful for practical init scripts; for example, I’d write runscript for app that uses database; this may be deployed on the same mechine, then the app should be automatically started _before_ the database, but if there’s no database (i.e. it uses remote one), then it should not fail 2016-03-12 00:03:32 eh, sorry, yeah, I mean s6-rc 2016-03-12 00:04:17 you mean after the database? 2016-03-12 00:04:32 eh, yeah 2016-03-12 00:05:31 do you switch from a local to a remote db often? 2016-03-12 00:05:40 (or the other way round) 2016-03-12 00:05:51 and even if the database is deployed on the same system, then I don’t want to stop the app when I stop the database (e.g. for short maintance), because database can live few seconds without DB, there’s no need to stop it 2016-03-12 00:06:11 in this precise case, you wouldn't use s6-rc to stop the db 2016-03-12 00:06:23 you would just stop it with s6-svc 2016-03-12 00:06:32 that’s not my point… the point is runscript distributed with distribution 2016-03-12 00:06:59 I am answering to your points :) 2016-03-12 00:07:19 when I write runscript myself, then it’s not a problem, but when I just use runscript from some package, then it is 2016-03-12 00:07:22 about short maintenance: you don't have to change the machine state, you can just temporarily stop individual services without s6-rc being aware 2016-03-12 00:07:47 aha 2016-03-12 00:08:23 about apps that shouldn't fail if there's no database : that's something we'll have to think about, but there are several solutions 2016-03-12 00:08:59 I think those cases are rare enough to not justify a special case in the service manager 2016-03-12 00:09:25 i disagree, it used on every Gentoo system… for net 2016-03-12 00:09:33 it’s 2016-03-12 00:10:27 hm, no, this isn’t the best example 2016-03-12 00:12:51 I’ll think about it later and summarize some use cases 2016-03-12 00:13:11 sure, I'm very interested in this 2016-03-12 00:13:23 another concern: I’ve looked at source code and if I remember it correctly, there are no tests 2016-03-12 00:13:37 there are no *automated* tests 2016-03-12 00:13:59 believe me, I've tested the shit out of this. That's why it took months to release. >.> 2016-03-12 00:14:01 this is like the same 2016-03-12 00:14:06 I know. 2016-03-12 00:14:16 tests are essential 2016-03-12 00:14:20 If you want to contribute test, I shall be eternally grateful. :) 2016-03-12 00:14:21 I mean automated tests 2016-03-12 00:14:23 tests* 2016-03-12 00:14:27 great :) 2016-03-12 00:14:39 There are important restrictions though. 2016-03-12 00:14:51 I don't want to have to install the kitchen sink to run the test suite. 2016-03-12 00:15:04 I should be able to run it on the exact same machine as the software runs on. 2016-03-12 00:15:17 That means, sometimes, an embedded i486 with 32 MB of RAM. 2016-03-12 00:16:01 I think that it’s really not necessary to run automated tests on machine that no one actually use 2016-03-12 00:16:24 This would still be my router if it did gigabit ethernet. 2016-03-12 00:16:29 I changed it last year. 2016-03-12 00:16:40 I still have it. 2016-03-12 00:16:56 great if s6 runs even on such an old machine, but it’d be *much* better to have tests that run only on some HW with kitchen sink around than no automated tests at all 2016-03-12 00:19:20 then it can be a separate package. Again, I'm not opposed to tests, at all. The only reasons I don't write them is 1. it bores me to tears, 2. I haven't found a test suite that I like. 2016-03-12 00:19:33 I understand 2016-03-12 00:20:08 I’m glad that you’re not against tests… I was a little afraid that you’re one of guys who thinks that they don’t need tests :) 2016-03-12 00:22:04 to be perfectly honest, I do believe that in the back of my mind - because I do a lot of manual tests before releasing. And working without a safety net forces me to pay very close attention to the code, which is the best quality guarantee you can get. However, this does not scale with the amount of code I write. 2016-03-12 00:22:23 and last year I have let slip an embarrassing number of bugs. 2016-03-12 00:22:41 > this does not scale with the amount of code I write. 2016-03-12 00:22:41 exactly, and it also doesn’t scale with more people in the project 2016-03-12 00:22:45 Because as the amount of code grows, the amount of interactions grow quadratically. 2016-03-12 00:23:19 nothing scales with more people in the project. Nothing. 2016-03-12 00:23:43 the only good reason to have more people in a project is to increase the bus factor. 2016-03-12 00:23:52 one person can’t do anything 2016-03-12 00:24:04 not a video player, and not a web browser, I agree. 2016-03-12 00:24:12 But system stuff? definitely yes. 2016-03-12 00:24:37 Look at musl. They are basically one and a half. 2016-03-12 00:24:53 And a libc is a completely insane project. 2016-03-12 00:24:58 well, but in long term it’s extremly risky to use software that only one person is able to maintain 2016-03-12 00:25:12 that's what I'm saying. The important thing is to have a large enough bus factor. 2016-03-12 00:27:45 and honestly, the bus factor of Alpine isn't very high. 2016-03-12 00:28:00 I'd be more concerned with it atm. :) 2016-03-12 00:28:56 since it became the preffered distro for official Docker images, I think that it’s gonna gain larger bus factor 2016-03-12 00:30:05 yes, I think it will. I only hope it won't come with the drawbacks of management by committee. 2016-03-12 00:30:49 and that Alpine won't become Docker's puppet, at some people already fear. 2016-03-12 00:30:50 btw I found it interesting that scan directory reminds me sysfs and procfs… and I bet that it’s not a coincidence… inspiration from Plan9? :) 2016-03-12 00:31:09 yeah, I have exactly the same concerns 2016-03-12 00:31:36 not only because I don’t like Docker at all 2016-03-12 00:32:01 you'd have to ask djb. It's possible he gained inspiration from Plan 9 - but he never said so afaik. The scandir design is his. 2016-03-12 00:32:14 aha 2016-03-12 00:41:41 skarnet: it’s late here, I go to sleep; good night 2016-03-12 00:42:02 it's late here too. gn! 2016-03-12 00:42:37 ah, you’re from France :) 2016-03-12 00:42:42 yes :) 2016-03-12 00:43:06 so GMT+1 as here 2016-03-12 00:43:14 where are you located? 2016-03-12 00:43:20 Czech Republic 2016-03-12 00:43:31 indeed, GMT+1 too :D 2016-03-12 02:33:10 thank you all for your comments; I do appreciate them. SO RELIEVED TO HEAR THIS GOOD NEWS. I will commence a VM test of Alpine immediately. Once I am familiar with it, and feel comfortable deploying it as my "main" (server) platform, I will do so. But there is one last concern I have. 2016-03-12 02:33:39 How does VirtualBox play on Alpine? That could be an issue, as I have some investment in VB VM's 2016-03-12 02:33:57 anyone have experience with VBox on Alpine? (Thanks, tia) 2016-03-12 02:38:38 Also, about how much memory and disk space would I need for Alpine itself (I have a good idea what I need for the VMs outside of Alpine itself) 2016-03-12 02:39:16 I'm going to try 1GB memory and 8GB hard disk to start. This will only be a test environment, so I can change it as needed. 2016-03-12 02:45:04 (I guess I should have asked, also, how does VB play INSIDE of Alpine? I'll be doing both -- Alpine as VBox client and as VBox server) 2016-03-12 03:16:38 sorry, got knocked off by a reboot 2016-03-12 03:17:07 (I was asking about VirtualBox and hardware requirements for Alpine itself) 2016-03-12 03:45:44 OK, I think I already see why there is not much response to my inquiry about VBox. Heheheh 2016-03-12 03:45:47 yeah 2016-03-12 03:46:14 Alpine runs fine under VBox, from what I can tell so far. 2016-03-12 03:46:37 That is fine for testing purposes. But getting it to host virtualbox server will be a big complicated mess I think. 2016-03-12 03:47:07 So the obvious response will be something like why not use xen or another "built-in" kernel virtualizer? 2016-03-12 03:47:32 And my response will be that I've already got quite a bit of investment in VMs. It would be hard to migrate them to other platforms. 2016-03-12 03:47:58 Trust me. It is not as simple as just copying file systems, etc. There are a number of other considerations. 2016-03-12 03:49:40 I do appreciate everyone's time, though. And I'm relieved that at least one Linux community hasn't lost its marbles! 2016-03-12 03:49:47 thanks. 2016-03-12 07:51:58 bozonius: I don't think Virtual Box runs on Alpine. When you get the time, consider moving away from VB and onto qemu. Whatever the host platform, you'll be glad you did. 2016-03-12 07:54:30 Years ago, I DID use qemu. Back then, though, it seemed under-featured. I haven't tried it since then. This was probably over 10 years ago, because I've been using VBox for about the last 10 years 2016-03-12 08:02:23 bozonius: qemu is actually good now. :) Also, the main reason why you're not getting answers is probably that most Alpine people are EU-based, so you'll have more answers during EU daytime. 2016-03-12 08:27:15 no problems, skarnet. And ty for all the feedback! 2016-03-12 08:27:43 Is X11 or Wayland or the like available for Alpine? (forgive me if this is enumerated elsewhere) 2016-03-12 08:27:57 or is Alpine purely command line? 2016-03-12 08:28:09 Xorg works 2016-03-12 08:28:32 is it in a repo for alpine so I can just install it, or does it require a full build? 2016-03-12 08:29:11 bozonius: https://pkgs.alpinelinux.org/packages?name=xorg%25&repo=all&arch=x86_64&maintainer=all 2016-03-12 08:29:26 bozonius: this should help you answer other questions you have too :) 2016-03-12 08:29:38 ok, I'll go take a look at the offerings 2016-03-12 08:33:06 I see. So there is already quite a bit of software available to install. Good, and thanks again. 2016-03-12 08:33:40 it's day right now in, like, australia 2016-03-12 08:33:54 okay and most of asia 2016-03-12 09:38:17 bozonius: qemu+kvm is really much better than VB, alternative is lxc 2016-03-12 09:38:51 if you are the sole local user and don't need different kernel for the guest, I would suggest lxc 2016-03-12 09:53:42 thanks xming. I will look at those. (I'm still installing packages in my test vm) 2016-03-12 09:53:43 I notice that most times, the package manager won't let me give multiple arguments. 2016-03-12 09:53:43 So I can only install one package at a time, otherwise I get an error about "world[world]" being a dependency? 2016-03-12 09:53:56 this is a little confusing for this beginner... 2016-03-12 10:05:45 I don't have issues with that, care to paste your line? 2016-03-12 10:23:59 I may have to wait until I have another package to install 2016-03-12 10:24:38 but I have a more pressing concern at the moment. I installed xorg-server, awesome, and xfce4 but now it just hangs with a blank screen 2016-03-12 10:25:03 (startx had complained that xfce4 was missing, so I installed it) 2016-03-12 10:26:40 I'm rebooting it (hard reset) 2016-03-12 10:31:25 nvm. I failed to follow some of the directions... 2016-03-12 10:35:57 now I start to get a desktop, but I cannot get the mouse to respond 2016-03-12 17:06:46 skarnet: on our conversation about openrc, wouldn't it be better to maintain one init system and allow it to become more usable vs splitting off? Just so there is some singular hope of a (really) good alterative to systemd? 2016-03-12 18:07:49 while installing a package, I get "no space left on device." But when I look using df, there appears to be no file systems outof space 2016-03-12 18:08:02 the package was the llvm dependency for the video driver 2016-03-12 18:08:48 The VM has plenty of disk and memory (I started with 8GB disk and 1GB RAM) 2016-03-12 18:08:55 So I am puzzled. 2016-03-12 18:10:02 How big is llvm-libs? 2016-03-12 18:11:24 btw, am I in the correct alpine forum for these questions? 2016-03-12 18:11:40 this is the development channel 2016-03-12 18:12:15 right. Where SHOULD I be? 2016-03-12 18:12:27 #alpine-linux 2016-03-12 18:12:31 ah! 2016-03-12 18:12:33 thanks 2016-03-12 19:43:00 Adran: a good alternative to systemd is exactly what s6+s6-rc is :) 2016-03-12 19:43:31 skarnet: my thoughts though are why split again instead of working on improving what already exists? 2016-03-12 20:03:42 Adran: because there's no real improvement to make to OpenRC apart from changing the design entirely - which would be the exact same as writing something else. OpenRC works well for what it does, but it's limited by its design. 2016-03-13 14:36:43 hey :) 2016-03-13 15:20:49 clandmeter: ping 2016-03-13 15:32:22 clandmeter: your package htop is out of date. its on 2.0.0 but 2.0.1 is current 2016-03-13 19:38:52 ncopa: hey, just noticed that you updated firefox in testing/ would you mind also updating the version in community/? 2016-03-13 19:39:17 I would help out with the firefox package but appearently it takes hours to compile firefox on my hardware 2016-03-13 22:44:36 ncopa: if you have a minute, trying to use the iso builder and its not finding init 2016-03-14 05:13:56 ncopa: mkinitfs when used for the livecd does something special doesn't it? I added some extra files to it, but they don't show up once the livecd is booted. 2016-03-14 09:07:43 morning 2016-03-14 09:09:04 fabled: you about? just wondering if the raspi kernel includs the adafruit stuff (drivers for their hardware, like touchscreens etc) 2016-03-14 09:09:56 ScrumpyJack, i have tried to enable most of it, but it's possible some is missing. i personally use adafruit18 tft screen so that works for sure 2016-03-14 11:00:07 hey :) 2016-03-14 11:36:09 leo-unglaub: hey 2016-03-14 11:50:24 anyone familiar with strftime() ? I'm trying to format time in hours in decimal but not 24h clock so that i can show 50 hours for example 2016-03-14 11:51:49 afaik strftime won't allow you to do that 2016-03-14 11:51:54 :( 2016-03-14 11:52:14 what do you have, a number of seconds? 2016-03-14 11:52:44 if it's the case, it's pretty easy to take t%60, (t/60)%60 and t/3600 2016-03-14 11:53:03 thanks algitbot, but go back to sleep now 2016-03-14 11:53:54 skarnet: yeah, seconds 2016-03-14 11:55:06 honestly I'd just do printf("%u:%u:%u", t/3600, (t/60)%60, t%60) 2016-03-14 11:55:16 ACTION shoots algitbot 2016-03-14 11:55:57 i don't need seconds after the minute ;) 2016-03-14 11:56:20 left as an exercise for the reader. :P 2016-03-14 12:25:02 ncopa: which of the following two solution do you prefer for passing additional command line arguments to patch() in abuilds prepare function? http://sprunge.us/GNge or http://sprunge.us/aNHi ? 2016-03-14 12:42:42 Something in contrary to add redis support to syslog-ng? 2016-03-14 12:43:42 http://sprunge.us/hZHe 2016-03-14 12:43:50 fixes #5260 2016-03-14 13:12:39 fcolista: rsyslog has redis support to soo just go ahead and add it to syslog-ng as well 2016-03-14 13:17:40 *too 2016-03-14 13:18:18 nmeum: "${@:--p1}" surely? since otherwise it'll get wordsplit twice 2016-03-14 13:21:14 I think it might be more readable to just use "$@" and have a line earlier something like [ ${1+1} ] || set -- -p1 though 2016-03-14 13:25:31 Because "[ ${1+1} ] || set -- -p1" is perfectly readable :D 2016-03-14 13:26:19 ammunta: indeed if ncopa prefers that approach I will need to polish it a bit 2016-03-14 13:26:57 http://sprunge.us/FMRN is what I was thinking 2016-03-14 13:26:59 skarnet: well 2016-03-14 13:27:56 I feel like [ ${var+1} ] || is some kind of an idiom right ... 2016-03-14 13:28:39 it's more that I prefer to up-front setting default parameters in this case, though 2016-03-14 13:34:37 but that approach would only allow you to change the value of the -p flag, you couldn't pass additional flags since you qouted $@ 2016-03-14 13:34:58 ? it's $@ 2016-03-14 13:35:13 I think 2016-03-14 13:35:32 this is the sort of thing that makes people hate shell, should probably test that 2016-03-14 13:36:38 ^ 2016-03-14 13:38:03 if I run patch "-p1 -c" for example patch interprets '1 -c' as an argument to the -p flag instead of interpreting -c as a flag 2016-03-14 13:39:39 ~ $ f() { printf '%s\n' "${@:-hi world}"; }; f; f 1 2 2016-03-14 13:39:45 seems to have the expect result 2016-03-14 13:40:02 ("hi world\n1\n2\n") 2016-03-14 13:41:18 that's a different thing my point is that you can't pass more than one flag to patch if you qoute $@ 2016-03-14 13:41:43 that's the same thing 2016-03-14 13:41:49 it's getting passed in as two arguments 2016-03-14 13:44:48 ammunta: no, it's not. try to do the same thing with a command line flag that expects an argument like -p in patch(1) if you qoute the flag everything in the qoutes will be interpreted as an argument to the -p flag and thus you can't pass additional flags within that qoutes 2016-03-14 13:46:47 uhh maybe I am wrong 2016-03-14 13:47:22 that would be exceptionally strange 2016-03-14 13:47:36 like, even for sh 2016-03-14 13:47:59 (standards, yeah, I know) 2016-03-14 13:53:01 yeah sorry I was wrong 2016-03-14 13:56:05 http://pubs.opengroup.org/onlinepubs/9699919799/utilities/V3_chap02.html#tag_18_05_02 2016-03-14 13:56:08 > When the expansion occurs within double-quotes, and where field splitting (see Field Splitting) is performed, each positional parameter shall expand as a separate field 2016-03-14 13:58:48 I avoid thinking about it too hard 2016-03-14 14:01:03 nmeum: i think i prefer http://sprunge.us/GNge 2016-03-14 14:02:02 otherwise you have to call the function to override the args 2016-03-14 14:03:47 yep 2016-03-14 14:15:12 fcolista: https://github.com/getpatchwork/patchwork/releases/tag/v1.1.0 2016-03-14 14:25:13 ncopa: would you mind if I push this with a slight modification? http://sprunge.us/XQGJ Tested it with the pulseaudio aport, seems to work as expected 2016-03-14 14:25:40 clandmeter, i'm working on patch set 2016-03-14 14:26:18 fcolista: of? 2016-03-14 14:26:26 patchwork 2016-03-14 14:26:27 nmeum: LGTM 2016-03-14 14:26:36 push it to abuild repo 2016-03-14 14:26:40 fcolista: nice 2016-03-14 14:26:48 nmeum: thanks! 2016-03-14 14:26:50 ncopa: alright 2016-03-14 14:33:49 nmeum: what is your use case for this patch? 2016-03-14 14:35:04 looking at pulseaudio, you want to set a different -p level? 2016-03-14 14:35:31 clandmeter: some packages have patchset which require a different -p option readline pulseaudio or bash for example 2016-03-14 14:35:34 yes 2016-03-14 14:36:26 what if you have mixing -p switches? 2016-03-14 14:37:05 then you need to manually overwrite prepare() 2016-03-14 14:38:39 what if you can set a different p level per patch, like we do for saveas? 2016-03-14 14:39:04 i think its not worth it 2016-03-14 15:20:01 I'd rather unify patches to apply than write own prepare() 2016-03-14 15:21:34 oh boy 2016-03-14 15:21:43 dbus is so damn bad 2016-03-14 15:22:11 just for fun, run dbus-monitor and then start thunar 2016-03-14 15:22:25 you will be overwelmed with errors 2016-03-14 15:49:32 ncopa: what do you think about shipping opensshd with some more secure defaults 2016-03-14 15:49:58 something like: https://paste.debian.net/plain/415161 2016-03-14 15:50:09 that would make alpine even more secure by default 2016-03-14 15:50:17 sounds good to me 2016-03-14 15:50:41 unless it means that ppl are getting locked out from their boxes when they upgrade 2016-03-14 15:51:01 hmm, that is a good point 2016-03-14 15:51:31 if someone actually would use a dsa setup and he has not modified the config file apk would overwrite the sshd_config without asking and lock him out yes 2016-03-14 15:51:45 but on the other hand, he propobly deserves to be locked out in this case *g* 2016-03-14 15:51:46 i know for sure that rsa and ed25519 will not lock myself out 2016-03-14 15:51:54 dss is already obsolete 2016-03-14 15:52:09 i am already locked out from the boxes which uses dss 2016-03-14 15:52:12 dsa 2016-03-14 15:52:19 hehe, but you are right 2016-03-14 15:52:26 we should not piss of people by locking them out 2016-03-14 15:52:37 but i am a big fan of security out of the box 2016-03-14 15:52:43 i agree 2016-03-14 15:52:59 because the more config files i have to edit in order to be secure .. the more likley i am to fotgett one 2016-03-14 15:53:06 i wonder if we should check current config 2016-03-14 15:53:11 in pre-upgrade or simlar 2016-03-14 15:53:45 if there are weak hostkeys already, then echo warning 2016-03-14 15:53:49 hmmm, risky .. because would have to grep/sed the config because we cannot use the sshd binary to get the running config like in postfix or similar 2016-03-14 15:54:32 if lockout is 'unlikely' then i'm okish with it 2016-03-14 15:55:08 if we can use some tricks to check if there are a running config, and prevent that ppl are locked out, then that would be great 2016-03-14 15:55:20 and for people doing new installs, use secure default 2016-03-14 15:55:48 if user have changed the sshd_config then apk will not replace it 2016-03-14 15:55:58 but if they use the default, then it will be replaced 2016-03-14 16:28:41 ncopa: poke 2016-03-14 16:33:15 Adran: hi 2016-03-14 16:33:27 ncopa: see my messages about the iso builder? 2016-03-14 16:34:06 its probably confusion on my part but was hoping to have your input 2016-03-14 16:34:44 basically I can add files to the initfs but when the livecd boots those files aren't in their proper locations - or anywhere. but if you extract the initfs they're there. 2016-03-14 16:38:51 initfs is just a temp filesystem 2016-03-14 16:39:19 it will set up the rootfs and thwn switch_root to it 2016-03-14 16:39:36 what kind of files are you trying to ship with your iso? 2016-03-14 16:39:58 scripts? binaries? data files (images etc)? config files? 2016-03-14 16:40:43 I have some scripts and configs 2016-03-14 16:40:58 Ideally I want to load two packages in addition on the livecd + these scripts 2016-03-14 16:41:13 if its config then it probably belongs to an apkovl 2016-03-14 16:41:27 And when I use apkvol 2016-03-14 16:41:31 on a livecd, it looses its ability to do installs 2016-03-14 16:41:47 I have the apkovl available if you wish 2016-03-14 16:41:59 ok 2016-03-14 16:42:05 yes i can have a look at it 2016-03-14 16:43:39 you probably only need to add those services to your apkovl: http://git.alpinelinux.org/cgit/mkinitfs/tree/initramfs-init.in#n453 2016-03-14 16:44:30 things are also missing, like keyboard stuff, etc 2016-03-14 16:44:35 timezone 2016-03-14 16:47:09 what happens if you include the following symlinks in your apkovl: 2016-03-14 16:47:35 ln -s /etc/init.d/devfs /etc/runlevels/sysinit/devfs 2016-03-14 16:47:46 ln -s /etc/init.d/dmesg /etc/runlevels/sysinit/dmesg 2016-03-14 16:47:59 ln -s /etc/init.d/mdev /etc/runlevels/sysinit/mdev 2016-03-14 16:48:04 ln -s /etc/init.d/mdev /etc/runlevels/sysinit/mdev 2016-03-14 16:48:18 ln -s /etc/init.d/hwdrivers /etc/runlevels/sysinit/hwdrivers 2016-03-14 16:48:28 ln -s /etc/init.d/modloop /etc/runlevels/sysinit/modloop 2016-03-14 16:48:39 and the rest listed here: http://git.alpinelinux.org/cgit/mkinitfs/tree/initramfs-init.in#n453 2016-03-14 16:48:59 let me see 2016-03-14 16:49:01 oh 2016-03-14 16:49:07 you know what 2016-03-14 16:49:29 i think you can simly do: touch /etc/.default_boot_services 2016-03-14 16:49:41 and include an empty file like that in your apkovl 2016-03-14 16:50:07 if [ -f "$sysroot/etc/.default_boot_services" -o ! -f "$ovl" ]; 2016-03-14 16:50:09 won't that not trigger still? 2016-03-14 16:50:21 it should trigger 2016-03-14 16:50:45 if .default_boot_services exists *or* apkovl exist, then enable the dfault boot services 2016-03-14 16:50:48 oh -o is or not and 2016-03-14 16:50:50 oops 2016-03-14 16:51:05 alright 2016-03-14 16:51:06 yes, that should do it 2016-03-14 16:51:21 takes about 30 minutes to rebuild since it redownloads each time. :P 2016-03-14 16:51:28 wil let you know 2016-03-14 16:51:48 I made sure not to touch any config files that are generated on boot/automatically, so the system was quite broken 2016-03-14 16:51:57 you can set up a local reverse http proxy and use that as http://mirror 2016-03-14 16:52:21 you can probably simply enable apk cache 2016-03-14 16:52:24 yea, i mean i'm fine downloading each time 2016-03-14 16:52:33 its just slow :D so whether this works or not will take a few 2016-03-14 16:52:35 30mins is not fine... 2016-03-14 16:53:10 ln -s /var/cache/apk /etc/apk/cache 2016-03-14 16:53:14 and it should cache it locally 2016-03-14 16:53:47 unless its the local disk that is slow ofc :) 2016-03-14 16:54:47 its ssd 2016-03-14 16:54:49 so hopefully not 2016-03-14 16:54:54 lol 2016-03-14 16:55:02 also while your'e here 2016-03-14 16:55:08 how does X actually work on alpine. 2016-03-14 16:55:36 how do you mean? 2016-03-14 16:55:51 what differs alpine from other distros? 2016-03-14 16:56:07 otherwise, it works over unix socket like other distros :) 2016-03-14 16:56:09 I can move my mouse and intel works. ;) 2016-03-14 16:56:17 intel graphics 2016-03-14 16:56:20 xserver listens to a unix socket 2016-03-14 16:56:29 and xclients connects to the x server 2016-03-14 16:56:33 thats how it works :) 2016-03-14 16:57:03 Right, that part mostly functions, 2016-03-14 16:57:13 what does not work? 2016-03-14 16:57:21 hotplugging? 2016-03-14 16:57:24 but if I startx lxdm, I get a mouse in the center of the screen, cannot switch ttys, nothing moves 2016-03-14 16:57:34 and it doesn't finish loading (xfce guide on wiki) 2016-03-14 16:57:51 I've tried this in Virtualbox, VMware, and a Broadwell based computer. I'm really hoping its something I did (which it probably is) 2016-03-14 16:58:25 did you install the input driver for evdev or what its called? 2016-03-14 16:58:34 sounds like input driver is not working properly 2016-03-14 16:58:47 can you log in via ssh and look at your Xorg.log? 2016-03-14 16:58:53 uhh 2016-03-14 16:59:02 i didn't install ssh, let me redo it 2016-03-14 16:59:15 the wiki is probably a bit old 2016-03-14 17:07:11 ncopa: also btw, how would you add packages automatically to a live boot? 2016-03-14 17:07:19 its it with the apkovl? (even without internet?) 2016-03-14 17:16:07 the package needs to be in the .boot_repository 2016-03-14 17:16:14 normally in apks/ on boot media 2016-03-14 17:16:29 so you need to add the packages you want to the iso 2016-03-14 17:16:32 sure, but one its there, how does it get automatically setup? 2016-03-14 17:17:09 it will install the packages listed in /etc/apk/world that is shipped in the apkovl 2016-03-14 17:17:24 ok 2016-03-14 17:20:01 also 2016-03-14 17:20:12 you may want include your public key in the apkovl 2016-03-14 17:20:21 in /etc/apk/keys/ 2016-03-14 17:21:01 those appear to get automatically pulled in 2016-03-14 17:21:24 but i'll make sure 2016-03-14 17:23:02 ncopa: here's xorg attemping to start with lxdm https://gist.github.com/Ttech/8e106ca0bea36e00408c 2016-03-14 17:24:20 ncopa: bottom is Xorg -configure top is lxdm http://imgur.com/a/m6Rpj 2016-03-14 17:24:51 yes, Xorg -configure does not work i think 2016-03-14 17:25:05 due to no support of lazy loading of the modules 2016-03-14 17:25:39 so what do I need to do to get this to work? 2016-03-14 17:25:44 is udev running? 2016-03-14 17:26:03 Assertion failed: key->initialized (../include/privates.h: dixGetPrivateAddr: 122) XIO: fatal IO error 11 (Resource temporarily unavailable) on X server ":0" after 974 requests (974 known processed) with 2 events remaining. 2016-03-14 17:26:04 xinit: connection to X server lost 2016-03-14 17:26:07 2606 root 0:00 /sbin/udevd --daemon 2016-03-14 17:26:32 ok 2016-03-14 17:26:34 good 2016-03-14 17:26:45 udev is needed for hotplugging 2016-03-14 17:26:58 ncopa: here's my world for reference: https://gist.github.com/Ttech/ec9eb824bd521684f48e 2016-03-14 17:27:54 what alpine is it? 2016-03-14 17:27:58 version 2016-03-14 17:28:00 3.3? 2016-03-14 17:28:05 i think we use eudev nowdays 2016-03-14 17:30:02 3.3 2016-03-14 17:30:15 this is a clean install, wasn't sure if I should use "stable" or edge, but neither work any more then this 2016-03-14 17:34:43 ncopa: still getting that assertion error interstingly 2016-03-14 17:35:44 can you strace it? 2016-03-14 17:35:58 might be you are missing some permissions in /dev/ 2016-03-14 17:36:20 or that it wants access /sys 2016-03-14 17:36:45 https://gist.github.com/Ttech/2565d5838bfa45626f43 2016-03-14 17:38:08 thats just for xinit 2016-03-14 17:38:12 it forks X 2016-03-14 17:39:43 strace on what then? 2016-03-14 17:39:48 strace -f 2016-03-14 17:39:54 ok will do 2016-03-14 17:39:55 but i dont have time to help you analyze it :-( 2016-03-14 17:40:39 yea understandable 2016-03-14 17:40:43 my hope is I can get something working 2016-03-14 17:40:47 and make a better guide for the wiki 2016-03-14 17:40:49 and pull down the old stuff 2016-03-14 17:41:27 woudl be great 2016-03-14 17:41:42 maight be i can spin up a virtualbox later tonight 2016-03-14 17:41:45 and try myself 2016-03-14 18:31:13 ncopa, something in contrary in applying this: 2016-03-14 18:31:14 https://aur.archlinux.org/cgit/aur.git/tree/split_pane.patch?h=thunar-extended 2016-03-14 18:31:15 ? 2016-03-14 18:33:49 This patch was never applied in the upstream...dunno why since according with: https://bugzilla.xfce.org/show_bug.cgi?id=10232 this caused not issue... 2016-03-14 20:47:11 hm 2016-03-14 20:47:22 trhmmmmmmm 2016-03-14 20:47:36 fcolista: i'm slightly sceptic to that 2016-03-14 20:47:47 if upstream would have taken it, then no problem 2016-03-14 21:04:23 11:51 <@ncopa> i know for sure that rsa and ed25519 will not lock myself out 2016-03-14 21:04:30 plausible that someone might use ecdsa though 2016-03-14 21:04:42 who could get locked out 2016-03-15 07:41:23 clandmeter, patches in patchwork are specific for patchwork.alpinelinux.org 2016-03-15 07:41:36 does it make sense to have them in official repo? 2016-03-15 07:42:01 Or we can manage to apply them after, by ourselves? 2016-03-15 07:42:34 so far they are in official repo..but i dunno if is totally correct 2016-03-15 08:43:13 got ZenFone_2_ZE551ML, seems nice (intel atom based) 2016-03-15 09:02:07 it would be so cool to have a nice thin mips kernel 2016-03-15 12:15:46 good morning ! 2016-03-15 12:15:54 well, not really morning anymore 2016-03-15 12:16:06 depends which timezone i guess :) 2016-03-15 12:16:18 vienna ... 2016-03-15 12:16:43 for our american friends its pretty early 2016-03-15 12:16:58 i volunteered all night at the red cross 2016-03-15 12:17:12 and just woke up 2016-03-15 12:21:57 did you volunteered to install 2016-03-15 12:22:10 Alpine Linux onn all their servers? :p 2016-03-15 12:27:14 xming: https://twitter.com/LeoUnglaub/status/643182531588792320 2016-03-15 12:27:32 but last night i worked as a paramedic 2016-03-15 12:27:53 leo-unglaub: really cool 2016-03-15 12:28:26 anyone know how to easily modify a nested commit in a single PR for github? 2016-03-15 12:28:35 leo-unglaub: cool, retweeted 2016-03-15 12:31:06 clandmeter: what's a nested commit? git rebase -i should work... 2016-03-15 12:31:59 mutliple commits in a single pr, i need to modify one of them. 2016-03-15 12:32:27 a normal workflow would be to just add a commit, but thats not the case in a pr. 2016-03-15 12:33:52 checkout the change use git rebase -i and select the commit you want to modify 2016-03-15 12:33:55 *changes 2016-03-15 12:35:27 you can fetch it locally 2016-03-15 12:46:32 hi 2016-03-15 12:46:39 I have question about writing abuild - how I can declare dependency on any version of openjdk-jre, i.e. currently openjdk7-jre or openjdk8-jre? 2016-03-15 12:50:56 with similar file perms and symlinks I can access both symlinks via root login 2016-03-15 12:51:00 http://i.imgur.com/MMhiM4E.jpg 2016-03-15 12:51:13 but not on busybox ? 2016-03-15 12:51:28 I mean busyboxy+AL 2016-03-15 12:51:47 is this ok ? or a busyboxy bug ? 2016-03-15 12:58:08 nmeum: thx, first time for rebase -i for me. 2016-03-15 13:04:20 i'm working on llvm/clang/lldb 2016-03-15 13:04:40 currently we build all the tools in the llvm source tree 2016-03-15 13:04:48 and split out the packages 2016-03-15 13:05:00 i wonder if we should split it to separate apkbuilds 2016-03-15 13:05:31 looks like fedora builds clang separately 2016-03-15 13:05:42 @ncopa, could you please answer my question about abuild dependencies? 2016-03-15 13:06:04 jirutka1: i didnt see it sorry 2016-03-15 13:06:09 what was the question again? 2016-03-15 13:06:15 ncopa: I have question about writing abuild - how I can declare dependency on any version of openjdk-jre, i.e. currently openjdk7-jre or openjdk8-jre? 2016-03-15 13:06:37 i think that does not work currently 2016-03-15 13:07:05 i suppose we could add provides=openjdk-jre to them both 2016-03-15 13:07:15 and then you could add depends=openjdk-jre 2016-03-15 13:07:21 but currently i dont think its possible 2016-03-15 13:07:27 apk supports provides already? 2016-03-15 13:07:32 yes 2016-03-15 13:07:48 great! so I can just send patches to openjdk abuilds? 2016-03-15 13:07:54 ncopa: ^^^ http://i.imgur.com/MMhiM4E.jpg 2016-03-15 13:07:54 but i think you could just require openjdk8 now 2016-03-15 13:08:13 openjdk7 is no longer maintained upstream i think 2016-03-15 13:08:18 that’s true 2016-03-15 13:08:23 but we need it for bootstrapping openjdk8 2016-03-15 13:08:36 thats the only reason we keep it around 2016-03-15 13:08:42 but maven abuild still depends on openjdk7 2016-03-15 13:09:06 uh, I have a meeting, sy later 2016-03-15 13:09:11 k see u 2016-03-15 13:09:20 vkris: i dont know 2016-03-15 14:07:17 hi I have been trying to get alpine-linux on multiple machines to run xwindows in a vmware vm on both fusion and esxi with no luck ( I get a white underscore cursor in top left corner that doesn't blink multiple people have tried in vain, I have managed to get it working with xfce4 on this desktop that I am irc'ing from 2016-03-15 14:08:03 no virtualization involved in this desktop 2016-03-15 14:08:57 I have even tried mirroring the packages from this desktop to a vm and adding the open-vm-tools and xf86 packages for video and mouse same issue 2016-03-15 14:19:38 transhuman: alpine run fine in xen as a pv and hv guest. it also runs in vmware as hv guest, although i've not tried X in vmware. 2016-03-15 14:20:31 can you get alpine up and running without X on esxi? 2016-03-15 14:20:42 yes no problem at all 2016-03-15 14:20:52 it's just X right? 2016-03-15 14:20:59 same problem with fusion...just x 2016-03-15 14:22:43 crap, i don't know much about esxi graphics hw drivers. you've added xf86-video-vmware and loaded vmware_drv ? 2016-03-15 14:24:02 how do you load vmware_drv exactly? is that an exact module name? 2016-03-15 14:24:43 did load xf86-video-vmware and xf86-vmware-input or whatever its called 2016-03-15 14:25:00 modprobe vmware_drv 2016-03-15 14:25:59 how are you starting X? 2016-03-15 14:26:56 xf86-video-vmware xf86-video-vmware-doc xf86-input-vmmouse-doc xf86-input-vmmouse 2016-03-15 14:27:04 startx or startxfce4 2016-03-15 14:27:40 ok hold on let me try that modprobe command didnt try that 2016-03-15 14:29:54 as root echo "vmware_drv" >> /etc/modules 2016-03-15 14:30:12 (for persistance across reboots) 2016-03-15 14:30:37 once your module is loaded, run Xorg -retro and describe what you see 2016-03-15 14:32:26 module not found kernel headers are installed 2016-03-15 14:33:51 maybe it's an Xorg module. add it to /etc/X11/xorg.conf.d/20-modules.conf 2016-03-15 14:34:06 ok 2016-03-15 14:40:36 any joy? 2016-03-15 14:40:55 cant seem to find any reference anywhere to that module 2016-03-15 14:41:25 adding it anyway 2016-03-15 14:44:25 the kernel module is actually vmwgfx 2016-03-15 14:44:37 load that too for good measure 2016-03-15 14:46:35 oh i thought that was the vm sharing module 2016-03-15 14:46:36 ok thanks 2016-03-15 14:46:46 thats a big help cause the first didnt work 2016-03-15 14:47:19 you're loading the video driver vmware_drv in xorg too right? 2016-03-15 14:47:30 yes trying to 2016-03-15 14:47:37 1 more minute thanks 2016-03-15 14:49:51 modprobe fatal mdoule vmgfx not found in directgory lib/mdoules/4.1.18-0-grsec 2016-03-15 14:50:16 do i need the virtual grsec kernel? 2016-03-15 14:50:23 you typed vmgfx 2016-03-15 14:50:28 it's vmwgfx 2016-03-15 14:50:40 ah sorry 2016-03-15 14:50:41 thanks 2016-03-15 14:53:25 no joy hold on I will get you the info you asked for 2016-03-15 14:54:07 xorg retro launches a blank desktop that flashes a lot 2016-03-15 14:54:30 X for cursor but nothing else 2016-03-15 14:54:41 blank as in black? or sort of like an untuned TV 2016-03-15 14:55:09 "X for cursor but nothing else" <- good start, you're almost there 2016-03-15 14:55:22 can you move the X mouse cursor? 2016-03-15 14:55:36 no its grey can move cursor around 2016-03-15 14:55:50 its a big x 2016-03-15 14:56:15 it's a big X and you can move it? success. you have just got X11 working 2016-03-15 14:56:21 have to reboot cause a problem in vmware fling is that I cant use ctrl+ alt +f2 2016-03-15 14:56:25 doesnt work 2016-03-15 14:56:47 yeah it does :) 2016-03-15 14:56:58 no its picked up by host 2016-03-15 14:57:18 you now need to wrap it in xinit, and create a .xinitrc file with what you need 2016-03-15 14:57:39 do you have a terminal emulator installed? 2016-03-15 14:58:14 lxterminal I believe 2016-03-15 14:58:27 ok, and xinitrc? 2016-03-15 14:58:33 i mean xinit? 2016-03-15 14:59:31 echo "lxterminal" > .xinitrc 2016-03-15 14:59:41 then as the same use, run xinit 2016-03-15 14:59:58 you should get your term (no window manager yet) 2016-03-15 15:02:40 ok now I get a different cursor but a black screen 2016-03-15 15:03:10 oops, .xinitrc is a shell script 2016-03-15 15:03:26 start it with #!/bin/sh and make it executable 2016-03-15 15:07:47 same result blank desktop with differnet mouse, checking Xorg log 2016-03-15 15:09:54 I had the suggestion to disable /etc/X11/xorg.conf.d/20-modules.conf Disabled glx, dri and vesa 2016-03-15 15:10:02 let's take this elsewhere, lots of noice 2016-03-15 15:10:04 do you see merit in that 2016-03-15 15:12:36 srunge.us/WPgK 2016-03-15 15:12:46 sprunge.us/WPgK 2016-03-15 15:13:25 its in /root/.xinitrc with permissions rwx rx rx 2016-03-15 15:18:56 ncopa: ^ 2016-03-15 15:19:04 err ths isn't -linux 2016-03-15 15:29:48 ncopa: what do you think about splitting the bind-tools package into two packages. one containing the normal tools and then bind-tools-dnssec 2016-03-15 15:29:59 that would reduce the package size by 50% 2016-03-15 15:30:05 sounds good to me 2016-03-15 15:31:46 ncopa: btw, that . file did the trick. :) thanks for that little tip 2016-03-15 15:32:10 able to roll livecds in about 5 minutes now and they're fully functional 2016-03-15 15:32:36 nice :) 2016-03-15 15:32:42 i didnt know about that . file myself 2016-03-15 15:32:53 didn't you design that file? ;) 2016-03-15 15:32:56 i foudn it while reading the source 2016-03-15 15:32:57 i did 2016-03-15 15:33:04 #valueadd 2016-03-15 15:33:06 its probably me who did it :) 2016-03-15 15:33:17 but i cannot remember doing it 2016-03-15 15:33:26 well it makes sense 2016-03-15 15:39:35 i've updated teh wiki page for you 2016-03-15 15:43:30 leo-unglaub: ltm you'd save maybe 400k, but you'd still have to pull in the whole 2.4M of bind-libs 2016-03-15 15:55:03 ammunta: yeah, indeed. its progress but not as much as i would love to have 2016-03-15 15:56:17 alternatively doesn't busybox come with nslookup and friends available? 2016-03-15 16:09:37 argh! 2016-03-15 16:09:58 clang does not build on x86 2016-03-15 16:10:08 ../../../../lib/liblldbTarget.a(ThreadPlanCallFunction.cpp.o): In function `lldb_private::ThreadPlanCallFunction::ThreadPlanCallFunction(lldb_private::Thread&, lldb_private::Address const&, lldb_private::EvaluateExpressionOptions const&)': 2016-03-15 16:15:38 Adran: that's good enough for some of the stuff you might want to do, but not the same thign 2016-03-15 19:43:20 new aport waiting to be merged https://github.com/alpinelinux/aports/pull/36 :) 2016-03-16 01:07:52 clandmeter, are you around? 2016-03-16 08:13:50 so, ncopa, are you going to tell us about your new job? :) 2016-03-16 08:17:17 ScrumpyJack: you mean https://www.brianchristner.io/docker-is-moving-to-alpine-linux ? 2016-03-16 08:19:36 ScrumpyJack: i work for Docker Inc. now 2016-03-16 08:26:14 ncopa: that's exciting! 2016-03-16 08:30:50 ScrumpyJack: it actually is 2016-03-16 08:31:05 so far the job has been great! 2016-03-16 08:31:36 Docker is now starting to makes sense 2016-03-16 08:31:47 are you working from home? 2016-03-16 08:32:13 yes 2016-03-16 08:50:37 Oh docker, nice that's must be the reason why we saw an increase use of the word 'docker' here :D 2016-03-16 08:55:08 I'm thinking of trying EDE on alpine rpi as a window manager (it's fltk only, keeping other toolkits off the rpi) 2016-03-16 08:55:20 does anyone know anything about this? http://equinox-project.org/wiki/edeonalpinelinux 2016-03-16 09:39:25 hope the salary is great as well :) 2016-03-16 09:40:58 ScrumpyJack: a window manager looking like windows and that doesn't work with grsec kernel ? meh 2016-03-16 09:44:25 xming: the increase is due to popularity of alpine on containers even before ncopa was hired/company decided to move to alpine 2016-03-16 09:47:23 coredumb: I'm not fussed about the look. I like that it uses fltk. I wonder why it doesn't work with grsec though 2016-03-16 09:50:01 :) 2016-03-16 10:02:42 barthalion: you did noticed the smiley at the end, didn't you :p 2016-03-16 10:49:45 ncopa: shouldnt easy-rsa be linked/installed to usr/bin? 2016-03-16 10:55:39 bah its all relative 2016-03-16 10:59:06 abuild gives me getfattr: invalid regular expression "*" when stripping binaries 2016-03-16 12:34:37 @ncopa, could you please take a look? https://github.com/alpinelinux/aports/pull/36 2016-03-16 12:55:04 jirutka: looks good to me 2016-03-16 12:55:26 i wonder if we want use pkgver=9.2.15.20160212 2016-03-16 12:55:33 that the date is part of the version number 2016-03-16 12:56:16 hey :) 2016-03-16 13:01:58 hi 2016-03-16 13:05:21 o/ 2016-03-16 13:31:11 @ncopa, O think that this versioning schema (including date) is totally stupid, that’s why I stripped it off 2016-03-16 13:31:15 *O→ I 2016-03-16 13:31:51 there’s no version 9.2.15. it’s just redundand 2016-03-16 13:37:43 ok 2016-03-16 13:42:44 However, if you prefer to include date in the version number, then I’ll change it of course :) 2016-03-16 13:43:04 was just thinking on future updates 2016-03-16 13:43:10 i have this 'abump' script 2016-03-16 13:43:17 abump package-1.0.3 2016-03-16 13:44:19 abump jetty-runner-9.2.16.20160501 2016-03-16 13:44:41 but its okish with what you do i suppose 2016-03-16 13:47:21 this is good point, I didn’t know about abump script 2016-03-16 13:49:51 now to package abump 2016-03-16 13:50:02 just to write "abump abump" 2016-03-16 13:51:17 I wonder how to simply rewrite 9.2.16.20160501 to 9.2.16.v20160501 in APKBUILD 2016-03-16 13:51:53 ${pkgver/./.v/} would replace the first dot, ${pkgver//./.v/} all the dots, I don’t know how to replace just the last dot 2016-03-16 13:52:49 do you have to do it dynamically while the script is running? 2016-03-16 13:53:32 how else if we want to utilize abump script for simply bumping? 2016-03-16 13:54:15 I mean, do you need to convert from a.b.c.d to a.b.c.vd all the time? 2016-03-16 13:54:40 I don't understand what you're trying to accomplish here 2016-03-16 13:54:47 yes, because this must be used in URL to download the JAR file 2016-03-16 13:56:35 do it the pedestrian way, it's fine 2016-03-16 13:58:28 I’d like to use conventional versioning schema, e.g. 9.2.16, but Jetty uses uterly stupid 9.2.16.v20160501 – the date is redundat, 9.2.16 is itself unique; so I just used 9.2.16 for apk version, see https://github.com/alpinelinux/aports/pull/36/files#diff-4065196d3a3cf95925e6e1f500547912R5 2016-03-16 13:58:29 but ncopa noted, that then it can’t be simply bumped using abump script, that’s certainly a good point, so he proposed to use 9.2.16.20160501 (note missing „v“) 2016-03-16 13:59:51 I suppose that “v” can’t be here because of versioning scheme used by apk (I have the same “problem” in Gentoo ebuilds) 2016-03-16 13:59:54 pkgverdate=$(echo $pkgver | cut -f4 -d.) ; pkgverno=${pkgver%.$pkgverdate} ; pkgvervdate=$pkgverno.v$pkgverdate 2016-03-16 14:00:38 okay, I haven’t been clear… I know how to do that, but I don’t know how to do that elegantly 2016-03-16 14:00:39 _ver=${pkgver%.*} 2016-03-16 14:00:50 _date=${pkgver##*.} 2016-03-16 14:01:15 ah, of course! i’m stupid 2016-03-16 14:01:19 ^ or that when you're a shell grandmaster 2016-03-16 14:01:30 that’s much better 2016-03-16 14:01:38 _urlver=${_ver}.v${_date} 2016-03-16 14:02:08 meh, I like cut 2016-03-16 14:02:23 the shell language is overpowered. :P 2016-03-16 14:02:35 except it will fork/exec a shell + fork/exec cut 2016-03-16 14:02:41 ^ exactly 2016-03-16 14:03:08 because abuild scripts need to be efficient :P 2016-03-16 14:03:18 they actually do 2016-03-16 14:03:36 i have a script that will parse the entire aports tree 2016-03-16 14:03:42 parse all of the apkbuilds 2016-03-16 14:04:08 when you perform a compilation, I honestly doubt spawning one cut among the myriad of gccs will be your bottleneck 2016-03-16 14:04:08 only way to parse an APKBUILD file is to execute it 2016-03-16 14:04:39 so if every APKBUILD did 1 extra fork/exec, it would be painfully slow to calculate the build time deps 2016-03-16 14:05:02 for the compiling itself it doesnt matter 2016-03-16 14:05:02 (I'm not pretending your syntax isn't the right way to do it) 2016-03-16 14:05:17 but to caclulate the build time deps, it matters 2016-03-16 14:05:22 and that’s totally okay when APKBUILDs are just plain shell scripts with just variables and function on top level 2016-03-16 14:05:42 "totally okay", i dont know :) 2016-03-16 14:05:47 i'd say "okish" :) 2016-03-16 14:06:12 ok, so you actually run the APKBUILDs without calling their functions sometimes 2016-03-16 14:06:12 but similar to gentoo, i want avoid fork/exec in global scope 2016-03-16 14:06:18 yup 2016-03-16 14:06:27 makes sense 2016-03-16 14:06:48 same goes for openrc scripts actually 2016-03-16 14:07:03 yeah, but that's a giant hack 2016-03-16 14:07:16 they are parsed to calculate the deps 2016-03-16 14:07:17 +1 2016-03-16 14:07:32 thats one of the reasons i want get rid of openrc 2016-03-16 14:07:53 and this is also one of the reasons i have been talking about replace the APKBUILD format 2016-03-16 14:08:02 well… I usually prefer expressive turing-complete lang over plain declarative configuration format, because it always leads to actually much simpler solutions 2016-03-16 14:08:08 let's fix the world one piece at a time 2016-03-16 14:09:41 get rid of openrc???? 2016-03-16 14:09:53 systemd incomming? 2016-03-16 14:10:05 ACTION stares at leo-unglaub 2016-03-16 14:11:58 leo-unglaub: i was more thinking of s6 2016-03-16 14:12:02 for example, OpenRC demonstrates that you can use shell for simple declarative-only configuration (you just define few variables) that allows you to use imperative approach when you need to, to solve special use cases — instead of creating extremly complex declarative language that still can’t solve every use case 2016-03-16 14:12:58 hmmm 2016-03-16 14:13:06 jirutka: yes. there is a reason why i havent already switched 2016-03-16 14:13:14 its difficult to catch all cornercases 2016-03-16 14:13:32 i personally dont see an issue with openrc 2016-03-16 14:13:45 leo-unglaub: try stracing it :D 2016-03-16 14:14:05 openrc is okish 2016-03-16 14:14:09 APKBUILD is the same… you just define few variables and then build functions, but if you need to solve some special use case, then you can generate values of some variables dynamically 2016-03-16 14:14:58 I think that if we design good configuration format (skarnet calls it user interface) for s6, then it’ll be much better than OpenRC 2016-03-16 14:15:26 the issue is much more complex than that 2016-03-16 14:15:32 leo-unglaub: http://skarnet.org/software/s6/why.html 2016-03-16 14:15:55 jirutka: yes, the APBKUILD format allows us to generate things dynamically 2016-03-16 14:15:59 which is very handy at times 2016-03-16 14:16:19 well, “configuration format”… I still thinks that we can use just shell, in similar way as in OpenRC, to generate s6 scan directory (or how it that called) 2016-03-16 14:16:39 configuration is one of the aspects, but service manager config isn't the same as package manager config 2016-03-16 14:16:43 there are similarities 2016-03-16 14:16:57 i know 2016-03-16 14:16:58 but I don't think the same reasoning applies 2016-03-16 14:17:10 hmmm 2016-03-16 14:17:13 i have not come up with anything better than APKBUILD yet 2016-03-16 14:17:22 was thinking of embeddin shell in yaml or similar 2016-03-16 14:17:35 ncopa: we can brainstorm about it in 2 years if you want :D 2016-03-16 14:17:40 or have partially yaml and paritially shell 2016-03-16 14:17:48 but as I see it now from user-perspective, missing sane “configuration format” is the main shortage of s6 2016-03-16 14:18:19 well, I like YAML, but it’s *very* complex 2016-03-16 14:18:45 http://yaml.org/spec/1.2/spec.html 2016-03-16 14:18:59 i know :-/ 2016-03-16 14:19:16 and please avoid Ansible way… it’d be horrible mistake 2016-03-16 14:19:42 do you know that Ansible really is? scripting in YAML and Jinja2 (simple template lang for HTML) 2016-03-16 14:20:08 hmmm 2016-03-16 14:20:20 how urgent is that openrc switch? 2016-03-16 14:20:37 because in my opinion the libressl switch would be much more important 2016-03-16 14:20:52 agree 2016-03-16 14:21:18 and then of course rust :) 2016-03-16 14:21:41 ^ yeah! 2016-03-16 14:22:54 i am sure skarnet is a good developer and s6 is awesome 2016-03-16 14:23:01 but openrc works and is very small 2016-03-16 14:24:20 don't worry, openrc will still be available for those who like it 2016-03-16 14:24:33 my goal is to make the service manager switchable 2016-03-16 14:24:39 btw I’ve just updated the PR with jetty-runner 2016-03-16 14:24:39 not to enforce anything 2016-03-16 14:26:12 leo-unglaub: if you have some time, read pages on http://skarnet.org/software/s6/, it’s interesting and inspiring 2016-03-16 14:26:29 i know it 2016-03-16 14:26:40 skarnet and me talked about it a while ago 2016-03-16 14:27:43 switching to another mandatory service manager would be a detrimental move to FOSS overall 2016-03-16 14:28:03 even if the new service manager is way better :P 2016-03-16 14:28:16 it would divide the community and split resources 2016-03-16 14:28:45 and confirm the idea that service manager is pure policy and distros have to stick to one 2016-03-16 14:28:59 That's the opposite of what I'm after. 2016-03-16 14:29:03 :) 2016-03-16 14:31:03 > and confirm the idea that service manager is pure policy and distros have to stick to one 2016-03-16 14:31:04 I call it Lennart way, that’s how poor programmers with evil company in back takes control over FOSS 2016-03-16 14:31:47 well that one isn't Lennart's fault, even if systemd thrives on this idea and takes it to the next level 2016-03-16 14:31:58 the idea was there even with sysvinit 2016-03-16 14:32:21 the first problem is that systemd is *not* a service manager 2016-03-16 14:32:31 the first problem is systemd 2016-03-16 14:32:34 yeah 2016-03-16 14:32:50 but that's not my point here 2016-03-16 14:33:02 barthalion: you about? 2016-03-16 14:33:07 sysvinit is old, bad, but it do just one thing 2016-03-16 14:33:15 look at what distros did before systemd even was a thing 2016-03-16 14:33:23 you had sysv-rc and that was it 2016-03-16 14:33:51 and then, people tried replacing "init", which in their mind had to be init + service manager 2016-03-16 14:34:15 you had bunch of very poorly written shell scripts called “rc script” and sysv-rc in behind 2016-03-16 14:34:27 so Upstart was born and Ubuntu supported *only* upstart 2016-03-16 14:34:41 that’s not true 2016-03-16 14:34:53 oh? 2016-03-16 14:35:08 you had a sysv-rc compatibility layer, but that's all it was 2016-03-16 14:35:37 Ubuntu supported both, the problem is that some packages had upstart scripts, some sysv init… and you can’t declare dependency between them 2016-03-16 14:36:04 fun 2016-03-16 14:36:10 no wonder it couldn't work 2016-03-16 14:36:27 I wouldn't call that "support" honestly 2016-03-16 14:36:29 and also Upstart is clearly broken by design 2016-03-16 14:36:35 and it’s even written in manual 2016-03-16 14:36:50 all service managers so far are, one way or another :P 2016-03-16 14:37:33 then came systemd and it's obvious that a systemd distro can *only* support systemd 2016-03-16 14:37:37 upstart can bring the system up and down, nothing between… if you want to stop service and then run it again, you’re quite out of luck, the dependency sytem doesn’t work in this scenario 2016-03-16 14:37:44 then came OpenRC and Gentoo only supports OpenRC, afaik 2016-03-16 14:37:50 nope 2016-03-16 14:37:57 Gentoo supports both OpenRC and systemd 2016-03-16 14:38:16 so they maintain double service scripts? 2016-03-16 14:38:20 many packages have both OpenRC runscript and systemd unit files 2016-03-16 14:38:24 yeah 2016-03-16 14:38:25 does s6 have the same runlevels than openrc and then simply softlinks to the script? 2016-03-16 14:38:31 because that idea is awesome 2016-03-16 14:38:48 leo-unglaub: softlinks, no, because the architecture is different, but infinite runlevels, yes 2016-03-16 14:38:59 and that's s6-rc, not s6 ;) 2016-03-16 14:39:10 jirutka: what of packages that don't provide both? 2016-03-16 14:39:45 well, then you can write yourself… as I know, every package support OpenRC, just some doesn’t support systemd 2016-03-16 14:40:00 * don’t support 2016-03-16 14:40:36 then I think it's inaccurate to say "Gentoo supports systemd" if it's not 100% 2016-03-16 14:41:00 it’s just playing with words 2016-03-16 14:41:35 clarity and precision are important. Users need to know whether they can switch service managers without breaking anything! 2016-03-16 14:41:39 however, I really don’t care about systemd support, this cancer is not allowed to infect my systems 2016-03-16 14:42:11 I don't care about systemd either, but I do care about not breaking anything 2016-03-16 14:43:03 anyway, my point is that supporting several service managers is *hard* 2016-03-16 14:43:12 yeah, that is 2016-03-16 14:44:09 and that's why I need to do some serious analysis work on Alpine package dependencies to the service manager before anything else 2016-03-16 14:44:19 @ncopa, could you please merge jetty-runner abuild? I’ve already updated pkgver 2016-03-16 14:44:53 and maybe even a configuration generator, which would please you, jirutka 2016-03-16 14:45:34 take a config file, and autogenerate both a s6-rc definition directory and an openrc script 2016-03-16 14:45:41 that would be awesome 2016-03-16 14:45:53 but I'm not sure whether it's even possible 2016-03-16 14:45:54 yeah, that would be indeed awesome! 2016-03-16 14:46:21 I think that it’s not possible in general for every use case, but for most of them, I think that it should be possible 2016-03-16 14:46:54 I will need to think about that, after the dependencies analysis 2016-03-16 14:50:01 skarnet: waaaaa, please no 2016-03-16 14:50:11 autogenerate config files is always how the madness starts 2016-03-16 14:50:29 call me old school, but a config file is a config file and anways static hand written 2016-03-16 14:50:38 that’s very true 2016-03-16 14:50:46 uh, yes? 2016-03-16 14:51:15 jirutka: my tree is not clean 2016-03-16 14:51:19 i have llvm there 2016-03-16 14:51:21 however, I don’t consider s6-rc scan directory to be configuration files that should be written by hand 2016-03-16 14:51:28 and i messed up my other computer 2016-03-16 14:51:33 and i have to run now sorry 2016-03-16 14:51:39 and I consider autogenerating OpenRC runscripts to be just a temporary solution 2016-03-16 14:51:47 okay, no problem 2016-03-16 14:51:56 barthalion or nmeum can you please: curl https://github.com/jirutka/alpine-aports/commit/a21e1550240842979ab571857bec11a9b5b72968.patch | git am 2016-03-16 14:52:08 skarnet: one example for all: authells (autotools) 2016-03-16 14:52:09 and push 2016-03-16 14:52:12 thanks! 2016-03-16 14:52:30 or rnalrd 2016-03-16 14:52:43 I knew you were thinking about autoconf, and I'll just say: guys. Give me some credit, puh-lease. 2016-03-16 14:52:45 or clandmeter 2016-03-16 14:53:12 ncopa: curlbashing much? 2016-03-16 14:53:59 @ncopa: you know git stash, don’t you? 2016-03-16 14:54:15 https://curlpipesh.tumblr.com/ 2016-03-16 14:54:48 skarnet: that’s no arbitrary code 2016-03-16 14:54:59 I know, and git am isn't sh :D 2016-03-16 14:55:02 but still 2016-03-16 14:55:11 I think that it’s totally okay in this case 2016-03-16 14:55:37 that's not directly dangerous 2016-03-16 14:59:46 ACTION loves this openrc discussions *g* reminds me of my 4chan days *g* 2016-03-16 15:27:21 skarnet: i think part of the problem with that philosophy "there's another option" implies software packages will be maintained for both systems, with isnt realistic 2016-03-16 15:28:39 it's realistic if it's done right. Maintaining an init script should be a negligible effort compared to the effort of maintaining the mechanism package in the first place. 2016-03-16 15:28:53 thats very optimistic 2016-03-16 15:28:59 But "doing it right" will require a *lot* of effort up front. 2016-03-16 15:29:54 skarnet: the service manager approach seems very complicated imo, with directories and such 2016-03-16 15:30:28 having to drop multiple files in diffferent locations to start one service 2016-03-16 15:30:28 Adran: that's a user interface issue. It can be solved with tools. 2016-03-16 15:32:03 can be, but this is where systemd fails as well - there is a ctl manager for every thing , but none of them are very good at debugging anything 2016-03-16 15:33:36 I very much doubt that what I have in mind is remotely comparable to anything systemd does. 2016-03-16 15:35:34 anyway it's much too early to talk about specifics. I need to finish working on my contract first, then (around next November) do the analysis pass and gather information on exactly what is needed, and then we can discuss specifics 2016-03-16 15:36:13 tbh 2016-03-16 15:36:47 this probably should be in #s6 :P 2016-03-16 15:37:28 maybe, maybe not. If it's Alpine-specific, it belongs here :P 2016-03-16 15:40:37 i mean is alpine is going to be the defacto standard for docker and thats all that matters about it, a service manager makes complete sense like this, but if the goal is also embedded and lightweight actual systems, I don't know honestly :/ maybe my opinion but the more parts there are to an embedded system the less its likely used 2016-03-16 15:48:09 Adran: that's another reason why small package granularity is important 2016-03-16 15:48:42 not sure what you mean 2016-03-16 15:49:53 For embedded and lightweight, it's important that you pull in *only* what you need 2016-03-16 15:50:05 oh yes, that is true 2016-03-16 16:19:49 well this is promising 2016-03-16 16:20:19 the 8.0 version of ghc hasn't died yet on an arm build 2016-03-16 16:21:33 also llvm 3.7.1 is really slow at compiling compared to 3.5 2016-03-16 16:55:54 ncopa: pushed 2016-03-16 16:56:14 jirutka: ↑ 2016-03-16 16:56:21 thanks! 2016-03-17 01:31:15 https://github.com/openresty/lua-nginx-module#lua_transform_underscores_in_response_headers 2016-03-17 13:50:18 . 2016-03-17 14:09:06 i test rebooted rpi3 on alpine just now 2016-03-17 14:10:08 [ 0.000000] Machine model: Raspberry Pi 3 Model B Rev 1.2 2016-03-17 14:15:09 fabled: That sounds cool, I have one myself, be happy to test some images if you want (with configurations like i3wm firefox kiosk and lxc) 2016-03-17 14:40:04 skazz: what do you use for kiosk? 2016-03-17 14:40:28 clandmeter: i3wm, firefox, r-kisok plugin for firefox 2016-03-17 14:40:46 runs a web interface for the public to use 2016-03-17 14:41:18 but there is a requirement to run video, and the pi2 cannot run the video smoothly (because firefox does not have hw accelerated video playback) 2016-03-17 14:41:45 Going to have to use an oDroid with debian jessie in the short term 2016-03-17 14:42:03 only because I don't know where to start getting alpine working on the odroid u4 or c1 2016-03-17 15:24:58 ncopa: just noticed that the default abuild prepare function actually relies on a variable prefixed with an underscore (_builddir). Hasn't the rule always been not to use global variables prefixed with underscores in abuilds default functions? 2016-03-17 15:27:12 yes 2016-03-17 15:27:15 thats a bug 2016-03-17 15:27:22 hm 2016-03-17 15:27:42 abuild should not use _ prefixed vars/functions 2016-03-17 16:00:24 ncopa: We could remove the underscore from the variable name but we would have to change a bunch of existing APKBUILDs with the next abuild release then 2016-03-17 16:07:24 yes 2016-03-17 16:07:29 thats a problem 2016-03-17 16:07:43 i say, lets remove it asap 2016-03-17 16:07:47 and let the apkbuilds break 2016-03-17 16:08:21 they can be fixed with: builddir="$_builddir" 2016-03-17 16:42:49 so should i nuke any use of _ prefixed variables in general? 2016-03-17 22:48:14 oh whoops 2016-03-17 22:48:48 I think I basically just copied the code that used $_builddir out of newapkbuild scaffolding 2016-03-17 22:50:11 mitchty: no 2016-03-17 22:59:48 ok, so just _builddir then? 2016-03-17 23:13:12 mitchty: you can keep using it for now but if we decide to use it in functions supplied by abuild itself it should no longer be prefixed with an underscore 2016-03-18 07:23:18 fabled: should #5261 be cherry-picked for 3.3-stable? 2016-03-18 07:46:00 ncopa, probably yes 2016-03-18 08:45:25 good morning climbers 2016-03-18 08:45:28 almost weekend! 2016-03-18 08:55:56 yes, happy friday! 2016-03-18 09:52:49 fabled: #5197 2016-03-18 09:52:56 we set target 3.3.3 or 3.4.0? 2016-03-18 09:53:01 its just 1 month for 3.4 2016-03-18 09:53:16 it works on 3.3-stable already i think 2016-03-18 09:53:22 it's missing the wlan firmware though 2016-03-18 09:53:24 but it'll boot 2016-03-18 09:53:32 so its resolved? 2016-03-18 09:53:39 it should be resolved yes 2016-03-18 09:53:53 i set target 3.3.3 2016-03-18 09:53:59 and ppl will have to test 3.3.2 2016-03-18 09:54:23 yes, the linux-rpi 4.1.19 has the patches, and the latest alpine-iso changes will make it book 2016-03-18 09:54:25 boot* 2016-03-18 09:54:38 wlan firmware is out-tree, and not included yet 2016-03-18 09:54:44 so i target 3.3.2 again and close it? 2016-03-18 09:54:45 but the kernel module is compiled 2016-03-18 09:54:52 that's ok too 2016-03-18 09:54:54 i feel wlan should be different ticket 2016-03-18 09:55:42 yes 2016-03-18 09:56:00 you can comment to create separate tickets if wlan/bt is non-functional or require additional firmware files 2016-03-18 10:08:15 ncopa: can you tell me about alpine-desktop? 2016-03-18 10:09:24 is it a meta package that is just the stuff you like? 2016-03-18 10:13:55 hm, we now do have php7 specific packages I thougt the goal was to replace main/php with testing/php7, wasn't it? 2016-03-18 10:14:57 ScrumpyJack: i think alpine-desktop can be removed 2016-03-18 10:15:20 the idea was a meta package for settin up xfce 2016-03-18 10:15:26 but its wronga pproach 2016-03-18 10:15:41 better write a setup-desktop script 2016-03-18 10:16:51 i just stumbled upon it when trying to get X working in virtualbox/vmware etc. really struggling with that. we'll find what's wrong in the end 2016-03-18 10:17:37 lots of folks trying out alpine as a guest os, and expecting some sort of desktop experience. 2016-03-18 10:17:44 yes 2016-03-18 10:17:47 we need fix tht 2016-03-18 10:17:58 virtualbox and vmware experience 2016-03-18 10:18:20 i think in qemu it works? 2016-03-18 10:18:56 i'm not suggesting a default desktop environment, i think folk should tinker with that 2016-03-18 10:19:16 yes 2016-03-18 10:19:18 yes, if there was a way to get the mouse to work in virtualbox, that would be a great thing. 2016-03-18 10:19:18 agree 2016-03-18 10:19:18 qemu with std vga isn't working yet 2016-03-18 10:19:31 qemu vmware video work iirc 2016-03-18 10:19:46 i got it working with some other video adapter 2016-03-18 10:19:46 nope :( 2016-03-18 10:20:04 vbox video works great and I can run apps remotely to X but no input, neither -input-evdev nor -input-{keyboard|mouse} work :( 2016-03-18 10:20:32 hm 2016-03-18 10:20:44 i wonder why 2016-03-18 10:21:13 here is what i have so far. X works everywhere, i can run X with twm or mwm and a x client like mrxvt, mouse works etc. but as soon as you try a more complex window manager, you have issues 2016-03-18 10:22:31 everywhere = baremetal, qemu with std vga and vmware display. not tried virtualbox yet, but it should also work there 2016-03-18 10:25:59 xfce4 works just fine here... 2016-03-18 10:26:15 bare-metal right? 2016-03-18 10:26:18 yes 2016-03-18 10:26:48 but as soon as you guest it, nothing 2016-03-18 10:31:42 wwwtest.alpinelinux.org 2016-03-18 10:31:49 http://wwwtest.alpinelinux.org/posts/Alpine-3.3.2-released.html 2016-03-18 10:31:51 looks ok? 2016-03-18 10:32:06 i cannot do much about those long commit messages 2016-03-18 10:34:43 looks good, except i didn't make it to the shortlog ;p 2016-03-18 10:36:00 maybe we should have an offical blog, would be nice to write some articles 2016-03-18 10:46:16 do we have someone who wants to do the writing 2016-03-18 10:47:53 we could set up a blog aggregator 2016-03-18 10:47:58 well, anyone potentially. i'd like to do one on the "unix philosophy", and how alpine linux is the closest you'll get to it :) 2016-03-18 10:48:03 if we had writers 2016-03-18 10:48:33 eeeeh 2016-03-18 10:49:44 ncopa: writers aren't hard to find. we all know enough people we can ask who might contribute 2016-03-18 10:51:25 what's hard to find is time 2016-03-18 10:51:37 I'd love to write a few articles, but that's a non-negligible time investment 2016-03-18 10:52:03 i have this working in a guest. alpine-setup, reboot. setup-xorg-base && apk add twm mrxvt; echo "mrxvt & \ twm > .xinitrc && xinit 2016-03-18 10:52:48 but s/twm/jwm and you'll see the problem 2016-03-18 11:28:30 hi there 2016-03-18 11:40:20 hi leo-unglaub 2016-03-18 11:47:49 hi leo-unglaub 2016-03-18 11:49:43 hello. 2016-03-18 12:09:34 ncopa: using lua was a bad idea *g* its not as easy as i thought when the programm gets bigger ... small programms sure ... but this letsencrypt stuff is a little bit bitter than i thought *g* 2016-03-18 12:10:06 ACTION whistles softly 2016-03-18 12:10:19 leo-unglaub: where’s the problem with Lua? 2016-03-18 12:11:38 and what do you consider to be “bigger”? 2016-03-18 12:15:22 Lua is quite flexible and supports high-order functions, so it’s just up to you how do you structure your program to be sane when it gets bigger 2016-03-18 12:20:00 jirutka: yeah, its propobly a structure problem 2016-03-18 12:20:12 do you have your code somewhere online? 2016-03-18 12:20:13 those 8000 lines in one file are fucking unreadable 2016-03-18 12:20:18 WHAT?! 2016-03-18 12:20:41 then it’s definitely not problem in Lua, but between monitor and chair… 2016-03-18 12:21:21 if you're going to implement an acme client, it's going to be big no matter what language you use 2016-03-18 12:21:36 there are simple scripts for let’s encrypt written in few lines of bash or python, so why the hell is this 8k lines?! 2016-03-18 12:32:00 skarnet: well, the client itself is quit simple 2016-03-18 12:32:22 the overhead comes from argument parsing, config file parsing and the cron stuff, the locking, ... 2016-03-18 12:36:16 oh, you mean the stuff you shouldn't be doing? 2016-03-18 12:36:21 exactly 2016-03-18 12:36:52 there are alreay Lua libs for argument parsing and config parsing 2016-03-18 12:37:04 cron stuff? eh, just use normal cron? 2016-03-18 12:37:35 https://luarocks.org/ is your friend 2016-03-18 12:40:56 jirutka: well, if i have to add 20 dependencies, then i could use the official client as well 2016-03-18 12:41:01 i want it to be as small as possible 2016-03-18 12:42:49 you don’t need 20 dependencies for arguments and config file parsing 2016-03-18 12:43:24 and if you prefer copy&paste programming or reimplementing the wheel… well… enjoy… i personnally don’t wanna use such software 2016-03-18 12:44:38 morever, config file may be very simple, like 10 lines https://github.com/jirutka/ssh-getkey-ldap/blob/master/ssh-getkey-ldap#L52-L59 2016-03-18 14:28:22 hi. I see that there are a lot of Python APKs, but only for Python 2 and not 3 (except for the interpreter itself). 2016-03-18 14:28:47 are there any plans to introduce packages built for Python 3? 2016-03-18 14:29:22 good point, we should definitely fix this, I’m with you 2016-03-18 14:30:02 also, some packages might already work with Python 3, but depend on the 'python' package, which means installing a second interpreter which can cause issues and wastes space on preferably minimal Docker images. 2016-03-18 14:31:05 upgrading anything from Python 2 to 3 probably is not an option at all, so a separate set of py[thon]3-* packages might be the obvious thing to do 2016-03-18 14:31:49 maybe we can fix it using directive ”provides” – add e.g. provides=python23 to both python and python3 packages (interpreters) and add depends=python23 to packages that works both with python2 and python4 2016-03-18 14:31:51 *python3 2016-03-18 14:32:11 and while pure-Python packages are rather lightweight to install via pip, there are some that take quite a while to compile and have quite some dependencies, e.g. lxml and Pillow 2016-03-18 14:32:31 it’d be much better to use just “python” for those that works with both, but I’m afraid that ppl might not like breaking change with name, from python → python2 2016-03-18 14:33:09 so currently I have a docker file which installs a bunch of -dev packages and drops them afterwards. this is stuff that should be only done once and result in a package that prevents wasting lots of developer time :) 2016-03-18 14:33:46 jirutka: I agree. python2 would be a lot clearer. 2016-03-18 14:35:12 so you're suggesting a 'python' meta package that is either provided by python2 or python3 (in Debian terms)? 2016-03-18 14:38:22 yeah 2016-03-18 14:40:11 we should discuss this with @ncopa 2016-03-18 14:44:16 ncopa has already suggested few solutions, here: http://lists.alpinelinux.org/alpine-devel/4666.html 2016-03-18 14:44:25 Y0Gi: ^ 2016-03-18 14:45:14 ah, I've only taken a look at the issue tracker so far 2016-03-18 14:46:54 scadu: thanks for posting. 2016-03-18 14:47:16 Y0Gi: np 2016-03-18 14:47:36 dropping py2 would be fine with me personally. didn't know ncopa would consider that. 2016-03-18 14:49:00 are there some more recent statements about the matter? 2016-03-18 14:51:08 Y0Gi: as far as I know, no. however I haven't been here for some time. 2016-03-18 14:51:49 ok, thanks 2016-03-18 14:59:46 hi 2016-03-18 15:00:23 what i would have preferred was if we could replace python2 with python3 2016-03-18 15:00:51 wow 2016-03-18 15:01:00 a year later we still have not upgraded python :) 2016-03-18 15:02:09 guess it's time :) 2016-03-18 15:02:45 are virtual/transitional/meta packages possible in alpine? 2016-03-18 15:11:45 Y0Gi: meta and subpackages are supported. 2016-03-18 15:15:02 so what about going with python3-* for now, renaming py-* (i.e. py2) later to python2-*, then re-introducing python-* as aliases for python3-*, and finally remove python3-*? 2016-03-18 16:06:24 ncopa: could you take a look at %1705? 2016-03-18 17:33:27 nmeum: i'm okish with it. please check setup-* scripts and lxc-create in case one of those uses cron instead of crond 2016-03-18 17:35:56 ok, will do so 2016-03-18 23:15:11 is arm builder 3.3 testing still down? 2016-03-18 23:16:55 there was a commit that the x86_64 build was built, dunno if when arm builder is back up will it go back and build the queue? 2016-03-18 23:16:58 http://git.alpinelinux.org/cgit/aports/commit/?id=fa39deeaf3eb163a4ee378ae67f933d24a96623a 2016-03-19 05:20:31 hello 2016-03-19 06:06:48 hello folks 2016-03-19 06:07:56 latest version of alpine-linux does not boot 2016-03-19 06:07:58 on 2016-03-19 06:08:02 raspberry pi 3 2016-03-19 06:08:24 but the os on their page does boot 2016-03-19 06:08:30 i get blank screen 2016-03-19 06:08:46 with our alpine linux 2016-03-19 13:18:29 hey :) 2016-03-19 18:33:58 hello 2016-03-19 18:34:08 the latest distro for rpi is not booting in raspberry pi 3 2016-03-19 18:34:23 can someone please check why its not booting? 2016-03-19 22:55:09 ncopa: I renamed the cron service and adjusted alpine-conf accordingly 2016-03-19 22:55:33 ubermacin: testing is not built on released versions, only for edge 2016-03-20 01:52:41 we should really fix the testing/php7 abuild on x86 and move it to main/ (effectivly upgrading main/php) otherwise people will submit even more php7 specific packages 2016-03-20 02:28:47 fcolista: could you take a look at the package conflict between perl-test-simple 2016-03-20 02:29:13 and perl-test-tester (you are the maintainer of the latter) 2016-03-20 02:58:00 Not sure if this sort of thing belongs in the regular irc or here, but.. Is there any status on official support (or any) for EFI systems? 2016-03-20 03:02:38 Welp, I'm smart. https://bugs.alpinelinux.org/issues/4424 2016-03-20 03:06:48 is php7 supposed to replace whatever old version of php? or 2016-03-20 03:09:29 is it ~stable/testing or something 2016-03-20 06:11:20 http://sprunge.us/ihhO er, this seems buggy 2016-03-20 06:11:35 if that's so, surely there's no point in splitting the packages 2016-03-20 06:15:02 unless I'm fundamentally misunderstanding something 2016-03-20 06:15:13 about this package hierarchy 2016-03-20 07:16:04 hello anyone here 2016-03-20 07:16:26 ncopa: cant boot rpi3 with latest release of alpine 2016-03-20 07:16:38 any ideas why its could be happening 2016-03-20 10:49:35 anyone here 2016-03-20 10:49:55 can someone check this bug please 2016-03-20 10:49:55 http://bugs.alpinelinux.org/issues/5296 2016-03-20 10:50:25 wake up people...please someone 2016-03-20 10:50:28 wake up 2016-03-20 10:52:42 people should have the right to sleep in on Sundays 2016-03-20 10:52:53 *even* when something is broken 2016-03-20 10:52:54 yeah man. it's sunday. relax :3 2016-03-20 10:53:24 :-D 2016-03-20 10:53:29 glad someone is here 2016-03-20 10:53:35 i thought i was alone 2016-03-20 10:53:39 it's Rainday here :( 2016-03-20 13:08:26 http://comyo.net/FS9Ko 2016-03-20 14:08:36 would be nice to have http://github.com/Benjamin-Dobell/Heimdall.git 2016-03-20 14:09:23 is similar to fastboot, but for samsung 2016-03-20 14:44:48 vkris: can't you use fastboot on samsung devices? 2016-03-20 16:31:34 ncopa: How can I retrigger the owncloud build on the x86_64 builder for 3.3? #5298 2016-03-20 16:31:48 According to the builddlog abuild couldn't fetch the source tarball http://dev.alpinelinux.org/buildlogs/build-3-3-x86_64/main/owncloud/owncloud-8.2.3-r0.log 2016-03-20 17:07:48 hello people 2016-03-20 17:25:51 nmeum: bump its pkgrel 2016-03-20 17:26:04 the fastest way for us, minions without further access 2016-03-20 17:26:20 oneinsect: sup 2016-03-20 17:27:00 has anyone tried the latest alpine release on raspberry pi 3 2016-03-20 17:27:26 barthalion: 2016-03-20 17:28:05 I don't own even the second 2016-03-20 17:28:25 but I saw someone else with same issue on mailing list 2016-03-20 17:28:36 I guess fabled will take a look at this tomorrow 2016-03-20 17:28:51 oooh thank god i am not alone 2016-03-20 17:30:47 thanks barthalion: 2016-03-20 18:09:19 oneinsect: see #5297 2016-03-20 18:10:09 nmeum: Raydavis is me 2016-03-20 18:10:25 those are the tickets i raised 2016-03-20 18:10:27 :-P 2016-03-20 18:10:30 ah, alright ^^ 2016-03-20 18:10:56 http://bugs.alpinelinux.org/issues/5296 2016-03-20 18:12:22 infact 2016-03-20 18:12:23 http://lists.alpinelinux.org/alpine-devel/5218.html 2016-03-20 18:12:48 had a try to boot the new Raspberry Pi 3 with this new 3.3.2 version, but unfortunately it does not work. Whereas if I get the official kernel.img from https://github.com/raspberrypi/firmware/blob/master/boot/kernel.img and modify config.txt to use it instead of the kernel provided by Alpine Linux, it does boot. 2016-03-20 18:13:04 official kernel.img works when I modify config.txt 2016-03-20 18:13:24 i mean it does boot to a emergency shell 2016-03-20 18:14:49 i changed [pi3] kernel=boot/kernel7.img initramfs boot/initramfs-rpi2 0x08000000 2016-03-20 18:14:56 in config.txt 2016-03-20 18:15:04 soo rpi3 hardware is good 2016-03-20 18:15:29 have to see if vmlinuz-rpi2 or something could be issue 2016-03-20 18:36:53 nmeum: not tried on newer devices, but did not work on "Galaxy Star" 2016-03-20 18:37:11 but Heimdall detected it 2016-03-21 07:02:55 i'm trying to debug the rpi issues reported on ml/tracker 2016-03-21 07:02:55 then it fails to mount boot media 2016-03-21 07:02:55 seems gpu_mem needs to be now specified or there's no fbcon picture 2016-03-21 07:02:55 :/ 2016-03-21 07:07:24 seems nlplug-findfs does not find repositories 2016-03-21 07:12:51 seems it's not mounted, since the device mount point in /media is not created 2016-03-21 07:18:38 ncopa, any ideas why nlplug-findfs would've stopped working? 2016-03-21 07:22:03 seems that it finds mmcblk0 2016-03-21 07:22:07 but not mmcblk0p1 2016-03-21 07:37:28 morning 2016-03-21 07:37:39 http://hubicka.blogspot.no/2016/03/building-libreoffice-with-gcc-6-and-lto.html?m=1 2016-03-21 07:38:08 they say that alpine builds libreoffice with musl in a comment :) 2016-03-21 08:35:28 good morning 2016-03-21 08:35:43 i hate nothing more than waking up to a phone call saying: "the server is not working" 2016-03-21 08:41:25 leo-unglaub: in android there is a function called "Silent hours", you should try it. 2016-03-21 08:41:44 clandmeter: hahaha 2016-03-21 08:41:48 :) 2016-03-21 08:41:48 good idea 2016-03-21 08:42:13 but this incident is a perfect example for our discussion about s6-process supervsion 2016-03-21 08:42:41 the mariadb server crashed ... and it did try to restart .. but the reson it crashed was still there 2016-03-21 08:42:51 so it restarted and restarted over and over again 2016-03-21 08:43:21 alpine server? 2016-03-21 08:43:58 no, my last non-alpine server 2016-03-21 08:44:37 but it would not have beed a difference ... because this was just a case where process supervison simply did more harm than good 2016-03-21 08:44:46 because it restarted and restarted over and over again 2016-03-21 08:45:04 just on of the cases to keep in mind if you decide to switch to a new software 2016-03-21 08:49:41 there is a deter bruteforce feature in grsecurity kernel 2016-03-21 08:49:59 which means kernel will prevent service to start again 2016-03-21 08:50:15 it will allow it after a while 2016-03-21 08:50:48 but yeah, beter fix the bug instead of just restart the service 2016-03-21 08:51:44 thats good to know 2016-03-21 08:51:50 did not know that about grsecurity 2016-03-21 08:52:50 its redmine monday 2016-03-21 08:53:27 ncopa: you can ignore the mail ive send to you 2016-03-21 08:53:42 clandmeter: i always do 2016-03-21 08:53:44 :-P 2016-03-21 08:53:57 (sorry, i could not resist) 2016-03-21 08:53:58 ACTION slaps ncopa with a cat 2016-03-21 08:54:22 MEEOOUU 2016-03-21 08:56:07 hahaha 2016-03-21 08:56:37 i have just pointed rsync.alpinelinux.org to the scaleway server 2016-03-21 08:56:49 meaning that master mirror is now hosted by scaleway 2016-03-21 08:56:56 i also need reconfig the builders a bit 2016-03-21 08:58:15 does the scaleway server do any logical stuff? or it it just a mirror? 2016-03-21 09:00:08 its just rysnc+lighttpd 2016-03-21 09:00:23 just mirror 2016-03-21 09:02:10 we have scaleway server? 2016-03-21 09:02:15 yes 2016-03-21 09:02:16 how long was nl rysnc? i should prepare the invoice :p 2016-03-21 09:02:37 as primary one now? 2016-03-21 09:02:38 10 years? 2016-03-21 09:02:50 could be 2016-03-21 09:03:03 barthalion: yes, scaleway is now primary mirror 2016-03-21 09:03:09 the rsync.alpinelinux.org 2016-03-21 09:03:17 its a cname to fr.alpinelinux.org 2016-03-21 09:03:19 for as long their peering allows it... 2016-03-21 09:03:20 cool, I didn't know 2016-03-21 09:03:40 clandmeter managed to get them sponsor it (i think?) 2016-03-21 09:03:49 yeah, I assumed so 2016-03-21 09:03:53 clandmeter: or do we pay for it? 2016-03-21 09:03:55 btw, its sponsored by scaleway 2016-03-21 09:04:00 :) 2016-03-21 09:04:02 nice 2016-03-21 09:04:20 i think we should have a logo on the downloads page or similar? 2016-03-21 09:04:30 and fastly 2016-03-21 09:07:01 fastly also sponsors us? 2016-03-21 09:07:25 but yeah, logo would be nice 2016-03-21 09:11:00 ncopa: any idea why i get these errors on some containers? 2016-03-21 09:11:00 lxc-attach: attach.c: lxc_attach_to_ns: 270 Operation not permitted - failed to set namespace 'pid' 2016-03-21 09:11:03 lxc-attach: attach.c: lxc_attach: 926 failed to enter the namespace 2016-03-21 09:11:53 clandmeter, what does dmesg say? 2016-03-21 09:12:11 i dont know 2016-03-21 09:12:35 grsec: From 172.16.5.6: use of CAP_SYS_ADMIN in chroot denied for 2016-03-21 09:13:03 fcolista: good catch 2016-03-21 09:13:10 echo 0 > /proc/sys/kernel/grsecurity/chroot_deny_chmod 2016-03-21 09:13:10 echo 0 > /proc/sys/kernel/grsecurity/chroot_caps 2016-03-21 09:13:11 i recently updated to grsec 2016-03-21 09:14:11 i prefer lxc-attache instead of lxc-console. 2016-03-21 09:15:03 fcolista: do we have something proper to set that at boot time? 2016-03-21 09:16:21 clandmeter, : sysctl ? 2016-03-21 09:17:38 One sec that i'll be more precise. 2016-03-21 09:18:23 fcolista: right 2016-03-21 09:18:42 kernel.grsecurity.chroot_caps=0 2016-03-21 09:18:43 kernel.grsecurity.chroot_deny_chmod=0 2016-03-21 09:18:51 maybe we should provide a sysctl.d? 2016-03-21 09:18:55 for lxc 2016-03-21 09:19:07 just add this to sysctl.d/00-lxc.conf 2016-03-21 09:19:24 that would be enough. 2016-03-21 09:20:00 i'm not sure we want enable that by default 2016-03-21 09:20:23 ncopa: we could rename it 2016-03-21 09:20:46 but who doesnt want to use lxc-attach? 2016-03-21 09:21:19 for what price 2016-03-21 09:21:40 well 2016-03-21 09:21:41 if you install lxc, then i presume you want it. 2016-03-21 09:21:48 a simple solution can be use a wrapper 2016-03-21 09:21:54 that enable that at runtime 2016-03-21 09:21:54 i have never needed it 2016-03-21 09:21:58 lxc-att 2016-03-21 09:22:08 ncopa: cause you use console? 2016-03-21 09:22:12 the problem with that setting is that it is global 2016-03-21 09:22:16 yes 2016-03-21 09:22:26 who enable it, you do $hwatever with lxc-attach, the disable it 2016-03-21 09:22:34 *then 2016-03-21 09:22:46 atleast we can provide a sample file 2016-03-21 09:22:51 but basically, the reason i use lxc-console is that i never made lxc-attach work :) 2016-03-21 09:22:52 yes 2016-03-21 09:23:01 we can provide a sample file 2016-03-21 09:23:03 good idea 2016-03-21 09:23:04 lol :) sounds like a good idea 2016-03-21 09:23:59 if you search on that error, alpine is one of the only results (iirc) 2016-03-21 09:25:01 +1 to ncopa 2016-03-21 09:25:10 I always use console because attach was broken to me :) 2016-03-21 09:25:22 and I doubt I'd use attach now anyway 2016-03-21 09:26:07 ACTION slaps ppl that work around alpine issues 2016-03-21 09:26:14 :p 2016-03-21 09:26:36 if I've had a penny for every workaround I did, I'd be a millionaire 2016-03-21 09:26:46 :D 2016-03-21 09:27:08 life's too short to remove usb safely 2016-03-21 09:31:25 hi, I got a patch file I was told I should try on openldap. What is the correct way to do that? I can see the APKBUILD in git and can guess how would one add another patch, but what should be run then? apk xxx? 2016-03-21 09:32:15 firefox 45.0.1 is available 2016-03-21 09:32:26 xificurC: we use abuild to build our packages. 2016-03-21 09:34:41 clandmeter: so I can put in the patch manually and run abuild from that directory? 2016-03-21 09:35:23 xificurC: you need a build enviorment for it. 2016-03-21 09:35:31 i think its written on the wiki how to do it. 2016-03-21 09:37:06 something like, apk add alpine-sdk, git clone aports, create abuild keys (to sign packages), set some group permissions.... i problably missed a few. 2016-03-21 09:45:39 clandmeter: found the wiki page, thank you 2016-03-21 09:47:58 ncopa: did you add a rsync hook to update nl.a.o? 2016-03-21 09:56:16 clandmeter: no 2016-03-21 09:56:22 we should do so 2016-03-21 10:02:15 leo-unglaub: don't blame supervision for what it's not responsible for. 2016-03-21 10:02:35 If a server crashes repeatedly, then there's a serious error that should be investigated right away. 2016-03-21 10:02:57 supervision cannot, and does not aim to, replace correct in-service monitoring and alerting procedures. 2016-03-21 10:03:10 It can help, but it can't substitute itself to it. 2016-03-21 10:03:36 ncopa: pkgs will also be out of date because of it. 2016-03-21 10:03:47 If your server kept failing and restarting all night, then it's your alerting system that's in question, not the supervision framework. 2016-03-21 10:06:22 skarnet: then please tell me what the usecase of supervison is 2016-03-21 10:06:30 because if there is no problem a daemon does not crash 2016-03-21 10:06:47 and if it crashes it usualy crashes because there is a non trivial problem 2016-03-21 10:07:02 there are lots of reasons why a daemon could crash... 2016-03-21 10:07:20 the thing is, it won't always crash repeatedly. 2016-03-21 10:07:25 i cannot think of one reason that exists where a daemon crashes and simply restarting hin helps 2016-03-21 10:07:28 Or it could serve for some time before crashing again. 2016-03-21 10:07:37 you cannot? think harder. 2016-03-21 10:07:51 every time a daemon crashed i had to fix something manually 2016-03-21 10:08:05 how about a temporary shortage of memory because some other daemon took more than usual? 2016-03-21 10:08:15 then when the memory's freed, your server can work normally again. 2016-03-21 10:08:18 and that's just one example. 2016-03-21 10:08:32 that does not crash the daemon .. that simply swaps it out 2016-03-21 10:08:43 ... 2016-03-21 10:09:10 if a malloc() returns NULL, daemons usually die. 2016-03-21 10:09:28 not if you catch it correctly 2016-03-21 10:09:41 then you for example dont fork a new instance 2016-03-21 10:09:56 oh, so all daemons should implement backoff and retry policies in case of oom? 2016-03-21 10:10:12 (sorry if I'm being abrasive - didn't get much sleep) 2016-03-21 10:10:25 few daemons actually do this. 2016-03-21 10:10:37 of course, a software that does not handle a failing malloc correctly is written wrong 2016-03-21 10:10:52 and of course you have to fix something manually when a daemon crashes. You have to fix the thing that made it possible in the first place so the situation doesn't happen again. 2016-03-21 10:10:59 the same with a full disk, if a software does not handle it corretly the software is written wrong 2016-03-21 10:11:17 in many cases, the correct way to handle a failing malloc is to exit. 2016-03-21 10:11:26 Because there's simply no way to proceed on. 2016-03-21 10:11:42 And you'll see *lots* of examples around. 2016-03-21 10:12:21 anyway, very often, it will take some time for a daemon to get to the point where it crashes again. 2016-03-21 10:12:40 *In any case*, as soon as it *first* crashes, the admin should be notified and should investigate. 2016-03-21 10:12:57 But in the meantime, a supervision framework will make it so your daemon serves, or attempts to serve. 2016-03-21 10:13:08 Which is better than being down all night. 2016-03-21 10:13:11 imho, there are other reasons why supervision is a good idea 2016-03-21 10:13:30 yes, there are other reasons, but I wanted to address this point because it's a common strawman 2016-03-21 10:13:30 restarting services is not the only reason 2016-03-21 10:14:19 sure, there are other advantages (and you can even use a supervision framework *without* restarting the services if they crash) 2016-03-21 10:14:54 i suppose in some situation you want it to restart and in some not 2016-03-21 10:14:59 but in general, people saying "supervision isn't good because I don't want my daemon to crash over and over again, I want to fix the problem instead" simply don't realize that both things aren't mutually exclusive 2016-03-21 10:15:10 so its a config option 2016-03-21 10:15:12 you *should* fix the problem as soon as you get a crash 2016-03-21 10:15:19 +1 2016-03-21 10:15:25 but restarting is most of the time still better than not restarting 2016-03-21 10:16:09 i think in many real life situations, restarting the service is a way to not fix the real problem 2016-03-21 10:16:36 in some peoples eyes "restart service = problem solved" 2016-03-21 10:16:36 that's a sysadmin failure, not a framework failure. 2016-03-21 10:16:41 it is 2016-03-21 10:17:59 i think it is a common sysadmin failure, which has led to the "supervision isn't good because I don't want my daemon to crash over and over again" 2016-03-21 10:18:01 and if leo-unglaub's service had really been critical, his company would have had a monitoring system, and a night team with a pager, who would have worked on the problem as soon as it occurred 2016-03-21 10:18:38 skarnet: it was very critical 2016-03-21 10:18:45 central mariadb crashed 2016-03-21 10:19:06 would be a good idea if 2 more comands were available buildmk buildpkg ? 2016-03-21 10:19:06 this is just split of `abuild build` 2016-03-21 10:19:25 vkris: ? 2016-03-21 10:19:44 leo-unglaub: then why wasn't it solved when you woke up? don't you have an oncall system? 2016-03-21 10:19:55 I mean I wanted to do 'abuild build' but not populate $pkgdir 2016-03-21 10:20:06 vkris: you can do that now 2016-03-21 10:20:08 skarnet: because i an the guy who gets woken up to fix stuff 2016-03-21 10:20:17 ah, you were the oncall 2016-03-21 10:20:21 i get called automatically when a crytical daemon dies 2016-03-21 10:20:35 just like 'make' , but no fakeroot yet 2016-03-21 10:20:45 vkris: abuild build does that right now 2016-03-21 10:20:56 abuild rootpkg will do the fakeroot 2016-03-21 10:20:56 so you got woken up by an alert, and went to work on the problem? sounds like it's working as intended then, supervision or not :) 2016-03-21 10:21:18 yes 2016-03-21 10:21:24 "Compile AND install package into $pkgdir" 2016-03-21 10:21:34 leo-unglaub: so you woke up early today. we understand that you may be grumpy :) 2016-03-21 10:21:50 m point is that every time something like that happens supervision simply would not have helped in my cases because i had to do something to fix it besides restarting 2016-03-21 10:22:00 you always have to 2016-03-21 10:22:10 supervision isn't a magical fix that will allow you to slack :P 2016-03-21 10:22:32 but if the problem is temporary, at least it keeps your services kinda running while you're working on the issue. 2016-03-21 10:23:07 or just limit the number of retries 2016-03-21 10:23:12 might be easier to analyze the real problem if the service is not running 2016-03-21 10:23:22 does "testing/android-tools" still build ? 2016-03-21 10:23:24 apologies for chipping in my 5 cents here, but why would want to wake up (and pay) an engineer to "work on a problem" before automatic remediation has failed (flap detection on the supervisor is pretty basic) ? 2016-03-21 10:23:33 ncopa: no, skarnet is grumpy ... i am .. well ... i am grumpy to ... but that has nothing todo with getting up so early ... more with having to work all the time and not getting payed .. 2016-03-21 10:23:36 hehe 2016-03-21 10:23:57 not getting paid is a more serious problem than mariadb failing imo :P 2016-03-21 10:24:13 +1 2016-03-21 10:24:15 someone hit an edge case in your application, and it goes down for the duration until an engineer has "looked at the problem" is bananas imho 2016-03-21 10:24:33 also, high availability mate 2016-03-21 10:24:34 well, we all got fired ... but .. still every time a daemon crashes i feel like i have to fix it 2016-03-21 10:24:40 still feels like my server 2016-03-21 10:24:49 thorduri: depends on the problem 2016-03-21 10:24:56 thorduri: but normally yes 2016-03-21 10:25:06 leo-unglaub: keep your Red Cross mind out of your job 2016-03-21 10:25:11 else you'll never get any sleep 2016-03-21 10:25:16 ncopa: of course 2016-03-21 10:25:50 if it were me... oh, I got fired? then the server can crash and burn and that's too bad 2016-03-21 10:26:06 but I'm not exactly a philanthropist 2016-03-21 10:26:58 ncopa: buildmk == make and buildpkg == install package into $pkgdir 2016-03-21 10:27:04 ACTION takes off the hat for leo-ungalub 2016-03-21 10:27:06 hope making sense 2016-03-21 10:27:56 vkris: 'abuild build' will only call the build() { ... } function in APKBUILD, not the 'install' 2016-03-21 10:28:08 ah, ok 2016-03-21 10:28:20 thanks 2016-03-21 10:28:27 'abuild rootpkg' will fakeroot, install files in $pkgdir and create .apks 2016-03-21 10:28:35 skarnet: well, i developed the server infrastructure over the last 8 years .. so i still feel responsible for it ... and since i dont find another job ... i kind of do that 2016-03-21 10:29:10 ah, it's your baby. I understand. But why the heck did they fire you if you're the one with the knowledge o.O 2016-03-21 10:29:39 skarnet: they want to go to the usa and do some startup shit ... 2016-03-21 10:29:45 and i dont want to go to america 2016-03-21 10:30:00 and i hate apps ... i am a real softwar developer ... not some app hipster 2016-03-21 10:30:17 can I high-five you even though it's a typical American thing? XD 2016-03-21 10:30:33 sure bro *g* 2016-03-21 10:30:46 ACTION high-fives skarnet 2016-03-21 10:30:46 ACTION high-fives leo-unglaub 2016-03-21 10:31:06 ACTION high-fives leo-ungalub 2016-03-21 10:31:29 ncopa: didn't you just get hired by a US company? /grin 2016-03-21 10:31:39 there are real life situation when you get a close source piece of crap which is very crash prone, the only way to keep the service alive is to monitor/supervise and restart automatically 2016-03-21 10:31:50 i got hired, but i didnt have to go to usa 2016-03-21 10:32:16 ncopa: best of both worlds :D 2016-03-21 10:32:23 yeah, they're bit smarted than the average company 2016-03-21 10:32:28 a bit smarter* 2016-03-21 10:32:31 I can't keyboard today 2016-03-21 10:32:41 try a touchscreen 2016-03-21 10:32:49 ew 2016-03-21 10:32:51 lol 2016-03-21 10:33:37 haha, i really doubt that the same company as ncopa is going to hire me ... not if they read what i have written about there software on twitter *g* 2016-03-21 10:33:53 hehehe 2016-03-21 10:34:11 hehe docker will be better now ncopa is working for them :D 2016-03-21 10:34:40 if Docker chose to switch to Alpine, they can't be *all* bad :) 2016-03-21 10:34:49 ncopa: who would be 'soren' ? 2016-03-21 10:34:55 docker gets many things right 2016-03-21 10:34:58 I mean, Google is switching to musl :) 2016-03-21 10:35:13 things are getting better! 2016-03-21 10:35:15 I mean on irc 2016-03-21 10:35:35 vkris: its nmeum 2016-03-21 10:35:58 skarnet: really? 2016-03-21 10:36:06 Google is switching to musl? 2016-03-21 10:36:19 _what_ is actually switching? 2016-03-21 10:36:26 nmeum: could pls have look at testing/android-tools if it still builds 2016-03-21 10:36:33 chromeOS? Gentoo + musl? 2016-03-21 10:36:35 I'll have to check what team exactly is 2016-03-21 10:37:17 but one team is interested and there are long discussions to make musl licensing Google-lawyer compatible 2016-03-21 10:37:27 uff 2016-03-21 10:37:28 lol :) 2016-03-21 10:37:45 lmao 2016-03-21 10:37:59 it's matter of patching their entire lawyer's git tree 2016-03-21 10:38:19 MIT is not good for them? 2016-03-21 10:38:26 it's more complicated than that 2016-03-21 10:38:31 mostly has to do with public domain files 2016-03-21 10:38:54 PD isn't recognized in all countries, lawyers feel people could claim rights on them, etc. etc. 2016-03-21 10:39:20 check the musl ML archives - the signal to noise ratio has tanked these days, unless legal discussions is signal to you 2016-03-21 10:39:46 yes, the license discussion on ML was very interesting 2016-03-21 10:40:00 i normally hate that kind of discussions 2016-03-21 10:40:12 but whats interesting is that google wants to use musl 2016-03-21 10:40:14 first grsec, now musl ;(( 2016-03-21 10:40:31 vkris: no, musl won't go the way of grsec 2016-03-21 10:41:05 musl is not having thesame faith as grsec 2016-03-21 10:41:08 the worst thing that can happen is that Google won't use it 2016-03-21 10:41:17 because the musl maintainer is a hardliner on open source software 2016-03-21 10:43:07 https://twitter.com/RichFelker/status/711211449109110784 2016-03-21 10:43:17 this guy is worth following on twitter! 2016-03-21 10:43:29 he's worth following anywhere :P 2016-03-21 10:43:38 there is something worth following in Twitter? :) 2016-03-21 10:43:51 definitely. I like Twitter. :) 2016-03-21 10:44:07 (the users, not the company.) 2016-03-21 10:48:08 skarnet: wow, he follows you on twitter to 2016-03-21 10:48:18 maybe he is a s6-* fanboy *g* 2016-03-21 10:48:39 I like to think he values my technical thinking :P 2016-03-21 10:51:55 i have to go ... i have to meet with an old customer who complains that her iphone is not syncing her calendar over itunes in outlook .. like thats my fault if apple and microsoft fuck up ... 2016-03-21 10:51:58 see you guys later 2016-03-21 10:52:25 try to get some money out of this kind of thing! :D 2016-03-21 10:52:48 i dont ... but they always offer me a fresh coca cola ... 2016-03-21 10:52:58 leo-unglaub: be happy if that gets income 2016-03-21 10:53:19 why do you think i drive so much night shifts as a paramedic at the red cross? because there are free donougts there ;) 2016-03-21 10:53:21 or food ;) 2016-03-21 10:53:25 once I have a .apk file how can I "install" it? 2016-03-21 10:53:54 i.e. one builds a package locally and wants the equivalent of apk add mypackage 2016-03-21 10:56:58 xificurC: you can do "apk add xxx.apk" or add your build directory to /etc/apk/repositories 2016-03-21 10:58:00 i cannot resist ... skarnet .. you have to excuse me but i have to do it .... "i would love to use s6-supervision if it where written in rust" ... sorry, i had to do that joke ;) 2016-03-21 10:58:04 kunkku: I see, thanks 2016-03-21 10:58:10 ncopa will get it ;) 2016-03-21 10:58:38 xificurC: if you use the latter method, you may want to copy your public key to /etc/apk/keys 2016-03-21 10:59:00 :D 2016-03-21 11:01:28 I like metal, so I'm an opponent of rust 2016-03-21 11:03:20 Rust is system programming language, it keeps you as close to metal as C, but with safety 2016-03-21 11:04:02 hmmm... like flaky stuffs ;) 2016-03-21 11:04:35 so you need "Servo" 2016-03-21 11:04:59 wonder autobile industry had anything to do with this development 2016-03-21 11:05:34 jirutka: metal (music). metal (matter). rust, eating metal (matter). *cough* 2016-03-21 11:05:53 ACTION wonders why he has to keep explaining bad jokes 2016-03-21 11:05:59 so at sevice center if technician say "Replace Servo" 2016-03-21 11:06:25 the other would say Browser or Oil ? 2016-03-21 11:08:46 nmeum: when at android-tools, could you see if there is any 'unmkbootimg' and if possible to include 2016-03-21 11:11:46 then the 1st technician say, ah, both got dirty, replace em` all 2016-03-21 11:12:19 safe metal! 2016-03-21 11:18:32 clandmeter: we let nl.a.o sync every hour? 2016-03-21 11:18:52 every 5 min 2016-03-21 11:18:54 :) 2016-03-21 11:19:20 then probably better sync on mqtt messages 2016-03-21 11:19:27 +1 2016-03-21 11:20:02 other option is to set it up as reverse caching proxy 2016-03-21 11:20:29 lots of ppl use nl.a.o i guess 2016-03-21 11:20:52 which is why i think it might be an idea to let it sync hourlyl 2016-03-21 11:21:13 ok but pkgs will be one hours off 2016-03-21 11:21:16 people normally uses nl.a.o for instant 2016-03-21 11:21:21 hm 2016-03-21 11:21:27 1+(+/-15min) 2016-03-21 11:31:53 clandmeter: it should now sync every 15 mins 2016-03-21 11:31:58 keep an eye on it 2016-03-21 12:39:10 and now there is 'redox' rust-written microkernel 2016-03-21 13:33:57 is Leonardo Arena in the irc channel? 2016-03-21 13:36:49 hi leo-unglaub 2016-03-21 13:37:26 rnalrd: are you Leonardo Arena? 2016-03-21 13:37:29 yep 2016-03-21 13:37:56 great :) i have question about your owncloud package. does it allow multiple instances at once? 2016-03-21 13:38:09 for example one owncloud for me an one for a company? 2016-03-21 13:39:43 currently supports only one instance because its config file is in a fixed location 2016-03-21 13:40:11 /etc/owncloud 2016-03-21 13:40:17 so it goes into lbu 2016-03-21 13:40:33 supporting multiple instance means going back to default location 2016-03-21 13:40:59 so when i install owncloud i have just one setup on one server? 2016-03-21 13:41:36 y 2016-03-21 13:42:18 i see, thanks! 2016-03-21 13:43:02 yw 2016-03-21 13:45:05 leo-unglaub, they config location it's just a symlink 2016-03-21 13:45:24 so you can easy point different installations on different configs 2016-03-21 13:46:19 the package version needs to be the same across all installation on the same server of course 2016-03-21 13:46:40 hmm, indeed. if the version is the same i coulf softlink everything i need into the web root 2016-03-21 13:46:50 y 2016-03-21 13:47:07 i think i will do that 2016-03-21 13:49:11 need to tweak db name upon installation, if not using sqlite 2016-03-21 13:55:00 ncopa: could you please bump firefox? thanks :) 2016-03-21 14:06:03 leo-unglaub: will do but not right now 2016-03-21 15:22:53 good day, can someone help with questions about xen? 2016-03-21 15:25:52 xen channel too quiet? :P 2016-03-21 15:26:06 give it a shot 2016-03-21 15:26:23 i mean give us your question 2016-03-21 15:26:32 i'm interested to know if xsm was compiled into the current xen in the alpine distro 2016-03-21 15:27:41 the build log is not present in the current xen package 2016-03-21 15:33:27 it doesn't look like XSM is enabled 2016-03-21 15:36:23 it's disabled by default, and needs to be enabled at configure time, and i see no evidence that it is done at package build time 2016-03-21 15:37:46 if you'd like to see it, create a feature request at http://bugs.alpinelinux.org/projects/alpine/issues 2016-03-21 15:41:51 I see, 2 more questions if I may 2016-03-21 15:42:19 how long will it take to add such feature? 2016-03-21 15:43:31 can I do it by myself locally by compiling xen by myself? 2016-03-21 15:44:04 what is xsm? 2016-03-21 15:44:32 http://wiki.xen.org/wiki/Xen_Security_Modules_:_XSM-FLASK 2016-03-21 15:46:41 i dont think we want to push that for 3.3-stable 2016-03-21 15:47:02 but i think its ok to ship alpine 3.4 with it 2016-03-21 15:47:40 when should it be? 2016-03-21 15:47:51 beginning of May 2016-03-21 15:48:07 but we might need help testing it before that 2016-03-21 15:49:28 can you explain how can I compile xen locally with this feature enabled? 2016-03-21 15:49:38 and I might be able to help with testing 2016-03-21 15:50:05 i think: git clone git://git.alpinelinux.org/aports 2016-03-21 15:50:11 edit main/xen/APKBUILD 2016-03-21 15:50:30 abuild-keygen -a -i 2016-03-21 15:50:58 and: cd main/xen && abuild -r 2016-03-21 15:51:05 should give you the built package 2016-03-21 15:51:41 is there any meta-package for development env? 2016-03-21 15:56:24 alpine-desk 2016-03-21 15:56:27 sorry 2016-03-21 15:56:29 alpine-sdk 2016-03-21 15:56:38 should pull in git, abuild gcc etc 2016-03-21 15:58:44 ncopa: I have found a weird bug in busybox sh 2016-03-21 15:58:58 and I'm having a very hard time coming up with a minimal case 2016-03-21 15:59:06 sounds fun :) 2016-03-21 15:59:16 reminds me 2016-03-21 15:59:23 i should update busybox 2016-03-21 15:59:33 well, it also means I can't build my stuff on a pristine alpine box :P 2016-03-21 15:59:49 what happens? 2016-03-21 15:59:59 Guest43: create a patch file for config.mk or whatever that changed the value of XSM_ENABLE or FLASK_ENABLE (or whatever) then submit the patch to the alpine-aports mailing list 2016-03-21 16:00:47 ncopa: ./configure for s6-dns (and only this one) will write the wrong CC into config.mak when cross-compiling, because ${CC_AUTO##${cross}} doesn't remove the prefix correctly 2016-03-21 16:01:21 what is the general opinion on toybox as a serious alternative to busybox (purely for religious reasons)? 2016-03-21 16:01:24 and my other packages *work*, with the exact same ./configure script except one completely irrelevant line 2016-03-21 16:01:38 ScrumpyJack: Ok, I'll try. Going over APKBUILD to understand where to change/add such feature 2016-03-21 16:02:07 add your patch to the sources variable 2016-03-21 16:02:59 ScrumpyJack: some say toybox is technically better. It's also not quite complete. afaict the difference is mostly religious. 2016-03-21 16:03:10 skarnet: what happens if you do: ${CC_AUTO##"${cross}"} 2016-03-21 16:03:34 can you also echo "'$cross'" 2016-03-21 16:03:48 to see if you have any leading spaces or similar 2016-03-21 16:04:26 I don't 2016-03-21 16:04:32 everything appears to work 2016-03-21 16:04:42 it just doesn't work in that particular instance of the script 2016-03-21 16:04:56 and only this one 2016-03-21 16:05:05 huh 2016-03-21 16:05:09 thats is weird 2016-03-21 16:05:18 that's the first thing I said :P 2016-03-21 16:05:37 I have quadruple-checked my parameters, they are valid 2016-03-21 16:05:51 so i agree. it is weird :) 2016-03-21 16:05:58 all my other packages with the exact same configure script save one irrelevant line work without trouble 2016-03-21 16:06:13 this one gets the same parameters, the same operations, and ## fails 2016-03-21 16:06:25 can you #!/bin/sh -x it? 2016-03-21 16:06:39 so you can see the expansion? 2016-03-21 16:06:55 I've done it. It appears to work. The ## operation happens in a here document though, so I can't see it 2016-03-21 16:07:33 what is the value of ${cross}? 2016-03-21 16:07:43 and CC_AUTO 2016-03-21 16:08:16 CC_AUTO=x86_64-linux-musl-muslgcc (yes, the muslgcc is intentional) and cross=x86_64-linux-musl- 2016-03-21 16:08:35 if you try the ## by hand, it works 2016-03-21 16:08:44 it works in all my other configure scripts, too. But not this one. 2016-03-21 16:08:49 and if you do it in a heredoc? 2016-03-21 16:09:00 it works, too. 2016-03-21 16:09:08 the other configure scripts do it in a heredoc. 2016-03-21 16:09:15 it's insane. 2016-03-21 16:09:18 except in that specific configure script 2016-03-21 16:09:22 yes. 2016-03-21 16:09:26 what does it expand it to? 2016-03-21 16:09:54 to the whole $CC_AUTO without removing the prefix 2016-03-21 16:10:35 I've run the configure script with bash, it works 2016-03-21 16:11:03 need to figure out how to reproduce it then 2016-03-21 16:11:13 and make a busybox testcase of it 2016-03-21 16:11:13 I know how to reproduce it 2016-03-21 16:11:21 I just don't know how to minimize the test case 2016-03-21 16:11:50 i think you should take it up in busybox mailing list 2016-03-21 16:12:04 does it happen on x86_64 only? or on 32bit too? 2016-03-21 16:12:22 I haven't tested on other archs tbh 2016-03-21 16:13:06 and if you expand it to other variable outside the heredoc? 2016-03-21 16:13:19 I'll try that 2016-03-21 16:13:41 _tmp_CC="${CC_AUTO##${cross}}" 2016-03-21 16:13:47 and use $_tmp_CC in the ehredoc 2016-03-21 16:14:23 getting the script to a point where it works isn't the problem 2016-03-21 16:14:33 also, test with ${CC_AUTO#${cross}} 2016-03-21 16:14:46 exhibiting the bug without having to run that configure script is XD 2016-03-21 16:14:47 figuring out minimal test case is your problem, i understand 2016-03-21 16:23:17 i'm pushing 4.4 kernel now 2016-03-21 16:24:53 oh jeez oh jeez 2016-03-21 16:25:12 this configure script is haunted 2016-03-21 16:25:27 I'm running it and it blocks in D state now 2016-03-21 16:58:44 oh nice, I've been waiting for 4.4 2016-03-21 17:18:19 github is only serving me unicorns now 2016-03-21 17:18:21 i think its down 2016-03-21 17:33:31 on some pages, on others works fine 2016-03-21 17:35:56 its back 2016-03-21 17:54:15 yes 2016-03-21 18:20:08 4.4 nice !! 2016-03-21 18:23:45 what if drbd9-grsec moved to community 2016-03-21 18:24:00 if its working okish 2016-03-21 18:25:06 now feels like edge ;)) 2016-03-21 19:05:39 Timo Thanks 2016-03-21 19:05:53 for patching the non booting issue with RPI 2016-03-21 19:12:25 ncopa: still here? 2016-03-21 19:13:50 depends... 2016-03-21 19:13:51 :) 2016-03-21 19:14:15 :) 2016-03-21 19:15:05 it seems mariadb ships its initd in mariadb-common 2016-03-21 19:15:29 mariadb-libs pulls in mariadb-common 2016-03-21 19:15:53 while i already doubht thats needed, there should be no need for the initd file. 2016-03-21 19:16:19 and openrc will complain as this initd file depends on some mariadb bins 2016-03-21 19:21:10 it complains about my_print_defaults? 2016-03-21 19:21:23 i think so 2016-03-21 19:21:45 mariadb-common mentions its for client and server 2016-03-21 19:21:54 then why its a dep of libs 2016-03-21 19:22:30 i think the dev wanted to move the my.cnf to common, and didnt think about it also moves thei nitd 2016-03-21 19:22:36 can you try: http://tpaste.us/AaEo 2016-03-21 19:23:27 oh 2016-03-21 19:23:59 the init.d should not be shipped with -common 2016-03-21 19:24:07 right 2016-03-21 19:24:21 but what about all those files in -libs? 2016-03-21 19:24:28 but the init.d script should be fixed too 2016-03-21 19:24:32 shouldnt the so file be the only needed thing? 2016-03-21 19:24:54 might be that the libraries needs data files 2016-03-21 19:25:53 ok 2016-03-21 19:26:23 looks like it 2016-03-21 19:27:51 ncopa: dont you ever sleep 2016-03-21 19:28:14 oneinsect_raydav: what are you talking about? i am sleeping now... 2016-03-21 19:28:23 lol 2016-03-21 19:28:24 :) 2016-03-21 19:28:50 i am amazed at the kind of energy you guys bring in to alpine 2016-03-21 19:30:11 clandmeter: can you please test this: http://tpaste.us/3qV9 2016-03-21 19:30:53 good evening 2016-03-21 19:31:22 what a shit day ... took me 8 hours to get an iphone calendar over itunes to sync calendars to outlook in windows ... 2016-03-21 19:31:37 here is the bug .. itunes hangs when it uses an usb3 port 2016-03-21 19:31:44 with usb2 it works sometimes .. 2016-03-21 19:31:53 took me 8 hours to debug that 2016-03-21 19:32:19 leo-unglaub: dare to invoice full 8 hours? :) 2016-03-21 19:32:39 i got ice tea and a big chockolate bar !!1! 2016-03-21 19:32:46 :D 2016-03-21 19:32:58 you should have a beer minimum for that... 2016-03-21 19:34:05 hehe, i want to fotgett that entire day ... just remember: windows 10, itunes + usb3 = broken! 2016-03-21 19:34:52 i want to forgett the rest ... for example that other people get for similar work 2000€ a month ... and i got chockolate ... 2016-03-21 19:34:54 try with wine 2016-03-21 19:34:58 should be much better 2016-03-21 19:35:03 :p 2016-03-21 19:35:06 or scotch 2016-03-21 19:37:11 leo-unglaub: you are always complaining about bad software, and now you talk about itunes? 2016-03-21 19:37:45 clandmeter: yeah, what can i do if my customer uses that crap 2016-03-21 19:38:05 or does the red cross give away free iphones? 2016-03-21 19:38:21 clandmeter: that has nothing todo with the red cross 2016-03-21 19:38:31 all my red cross stuff is just as a voulenteer 2016-03-21 19:38:43 this customer is from my dayjob i used to have 2016-03-21 19:38:56 and volunteering is exactly what you did today :) 2016-03-21 19:39:30 i know what leo-unglaub deserves after a day like this 2016-03-21 19:39:34 rust! 2016-03-21 19:39:42 hehe 2016-03-21 19:39:45 muuaaaahhhhhhhhhhh 2016-03-21 19:39:49 hahahaha 2016-03-21 19:39:51 it actually makes sence in dutch 2016-03-21 19:40:07 or was it sense 2016-03-21 19:40:08 :) 2016-03-21 19:40:17 sense :) 2016-03-21 19:40:31 can someone please give leo-unglaub a rust? 2016-03-21 19:41:00 ACTION would, but is rust-free 2016-03-21 19:41:43 i would too, but i am sleeping 2016-03-21 19:42:05 ACTION gives ncopa 1000 liters of coffee 2016-03-21 19:42:38 that will rust anything :/ 2016-03-21 19:43:53 http://www.cutecaptions.com/2012/08/im-so-sleepy-even-coffee-dont-help/ 2016-03-21 19:45:04 rust is just metal burning really slowly :) 2016-03-21 19:45:16 cats everywhere 2016-03-21 19:45:32 mitchty: it's a waste of allomantic power! 2016-03-21 19:47:37 leo-unglaub | clandmeter: yeah, what can i do if my customer uses i that crap < change customer ? 2016-03-21 19:48:11 coredumb: hehe, well officially they are not my customers anymore 2016-03-21 19:48:24 so just let them fuck themselves 2016-03-21 19:48:44 i can't ... 2016-03-21 19:48:52 why not ? 2016-03-21 19:48:58 i feel sorry for them not having there it under control 2016-03-21 19:49:32 ok ... so ? 2016-03-21 19:49:41 I feel sorry for a lot of people 2016-03-21 19:50:34 i have a lot of time .. because i am withot a job ... so i thought i help them ... because it was my job ... 2016-03-21 19:50:42 okay, maybe its just me but that sounds stupid 2016-03-21 19:51:20 don't worry it's not just you :P 2016-03-21 19:51:44 doing your paid job for free 2016-03-21 19:51:55 yeah sounds a bit stupid indeed 2016-03-21 19:52:06 that's just generous volunteering. 2016-03-21 19:52:26 ohhhhh 2016-03-21 19:52:34 or stupid 2016-03-21 19:52:36 :P 2016-03-21 19:52:42 it's not stupid, just too altruistic. 2016-03-21 19:52:54 Don't call generous people stupid. We need more generous people. 2016-03-21 19:53:02 ACTION grabs a dictionary 2016-03-21 19:53:27 that's a thin line there 2016-03-21 19:54:18 altruism in excess only hurts the altruistic person. Stupidity in excess can hurt *a lot* of people. 2016-03-21 19:54:45 you may be true 2016-03-21 19:54:47 to drink and drive is stupid. That's the difference. :) 2016-03-21 19:56:47 well as long as you can pay the bills at the end of the month why not work for free 2016-03-21 20:46:26 vkris: I wanted to refactor the android-tools aport anyhow. It's pretty outdated but building android-tools is a pain in the ass therefore I didn't do any further work on it so far 2016-03-21 21:06:50 ncopa: redmine is running fine i think, so we can switch tomorrow (just need to check mqtt and smtp). 2016-03-21 23:12:25 hm are there cases where apk can leave the system in an unfixably inconsistent state 2016-03-22 07:37:25 good morning 2016-03-22 07:37:28 please git pull git://git.alpinelinux.org/user/fab/aports 2016-03-22 07:37:42 just a couple of updates 2016-03-22 07:59:48 hi 2016-03-22 08:00:06 A question about package compilation workflow: 2016-03-22 08:00:43 sup 2016-03-22 08:00:59 I want to add a feature in configure file. abuild fetch -> unpack -> prepare -> package ? Is this the right way? 2016-03-22 08:01:20 ncopa, can i specify from syslinux which overlay alpine needs to load? 2016-03-22 08:01:36 if i want to support multiple overlya, e.g. 2016-03-22 08:01:40 *overlays 2016-03-22 08:02:17 fcolista: i think currently not 2016-03-22 08:03:26 ok 2016-03-22 08:03:41 i saw something like that in b.a.o 2016-03-22 08:03:54 probably it's a feature request 2016-03-22 08:04:43 MazeWalker: sounds like that, but I guess I would just patch it from APKBUILD 2016-03-22 08:04:58 abuild -r will do the magic 2016-03-22 08:05:21 barthalion , when I do that, the build overwrites my edits 2016-03-22 08:05:39 and I still do not know how to create the patch, the right way 2016-03-22 08:05:47 ah, I see 2016-03-22 08:05:52 MazeWalker: i think what you want is: abuild deps fetch unpack prepare 2016-03-22 08:05:58 edit the stuff 2016-03-22 08:06:01 and then abuild build 2016-03-22 08:06:09 once it builds you can do: abuild rootpkg 2016-03-22 08:06:38 if you want fix how files are put in subpackages, then you can rm -r pkg/ and again do abuild rootpkg 2016-03-22 08:06:42 ncopa, abuild rootpkg will only prepare the final package without installing it? 2016-03-22 08:06:47 any idea why numpy is not build for armhf? I triggered a rebuild, but looks like arm builders are busy with other packages 2016-03-22 08:07:18 and it has arch=all 2016-03-22 08:20:46 MazeWalker: correct. will build the .apks 2016-03-22 08:21:44 @ncopa, thanks. Another one. How can I prepare a patch file and test it? Using git? 2016-03-22 08:21:54 yes 2016-03-22 08:22:14 i sometimes do cd src/foo-1.0/ && git init; git add .; git commit -minit 2016-03-22 08:22:23 and create patch from git diff 2016-03-22 08:22:39 and sometimes i just cp path/to/file.c path/to/file.c.orig 2016-03-22 08:22:57 and diff it with: diff -u ./path/to/file.c.orig ./path/to/file.c 2016-03-22 08:23:51 and sometimes i pull the sources from latest upstream git master, make it build there, and then backport the patch to the release i try build package for 2016-03-22 08:23:56 and submit fix to upstream 2016-03-22 08:24:48 @ncopa, more then one way to do it :-) 2016-03-22 08:25:19 yup :) 2016-03-22 08:25:28 generally its a good idea to work with upstream 2016-03-22 08:25:34 as early as possible 2016-03-22 08:25:56 and the backport whatever fixes they accept 2016-03-22 08:28:02 build is done, if everything is fine, I'll submit a feature request with the patch 2016-03-22 08:32:03 i think b.a.o is going down for maintenance 2016-03-22 08:35:00 yeah sorry, was watching the news and trying to migrate at the same time. 2016-03-22 08:39:07 these are interesting times we live in 2016-03-22 08:40:49 very sad :-( 2016-03-22 09:21:04 b.a.o should be online again. 2016-03-22 09:31:46 is it just me or is it snappier now? 2016-03-22 09:34:50 @ncope, I have package which needs selinux libs but I can't find such in aports. Is there a reason for that or selibs have other names? 2016-03-22 09:35:42 we dont really support selinux 2016-03-22 09:35:43 ncopa: should be faster 2016-03-22 09:35:56 better disk io on db 2016-03-22 09:36:08 maybe some performance increase by puma 2016-03-22 09:36:09 but it should be possible to build libselinux 2016-03-22 09:36:19 clandmeter: great! 2016-03-22 09:36:45 @ncopa: Do I need to create apk for that? 2016-03-22 09:36:52 but redmine has always been kind of light on db queries 2016-03-22 09:37:26 a lot of modules have been updated, so exepct some functions to fail. 2016-03-22 09:37:44 MazeWalker: yes please 2016-03-22 09:41:41 @ncopa: should i do it in testing? 2016-03-22 09:41:47 yes please 2016-03-22 11:08:50 519798376 2016-03-22 11:08:54 oops 2016-03-22 11:10:40 time to change password... :) 2016-03-22 11:14:49 or telephone number :P 2016-03-22 11:21:03 using telephone number as password is a bad idea :D 2016-03-22 11:22:15 good morning 2016-03-22 11:22:22 morning leo-unglaub 2016-03-22 11:23:10 last nights updates where very awesome .. now my boot resolution is at maximum 2016-03-22 11:23:13 very nice :) 2016-03-22 11:45:16 nice 2016-03-22 11:45:33 i must be something new in the kernel 2016-03-22 11:45:40 but its very nice, i like it! 2016-03-22 12:10:07 thinking to pkg this https://github.com/jpmens/jo 2016-03-22 12:10:19 any thing similar yet ? 2016-03-22 12:11:15 jq? 2016-03-22 12:12:33 hmm shell variables to json ? 2016-03-22 12:12:39 let me check 2016-03-22 12:15:33 @ncopa could you please take a look at https://github.com/alpinelinux/aports/pull/37 ? 2016-03-22 12:24:12 >>> maven: Entering fakeroot... 2016-03-22 12:24:12 install: can't create '/home/ncopa/aports/testing/maven/pkg/maven/usr/share/java/maven-3/bin/m2.conf': No such file or directory 2016-03-22 12:24:34 there are also whitespace damage 2016-03-22 12:24:37 + install -m644 -Dt "$destdir/bin" ./bin/*.conf || return 1 2016-03-22 12:24:38 + install -m644 -Dt "$destdir/boot" ./boot/*.jar || return 1 2016-03-22 12:24:38 + install -m644 -Dt "$destdir/lib" ./lib/*.jar || return 1 2016-03-22 12:24:38 + cp -Rp ./conf "$destdir/" || return 1 2016-03-22 12:24:44 has spaces before tabs 2016-03-22 12:25:57 has someone an thunderbird apk? 2016-03-22 12:26:15 also quoted pkgver hurts my eyes 2016-03-22 12:27:46 thanks, I’ll take a look; and I’m sorry whitespace issues, this usually don’t happen to me 2016-03-22 13:11:51 :highlight ExtraWhitespace ctermbg=red guibg=red 2016-03-22 13:11:52 :match ExtraWhitespace /\s\+$/ 2016-03-22 13:11:52 :match ExtraWhitespace /\s\+$\| \+\ze\t/ 2016-03-22 13:34:12 ncopa: visual studio also has a function to highlight extra whitespace 2016-03-22 13:34:44 Hi there! 2016-03-22 13:34:50 so i no longer need to use vim? :) 2016-03-22 13:35:02 1st, thanks for such amazing distro! 2016-03-22 13:35:06 no sir .... microsoft got us covered ! 2016-03-22 13:35:09 hi kikitux 2016-03-22 13:35:37 http://patchwork.alpinelinux.org/patch/1776/ 2016-03-22 13:36:00 I did open the bash patch, for some reason the hash are different between the build file and what is on the ftp 2016-03-22 13:36:16 kikitux: the question was why is the patch different 2016-03-22 13:36:25 what changed in the patch 2016-03-22 13:36:49 the hash did change upstream, the patch is to update md5/sha256/sha512 otherwise it doesn't pass checksum check 2016-03-22 13:37:07 understand 2016-03-22 13:37:13 something changed upstream 2016-03-22 13:37:19 i wonder what changed 2016-03-22 13:37:21 and why 2016-03-22 13:37:32 its not first time a project silently changes things 2016-03-22 13:38:06 you can find the original patch on http://distfiles.alpinelinux.org/distfiles/bash043-042 2016-03-22 13:38:15 cool, I was thinking I was explaining poorly. No idea why they did republish bash43-042 2016-03-22 13:38:19 I can check what change 2016-03-22 13:38:28 i think exact url is http://distfiles.alpinelinux.org/distfiles/bash43-042 2016-03-22 13:38:44 yes, please run a diff 2016-03-22 13:39:12 this is exactly why we do checksums 2016-03-22 13:39:20 so people cannot change things without we notice 2016-03-22 13:44:43 ncopa: yeah, very important ... there is way to much tempering with existing releases going on these days 2016-03-22 13:45:18 i am sorry, it seems it created a new ticket instead to append to the existing one 2016-03-22 13:45:26 http://patchwork.alpinelinux.org/patch/1778/ 2016-03-22 13:47:50 -*** ../bash-4.3-patched/parse.y 2015-05-18 19:27:05.000000000 -0400 2016-03-22 13:47:52 +*** ../bash-4.3-patched/y.tab.c 2015-05-18 19:27:05.000000000 -0400 2016-03-22 13:47:54 ncopa: this is the diff 2016-03-22 13:48:01 understand 2016-03-22 13:48:18 i'm just confused 2016-03-22 13:48:22 why would they do that 2016-03-22 13:48:30 and how do we fix it 2016-03-22 13:48:31 because they put parse.y twice 2016-03-22 13:48:53 when they meant y.tab.c 2016-03-22 13:49:01 but result when applying the patch should be the same? 2016-03-22 13:49:04 the patch updates the hash, after that the builds works just fine and compiles the package without further changes 2016-03-22 13:49:12 ^ alpine patch 2016-03-22 13:49:17 kikitux: understand 2016-03-22 13:49:30 the problem is that the sourcefile is cached 2016-03-22 13:50:33 ncopa: no, it is just a coincidence that it applies anyway 2016-03-22 13:59:56 awilfox: so the resulting binary is different? 2016-03-22 14:00:06 because it patches a different file 2016-03-22 14:00:27 i wonder if i should bump pkgrel or not 2016-03-22 14:26:02 barthalion: whats requirements for numpy (hardware related, if any) ? 2016-03-22 14:32:03 fabled: are we good for v3.3.3 now? 2016-03-22 14:45:39 woh ! what happened to b.a.o, "Red Bull" 2016-03-22 14:46:05 red bull? 2016-03-22 14:46:28 "Red Bull" gives wings 2016-03-22 14:46:37 :)) 2016-03-22 15:02:40 ncopa, i have musl fix in queue 2016-03-22 15:03:13 but it's not critical, so perhaps postpone that. i think 3.3.3 is ok to be tagged 2016-03-22 15:11:17 ok 2016-03-22 15:14:26 ncopa, pushed musl fix to edge just now 2016-03-22 15:15:20 though, 3.3-stable is a bit earlier musl. should cherry-pick all important fixes 2016-03-22 15:16:06 probably yes 2016-03-22 16:24:00 ncopa: around? 2016-03-22 16:25:42 hi 2016-03-22 16:35:29 can anyone confirm that domain is ok now? 2016-03-22 16:35:53 domain? 2016-03-22 16:36:10 it expired 2016-03-22 16:36:25 i had not put enough funds there 2016-03-22 16:36:38 so autorenew didnt kick in 2016-03-22 16:36:51 and the email warning didnt get to me 2016-03-22 16:37:06 or its amont the 2900+ unread emails... 2016-03-22 16:40:35 you seriously need to do something about your mailbox... 2016-03-22 16:43:25 i only have 2746 2016-03-22 16:45:37 domain looks correct from here 2016-03-22 16:45:39 but 2016-03-22 16:45:51 ;; ANSWER SECTION: 2016-03-22 16:45:51 www.alpinelinux.org. 997 IN A 88.159.20.184 2016-03-22 16:46:01 ;; ANSWER SECTION: 2016-03-22 16:46:01 alpinelinux.org. 233 IN A 209.99.40.225 2016-03-22 16:46:02 its fixed 2016-03-22 16:46:09 i dont know for how long it may be ngatively cached 2016-03-22 16:46:21 xming: it should be fixed in 233 seconds 2016-03-22 16:46:28 but could take some time, normal dns ttl is 24h right? 2016-03-22 16:46:37 there's no "normal dns ttl" 2016-03-22 16:46:46 every cache does as it pleases 2016-03-22 16:46:57 i think we set ttl to 1 hour or so 2016-03-22 16:46:58 and some buggy caches will keep the old value for a week 2016-03-22 16:47:13 ncopa: we did, but what did they other dns servers do? 2016-03-22 16:47:17 what's important is that the data is in place 2016-03-22 16:47:27 alpinelinux.org. 233 IN A 209.99.40.225 2016-03-22 16:47:34 the 233 there is the TTL i think 2016-03-22 16:47:41 ncopa: that was a fresh query from the browser 2016-03-22 16:47:52 alpinelinux.org. 299 IN NS dns15.parkpage.foundationapi.com. 2016-03-22 16:47:55 alpinelinux.org. 299 IN NS dns16.parkpage.foundationapi.com. 2016-03-22 16:48:46 clandmeter: that's wronly cached data 2016-03-22 16:48:59 (/mnt/tmpfs/dev/pts/0) ska@alyss:~/cacaoweb/cacaolinux/longhorn-bootstrap > s6-dnsq ns alpinelinux.org `s6-dnsip4 d0.org.afilias-nst.org` 2016-03-22 16:48:59 101 bytes, 1+0+2+2 records, response, noerror 2016-03-22 16:48:59 query: 2 alpinelinux.org. 2016-03-22 16:48:59 authority: alpinelinux.org. 86400 NS ns1.alpinelinux.org. 2016-03-22 16:48:59 authority: alpinelinux.org. 86400 NS ns2.alpinelinux.org. 2016-03-22 16:48:59 additional: ns1.alpinelinux.org. 86400 A 74.117.189.114 2016-03-22 16:48:59 additional: ns2.alpinelinux.org. 86400 A 88.159.20.184 2016-03-22 16:49:15 yup 2016-03-22 16:49:29 googles cache, 8.8.8.8 gives that 2016-03-22 16:49:31 of course its cashed 2016-03-22 16:49:35 thats the issue 2016-03-22 16:49:51 the right data will slowly infuse into caches 2016-03-22 16:49:55 and its not cached, it was the query i run when it was down :) 2016-03-22 16:49:55 $ dig alpinelinux.org @8.8.8.8 2016-03-22 16:50:00 there's nothing you can do about it 2016-03-22 16:50:04 ;; ANSWER SECTION: 2016-03-22 16:50:04 alpinelinux.org. 2854 IN A 88.159.20.184 2016-03-22 16:50:11 caches will update when they please. 2016-03-22 16:50:16 google dns is correct now 2016-03-22 16:50:29 only every N-th request 2016-03-22 16:50:38 orly? 2016-03-22 16:50:54 ah you are right 2016-03-22 16:50:55 have a nice evening climbers 2016-03-22 16:51:00 even Google doesn't get DNS right :/ 2016-03-22 16:51:05 :) 2016-03-22 16:51:19 their caches aren't in sync :/ 2016-03-22 16:51:39 that means they don't check for the ttl 2016-03-22 16:51:54 if they did, they would all give the right answer 2016-03-22 16:52:01 the bad one still has 200s to go 2016-03-22 16:52:16 they probably check it, but their caches may ask other cache, i dunno 2016-03-22 16:52:21 not neccessary 2016-03-22 16:52:38 no, Google does the resolution itself 2016-03-22 16:52:48 they don't rely on anything else than themselves 2016-03-22 16:52:51 if they queried at different times and do not have a central cache then they can be off 2016-03-22 16:53:07 probably yes 2016-03-22 16:54:07 then if they honor the ttl, it means that the grace period isn't expired yet and it's normal that stuff breaks :D 2016-03-22 16:54:21 idd 2016-03-22 16:54:30 it's all normal 2016-03-22 16:55:21 there are at least 6 diff servers answering queries here, 4 of them aren't updated yet 2016-03-22 16:57:41 dns15.parkpage.foundationapi.com. 2016-03-22 16:57:50 $10.69 a year and $61 for restoring from quarantine, pretty steep pricing 2016-03-22 16:58:19 those dns servers does not give correct answer 2016-03-22 16:59:17 ncopa: its ttl doen't decrease :( 2016-03-22 17:00:30 it is strange, my cache has cached the ns#.alpinelinux.org to the parked IP, so my resolvers try to go to the parked IP for the name, which gives the parked IP always 2016-03-22 17:00:56 so it would seem, the real TTL we will wait for is the one for the nameservers (86400) 2016-03-22 17:01:01 https://members.nearlyfreespeech.net/faq?q=ExpiredPark#ExpiredPark 2016-03-22 17:01:02 ncopa: ah they thing they are authoritive for alpinelinux.org 2016-03-22 17:01:22 s/thing/think 2016-03-22 17:01:33 the glue record hasn't updated yet 2016-03-22 17:02:14 sorry the glue record is updated 2016-03-22 17:03:14 anyone has dns15.parkpage.foundationapi.com cached as SOA is getting wrong answer 2016-03-22 17:12:21 sigh 2016-03-22 17:12:24 this was bad 2016-03-22 17:12:29 *really* bad 2016-03-22 17:12:33 i should fire myself 2016-03-22 17:18:26 what was bad? 2016-03-22 17:21:20 not renewing the domain 2016-03-22 17:21:30 in time 2016-03-22 17:21:42 don't fire yourself, find a way to catch up with your inbox instead :P 2016-03-22 17:31:23 oh great, we made it to zen.spamhaus.org too 2016-03-22 17:32:00 or? 2016-03-22 17:33:59 no, its gmail that got blacklisted in spamhaus 2016-03-22 17:35:31 skarnet: i think my catch-up-inbox problem is solved. new emails does not seem to be delivered 2016-03-22 17:35:33 :) 2016-03-22 17:36:06 no, that only solves the problem of new emails 2016-03-22 17:36:12 you still have a batch to catch up on :P 2016-03-22 17:38:41 friends 2016-03-22 17:40:14 anyone still here 2016-03-22 17:41:14 we are here 2016-03-22 17:41:26 oneinsect: always! 2016-03-22 17:41:38 how much is a domain cost for paying for 10 years of 15 years 2016-03-22 17:41:42 in one shot 2016-03-22 17:42:14 depends on the tld 2016-03-22 17:42:14 i would like to donate for next month from my salary..to make alpinelinux.org 2016-03-22 17:42:17 doesnt go down 2016-03-22 17:42:28 but around 30€ a year 2016-03-22 17:42:54 fine I will pay in april...for 10 or 15 years 2016-03-22 17:43:35 keep your money, just teach ncopa multitasking management :D 2016-03-22 17:43:47 and task delegation! 2016-03-22 17:43:47 lol 2016-03-22 17:44:37 just give him a bette email client 2016-03-22 17:44:37 *g* 2016-03-22 17:44:47 no wonder stuff gets lost if you not using thunderbird *g* 2016-03-22 17:45:46 oh, stuff can get lost with Thunderbird all right 2016-03-22 17:46:09 yeah, thats why mutt exists *g* 2016-03-22 17:46:23 as they say for John D Carmack - When the god speaks the industry listens....i guess it goes for the same for God Ncopa 2016-03-22 17:46:48 ncopa: but you are in good company *g* http://www.cnet.com/news/what-did-google-pay-the-guy-who-briefly-owned-google-com/ 2016-03-22 17:46:52 we listen when he's talking about Alpine 2016-03-22 17:46:58 lol 2016-03-22 17:47:01 for the rest, you better not listen to him too much :D 2016-03-22 17:47:10 the problem is not the $ 2016-03-22 17:47:35 the stupid thing is that all they years i didnt have $, i managed to get it renewed always 2016-03-22 17:48:20 as soon as i got the $ i failed 2016-03-22 17:48:47 i only renewed it for 1 year this time too, because if i do 10 years, i will forget how to to do it next time 2016-03-22 17:49:13 I kinda hope you'll have learned how to manage your email in 10 years 2016-03-22 17:49:28 heheh 2016-03-22 17:49:36 (also, we'll have conquered the world, so that won't be an issue) 2016-03-22 17:50:22 ACTION is going to register mountainlinux.org ... just in case the project needs a new domain next year *g* 2016-03-22 17:50:34 lol 2016-03-22 17:51:17 is there any wiki page that can help guide newbies like us to build the latest iso from git right from the start command git clone.....etc? 2016-03-22 17:51:51 oneinsect: http://wiki.alpinelinux.org/wiki/How_to_make_a_custom_ISO_image 2016-03-22 17:52:37 aha lovely 2016-03-22 17:53:26 but building the iso is in 99% of the cases not needed 2016-03-22 17:53:32 i am gonna try out the rpi 3 builds (Timo/Ncopa have been really helping with the bugs i filed under Ray Davis) 2016-03-22 17:53:34 just use the existing iso and upgrade to edge 2016-03-22 17:53:49 ooh 2016-03-22 17:54:21 the rpi3 firwmare issue? 2016-03-22 17:54:35 yes 2016-03-22 17:54:40 1 question is: are we allowed to redistribute it? 2016-03-22 17:54:47 what license is it 2016-03-22 17:55:05 Timo seems to have updated the ticket with a license link 2016-03-22 17:55:15 https://github.com/RPi-Distro/firmware-nonfree/blob/master/brcm80211/LICENSE 2016-03-22 17:57:28 i think i made it possible to add a directory named firmware/ on boot media 2016-03-22 17:57:48 so it should be possible to put non-free firmware there 2016-03-22 17:58:07 but i don't know if it works, since i havent tested it as far i can remember 2016-03-22 17:58:15 hmmm 2016-03-22 17:59:12 yes 2016-03-22 17:59:17 it should work 2016-03-22 17:59:57 if the directory /firmware exists on boot media then it should be copied to /lib/firmware 2016-03-22 18:00:50 great 2016-03-22 18:01:29 do the build instructions for the above wiki page work for raspberry pi on an x86? do we need to have alpine installed? 2016-03-22 18:01:59 i think for rpi it will generate a tar.gz 2016-03-22 18:02:05 but i dont know if it will work from x86 2016-03-22 18:02:11 i gotta go 2016-03-22 18:02:13 hmmm 2016-03-22 18:02:14 sure 2016-03-22 18:02:44 leo-unglaub: 2016-03-22 18:02:59 will rpi build work 2016-03-22 18:03:03 from x86 machine? 2016-03-22 18:03:17 will it crosscompile? 2016-03-22 18:12:04 uff, good question. i am not really a rpi expert ... ask ncopa .. he is our tech god around here 2016-03-22 18:12:58 hmmm 2016-03-22 18:17:31 fabled: 2016-03-22 18:17:38 are you here 2016-03-22 18:18:27 Timo 2016-03-22 18:19:08 oneinsect, sort of busy, but reading 2016-03-22 18:19:41 will git cloning the alpine repo and running it on an x86 alpine machine build 2016-03-22 18:19:46 rpi tar ball 2016-03-22 18:21:25 since alpine-rpi.conf.mk and alpine-rpi.packages already present 2016-03-22 18:21:33 cant i just build it on an x86 machine? 2016-03-22 18:21:36 will it work? 2016-03-22 18:27:08 okie got it 2016-03-22 18:27:29 "make PROFILE=alpine-rpi iso" must done on a raspberry pi only 2016-03-22 18:27:38 seems its not too happy cross compiling 2016-03-22 18:28:11 am i correct fabled: 2016-03-22 18:28:14 ? 2016-03-22 18:31:34 or this? make PROFILE="alpine-rpi" rpi 2016-03-22 18:36:58 it's done on arm cpu on the official builders 2016-03-22 18:37:57 i think it currently needs to run on arm; the only main reason is that initramfs/modloop building needs it 2016-03-22 18:38:18 i think it could be made to be cross built with some work 2016-03-22 18:38:52 currently it just uses apk to install the packages, and that requires native hw for the post-install scripts and triggers. also depmod is required to be run 2016-03-22 18:39:25 so i need to have a working alpine running on rpi to build a newer alpine version 2016-03-22 18:39:28 isnt it? 2016-03-22 18:39:49 currently yes, it's self-hosted 2016-03-22 18:40:00 hmmm 2016-03-22 18:40:26 i wonder how you first built it when you never had alpine linux for rpi at all 2016-03-22 18:45:37 i see you there 2016-03-22 18:45:38 https://www.raspberrypi.org/forums/viewtopic.php?f=56&t=60569 2016-03-22 18:45:42 ncopa: algibot stopped logging ? 2016-03-22 18:45:48 if i am not wrong 2016-03-22 18:46:02 so got it...!! 2016-03-22 18:46:07 thanks fabled: 2016-03-22 18:46:46 algitbot \o/ 2016-03-22 18:46:52 well it's still there 2016-03-22 18:57:09 oneinsect, we build arm stuff with wandboard. the first image was cross built manually. 2016-03-22 18:57:35 aha 2016-03-22 18:57:38 got it fabled: 2016-03-22 18:57:44 the freescale imx series 2016-03-22 18:59:13 so when is the new release planned 2016-03-22 18:59:17 3.3.3 2016-03-22 19:05:43 i was hoping ncopa would've tagged it today 2016-03-22 19:16:29 hmmm 2016-03-22 20:58:28 @ncopa I don’t understand what’s wrong, it works on my system: 2016-03-22 20:58:29 ~/aports/testing/maven $ abuild 2016-03-22 20:58:29 >>> maven: Checking sanity of /home/jirutjak/aports/testing/maven/APKBUILD... 2016-03-22 20:58:29 >>> maven: Analyzing dependencies... 2016-03-22 20:58:29 >>> maven: Cleaning temporary build dirs... 2016-03-22 20:58:29 >>> maven: Checking sha512sums... 2016-03-22 20:58:30 apache-maven-3.3.9-bin.tar.gz: OK 2016-03-22 20:58:30 >>> maven: Unpacking /var/cache/distfiles/apache-maven-3.3.9-bin.tar.gz... 2016-03-22 20:58:31 >>> maven: Entering fakeroot... 2016-03-22 20:58:31 >>> maven*: Running postcheck for maven 2016-03-22 20:58:32 >>> maven*: Preparing package maven... 2016-03-22 20:58:32 >>> maven*: Tracing dependencies... 2016-03-22 20:58:33 >>> maven*: Package size: 9.2 MB 2016-03-22 20:58:33 >>> maven*: Compressing data... 2016-03-22 20:58:34 >>> maven*: Create checksum... 2016-03-22 20:58:34 >>> maven*: Create maven-3.3.9-r0.apk 2016-03-22 20:59:15 @ncopa: I don’t understand what’s wrong, it works on my system (see log http://haste.fit.cvut.cz/oxuromi.log) 2016-03-22 21:01:19 > can't create '/home/ncopa/aports/testing/maven/pkg/maven/usr/share/java/maven-3/bin/m2.conf': No such file or directory 2016-03-22 21:01:19 this directory should be created by install -D 2016-03-22 21:03:33 /home/ncopa really? 2016-03-22 21:04:12 it’s from ncopa’s log… what’s wrong about it? 2016-03-22 21:04:35 oh, sorry, I had reversed the meaning of your sentence 2016-03-22 21:04:41 been a long day >.> 2016-03-22 21:04:53 np :) 2016-03-22 21:47:40 What is preferred approach for packaging Python web applications with many, many dependencies? 2016-03-22 21:48:34 Is it okish to bundle dependencies and install them with the application in venv? 2016-03-22 23:09:14 jirutka: bundling dependencies and packaging in one sentence is a no-go 2016-03-22 23:09:31 so what I should do? 2016-03-22 23:09:42 package each dependency 2016-03-22 23:09:46 ha ha 2016-03-22 23:09:53 there are about 100 dependencies 2016-03-22 23:10:09 it’s insane to package all of them and maintain it 2016-03-22 23:10:30 if you are talking about package for personal usage, venv is probably okay, but I wouldn't package it at all then 2016-03-22 23:10:33 I’m talking about Python modules; just some of them have native extensions 2016-03-22 23:10:45 packaging is insane 2016-03-22 23:10:47 welcome to hell 2016-03-22 23:11:28 as long as something can be reused, it deserves a standalone package 2016-03-22 23:11:40 abuild package is good to disquish between build and runtime dependencies, so I don’t need to install build dependencies to the target system 2016-03-22 23:12:38 I’m thinking about packaging python modules that have native extensions and bundle all others 2016-03-22 23:13:06 only run-time dependencies should appear in a package 2016-03-22 23:13:14 build-time dependencies should appear in the -dev package 2016-03-22 23:13:30 obviously… 2016-03-22 23:13:57 the point is that package is built on build machine, only here is needed to install build dependencies 2016-03-22 23:14:21 indeed 2016-03-22 23:14:36 then I get binary package that I can just unpack on the target system (and install runtime deps of course) 2016-03-22 23:17:04 Python module provides list of all its (python) dependencies, required versions etc. and Python has package manager… it’s non-sense to throw it away, create custom abuild for all the dependencies and manage it by hand 2016-03-22 23:18:11 also you quite often need to install one python module in different versions… easy with bundles, but difficult with global installation 2016-03-22 23:19:21 a language that makes distributions' jobs harder instead of easier because it has the hubris of doing package management itself gets very little respect from me 2016-03-22 23:19:29 Gentoo devs are strictly against bundling any dependencies, I don’t know how Alpine devs 2016-03-22 23:19:48 funny 2016-03-22 23:20:11 why we have so-called distributions? just because of insane C legacy that doesn’t know package management 2016-03-22 23:21:03 no, because binary installations save time for users. 2016-03-22 23:21:21 package management is a consequence of the need to have distributions, not a cause. 2016-03-22 23:21:42 when I want to release some python module, I describe all its dependencies and additional metadata in one format and every other can use it, regardles of operation system 2016-03-22 23:22:03 oh, yes, Python dependency management works well in a void. 2016-03-22 23:22:15 Unfortunately, it does not exist in a void. 2016-03-22 23:22:33 Every language has its dependency manager, except C/C++. 2016-03-22 23:22:43 Not playing well with existing systems is a sign of hubris and asshatery. 2016-03-22 23:22:51 golang! 2016-03-22 23:23:00 yeah, that one too. 2016-03-22 23:23:10 its even worse 2016-03-22 23:23:16 come on, we’re in decent society, don’t talk about this shir 2016-03-22 23:23:40 we never bundle deps 2016-03-22 23:24:16 except for golang of course 2016-03-22 23:24:26 > Not playing well with existing systems is a sign of hubris and asshatery. 2016-03-22 23:24:26 That’s not true. For example emerge plays well with Python. 2016-03-22 23:24:58 does emerge play well with Python, or does Python play well with emerge? 2016-03-22 23:25:10 because if emerge has to adapt to Python, that's my point. 2016-03-22 23:25:23 And it’s not fault of every language that has its dependency manager that C guys doesn’t care about it at all, so you can’t describe your C dependencies. 2016-03-22 23:25:46 Dependencies *should not* be a property of a language! 2016-03-22 23:25:54 C actually does it right. 2016-03-22 23:26:01 No, it’s not. 2016-03-22 23:26:25 Dependencies are a property of a *software system*, no matter what language it's written in. 2016-03-22 23:26:39 Language is just the wrong classification here. 2016-03-22 23:26:39 Because it’s totally insane to ”package” each software in the world for every Linux distributin and every Unix system 2016-03-22 23:26:54 I agree with that. 2016-03-22 23:27:21 Tarballs should be enough, and installed with the default value, all the time. 2016-03-22 23:27:23 I agree that dependencies should not be a property of a language, in ideal world. 2016-03-22 23:27:34 Tarballs doesn’t describe its dependencies 2016-03-22 23:27:43 *don’t 2016-03-22 23:27:48 they should. 2016-03-22 23:27:51 how? 2016-03-22 23:28:00 tarball is just a tar, isn’t it? 2016-03-22 23:28:00 hey 2016-03-22 23:28:01 oops 2016-03-22 23:28:30 A text file in the tarball listing the build-time and run-time requirements. Some packages do it. Most don't. 2016-03-22 23:28:40 The huge majority doesn't even know the difference. 2016-03-22 23:28:43 Text file in which format? 2016-03-22 23:28:47 Who cares? 2016-03-22 23:28:54 It's for humans to read. 2016-03-22 23:29:07 I agree, it's a pest to find out all the deps of a package! 2016-03-22 23:29:10 Oh, you wanted an automated system? 2016-03-22 23:29:13 This isn’t package management. 2016-03-22 23:29:22 It must be machine readable and reasonably standardized. 2016-03-22 23:29:45 You want an automated package management portable to every Unix ? 2016-03-22 23:29:53 now THAT is some dream 2016-03-22 23:29:58 I really don’t want to read text file to manually resolve all transitive dependencies and install them by hand. 2016-03-22 23:30:11 then - tada! what you want is a distro 2016-03-22 23:30:32 That is what almost every language provides… by just for itself. 2016-03-22 23:30:44 and it's wrong. 2016-03-22 23:30:52 No, that is definitely not what I want. 2016-03-22 23:31:11 It MUCH better than to do nothing like C 2016-03-22 23:31:35 absolutely not. Doing nothing is much, much better than doing the wrong thing. 2016-03-22 23:31:49 At least, when you do nothing, other people can do the work. 2016-03-22 23:31:59 When you're doing the wrong thing, it makes it HARDER for other people to do the work. 2016-03-22 23:32:04 Cue: python dependencies. 2016-03-22 23:32:11 When you have even lang-specific dependency management, then you can easily declare cross-lang dependencies. The only problem is that f*cking C 2016-03-22 23:32:26 Because at least you know the name of the package 2016-03-22 23:32:53 Oh? How do you declare cross-lang dependencies? how do you standardize things across the N package managers for your N languages? 2016-03-22 23:33:02 You know when you can download it etc. You know absolutely nothing (machine readable) about C/C++ packages, because, eh, there are no packages! 2016-03-22 23:35:16 You may be thinking about automated download and installation 2016-03-22 23:35:36 Well, let’s say that my software written in lang X needs e.g. Python modules A, B, and C. Then I can just write names and required versions of these modules in some X’s package manager format, just add metainfo that these are Python modules. Then it’s trivial to program X’s package manager to run pip to install these dependencies. 2016-03-22 23:35:55 And what we do in every linux distro? Automated download and installation. 2016-03-22 23:36:15 And what’s more important, automated dependencies graph resolving! 2016-03-22 23:36:41 We do this when the admin actually requests it explicitly. 2016-03-22 23:37:08 See the difference… how can i specify what C software my package need? I can’t even name them, because there’s no repository like PyPI, RubyGems, LuaRocks, Maven Central, … 2016-03-22 23:37:22 ? 2016-03-22 23:38:48 what you want is a generic package management system, and I agree that it would be nice 2016-03-22 23:39:09 but the segregation between languages is absolutely unnecessary 2016-03-22 23:39:13 And if we go deeper, all the problems stars with dynamic linking… it adds extreme complexity that really isn’t worth it 2016-03-22 23:39:31 you could have that package management system without the language idiosyncrasies 2016-03-22 23:39:46 Indeed, but we don’t have any :( 2016-03-22 23:40:07 it's only a property of the language because every language thinks it its job to do package management, which is nothing else than hubris 2016-03-22 23:40:20 And then Docker… the main problem that Docker tries to solve is problem with lack of this package manager. 2016-03-22 23:41:03 So instead of solving this primary problem, we just bundle our software with whole f*cking Ubuntu and deploy it… great, really great… 2016-03-22 23:41:54 you want to solve the problem? then solve it. Do that generic package manager. Good luck with that. 2016-03-22 23:42:15 But don't pretend the problem comes from C because it's lacking a dependency language. The problem doesn't come from there. 2016-03-22 23:42:33 It would be very easy if every language have package management… 2016-03-22 23:42:43 Yes, it does 2016-03-22 23:42:51 We'll just have to disagree on that. 2016-03-22 23:43:06 Please read what I’ve writtens few moments ago. 2016-03-22 23:43:40 Have you seen the guix package manager? 2016-03-22 23:43:42 Today, you have grouping by distributions. I can't say I like it, but it's miles better than the grouping by languages that your proposal would get. 2016-03-22 23:43:51 or Nix 2016-03-22 23:43:57 it's quite nice 2016-03-22 23:44:26 I can easily describe Python, Ruby, Lua, … dependencies and write some meta package manager that will just utilize manager of eeach of these langs… I can’t do that for C software… and that’s the whole point 2016-03-22 23:45:04 Grouping by languages is IMO much less bad then grouping by linux distributions. 2016-03-22 23:45:41 If I have say Python module, I can install it on every operating system in the same way, including all the transitive dependencies. 2016-03-22 23:46:09 thats not true 2016-03-22 23:46:15 what if you are using python gtk 2016-03-22 23:46:34 it breaks across ffi boundaries 2016-03-22 23:46:34 If I have some C software, I can’t do that… I have to write package for every linux distribution and every operating system… 2016-03-22 23:46:49 It works for pure python modules 2016-03-22 23:46:59 yeha 2016-03-22 23:47:25 It doesn’t work for modules with native extensions… and the only reason that it doesn’t work for them is that you can’t even name C dependencies! 2016-03-22 23:47:40 #firstworldproblems 2016-03-22 23:48:00 because C doesn’t have any dependency manager and it’s named differently in every linux distribution 2016-03-22 23:48:40 because C was meant to program, not to fetch modules in a centralized repository and install them like your millenial devops 2016-03-22 23:48:50 that's just out of scope! 2016-03-22 23:49:07 So instead of installing it with just pip install… you must read some readme, there is somehow described what it needs, usually as names of Debian packages… so I have to find how it’s named in e.g. Alpine or Gentoo… and install them by hand… 2016-03-22 23:49:08 like it should be out of scope for any fucking flavour of the month language out there 2016-03-22 23:49:47 So what these languages shoud do? Copy&paste files like in f*cking PHP few years ago? 2016-03-22 23:50:10 "Frankly, my dear, I don't give a damn." 2016-03-22 23:50:22 please be pragmatic… 2016-03-22 23:51:08 I know that this situation, where almost every lang have its own package manager is far from ideal… but it much MUCH better than nothing 2016-03-22 23:51:59 maybe instead of a package manager they could just have a easily parsable format which all the information there 2016-03-22 23:52:23 This is about code reusability… when you have proper package manager, you can reuse small, simple modules… it’s no problem when your software transitively depends on hundreds of modules… but whiout any package manager, it would be total nightmare 2016-03-22 23:52:32 but they have! 2016-03-22 23:52:39 I've been compiling and installing my Linux machines for 16 years, without relying on distributions. Reading READMEs and installing stuff by hand is my bread and butter. Sorry, I'm the wrong person to go to for sympathy here. 2016-03-22 23:53:06 all the packages have unique names, there are repositories like PyPI where you can find them 2016-03-22 23:53:44 And don’t you think that it’s totally insane to manually traverse dependency graph, find tarballs all over the Internet and install them? 2016-03-22 23:54:23 I think it's insane to do it more than once. 2016-03-22 23:54:34 I also think it's absolutely necessary to do it once. 2016-03-22 23:55:00 What do you mean with “once?” 2016-03-22 23:55:07 Once per life just to try it? 2016-03-22 23:55:28 On one machine. Not just to try it, but to understand the complexity and the hidden costs of dependencies. 2016-03-22 23:55:41 How many machines do you manage? 2016-03-22 23:55:57 Depends on whether you count VMs or not. 2016-03-22 23:56:04 I count them 2016-03-22 23:56:27 at the moment, five. 2016-03-22 23:56:32 lol 2016-03-22 23:57:08 Well, then you can do that… I have over 30 systems… and I’d like to also do other things then just resolve and install dependencies by hand all over the time 2016-03-22 23:57:09 what, lol? I think that's a lot, probably too much, for someone who's not being paid for this., 2016-03-22 23:57:36 okay, but with just few machines you can’t understand how insane this approach is 2016-03-22 23:57:36 When I was being paid for this, I was managing around a million machines. 2016-03-22 23:58:01 okay and you’ve installed dependencies manually on all of them, one by one? 2016-03-22 23:58:15 As I said, it's insane to do it more than once. 2016-03-22 23:58:45 Maybe I just still don’t understand what do you mean with once, but let’s move 2016-03-22 23:58:49 I obviously used the package manager that they had developed. But had I stayed, I would have made a point to dive into the dependency system and understand it. 2016-03-22 23:59:05 I totally agree with that 2016-03-22 23:59:25 but I think that it’s not substantial in this discussion 2016-03-22 23:59:45 But that's exactly what I'm talking about. 2016-03-22 23:59:46 One should always know what he’s doing and also what software that he use is doing 2016-03-22 23:59:56 Be honest. How often is it the case? 2016-03-23 00:00:04 Well… 2016-03-23 00:00:39 we both know 2016-03-23 00:00:55 but again, it has nothing to do with this topic 2016-03-23 00:01:49 It does. That's an _excellent_ reason why C software has no central repository and should not have one. 2016-03-23 00:02:06 I don’t understand 2016-03-23 00:02:21 C guys are still learning how dependencies work or what? 2016-03-23 00:02:32 yes, as are other guys. 2016-03-23 00:02:33 autoconf IS their package manager 2016-03-23 00:02:54 autoconf is insanity 2016-03-23 00:03:13 Someone who does not know the hidden cost of dependencies will very soon make a package that you can't bootstrap. 2016-03-23 00:03:15 it’s mirror of C world… 2016-03-23 00:03:37 Do you know how hard it is to bootstrap a system already? It's borderline insane. 2016-03-23 00:03:43 I know 2016-03-23 00:04:04 Take your average package in *any* language you want that has a central repository 2016-03-23 00:04:06 but that’s not a reason to not provide any package management 2016-03-23 00:04:09 take one at random 2016-03-23 00:04:21 and try to track down dependencies to the very end 2016-03-23 00:04:25 easy 2016-03-23 00:04:29 You'll likely pull half of the repository. 2016-03-23 00:04:34 nope 2016-03-23 00:04:49 and I can at least resolve all of them, with single press of key 2016-03-23 00:04:56 this is about “managing” dependencies 2016-03-23 00:05:01 sure, you can resolve them, but can you compute the complexity? 2016-03-23 00:05:21 and packages are all about code reuse that is much better thing then reinventing the well all over again or copy pasting code 2016-03-23 00:05:26 You have an approach that's valid for _application_ software 2016-03-23 00:05:35 My point of view is from a _system_ software guy 2016-03-23 00:05:44 so it's normal that we have different views. 2016-03-23 00:05:51 Ah 2016-03-23 00:06:03 well, now it starts to make a sense 2016-03-23 00:06:16 we’re talking about quite different worlds 2016-03-23 00:06:53 Yes we are. And still, it would do application devs a lot of good to understand the costs of dependencies. 2016-03-23 00:07:26 sure, fetch them from the repo at the press of a button, no problem. But understand that sometimes magic fails. 2016-03-23 00:08:02 anyway I won't be making much more sense at 1 am, so good night :) 2016-03-23 00:08:53 If you need some functionality, you have few options – you can write it yourself (you’ll be reinventing the wheal if someone else already solved it), find some existing solution and copy&paste code (totally wrong) or use that code as a dependency 2016-03-23 00:09:18 *wheel 2016-03-23 00:11:25 as a library developer you have two options – distribute one big module, to reduce dependencies, but then we get bloated software… or distribute many very small modules; then you will obvisouly get much more dependencies 2016-03-23 00:11:47 okay, good night 2016-03-23 06:32:04 ncopa, nlplug-findfs has still something fishy in it 2016-03-23 06:32:24 1 out of 5 boots fails on rpi 2016-03-23 06:32:30 or it's pretty random 2016-03-23 06:32:50 i wonder if it's timeout 2016-03-23 06:39:09 or some ordering issue 2016-03-23 06:48:21 seems that the mmc driver gets loaded and dmesg prints everything fine; but nlplug-findfs fails to find boot device 2016-03-23 06:55:55 nlplug-findfs says exit due to timeout (250) when things fail 2016-03-23 06:56:16 ideas? 2016-03-23 07:05:30 hmm... i may have used wrong mkinitfs version 2016-03-23 07:54:52 looks like i had wrong mkinitfs so it's good 2016-03-23 10:11:41 do we have any pending abuild changes/fixes? 2016-03-23 10:11:46 i want tag new abuild release 2016-03-23 10:19:28 wow you give them 7 minutes to reply? :) 2016-03-23 10:22:03 :) 2016-03-23 10:22:20 things will break 2016-03-23 10:22:39 the abuild provided prepare() function no longer uses _builddir 2016-03-23 10:22:52 so stuff that relies on that may break 2016-03-23 10:23:19 fix is to use $builddir instead of $_builddir 2016-03-23 10:24:19 didn't we plan to $builddir=$_builddir? 2016-03-23 10:24:28 (say, if not set) 2016-03-23 10:24:37 nah 2016-03-23 10:24:49 i dont think we want abuild use anything _prefixed 2016-03-23 10:25:12 what i did was default builddir to $rscdir/$pkgname-$pkgver if unset 2016-03-23 10:25:20 so i think in most cases it will not break things 2016-03-23 10:25:40 and this is also a relatively new feature so i dont think many apkbuilds acutally uses it yet 2016-03-23 10:59:43 hey friends! 2016-03-23 11:43:52 hey 2016-03-23 11:55:32 friends 2016-03-23 11:55:38 is the new release 2016-03-23 11:55:39 tagged? 2016-03-23 11:57:55 ncopa: 2016-03-23 11:59:03 oneinsect: i was thinking of waiting til tomorrow to let the dns caches catch up first 2016-03-23 11:59:25 great 2016-03-23 11:59:28 :-) 2016-03-23 11:59:39 but if its important i could do it now 2016-03-23 11:59:51 i will probably not have time to fix things if things break 2016-03-23 12:00:19 i mean i am just eager to test the rpi 3 and the new wifi and report any bugs...but its better you have time to fix thingsd 2016-03-23 12:00:31 better fix and release tomorrow 2016-03-23 12:00:42 yeah 2016-03-23 12:00:47 thanks for your patience 2016-03-23 12:00:54 most welcome 2016-03-23 12:34:16 ncopa: regarding abuild we could simply use sed to replace _builddir with builddir in all APKBUILDs that shouldn't break stuff, right? 2016-03-23 12:36:25 the only case I can think of where it would break stuff is when both _builddir and buildir are defined in the APKBUILD but I don't think that this is the case anywhere 2016-03-23 12:36:37 grep should tell that 2016-03-23 12:36:39 and yes 2016-03-23 12:36:43 we could use sed 2016-03-23 12:36:57 however, it will cause more work when backporting 2016-03-23 12:37:14 when backporting fixes to stable branches 2016-03-23 12:37:25 and i think in many cases it will just work 2016-03-23 12:37:29 in most cases 2016-03-23 12:37:42 so i think we can just deal with it when problem shows up 2016-03-23 12:38:09 alright, sounds good to me 2016-03-23 12:38:26 we can probably sed it for alpine 3.5 or 3.6 2016-03-23 12:48:26 ncopa: so is 3.5 getting kernel 4.4? 2016-03-23 12:48:44 alpine 3.4 will get kernel 4.4 2016-03-23 12:49:04 even better 2016-03-23 12:49:15 \o/ 2016-03-23 13:14:17 default builddir to $rscdir/$pkgname-$pkgver if unset : can you put that in the APKBUILD file? things like defaults if unset need to be obvious 2016-03-23 13:45:56 alpine does a very great job in making packages very small and clean. but i think we all could do a better job a defining better default configs 2016-03-23 13:46:05 just noticed it again in pure-ftpd 2016-03-23 13:46:11 the default paths are horrible 2016-03-23 13:54:37 leo-unglaub, i don't think pure-ftp is Alpine's favorite FTP server, given that's still in testing ;-) 2016-03-23 13:54:50 rnalrd: its just an example 2016-03-23 13:55:37 we tend to do not set defaults, unless absolutely necessary to make it start out-of-the-box 2016-03-23 13:56:53 i understand that ... but sometimes a more secure default for example would be good 2016-03-23 14:18:34 I have problem with uwsgi on Alpine… it doesn’t recognize installed plugins. 2016-03-23 14:18:34 When I run `uwsgi --plugin python`, it fails with “!!! UNABLE to load uWSGI plugin: Error loading shared library ./python_plugin.so: No such file or directory !!!” 2016-03-23 14:18:34 If I use absolute path and run `uwsgi --plugin /usr/lib/uwsgi/python_plugin.so` instead, then it works. 2016-03-23 14:18:34 It also works when I run it from /usr/lib/uwsgi, so it tried to load plugins from current working directory instead of /usr/lib/uwsgi. 2016-03-23 14:20:59 Seems to be a bug in the package http://projects.unbit.it/uwsgi/wiki/Guide4Packagers 2016-03-23 14:50:20 2016-03-23 14:50:20 Last modified 3 years ago 2016-03-23 14:50:56 yes, but it’s the exact problem, the package doesn’t set plugin_dir 2016-03-23 14:52:36 I’m gonna send patch in few minutes 2016-03-23 15:56:51 @ncopa, when I patch main/uwsgi, I must send PR to testing/uwsgi, right? 2016-03-23 16:16:43 @ncopa, okay, please take a look https://github.com/alpinelinux/aports/pull/38 2016-03-23 16:17:18 ncopa: can you cherrypick that mariadb fix to v3.3 2016-03-23 16:23:28 jirutka: for a fix you dont send it to testing/ 2016-03-23 16:23:55 I’ve sent it to main 2016-03-23 16:24:15 I’ve eventually figured out that it doesn’t make sense to send it to testing 2016-03-23 16:25:46 jirutka: I'll just leave this here: http://www.theregister.co.uk/2016/03/23/npm_left_pad_chaos/ 2016-03-23 16:26:24 (my intention is not to be antagonistic, but just informative) 2016-03-23 16:26:38 (I have to run an errand now) 2016-03-23 16:26:40 skarnet: thanks, I’ll read it later 2016-03-23 16:27:38 hm, I maybe should also add uwsgidecorators.py to the abuild 2016-03-23 16:29:44 ncopa: ? 2016-03-23 16:30:34 ncopa: 13d027613dc9bc51202eb338821c558580c031a4 + a bump of pkgrel 2016-03-23 16:30:46 good evening climbers. 2016-03-23 16:58:43 ncopa: hahaha, you just gave me a heart attack ... https://pkgs.alpinelinux.org/package/testing/x86_64/vim-rust 2016-03-23 16:59:18 why heart attack? o.O 2016-03-23 16:59:36 jirutka: he will know why ;) 2016-03-23 16:59:44 okey 2016-03-23 17:00:35 jirutka: back, sorry for dropping a bomb and running away 2016-03-23 17:01:01 no problem, I don’t have a time to discuss this at this moment anyway 2016-03-23 17:01:09 skarnet: do you have a moment? 2016-03-23 17:01:18 leo-unglaub: sure 2016-03-23 17:01:33 i am looking at s6 2016-03-23 17:01:48 jirutka: ok ^^ the lesson I get from this is that open source should not rely on company-driven platforms :/ 2016-03-23 17:02:02 leo-unglaub: if you have s6-specific questions, ask them in #s6 :) 2016-03-23 17:02:22 yes, its s6 speciffic 2016-03-23 17:02:30 because i am looking at it right now 2016-03-23 17:02:41 join #s6, we don't bite 2016-03-23 17:02:46 okay 2016-03-23 18:04:53 did we scare you away? XD 2016-03-23 19:36:47 skarnet: no, you did not 2016-03-23 19:43:31 glad to hear it! 2016-03-23 19:43:53 damn, there is lights out here ... 2016-03-23 19:43:57 see you tomorrow 2016-03-23 19:44:36 see you :) 2016-03-23 22:37:13 i am getting crazy ... server down ... 2016-03-23 22:37:15 best day ever 2016-03-23 22:39:25 leo-unglaub: https://goo.gl/e6Egfj 2016-03-23 22:39:52 hehe 2016-03-23 23:34:48 hmmm, how do i rewrite syslinux to sda when i am in chroot /mnt 2016-03-23 23:47:42 ah, i found the problem why i was not working ... the bind mount was broken .. 2016-03-23 23:57:12 this really makes me thing about going back to doing dd images .. 2016-03-23 23:57:31 then you literally can get back the entire system 2016-03-24 06:15:14 hi there 2016-03-24 06:15:21 does anyone knows how to run Alpine on Hyper-V? 2016-03-24 07:23:18 fabled: 2016-03-24 07:23:20 ncopa: 2016-03-24 07:23:34 is it possible to targe alpine for more embeded arm boards? 2016-03-24 07:24:53 oneinsect, yes. we just have not had more boards our selves. any upstream supported board should be simple as enabling the config options in the kernel and uboot 2016-03-24 07:25:24 great probably I can ship some boards to you fabled: 2016-03-24 07:25:30 if they require patching / separate kernel build, it needs to be thought out a bit more. but might be possible if someone is willing to maintain those kernels 2016-03-24 07:26:09 boards like orange pi once 2016-03-24 07:26:11 one* 2016-03-24 07:26:11 ncopa, can we have 3.3.3 now? 2016-03-24 07:26:25 morning 2016-03-24 07:26:26 yes 2016-03-24 07:26:27 oneinsect, iirc orange pi should be supported upstream with the exception of the multimedia stuff 2016-03-24 07:27:51 and more recently http://www.friendlyarm.com/index.php?route=product/product&path=69&product_id=109 2016-03-24 07:27:55 $11 2016-03-24 07:28:03 nano pi m1 quad core board 2016-03-24 07:28:17 probably without multimedia should be fine fabled: 2016-03-24 07:28:24 should i ship you some boards? 2016-03-24 07:29:37 but 2016-03-24 07:29:43 they are still based on old kernels 2016-03-24 07:29:46 3.4 2016-03-24 07:29:50 not like rpi 2016-03-24 07:37:45 fabled: kernel 4.6 is supposed to include lots of support for arm boards 2016-03-24 07:37:47 http://www.phoronix.com/scan.php?page=news_item&px=Linux-4.6-New-ARM-SoCs 2016-03-24 07:37:57 yes 2016-03-24 07:39:08 I could be some help with rk3288 boards 2016-03-24 07:40:05 I have Alpine running with self compiled kernel 2016-03-24 07:41:06 fabled: you once told be that alpine's uboot has custom patches, is that still the case? 2016-03-24 07:45:12 i think we got rid of most of them 2016-03-24 07:45:57 we seem to have few still, but those should be upstreamed. basically as long as the u-boot's board driver supports 'distroconfig' it's good 2016-03-24 07:49:03 xming: did you manually patch the alpine uboot executable after compiling it for rk3288 boards? 2016-03-24 07:50:02 oneinsect: I used custom uboot to have support for my board 2016-03-24 07:50:59 ooh 2016-03-24 07:51:07 fabled: where the git of alpine's uboot? 2016-03-24 07:51:43 oneinsect: upstream has support for the board but creates an image that was too large to flash :/ 2016-03-24 07:52:07 but I think that has been solved 2016-03-24 07:52:08 rk3288 has upstream support?? 2016-03-24 07:52:14 i didnt know about it 2016-03-24 07:52:43 yes, not 100% but basic stuffs are working 2016-03-24 07:53:46 I am (still) trying to get full Alpine working on rk3288 2016-03-24 07:55:50 the goal is to get rid of RK blobs/partitions/... 2016-03-24 08:00:12 xming: can you guide me sometime how to build alpine distro for say a particular board 2016-03-24 08:00:18 what are the steps like 2016-03-24 08:00:37 i want to compile one for nano pi m1 based on Allwinner H3 2016-03-24 08:00:46 when you are fre 2016-03-24 08:01:07 is that a armv6/armv7 soc? 2016-03-24 08:01:12 armv7 2016-03-24 08:01:26 http://wiki.friendlyarm.com/wiki/index.php/NanoPi_M1#Compile_U-boot 2016-03-24 08:01:30 well all the ARM packages work with it 2016-03-24 08:01:40 all you need is just uboot + kernel 2016-03-24 08:01:41 its ultra cheap $11 2016-03-24 08:01:44 quad core... 2016-03-24 08:02:06 well what are steps to generate a uboot + kernel for it? 2016-03-24 08:03:00 you need a cross-compiler first 2016-03-24 08:03:08 an x86 VM 2016-03-24 08:03:12 say ubuntu or deb 2016-03-24 08:03:32 if the boards has upstream support then you can just 'make config' and select what you need 2016-03-24 08:03:40 you don't really need a VM for that 2016-03-24 08:03:52 just a decent cross-compiler 2016-03-24 08:04:49 no upstream support ...just patched legacy kernel i guess 3.4.111 2016-03-24 08:07:34 https://github.com/friendlyarm/h3_lichee 2016-03-24 08:08:43 any ideas xming: 2016-03-24 08:12:48 have you checked https://linux-sunxi.org 2016-03-24 08:14:13 yes i know they are doing some mainline kernel integration for socs like h3 2016-03-24 08:17:57 well then you can only use their uboot + kernel and Alpine for userland 2016-03-24 08:19:33 $11 and $10 shipping, oh well 2016-03-24 08:20:06 yes' 2016-03-24 08:21:05 do you have tutorial anywhere on howto do that "compile their uboot + kernel 2016-03-24 08:21:13 and tie alpine for userland" 2016-03-24 08:21:18 sorry for being a noob 2016-03-24 08:22:13 for rpi i just copy to a fat32 format disk 2016-03-24 08:22:17 i dont use DD at all 2016-03-24 08:22:51 according to http://linux-sunxi.org/H3_Manual_build_howto 2016-03-24 08:23:11 should i just build and copy the boot0 and u-boot to fat32 disk 2016-03-24 08:24:40 3.3.3 is in the oven 2016-03-24 08:24:48 fabled: any suggestions for release notes? 2016-03-24 08:25:01 I don't have any allwinner boards and I don't know if they have some proprietary partitions scheme ala RK 2016-03-24 08:26:18 hmmm 2016-03-24 08:26:40 and I don't have any howtos, other questions might be better asked in #alpine-linux IMHO 2016-03-24 08:27:58 oneinsect: I was fooled by RK's proprietary stuff for a month 2016-03-24 08:28:26 if you use upstream uboot+kernel then it's not that different than on x86 2016-03-24 08:28:34 hmmm 2016-03-24 08:28:35 http://wiki.alpinelinux.org/wiki/Alpine_on_ARM 2016-03-24 08:28:52 my unfinished page :D 2016-03-24 08:30:00 http://wwwtest.alpinelinux.org/posts/Alpine-3.3.3-released.html 2016-03-24 08:30:07 should we mention rpi3 boot issues? 2016-03-24 08:30:17 but to my understanding we have not tested it yet? 2016-03-24 08:31:30 ncopa: maybe something like "there might be boot issues on rpi3, still under investigation" 2016-03-24 08:32:54 i think maybe fabled tested it? 2016-03-24 08:33:27 I only read people stating that it failed to boot 2016-03-24 08:33:46 yes it failed for me 2016-03-24 08:33:54 3.3.2 2016-03-24 08:33:55 can anyone test it *now*? http://fr.alpinelinux.org/alpine/v3.3/releases/armhf/ 2016-03-24 08:34:01 the 3.3.3 2016-03-24 08:34:08 rightaway 2016-03-24 08:34:16 great!!! 2016-03-24 08:34:52 i write that it fixes rpi3 booting but wait with pusblish til its confirmed :) 2016-03-24 08:36:50 "This is a bugfix release of the v3.3 musl based branch. It fixes a boot 2016-03-24 08:36:50 problem with rasperry pi 3." 2016-03-24 08:36:54 sounds ok? 2016-03-24 08:39:30 ncopa give me 30 minutes to download and test....yes sounds perfect 2016-03-24 08:44:27 hum 2016-03-24 08:44:37 looks like i never posted the news about the community repo 2016-03-24 08:44:47 from october 2015 2016-03-24 08:45:05 http://tpaste.us/AQzl 2016-03-24 08:54:00 ncopa: 2016-03-24 08:54:18 it booted fine but there is an error that shows up before openRC 2016-03-24 08:54:54 fbsplash: /media/*/fbsplash0.ppm: no such file 2016-03-24 08:54:58 or directory 2016-03-24 08:55:01 fabled: 2016-03-24 08:55:22 ok 2016-03-24 08:55:24 but it boots? 2016-03-24 08:55:28 http://alpinelinux.org/posts/Alpine-3.3.3-released.html page is broken 2016-03-24 08:55:31 (morning) 2016-03-24 08:55:33 yes it boots fine 2016-03-24 08:55:38 it boots fine 2016-03-24 08:55:38 ok good 2016-03-24 08:55:52 the slash error message is non-fatal 2016-03-24 08:55:59 ScrumpyJack: i have not yet published it 2016-03-24 08:56:19 yes its non-fatal 2016-03-24 08:57:15 http://wwwtest.alpinelinux.org/posts/Alpine-3.3.3-released.html 2016-03-24 08:57:50 typo in raspberry pi 2016-03-24 08:58:58 i am going to test wireless now in sometime 2016-03-24 08:59:00 and report back 2016-03-24 08:59:14 booting is a step forward :) 2016-03-24 08:59:18 thank you very much! 2016-03-24 08:59:38 my humble pleasure (bows down) 2016-03-24 09:00:43 everything ok now?http://wwwtest.alpinelinux.org/posts/Alpine-3.3.3-released.html 2016-03-24 09:00:56 "Raspberry Pi 3" spelled correct now right? 2016-03-24 09:01:38 yes 2016-03-24 09:01:42 ncopa: 2016-03-24 09:01:58 ok? 2016-03-24 09:02:31 ok to push release notes? 2016-03-24 09:04:51 yes 2016-03-24 09:04:55 please 2016-03-24 09:05:16 done 2016-03-24 09:05:29 thanks ncopa: 2016-03-24 09:06:18 left on my release-checklist: send announcement email, update irc channesl, celebrate! 2016-03-24 09:06:21 :) 2016-03-24 09:08:38 heeheee 2016-03-24 09:08:46 we need to pep up forums 2016-03-24 09:08:52 write more tutorials 2016-03-24 09:08:57 i will do so 2016-03-24 09:09:00 slowly 2016-03-24 09:09:09 after i get free this year 2016-03-24 09:09:25 make alpine more popular in embedded world (arm specially) 2016-03-24 09:09:43 there is room for improvement there yes 2016-03-24 09:10:17 i wonder if it would be an idea to ship a pre-configured sys-disk ext4 image for arm 2016-03-24 09:10:28 so you can simply dd it 2016-03-24 09:10:53 ncopa lot of us like the current setup 2016-03-24 09:10:59 good :) 2016-03-24 09:11:07 with tmpfs as root? 2016-03-24 09:11:28 coz its very easy to do fat32 format and just copy files 2016-03-24 09:11:43 especially if a microsd card has fewer write cycles 2016-03-24 09:11:45 yes 2016-03-24 09:11:57 in memory only operation is soo good 2016-03-24 09:12:13 cool :) 2016-03-24 09:12:19 i mean lbu ci is the best thing of this century 2016-03-24 09:12:31 you will only commit or write to disk if needed 2016-03-24 09:12:33 this milleium too 2016-03-24 09:12:36 otherwise dont touch it 2016-03-24 09:12:45 it means the sdcard will survive for years 2016-03-24 09:13:04 yes that was the idea 2016-03-24 09:13:17 it is good for rotating disks too 2016-03-24 09:13:35 fewer (literally) moving parts 2016-03-24 09:13:41 this has to be taken to the masses (this concept...i dont think people know so deep)...so next coming months 2016-03-24 09:13:48 we have to take all this to the masses 2016-03-24 09:14:02 i will need to get more active 2016-03-24 09:14:04 in forums 2016-03-24 09:14:07 write tutorials 2016-03-24 09:14:12 and take it to peeps 2016-03-24 09:14:41 yes you are right ncopa 2016-03-24 09:14:47 DD is like bad... 2016-03-24 09:14:48 somehow 2016-03-24 09:15:38 have to see if x-windows can run too in this mode 2016-03-24 09:15:48 i havent tested it 2016-03-24 09:16:49 fabled: do you have time for my ticket 2016-03-24 09:16:50 http://bugs.alpinelinux.org/issues/4978 2016-03-24 09:18:37 ncopa: we have add support for arm boards 2016-03-24 09:18:43 make it more popular 2016-03-24 09:18:56 more arm boards* 2016-03-24 09:19:34 oneinsect, gst-omx should probably be okish 2016-03-24 09:19:45 are omxh264enc and omxh264dec part of it, or somewhere else? 2016-03-24 09:20:21 i generally download and compile from here 2016-03-24 09:20:21 git clone git://anongit.freedesktop.org/gstreamer/gst-omx 2016-03-24 09:20:46 i think the current biggest problem for reaching the masses is the documentation 2016-03-24 09:21:55 yesss 2016-03-24 09:22:03 we need to start writing an ebook 2016-03-24 09:22:07 i can start one 2016-03-24 09:22:16 can anyone help me fill it if needed 2016-03-24 09:23:04 things like basics to advanced including developing packages for alpine 2016-03-24 09:23:08 all in one ebook 2016-03-24 09:23:10 for free 2016-03-24 09:23:22 compiling it for various platforms 2016-03-24 09:23:27 would be awesome 2016-03-24 09:23:45 great then let gather me a table of contents.... 2016-03-24 09:23:57 you can correct them...add or delete 2016-03-24 09:30:03 What number would you have if you took only the fifth digit of the number one hundred forty-seven million six hundred eighty-six thousand six hundred thirty-six? 2016-03-24 09:30:59 147686636 would it be six or 6 2016-03-24 09:31:04 i hate captchas 2016-03-24 09:31:28 rather 8 or eight 2016-03-24 09:32:54 *creates bot with a language parser especially for that captcha that parses the words as tokens* 2016-03-24 09:33:26 i mean it doesnt register me in the wiki 2016-03-24 09:33:30 however hard i try 2016-03-24 09:34:01 private browsing? (to clear all cache) 2016-03-24 09:34:15 could just have a odd cookie 2016-03-24 09:34:22 i think i am getting it wrong the number 2016-03-24 09:34:42 pray skazz: 2016-03-24 09:34:46 what is the output for 2016-03-24 09:34:46 What number would you have if you took only the fifth digit of the number five hundred fifty-nine million five hundred seventy-eight thousand eight hundred thirty-one? 2016-03-24 09:35:06 7 / seven 2016-03-24 09:35:44 oneinsect: ^ 2016-03-24 09:35:50 worst captcha ive seen was the one for varnish bug reports 2016-03-24 09:36:17 Yeah, how to keep a low bug count on your project... 2016-03-24 09:36:18 if i input seven Incorrect or missing CAPTCHA. 2016-03-24 09:36:52 this the alpine wiki oneinsect? 2016-03-24 09:36:55 yes 2016-03-24 09:38:00 finally 2016-03-24 09:38:07 it takes numerics aka 7 2016-03-24 09:52:15 ncopa: 2016-03-24 09:52:19 http://wiki.alpinelinux.org/wiki/User:Atlury 2016-03-24 09:52:38 The Comprehensive Guide to Using and Developing using Alpine Linux 2016-03-24 09:53:16 are those headings good? 2016-03-24 09:55:59 oneinsect: i'm ok with anything 2016-03-24 09:56:10 great 2016-03-24 09:56:11 i'll leave it to you 2016-03-24 09:56:17 i will start writing the book then 2016-03-24 09:56:37 how about asciidoc or similar? 2016-03-24 09:57:05 and tracking the changes in git instead of wiki 2016-03-24 09:57:12 that will be great 2016-03-24 09:57:20 but i dont have access to your git 2016-03-24 09:57:30 create a project on github? 2016-03-24 09:57:37 sure let me 2016-03-24 09:58:18 nice thing with asciidoc is that we can spit out pdf, epub, html from same sources 2016-03-24 09:58:39 yessss 2016-03-24 09:59:02 I actually prefer booting from other media (sd/usb) and then proceed the install from there (chroot), pretty much the same as x86 2016-03-24 09:59:03 and we could even use it as the official docs 2016-03-24 09:59:22 docs.alpinelinux.org 2016-03-24 09:59:25 but I don't know if all SoCs support that 2016-03-24 09:59:28 can you create one and invite me as a contributor 2016-03-24 09:59:29 https://github.com/atlury 2016-03-24 09:59:32 that is user id 2016-03-24 09:59:50 ncopa: is it too late up get opensmtpd-5.7.3-p2 into 3.3.3 ? it crashes when used with openssl 1.0.2f 2016-03-24 10:00:07 ncopa: where can I find the uboot git repo of alpine? 2016-03-24 10:00:16 3.3.3 is already out 2016-03-24 10:00:35 ScrumpyJack: please file a bug on bugs.a.o so we dont forget it. set target 3.3.4 2016-03-24 10:01:48 opensmtpd-5.7.3-p2 is already built in edge 2016-03-24 10:02:43 yes, but i use http://bugs.alpinelinux.org/projects/alpine/issues?fixed_version_id=106&set_filter=1&status_id=o as checklist for next relase 2016-03-24 10:02:55 so i dont forget anything 2016-03-24 10:03:39 ncopa: no problem, i'll raise a b.a.o 2016-03-24 10:03:41 ScrumpyJack: nevermind, i already cherrypicked it 2016-03-24 10:03:55 ncopa: that's very kind of you :) 2016-03-24 10:04:09 but for the next time, if there is something you think should go to *-stable then please file a ticket 2016-03-24 10:04:21 if you only mention it in IRC it will get forgotten 2016-03-24 10:04:42 ncopa: makes sense. will do. 2016-03-24 10:07:51 xming: http://git.alpinelinux.org/cgit/aports/tree/main/uboot-mkimage/APKBUILD 2016-03-24 10:08:45 thanks, it's in the aports, I was looking for it as a separate project, silly me 2016-03-24 10:10:15 hmm that's just upstream uboot 2016-03-24 10:10:35 xming: It's not referred to in the alpine-iso package, not sure how it fits togeather to become the final image ... I've been wanting to get alpine working on odroid but not sure really where to start 2016-03-24 10:11:13 fabled mentioned some patches, but I don't see them in $sources 2016-03-24 10:14:46 skazz: that's just for tools? 2016-03-24 10:33:07 anyone that had problems with the invalid http://alpinelinux.org/atom.xml ? 2016-03-24 10:34:15 found the validator 2016-03-24 10:34:29 https://validator.w3.org/feed/check.cgi?url=http%3A%2F%2Fwwwtest.alpinelinux.org%2Fatom.xml 2016-03-24 10:36:18 fabled: 2016-03-24 10:36:22 i have an error in wifi 2016-03-24 10:37:02 direct firmware load brcm/brcmfmac43430-sdio.bin failed with error 2 2016-03-24 10:37:29 yes, the wifi firmware is not there - it's open issue 2016-03-24 10:37:31 nice, fixed now: https://validator.w3.org/feed/check.cgi?url=http%3A%2F%2Falpinelinux.org%2Fatom.xml 2016-03-24 10:37:35 we have it in bug tracker 2016-03-24 10:37:51 but firmware is there in bcrm folder fabled: 2016-03-24 10:38:01 i see tose sdio bin files etc 2016-03-24 10:38:05 those* 2016-03-24 10:39:38 aah i think i dont see those 2016-03-24 10:39:57 yes let me download /firmware folder on sdcard and see if it picks up in the next boot 2016-03-24 10:40:02 to* 2016-03-24 10:45:47 okie wlan0 is shown in ifconfig -a 2016-03-24 10:46:08 ncopa: /firmware/bcrm is getting picked up 2016-03-24 10:46:11 thanks for the tip 2016-03-24 10:46:23 picked up from sdcard to lib/firmware folder 2016-03-24 10:47:04 congrats 2016-03-24 10:47:21 maybe write a short wiki article on how to make alpine work on rpi3? 2016-03-24 10:47:38 its a hot topic 2016-03-24 10:47:45 i will write 2016-03-24 10:47:54 just want to check if it works in ap mode 2016-03-24 10:47:59 with hostapd 2016-03-24 10:48:06 in from our repos 2016-03-24 10:48:17 ap mode will be more hotter 2016-03-24 10:48:23 for people who want to run routers 2016-03-24 10:48:31 :) 2016-03-24 10:49:07 what is the max speed for the wlan of the rpi3? 2016-03-24 10:49:28 will it keep up with a 100mbps internet connection? 2016-03-24 10:51:27 its supposed to be 2.4GHz 802.11n 2016-03-24 10:54:40 The results show average transfer speeds of 39.34Mbit/sec for the onboard Wireless LAN 2016-03-24 10:54:51 ncopa: http://blog.pimoroni.com/raspberry-pi-3/ 2016-03-24 10:55:09 ok, not enough for my home net then 2016-03-24 10:55:40 as router that is 2016-03-24 10:55:59 that's not the rpi 3, that's the wifi spec... 2016-03-24 10:56:03 i suppose i could have alot fun with it as wifi client though 2016-03-24 10:56:16 hmm 2016-03-24 10:56:27 you need 802.11ac if you need more than 56Mb/s 2016-03-24 10:56:28 pimoroni said it shows very similar performance to the Official Raspberry Pi USB WiFi dongle, with transfer speeds of up to about 40Mbit/sec 2016-03-24 10:57:03 and i'd like ipsec at that speed to... 2016-03-24 11:09:09 ncopa: is msg.a.o down ? 2016-03-24 11:09:28 should not 2016-03-24 11:10:03 vkris: might be you need to reconnect due to the dns issue we had 2016-03-24 11:10:10 ok 2016-03-24 12:01:49 friends 2016-03-24 12:02:15 can anyone tell me how to configure chrony to NOT set system clock to hwclock when shutting down 2016-03-24 12:02:17 runnig in docker container alpine:edge `update-ca-certificates` (after installing it of course) gives https://gist.github.com/xificurC/db8224ea5e8bf252024a 2016-03-24 12:02:30 something like SYSTOHC="no" FORCE_SYSTOHC="no 2016-03-24 12:02:35 is this a known issue or is the process of updating the certificates different? 2016-03-24 12:02:47 xificurC: try installing openssl 2016-03-24 12:03:22 barthalion: it is 2016-03-24 12:03:31 I installed curl and it pulled in both 2016-03-24 12:03:45 libcrypto1.0 as well? 2016-03-24 12:03:54 no 2016-03-24 12:04:07 I think libcrypto1.0 provides that binary 2016-03-24 12:04:28 barthalion: same error after installing that 2016-03-24 12:04:54 c_rehash sits in /etc/ca-certificates/update.d/ , as noted by the last line 2016-03-24 12:04:58 let me check locally 2016-03-24 12:06:10 weird 2016-03-24 12:07:10 clandmeter: https://pkgs.alpinelinux.org/contents?filename=c_rehash&path=&pkgname=&repo=all&arch=x86_64 that file is not there, unfortunately 2016-03-24 12:07:35 ? 2016-03-24 12:09:15 clandmeter: package search says that this file is in libcrypto1.0, but it's not; http://sprunge.us/CYZU 2016-03-24 12:09:37 the other thing is that it should be there 2016-03-24 12:10:29 I guess the one from libcrypto and the one from ca-certificates are different as symlinking it to /usr/bin hangs the `update-ca-certificates` call 2016-03-24 12:11:18 /etc/ca-certificates/update.d/c_rehash is just a hook or something like that to update-ca-certificates 2016-03-24 12:11:31 it tries to execute /usr/bin/c_rehash, which is not present, while it should 2016-03-24 12:11:49 ncopa: any idea what's going on? I'm out of the alpinist loop since a while 2016-03-24 12:12:28 http://sprunge.us/QjVa 2016-03-24 12:12:57 wut 2016-03-24 12:13:03 f78b44ebbbc514ac5ebc9f5dab4b8215abc6a1a6 2016-03-24 12:13:11 are you checking stable repo? 2016-03-24 12:13:16 we're talking about edge 2016-03-24 12:13:32 libcrypto1.0-1.0.2g-r3.apk 2016-03-24 12:13:43 libcrypto1.0-1.0.2g-r1 2016-03-24 12:13:52 r1 is yours 2016-03-24 12:14:01 outdated mirror then 2016-03-24 12:14:09 http://git.alpinelinux.org/cgit/aports/commit/main/openssl?id=f78b44ebbbc514ac5ebc9f5dab4b8215abc6a1a6 2016-03-24 12:14:23 yeah… 2016-03-24 12:14:37 dl-cdn is what? 2016-03-24 12:14:45 if it is needed for update-ca-certificates shouldn't it be a dependency? 2016-03-24 12:14:57 xificurC: it is a dependency… you have outdated package 2016-03-24 12:15:10 change your /etc/apk/repositories to nl.alpinelinux.org and do apk upgrade -U 2016-03-24 12:15:19 barthalion: nl is not the master mirror anymore 2016-03-24 12:15:21 or just update libcrypt1.0 this way 2016-03-24 12:15:29 clandmeter: well, it's at least up to date 2016-03-24 12:15:35 most are 2016-03-24 12:15:41 best is to take the cdn one 2016-03-24 12:15:51 its a cache of fr.a.o 2016-03-24 12:15:59 so what should I do -_- 2016-03-24 12:16:16 barthalion: we moved master mirror (rsync.a.o) to fr.a.o, (scaleway arm server) 2016-03-24 12:16:20 clandmeter: and this is what I had in /etc/apk/repositories 2016-03-24 12:16:22 we advise since recently to use dl-cdn.a.o 2016-03-24 12:16:24 which resulted in problem above 2016-03-24 12:16:36 it has better internet 2016-03-24 12:17:06 fastly is also sponsoring dl-cdn.a.o which is a caching reverse varnish proxy 2016-03-24 12:17:10 I can only agree on dl-cdn is better 2016-03-24 12:17:12 which uses rsync.a.o as backend 2016-03-24 12:17:23 for me fr.a.o is a lot worse 2016-03-24 12:17:26 is any of you actually reading what I wrote? 2016-03-24 12:17:33 no :) 2016-03-24 12:17:45 sorry, did you say someting? 2016-03-24 12:17:49 xificurC: change your mirror to nl or fr, both seems to be fine 2016-03-24 12:18:10 http://dl-cdn.alpinelinux.org/alpine/edge/main/x86_64/ and this give me 404 now, in my browser 2016-03-24 12:18:13 barthalion: ok, will there be a fix on edge for this at some point? 2016-03-24 12:18:28 xificurC: edge is fine, some mirrors are not 2016-03-24 12:18:48 ncopa: thats weird. 2016-03-24 12:18:49 barthalion: so edge should be updated to point at correct mirrors, no? 2016-03-24 12:18:54 barthalion: 404 is probably due to the domain expiration problem we had 2 days ago 2016-03-24 12:18:55 your cdn is wacky 2016-03-24 12:19:10 i think it should be good today 2016-03-24 12:19:24 xificurC: this is infra issue, I believe cdn was and will be fine at some point of time… 2016-03-24 12:20:38 xificurC: apk upgrade -U -a may help to make it pick up c_rehash 2016-03-24 12:22:28 yep changing to nl works 2016-03-24 12:22:33 thanks for now -_- 2016-03-24 12:22:35 ncopa: you cannot request the cdn to recache it? 2016-03-24 12:30:04 i think its the dns resolver cache thats broke 2016-03-24 12:30:27 not sure if we can request that 2016-03-24 12:51:09 okie AP mode works 2016-03-24 12:51:12 for rpi 3 2016-03-24 12:51:21 got dhcp also on mobile 2016-03-24 12:51:33 however two errors throw up 2016-03-24 12:51:56 Failed to create interface mon.wlan0: -95(Not supported) 2016-03-24 12:52:09 wlan0: Could not connect to kernel driver 2016-03-24 12:52:13 fabled: 2016-03-24 12:52:18 ncopa: 2016-03-24 12:53:23 i followed this 2016-03-24 12:53:24 https://gist.github.com/Lewiscowles1986/fecd4de0b45b2029c390 2016-03-24 13:07:12 hey :) 2016-03-24 13:09:11 hi leo 2016-03-24 13:09:26 leo-unglaub: 2016-03-24 13:29:29 fabled: 2016-03-24 13:29:56 updated the bug https://bugs.alpinelinux.org/issues/5297 2016-03-24 13:29:59 please check 2016-03-24 13:30:57 I'm back, I set the mirrors to nl and now running update-ca-certificates gives: WARNING: ca-certificates.crt does not contain exactly one certificate or CRL: skipping 2016-03-24 13:31:11 barthalion: ^ 2016-03-24 13:48:56 xificurC: it's just a warning 2016-03-24 13:49:24 okie wiki page is here 2016-03-24 13:49:25 http://wiki.alpinelinux.org/wiki/Raspberry_Pi_3_-_Configuring_it_as_wireless_access_point_-AP_Mode 2016-03-24 13:49:35 just incase anyone is interested 2016-03-24 13:49:40 correct if wrong please 2016-03-24 13:49:46 fabled: and ncopa: 2016-03-24 13:53:04 barthalion: my certs weren't getting through so I thought that might be the culprit, but it was just me having an unfortunate, silently ignored error in Dockerfile. Thanks 2016-03-24 13:53:41 no problem 2016-03-24 13:58:20 Anyone done WPA2-Enterprise with certificate auth? (Like one alpine as AP and one alpine as Client with own certificates) 2016-03-24 14:00:14 can try 2016-03-24 14:00:16 skazz: 2016-03-24 14:15:22 Hi there, could someone help me confirm a segfault on php7 on Alpine? 2016-03-24 14:15:44 Basically: apk add php7 php7-mcrypt then run php7 -r 'crypt("SomePassword", "Somehash");' 2016-03-24 14:16:02 any calls to crypt on php7 pukes solidly a SEGFAULT 2016-03-24 14:18:10 ncopa: 2016-03-24 14:18:21 suddenly i am seeing a missing package 2016-03-24 14:18:24 for armhf 2016-03-24 14:18:28 gst-plugins-ugly1 2016-03-24 14:18:42 its there for x86/64 2016-03-24 14:18:49 seems to have been in old version 2016-03-24 14:18:52 fabled: 2016-03-24 14:18:53 ? 2016-03-24 14:20:17 it includes the most lib 2016-03-24 14:20:18 libgstx264.so 2016-03-24 14:22:29 okie its missing in the edge 2016-03-24 14:22:33 http://mirrors.gigenet.com/alpinelinux/edge/main/armhf/ 2016-03-24 14:22:53 seems to be there in http://mirrors.gigenet.com/alpinelinux/v3.3/main/armhf/ 2016-03-24 14:23:05 why so? 2016-03-24 14:27:35 has anyone an idea why xorg is shutting down in this case? https://paste.debian.net/plain/418552 2016-03-24 14:44:08 leo-unglaub: disk full? 2016-03-24 14:45:35 ncopa: no, 100gb free 2016-03-24 14:46:10 its an acer aspire one D270, i have read that this special version does not need the xf86-video-intel, but the xf86-video-fbdev driver ... but its the same with both 2016-03-24 14:46:18 there must be something special with that version 2016-03-24 14:56:05 very wired .. the same driver works in debian ... hmmm .. 2016-03-24 15:42:43 according to google i would have to set the following: 2016-03-24 15:42:50 allowed_users=anybody 2016-03-24 15:42:50 needs_root_rights=yes 2016-03-24 15:42:56 but that seams very dangerous to me 2016-03-24 15:43:52 how to fix that is a question of policy, what does Alpine do with terminals? 2016-03-24 15:44:10 A common solution is to have a group that "owns" the console and terminals 2016-03-24 15:44:26 whichever user logs in to the console is added to that group 2016-03-24 15:44:48 and the terminals are set crw-rw---- root:consolegroup 2016-03-24 15:45:17 that would be one way 2016-03-24 15:45:20 It's not the only solution, but it's probably the easiest one. 2016-03-24 15:45:22 but x should not run as root 2016-03-24 15:45:37 so needs_root_rights=yes is stupid in my opinion 2016-03-24 15:45:56 because other drivers dont need root, its just this fbdev driver 2016-03-24 15:46:00 how do you expect X to take control of graphical devices if it doesn't run as root? 2016-03-24 15:46:42 maybe it has changed... (my experience of X is *severely* dated) 2016-03-24 15:47:00 but back in the day (tm) X was always root 2016-03-24 15:48:42 back in 2014 openbsd patched xorg to run without root privileges 2016-03-24 15:48:48 and that patch went upstream 2016-03-24 15:48:52 so it should be possible 2016-03-24 16:47:36 ncopa: 2016-03-24 16:48:07 gst-plugins-ugly1 is not being shown at https://pkgs.alpinelinux.org/packages?name=%25gst%25&repo=all&arch=armhf&maintainer=all 2016-03-24 16:48:17 fabled: 2016-03-24 16:48:19 anyone 2016-03-24 16:48:29 cladmeter: 2016-03-24 16:50:33 anyone still alive 2016-03-24 16:50:38 no 2016-03-24 22:43:11 hi everyone, does wnyone knows how to boot Alpine from Hyper-V ? 2016-03-24 22:54:26 all bots here? 2016-03-24 23:02:00 most Alpine devs are in Europe, it's night time there. 2016-03-24 23:02:10 You'll have better luck coming earlier. 2016-03-25 07:35:58 i'm working on the php crypt() issues 2016-03-25 07:36:04 seems it's upstream mess up 2016-03-25 07:58:55 ok the sha256/sha512 issue is also clear now. i'm doing test build and pushing soon 2016-03-25 07:59:49 goodmorning climbers 2016-03-25 07:59:52 almost weekend! 2016-03-25 08:03:01 almost monday 2016-03-25 08:07:01 already a day off here :P 2016-03-25 09:26:10 morning 2016-03-25 09:26:27 today is bank holiday 2016-03-25 09:26:33 too bad i dont work on a bank :) 2016-03-25 09:27:31 in NL you would sit on it when its that kind of day :) 2016-03-25 09:27:37 you work on a whale 2016-03-25 09:27:50 lol 2016-03-25 09:31:20 lol 2016-03-25 09:33:13 i work *in* a whale. my new office: http://offbeatla.com/wp-content/uploads/2015/12/ben-laguna4.jpg 2016-03-25 09:34:03 Natanael Jonah 2016-03-25 09:34:16 ha! 2016-03-25 09:36:55 i'm also thinking of moving in to a container: http://www.containerhomeplans.org/ 2016-03-25 09:38:43 I doubt that would work well in Norway 2016-03-25 09:40:22 the nice thing would be that i could just put my home on a ship and go to brazil or something part of the year :) 2016-03-25 09:41:02 shipping yourself to anywhere :D 2016-03-25 09:41:15 don't forget to put a barcode on your forehead 2016-03-25 09:41:19 lol 2016-03-25 09:41:40 should i remove the one i have in my neck then? :) 2016-03-25 09:41:58 ^_^ 2016-03-25 09:42:05 for the first time you would fear real pirates more than cyber ones 2016-03-25 09:43:48 nah. i'd be containerized, and watch the action from my kitchen window while drinking morning coffe :) 2016-03-25 09:44:43 # docker stop ncopa-home 2016-03-25 09:53:41 fabled: i wonder why we use -fno-PIC instead of -fno-PIE here http://git.alpinelinux.org/cgit/aports/plain/community/go/no-pic.patch 2016-03-25 09:54:20 without that patch go apps linking to C libs will segfault 2016-03-25 09:54:27 fabled, ncopa: can you point me to the custom patches against upstream uboot? 2016-03-25 09:54:32 but i wonder if -fno-PIE would be moce correct 2016-03-25 09:54:37 xming, no custom patches 2016-03-25 09:54:48 our u-boot is at: http://git.alpinelinux.org/cgit/aports/tree/main/u-boot 2016-03-25 09:55:01 the distroconfig-* are from upstream iirc 2016-03-25 09:55:13 and the include-sys-types.h is simple just to make it compile on musl 2016-03-25 09:55:33 fabled: so the patches you once told me about are all upstreamed? 2016-03-25 09:55:55 xming, we removed our patches when u-boot introduced the 'distroconfig' mode 2016-03-25 09:56:02 it serves our purposes perfectly 2016-03-25 09:57:25 IIRC you said something about tgz patches in uboot, or is my memory fading? 2016-03-25 09:57:52 ? 2016-03-25 09:58:11 we had custom patches for the u-boot script to load kernel+initramfs 2016-03-25 09:58:17 but that was ~1 year ago 2016-03-25 10:00:16 fabled: okay thanks, so if the board is supported by the upstream, all I need is just a right config, is that correct? 2016-03-25 11:49:22 hey :) 2016-03-25 11:50:35 hi leo-unglaub 2016-03-25 11:53:21 the rpi kernel is w/o gresec? 2016-03-25 11:53:57 yesturday i tryed to convince my girlfriend of linux on her laptop .. i installed alpine with xfce on it but she did not like it because wifi and suspend, .. did not work and she also did not like xfce ... not easy enought for her ... so i thought: lets install fedora with gnome3 .. that should be easy ... well, fedora crashes on x start *g* then i thought ... okay, give ubuntu a shot ... it installs until the bootloader .. 2016-03-25 11:54:39 xming, yes, the -rpi flavor does not include grsec. the grsec kernel shoud have the vanilla upstream support for rpi though 2016-03-25 11:55:42 if I want to add support for rk ARM SoCs, what's the convention? 2016-03-25 11:56:36 use linux-grsec and create a config-grsec-rk 2016-03-25 11:56:48 or a separate package linux-rk 2016-03-25 11:58:22 or adapt config-grsec.armhf to include support for more SoCs? 2016-03-25 12:00:25 leo-unglaub: your gf is Linux incompatible 2016-03-25 12:01:58 i think all other fancy installers just have become very instable 2016-03-25 12:02:19 thats why i love alpine ... its very easy .. and thats why it works 2016-03-25 12:02:29 :) 2016-03-25 12:02:33 except broken wifi and suspend 2016-03-25 12:02:35 ACTION hides 2016-03-25 12:02:53 my wife used alpine linux (for some things) for a while 2016-03-25 12:03:30 and xfce 2016-03-25 12:04:18 i got an old mac mini, which didnt boot. i replaced the disk with ssd and upgraded it to newest osx 2016-03-25 12:04:23 so now she uses mac... 2016-03-25 12:05:00 ouch ... mac hurts ... 2016-03-25 12:06:33 actually, suspend works on alpine 2016-03-25 12:06:46 and you can try mate desktop 2016-03-25 12:06:51 on alpine 2016-03-25 12:07:09 wifi is another issue 2016-03-25 12:07:13 which i'd like to fix 2016-03-25 12:07:33 well, give me rust ;) then i write a wifi manager *g* 2016-03-25 12:07:46 lol! 2016-03-25 12:08:10 im pretty sure you could bring peace on earth if you only got rust :) 2016-03-25 12:08:15 no, but on a serious note .. the problem is propobly not xfce for beginners ... its more the genreal workflow ... starting from the commandline, typing startxfce4 ... this stupp propobly keeps her away 2016-03-25 12:08:25 damn strait ! 2016-03-25 12:08:38 #rust4ever 2016-03-25 12:08:54 on my (now old) lappy i use lxdm 2016-03-25 12:09:25 so i get login prompt in x. no need to `startxfce4` 2016-03-25 12:09:39 leo-unglaub: simply install display manager? 2016-03-25 12:10:01 have you ever seen how windows or osx looks like? regular users don't care what's behind fancy UI 2016-03-25 12:10:16 starting anything from commandline is not a good user experience 2016-03-25 12:10:58 there is a clear difference between you and your partner or parents 2016-03-25 12:12:04 and to be honest, I also became too lazy over time to be fascinated by typing 'startx' and regular tty on my laptop 2016-03-25 12:12:27 ditto about using window manager vs entire desktop environment 2016-03-25 12:12:39 I looked into rust, it's a cross-compiling hell 2016-03-25 12:13:13 need corss gcc + cross llvm + cross rust :O 2016-03-25 12:14:18 xming: when does it need gcc? 2016-03-25 12:14:48 well at least here to bootstrap 2016-03-25 12:15:05 almost all DM are broken now 2016-03-25 12:15:43 hmmm, it should only need llvm 2016-03-25 12:16:12 the biggest problem is the 3 stages while compiling the compiler 2016-03-25 12:16:19 sddm doesn't support xdmcp, nor normal multiseat, lightdm removed normal multiseat, kdm is unmaintained, gdm grrr 2016-03-25 12:16:21 because they do this trust trust thing 2016-03-25 12:17:32 leo-unglaub: well I need cross gcc to bootstrap, I can build a rust which builds static bins 2016-03-25 12:18:01 statically linked to musl 2016-03-25 12:19:39 getting a static musl gcc-based toolchain (native or cross) was hard enough... I don't even want to imagine the work for rust 2016-03-25 12:20:36 now theoretically it should be possible to use that to generate a stage0 rust compiler and use it instead of the official one 2016-03-25 12:20:55 "theoretically" being the operative word 2016-03-25 12:21:12 theoretically, gcc compiles itself, too 2016-03-25 12:23:46 getting a cross musl gcc is not that hard, and it can produce static bins, getting a static gcc is hard, but that was not what I was talking about 2016-03-25 12:25:09 "is not that hard". Have you actually DONE it? 2016-03-25 14:33:31 ncopa: re: 'we delete old versions' ^^, 2016-03-25 14:33:31 another reason for not deleting(rolling over) new pgks on stable releases is that, they are 2016-03-25 14:33:34 1. stable releases 2. historic importance 2016-03-25 14:33:55 on edge it does not matter 2016-03-25 14:34:32 1. stable release (official) * 2016-03-25 14:38:01 vkris: i perfectly understand why it may be good to keep older versions 2016-03-25 14:38:25 the question is if its worth the cost 2016-03-25 14:39:16 lets say we push a big package 2016-03-25 14:39:18 we will let vkris pay for the extra storage ;-) 2016-03-25 14:39:31 an hour later we push a fix for it 2016-03-25 14:39:41 it will take the double space 2016-03-25 14:39:49 who will ever want the broken version? 2016-03-25 14:40:08 not only storage, also bandwith cost for syncing mirrors 2016-03-25 14:40:31 how should you apk add older version? 2016-03-25 14:40:41 how would the apkindex look like? 2016-03-25 14:40:46 an additional repo? 2016-03-25 14:40:56 we could keep older but if we do that we'd blow up apk index size 2016-03-25 14:41:15 meaning that things will be slower and bigger for everyone 2016-03-25 14:41:31 not only for the few people who actually want older version 2016-03-25 14:41:42 how double space ? 2016-03-25 14:41:49 its time to go home and hides the eggs. 2016-03-25 14:41:53 and those few people who wants older version, seldomly want older version of everyhing 2016-03-25 14:42:02 they will likely only need a specific older version 2016-03-25 14:42:05 just the new updates/ folder with relevant INDEX file 2016-03-25 14:42:40 its not always a broken version ! it may be a feature release 2016-03-25 14:43:06 not always broken not always feature release 2016-03-25 14:43:14 how bandwidth, only update folder gets pulled ? 2016-03-25 14:43:40 my point is you only want the older version at few times 2016-03-25 14:43:56 ok 2016-03-25 14:44:00 i will give you an example 2016-03-25 14:44:47 lets take kernel 2016-03-25 14:45:07 how many different versions of the linux-grsec package do we currently have? 2016-03-25 14:45:19 answer is 5 2016-03-25 14:45:28 4 stable branches + edge 2016-03-25 14:45:35 also one can never say "hey can i download AL v 3.0.0 as it was released" 2016-03-25 14:45:55 that is 5 copies of same package 2016-03-25 14:46:15 if we would save *every* package 2016-03-25 14:46:15 it creates a minor sense of instability ( uhem cough! ) 2016-03-25 14:46:42 the we'd have 479 copies of linux-grsec now 2016-03-25 14:46:47 minimum 2016-03-25 14:47:49 but then would you have the possibility to grab linux-grsec-2.6.31.5 2016-03-25 14:48:01 ok 2016-03-25 14:48:04 not fair comparison 2016-03-25 14:48:17 since 479 are all in git log 2016-03-25 14:48:39 lets say since last branch of support (eg 3.0-table branch) 2016-03-25 14:49:06 110 2016-03-25 14:49:17 we'd have 110 copies of linux-grsec.apk 2016-03-25 14:49:24 instead of 5 2016-03-25 14:50:43 but then you could apk add linux-grsec- since we branched 3.0-stable which is the oldest we support 2016-03-25 14:51:44 so, based on that, mirror size woudl become 22 x bigger than today 2016-03-25 14:51:51 apk index woudl become 22x bigger 2016-03-25 14:52:17 do we want that? 2016-03-25 14:52:50 one possibility would be that we dont provide all versions ever shipped in edge branch 2016-03-25 14:52:54 and only do this for stable branches 2016-03-25 14:53:33 would be smaller 2016-03-25 14:54:59 we,, I have full main/ for ALv3.3.0(x86_64) as pulled on 20151219 - size 5GB~ 2016-03-25 14:55:02 and 20160223 updates comes to ~1.5gb 2016-03-25 14:55:14 well, I* 2016-03-25 14:55:34 but then you dont have all versions inbetween do you 2016-03-25 14:56:00 not all version, only officiall tagged is needed ! 2016-03-25 14:57:08 but then you cannot pick any version 2016-03-25 14:57:16 unless we tag 2016-03-25 14:57:49 you will not be able to say, oh this last fix bokre things, i will get back one version of linux-grsec 2016-03-25 14:57:59 you will only be able to get back to a tag backwards 2016-03-25 14:58:42 wow, 3.0-stable has 203 kernel releases 2016-03-25 14:58:50 for official stable release only tagged /updates 2016-03-25 14:59:27 its not pick-n-choose pkg service ;) 2016-03-25 15:00:49 the tagged updates are difference between last update + new updates 2016-03-25 15:01:13 well, pick-n-choose pkg service was exactly what was asked for 2016-03-25 15:01:49 so with only making binary snapshots on tags 2016-03-25 15:02:04 we'd have 20 copies of linux-grsec instead of 5 2016-03-25 15:02:11 that is 4x 2016-03-25 15:03:10 but thats only for linux-grsec 2016-03-25 15:03:26 i susppose not all packages would be updated 2016-03-25 15:03:41 so making snapshots of tags is probably doable 2016-03-25 15:04:28 but still unsure if its worth the cost 2016-03-25 15:06:07 eg, if a,b,c is in v3.0.0, 2016-03-25 15:06:07 b got updated in 50times and on tag v3.1.0 only one goes in updates-3.1.0 folder 2016-03-25 15:06:10 c got updated after tag v3.1.0 (several times) and also b, so at tag v3.2.0 b+c goes in updates-3.2.0 folder 2016-03-25 15:07:40 the version at time of taging is official tagged released and has permanent existence 2016-03-25 15:08:52 yes, make snapshot of binaries on tag 2016-03-25 15:08:54 kind of 2016-03-25 15:09:26 could be implemented by hardlinks and tag dirs 2016-03-25 15:09:35 by full nature, AL is a rolling release development 2016-03-25 15:09:45 imo this would be nice to have for the purpose of having an archive if alpine had the resources to spare, but I don't think it would solve any operational problem atm. 2016-03-25 15:10:09 so any admin wants to keep the in-between releases , it not difficult to do locally 2016-03-25 15:10:37 someone asked for git-like 2016-03-25 15:10:45 apk add package@ 2016-03-25 15:10:49 or similar 2016-03-25 15:11:15 but that would be a thing 2016-03-25 15:11:31 vkris: imo any admin wants to have a local snapshot mirror and wants to only roll out updates after staging them anyways 2016-03-25 15:11:51 no responsible admin will want to update production systems straight from the official mirrors 2016-03-25 15:13:34 it would be nice to have historical archive though 2016-03-25 15:13:36 this is what i do, unionfs old release and rsync 2016-03-25 15:13:37 snapshots 2016-03-25 15:13:49 that gives me update relases (most of time) 2016-03-25 15:14:10 could be done similar to that ^^^ 2016-03-25 15:14:16 ok 2016-03-25 15:14:40 we would not need push the entire historical archive to the mirrors 2016-03-25 15:14:47 but it would be nice to have it some place 2016-03-25 15:14:51 some location 2016-03-25 15:15:33 and having one servers with a few TB is not an impossible thing 2016-03-25 15:15:53 one or two servers for backup/redundancy 2016-03-25 15:16:31 it could be implemented with filesystem or block device snapshots 2016-03-25 15:16:45 like layers with overlayfs 2016-03-25 15:16:47 or aufs 2016-03-25 15:16:56 yes, thats mostly I do 2016-03-25 15:17:07 *nod 2016-03-25 15:17:12 of just take a diff between current and old file lists 2016-03-25 15:17:14 or with zfs, brfs, or lvm 2016-03-25 15:17:32 diff method is more accurate 2016-03-25 15:17:57 then on git push notification 2016-03-25 15:18:05 or mirror synced notification 2016-03-25 15:18:14 aufs may get confused by rsync sometimes (timestamp issues) 2016-03-25 15:18:15 we could make new snapshot, and rsync 2016-03-25 15:18:54 that way would could even say that we only keep 1 month of history or similar 2016-03-25 15:18:57 if needed 2016-03-25 15:20:19 i think it is doable, in theory 2016-03-25 15:20:35 imo, use btrfs, it'll give you the benefit of atomic operations 2016-03-25 15:23:12 but note, updates folder has its own new generated index file, only for pgkgs in it 2016-03-25 15:26:04 clandmeter: btw, what eggs ! ;) 2016-03-25 15:26:24 you can put them in update basket 2016-03-25 15:27:40 just dont put all eggs in same update basket :) 2016-03-25 15:28:03 https://bugs.alpinelinux.org/issues/1607 is still here vsftpd 3.0.3-r0 and kernel 4.1.20-0-grsec x86_64 :( 2016-03-25 15:28:07 yeah ! tag the chicken 2016-03-25 15:30:44 this talk of eggs and chicken rings a bell 2016-03-25 15:31:07 pls don't ring hard, while eggs are being layed 2016-03-25 15:31:43 bells wont wake up the eggs, but it may wake up the chickens 2016-03-25 15:33:17 tru_tru: $ zcat /proc/config.gz | grep SECCOMP 2016-03-25 15:33:17 CONFIG_HAVE_ARCH_SECCOMP_FILTER=y 2016-03-25 15:33:17 CONFIG_SECCOMP_FILTER=y 2016-03-25 15:33:17 CONFIG_SECCOMP=y 2016-03-25 15:33:44 hum 2016-03-25 15:33:53 that was on 4.4 kernel 2016-03-25 15:34:13 tru_tru: can you please try: modprobe configs && zcat /proc/config.gz | grep SECCOMP 2016-03-25 15:40:28 CONFIG_HAVE_ARCH_SECCOMP_FILTER=y 2016-03-25 15:40:28 CONFIG_SECCOMP_FILTER=y 2016-03-25 15:40:28 CONFIG_SECCOMP=y 2016-03-25 15:41:03 so it is probably different issue 2016-03-25 15:41:10 what does dmesg say? 2016-03-25 15:41:31 let me rephrase, does dmesg give any hints? 2016-03-25 15:42:23 ACTION reverting the "seccomp_sandbox=0" change 2016-03-25 15:43:28 nothing logged in /var/log/messages 2016-03-25 15:43:48 from the client: $ ftp 192.168.1.22 2016-03-25 15:43:48 Connected to 192.168.1.22 (192.168.1.22). 2016-03-25 15:43:48 500 OOPS: priv_sock_get_cmd 2016-03-25 15:44:04 but I have nothing on the server side 2016-03-25 15:45:56 if I put back the line seccomp_sandbox=0 in /etc/vsftpd/vsftpd.conf and rc-service vsftpd restart, everything works fine 2016-03-25 15:46:26 $ ftp 192.168.1.22 2016-03-25 15:46:26 Connected to 192.168.1.22 (192.168.1.22). 2016-03-25 15:46:26 220 (vsFTPd 3.0.3) 2016-03-25 15:46:48 weird 2016-03-25 15:46:52 i dont know 2016-03-25 15:47:08 vsftpd.log: Fri Mar 25 15:46:14 2016 [pid 2] CONNECT: Client "192.168.1.64" 2016-03-25 15:47:26 and dmesg does not give any things? 2016-03-25 15:47:44 if its grsecurity doing somethign then it should have showed up in dmesg i think 2016-03-25 15:48:23 no, nothing is logged in /var/log/messages. I am not familiar with alpinelinux: any other place to look at? 2016-03-25 15:48:41 run the command `dmesg` 2016-03-25 15:49:47 nothing more than the usual boot logs 2016-03-25 15:49:54 ok 2016-03-25 15:50:03 this may be relevant: https://bugzilla.redhat.com/show_bug.cgi?id=845980 2016-03-25 15:51:19 does not look like an alpine specific problem 2016-03-25 15:52:07 you could try run vsftpd server in strace 2016-03-25 15:52:15 vsftpd + glibc issue, no? 2016-03-25 15:52:34 i think vsftpd + kernel 2016-03-25 15:52:46 alpine dont use glibc 2016-03-25 16:00:26 for archival, 4 to 8 sdhc (64gb, class 10) card mounted read-only on a fugal mboard+2gb ram can be 1 unit 2016-03-25 16:00:29 (hdd or dvd spin and have higher maintainence cost) 2016-03-25 16:00:56 more units can be add on need to basis 2016-03-25 16:03:55 skarnet: yes have gcc which can produce statically binaries 2016-03-25 16:04:27 incl. libc (musl) 2016-03-25 16:05:30 um, yes. It's actually harder to produce a toolchain that *cannot* produce static binaries. 2016-03-25 16:05:57 I was talking about static gcc and binutils binaries. 2016-03-25 16:06:08 But that's not what you wanted anyway. 2016-03-25 16:08:08 does anyone knows how to produce a stage0 rust compiler? 2016-03-25 16:10:22 are there any upstream rust ticket for it? 2016-03-25 16:14:12 ncopa: was also thinking ifthere was separate toolchain folder 2016-03-25 16:14:29 but thats far fetched 2016-03-25 16:14:57 it would be optimized only for features needed 2016-03-25 16:16:03 the thought comes from fact that we cannot have newer curl with http2.0 features coz thats part of toolchain 2016-03-25 16:16:42 this is against development 2016-03-25 16:19:46 guessing is 5pm :) and its weekend 2016-03-25 16:19:53 take care 2016-03-25 16:43:55 ncopa: strace -ff vsftpd /etc/vsftpd/vsftpd.conf 2>&1 |tee /tmp/strace-ff-vsftpd.log not much inside 2016-03-25 16:45:01 https://dpaste.de/22hZ/raw 2016-03-25 16:55:44 [pid 2503] prctl(PR_SET_SECCOMP, SECCOMP_MODE_FILTER, {len = 74, filter = 0x74fb701d6a70}) = 0 2016-03-25 16:55:45 [pid 2503] rt_sigaction(SIGALRM, {0x9b525d38749, ~[RTMIN RT_1 RT_2], SA_RESTORER, 0x692d0420c225}, NULL, 8) = 0 2016-03-25 16:55:45 [pid 2503] +++ killed by SIGSYS +++ 2016-03-25 16:55:52 i dont know really 2016-03-25 16:57:35 ACTION keeping the workaround 2016-03-25 17:00:18 btw, can v3.3 boot from PXE or ipxe? ( I have tried memdisk iso raw + alpine-3.3.3-x86.iso -> failed) 2016-03-25 17:01:09 there was a pending patch for 3.3.2 mentionned http://irclogger.com/.alpine-devel/2016-01-18#1453099968 2016-03-25 17:09:23 ncopa: gentoo also mentionned the issue and link back to redhat bugzilla (https://wiki.gentoo.org/wiki/Vsftpd#SECCOMP_Filtering_and_64-bit_Kernels_with_.3Dnet-ftp.2Fvsftpd-3.0.x) 2016-03-25 18:05:27 yay :) mariadb 10.1.13 is abailable 2016-03-25 19:23:43 hello people, anyone there? 2016-03-25 19:23:51 hi 2016-03-25 19:24:02 hi jirutka! 2016-03-25 19:24:33 did you ever try to run Alpine under hyper-v or the sort? 2016-03-25 19:24:37 having some trouble with that 2016-03-25 19:24:44 nope 2016-03-25 19:44:14 ncopa: 2016-03-25 19:44:18 fabled: 2016-03-25 19:45:09 anyone still awake 2016-03-25 19:46:34 no 2016-03-25 19:46:57 gst-plugins-ugly1 is not being shown at https://pkgs.alpinelinux.org/packages?name=%25gst%25&repo=all&arch=armhf&maintainer=all 2016-03-25 19:47:21 its missing all of a sudden from arm edge 2016-03-25 19:47:40 any specific reason> 2016-03-25 19:47:41 ? 2016-03-25 19:47:57 i wonder if it's result of the recent dns breakage. maybe clandmeter would know better. 2016-03-25 19:48:20 its there in current 3.3.3 but i dont see in the edge..and even pkgs search... 2016-03-25 19:48:31 clandmeter: 2016-03-25 19:49:33 hi 2016-03-25 19:49:51 i dont know 2016-03-25 19:50:13 fabled: not sure thats even possible? shouldnt it catch up after its working again? 2016-03-25 19:51:54 is it included in the edge at all fabled that package? 2016-03-25 19:51:58 fabled: 2016-03-25 19:53:32 i se ugly0.10 but not ugly1 2016-03-25 19:53:50 exactly 2016-03-25 19:53:58 infact ugly1 is there for x86 x64 2016-03-25 19:54:02 in edge 2016-03-25 19:54:50 check here fabled: 2016-03-25 19:54:52 https://pkgs.alpinelinux.org/packages?name=%25gst%25&repo=all&arch=x86&maintainer=all 2016-03-25 19:55:03 seems it's not built 2016-03-25 19:55:31 weird 2016-03-25 19:55:44 indeed 2016-03-25 19:56:11 how can such an error creep up.....i feel strange without any logs or anyone noticing at all 2016-03-25 19:56:15 the builder had a related bug 2016-03-25 19:56:21 oooh 2016-03-25 19:56:49 dont we have a system for following up those bugs? 2016-03-25 19:57:32 need to remind ncopa to write it as a bug report. 2016-03-25 19:57:47 btw, gstreamer 1.8 is out 2016-03-25 19:58:09 but now its weekend 2016-03-25 19:58:11 basically abuild works, but the parent script aports-build sometimes does not notice a rebuild is needed 2016-03-25 19:58:17 if it's missing one of the subpkgs 2016-03-25 19:58:24 or something similar 2016-03-25 20:12:25 ncopa: please write it as bug report 2016-03-25 20:12:38 i reminded him fabled: 2016-03-25 20:12:43 :-D 2016-03-25 20:13:55 i will get some sleep 2016-03-25 20:13:58 see you guys 2016-03-25 20:14:02 have fun... 2016-03-25 22:10:37 I created something like Gentoo Overlay for my aports using Travis CI as a build server, packages are synchronized with my server using SSHFS. I also wrote complete guide how to setup own repository. https://github.com/jirutka/aports#how-to-setup-your-own-repository /cc @ncopa 2016-03-25 22:11:44 It’s intended for abuilds that are not merged into the official repository yet or don’t adhere to Alpine polices (abuilds with bundled dependencies). 2016-03-25 22:38:12 geany 1.27 is out with a new plugin to perform some line operations: http://plugins.geany.org/lineoperations.html 2016-03-25 22:38:17 it does look quit interresting 2016-03-26 09:53:42 urgh. this gettext() libintl_snprintf is hairy 2016-03-26 11:25:36 hey :) 2016-03-26 11:27:10 hi 2016-03-27 16:50:41 is it “safe” to use ruby 2.1.7 from v3.1/main on v3.3 system? 2016-03-27 16:55:39 "is it safe?" 2016-03-27 16:56:20 ACTION absent-mindedly plays with a dentist's drill 2016-03-27 17:02:51 might be a better idea to check it out and build it 2016-03-27 20:26:25 jirutka1: theoretically yes, but why do so? 2016-03-27 20:26:50 barthalion: because GitLab still supports only 2.1 :/ 2016-03-27 20:27:16 oh 2016-03-27 20:27:28 they have apparently some problems with their bloated codebase on MRI 2.2+… 2016-03-27 20:27:45 so I believe that pinning v3.1 in /etc/apk/repositories and doing apk add ruby_stuff@pin could work 2016-03-27 20:27:59 but you could also package 2,1 separately, assuming it's still supported 2016-03-27 20:28:21 that’s what I’ve been thinking about 2016-03-27 20:30:09 packaging separately… e.g. Homebrew uses same approach for some packages (there’s package gcc43, gcc44, …), but… it’s quite clunky, miss Gentoo slots… 2016-03-28 04:22:08 hm in setup-disk native_disk_install, isn't $root_size always empty string by the time it gets to use 2016-03-28 18:56:42 finally back 2016-03-28 19:46:45 ls 2016-03-28 23:03:17 wait the isos no longer have enough stuff to offline install? 2016-03-28 23:04:29 <_mjones> The 'standard' small 83mb one wants network repos, correct. 2016-03-28 23:23:32 extended doesn't need network repos? 2016-03-28 23:24:15 <_mjones> ammunta: afaik, correct. 2016-03-28 23:24:51 <_mjones> I had an install problem with the standard one regarding bnx2 firmware the other day, that was fine with the full-sized 3.3.4. 2016-03-29 07:33:26 morning 2016-03-29 07:33:50 ncopa: back in his 'whey ale' office 2016-03-29 07:35:20 yeah, i took a day off 2016-03-29 07:35:29 long time since i did that last time 2016-03-29 07:35:38 found typo in email while working with aports db 2016-03-29 07:35:39 main/micro-tetris/APKBUILD 2016-03-29 07:35:39 main/lua-alt-getopt/APKBUILD 2016-03-29 07:36:21 ncopa, does newer kernel 4.2+ support stacking LSMs ? 2016-03-29 07:36:55 dunno 2016-03-29 07:37:01 http://youtu.be/iWay8SbfgdM 2016-03-29 07:37:10 https://github.com/smack-team 2016-03-29 07:37:10 Will having SECURITY_SMACK=y be useful, if it does not conflict with grsec ? 2016-03-29 07:37:30 if sounds nice can this be in AL v3.4 ? 2016-03-29 07:44:45 what do you want to use it for? 2016-03-29 07:45:17 what is the cost? how big is it? any preformance or memory cost? 2016-03-29 07:51:36 if its a separate module, then should not impact on performance 2016-03-29 07:51:59 only issue is, will is conflict with grsec when enabled 2016-03-29 07:52:30 i think mobile devices use them 2016-03-29 07:52:40 smart ones 2016-03-29 07:53:44 code is in, security/smack of kernel pkg 2016-03-29 07:54:25 what do you want to use it for? 2016-03-29 07:55:37 hmm, not fully sure on it ;), saw video, looked at code and seems nice 2016-03-29 07:55:44 so thought of asking 2016-03-29 07:56:34 since we have kernel v4.4+ maybe use some newer features 2016-03-29 08:03:05 vkris | if sounds nice < it sounds nice, but it's still really limited to other MAC systems 2016-03-29 08:04:05 ncopa: it should be pretty lightweight btw, it's the lightest MAC system there is in the kernel 2016-03-29 08:04:19 light in functionnalities as well :) 2016-03-29 08:06:48 its fs features uses xattrs 2016-03-29 08:06:52 but I must admit I've not used it for a long time :) 2016-03-29 08:08:20 so maybe not useful for AL for now, or unless someone comes with nice usecase 2016-03-29 08:09:47 Hey seems they have a better userland now 2016-03-29 08:11:07 I remember having to code stuff myself to change file attributes 2016-03-29 08:14:37 the main advantage is it's simplicity 2016-03-29 08:14:55 and it's its main disavantage as well... 2016-03-29 08:15:19 you start getting into complex rules it's possibilities aren't sufficient 2016-03-29 09:21:39 hey :) 2016-03-29 09:23:06 howdy 2016-03-29 09:39:24 hi 2016-03-29 09:42:58 \o/ 2016-03-29 09:44:10 i think its kind of funny how people always invent new stuff to secure other stuff, but never fix the main problem ... 2016-03-29 09:44:18 best example: DNSSEC, DANE ... 2016-03-29 09:44:31 even with both of those things, DNS is still insecure as shit .. 2016-03-29 09:46:22 :) 2016-03-29 10:02:55 use dnscurve :) 2016-03-29 10:07:20 skarnet: i like djb but dnssec has two problems: A: the key distribution and B: no one is using it 2016-03-29 10:09:18 yes, that's DNSSEC. 2016-03-29 10:09:33 DNSCurve is not DNSSEC. 2016-03-29 10:12:17 skarnet: i meant dnscurve 2016-03-29 10:12:21 misstyped 2016-03-29 10:16:08 well, it's a protocol that fixes the main problem. The fact that nobody's using it can't be blamed on the author of the protocol (although, it's djb, so yeah, maybe it can :D) 2016-03-29 10:29:20 clandmeter: what was the url for packages that are listed as out of date? 2016-03-29 10:30:11 flagged 2016-03-29 10:32:00 thanks 2016-03-29 12:08:35 <_mjones> ncopa: I've been asked to invite you to speak in new york city. What's your preferred email address to where I can send the invitation? 2016-03-29 12:09:25 There are 4 pull requests from me waiting for review and merge https://github.com/alpinelinux/aports/pulls :) 2016-03-29 15:07:11 clandmeter: could you please do the same here: https://pkgs.alpinelinux.org/contents 2016-03-29 15:07:30 that way someone can start searching directly or tab to another input field 2016-03-29 15:33:20 @ncopa: are you here? 2016-03-29 15:44:19 jirutka: im here 2016-03-29 15:44:31 There are 4 pull requests from me waiting for review and merge https://github.com/alpinelinux/aports/pulls :) 2016-03-29 15:44:48 i know... sorry for hving full hands :-/ 2016-03-29 15:45:12 no problem, I just wasn’t sure if you watch PRs on GitHub 2016-03-29 15:45:32 i get notifications on github 2016-03-29 15:46:00 btw is there anyone else who can and is responsible for reviewing and merging aports? I know just about you, but don’t wanna overload you :) 2016-03-29 15:49:56 ncopa loves being overloaded 2016-03-29 15:51:25 ah, okay XD 2016-03-29 15:52:32 I have also question about abuilds… why we don’t use `set -e` and instead most commands in build and package functions use `|| return 1`? 2016-03-29 15:53:07 jirutka: only reason is that i havent been able to actually implement it 2016-03-29 15:53:12 and clean up after it 2016-03-29 15:53:29 i want to use set -e too 2016-03-29 15:53:39 unfortunally `set -e` cannot be enabled inside a function and I’m not sure if I can enable it for whole script 2016-03-29 15:53:54 it will break things for sure 2016-03-29 15:54:10 i can do a wrapper for functions 2016-03-29 15:54:14 do you have some ideas how to solve this? 2016-03-29 15:54:25 call_func() { ( $1 ); } 2016-03-29 15:54:38 call_func() { (set -e; $1 ); } 2016-03-29 15:55:23 yay for forking a shell in every function 2016-03-29 15:55:43 its only 3 functions, prepare, build, install 2016-03-29 15:55:46 or similar 2016-03-29 15:55:53 gotcha, I’ll experiment with it 2016-03-29 15:56:15 probably better to clean up the abuild code to deal with set -t 2016-03-29 15:56:17 IMO there is no valid reason not to use -e in any non-trivial script whatsoever 2016-03-29 15:56:17 set -e 2016-03-29 15:56:26 well you need to do this recursively, else a set -e can lie to you 2016-03-29 15:56:59 i think the best solution is to make abuild #!/bin/sh -e 2016-03-29 15:57:06 yes 2016-03-29 15:57:07 me too 2016-03-29 15:57:19 its probably not too big project 2016-03-29 15:57:31 but there are a few lines to read... 2016-03-29 15:58:01 btw if anyone interested: I created something like Gentoo Overlay for my aports using Travis CI as a build server, packages are synchronized with my server using SSHFS. I also wrote complete guide how to setup own repository. https://github.com/jirutka/aports#how-to-setup-your-own-repository 2016-03-29 15:59:09 It’s intended for abuilds that are not merged into the official repository yet or don’t adhere to Alpine polices (abuilds with bundled dependencies, I call them bundles). 2016-03-29 16:07:04 nice 2016-03-29 16:07:32 what would be awesome was if we could use travis to test build github pull requests 2016-03-29 16:07:51 this is possible 2016-03-29 16:08:12 i think that has higher prio than abuild run as set -e 2016-03-29 16:08:28 as it would offload patch reviewing 2016-03-29 16:08:36 it would filter out the garbage 2016-03-29 16:08:48 okay, I’ll look at it, this would be quite easy 2016-03-29 16:09:03 would be awesome 2016-03-29 16:09:25 maybe use git diff --name-only to get a list of modified APKBUILDs 2016-03-29 16:09:41 yeah, that’s exactly what I’m thinking about 2016-03-29 16:09:52 the you can use 'ap builddirs pkg1 pkg2' to get the build order 2016-03-29 16:11:39 so the goal is to build modified/new packages, just to ensure that they are actually “buildable”? 2016-03-29 16:11:50 yes 2016-03-29 16:11:52 okay 2016-03-29 20:20:34 hmmm 2016-03-29 20:20:37 very strange ... 2016-03-29 20:20:46 install issues again 2016-03-29 20:20:54 i hate debugging install issues ... 2016-03-29 22:18:25 hello 2016-03-29 22:18:33 I've install virt-manager from community, but it complain about: libintl_fprintf: symbol not found 2016-03-29 22:18:45 by the way libintl is installed 2016-03-29 22:19:29 should I need to report the issue, or there's a fix? 2016-03-30 05:17:50 folks 2016-03-30 05:18:09 does alpine have any audio web stream playing client 2016-03-30 05:23:50 GUI: audacious 2016-03-30 05:23:54 CLI: mpv 2016-03-30 05:57:02 thanks 2016-03-30 05:57:06 kaniini: 2016-03-30 06:10:15 Hello 2016-03-30 06:10:38 Is anyone here? 2016-03-30 06:10:59 Yes! 2016-03-30 06:12:54 Does anybody here know about an issue with apk add pycffi? 2016-03-30 06:14:17 hey faffolter are you a dev on alpine linux? We have an issue with py-cffi package right now while using alpine linux on docker 2016-03-30 06:17:22 hi 2016-03-30 06:17:25 SeanK_: what's the problem? asking specific question allows everyone to help, really 2016-03-30 06:18:43 So I currently use alpine linux as a docker image. When I try to install anything in the container that has a dependency of py-cffi it fails 2016-03-30 06:19:07 it tries to install py-cffi-1.3.0, but when I go to the alpine linux package management site it lists py-cffi as 1.4.2 2016-03-30 06:19:17 If you have docker, I can walk you through the steps to recreate 2016-03-30 06:19:25 SeanK_: would be good 2016-03-30 06:19:29 for if you paste a dockerfile 2016-03-30 06:19:35 or* 2016-03-30 06:20:06 (12/12) Installing py-cffi (1.3.0-r0) 2016-03-30 06:20:13 it installed py-cffi for me 2016-03-30 06:20:55 the pkgs.a.o shows the packages from 'edge', the rolling, development branch (git master) 2016-03-30 06:20:59 So I'm running a fresh image pulled from the Dockerhub right now, how old is your image? 2016-03-30 06:21:55 CREATED ... 3 weeks ago 2016-03-30 06:22:15 stable repos don't necessarily have the same version as edge 2016-03-30 06:22:23 what do you mean exactly by fails 2016-03-30 06:22:29 So I'm running from no saved image. I run the command docker run -it alpine ash 2016-03-30 06:22:43 Which gets me into an ash shell inside of a freshly pulled alpine image from dockerhub alpine:latest 2016-03-30 06:22:53 I then run apk update 2016-03-30 06:22:58 that is alpine 3.3 2016-03-30 06:23:16 Right 2016-03-30 06:23:20 Then after I run apk update 2016-03-30 06:23:25 I run apk add py-cffi 2016-03-30 06:23:27 and I get this: 2016-03-30 06:23:33 ERROR: py-cffi-1.3.0-r0: package mentioned in index not found (try 'apk update') 2016-03-30 06:23:53 Same errors occurs for any apk add package with py-cffi as dependency 2016-03-30 06:23:57 what does the apk update uptput say? 2016-03-30 06:24:08 my output here is: 2016-03-30 06:24:11 v3.3.3-8-gaa90166 [http://dl-cdn.alpinelinux.org/alpine/v3.3/main] 2016-03-30 06:24:11 v3.3.3-9-gfc38db2 [http://dl-cdn.alpinelinux.org/alpine/v3.3/community] 2016-03-30 06:24:27 v3.3.3-8-gaa90166 [http://dl-cdn.alpinelinux.org/alpine/v3.3/main] 2016-03-30 06:24:34 v3.3.3-9-gfc38db2 [http://dl-cdn.alpinelinux.org/alpine/v3.3/community] 2016-03-30 06:24:36 Same 2016-03-30 06:25:13 and still you get that error on `apk add py-cffi`? 2016-03-30 06:25:43 Correct 2016-03-30 06:26:04 My partner is running the exact same thing and occuring the same error 2016-03-30 06:26:13 it's happening on both our local machines as well as our digital ocean droplets 2016-03-30 06:26:49 SeanK_: ok, that is weird 2016-03-30 06:27:19 but it sounds like it does find the package in the index 2016-03-30 06:27:34 can you try: `apk add --simulate py-cffi` 2016-03-30 06:28:12 will that give error or success? 2016-03-30 06:28:17 That worked when I run apk add --simulate py-cffi 2016-03-30 06:28:21 ok 2016-03-30 06:28:22 so 2016-03-30 06:28:27 but when I try to install a package with py-cffi it still breaks 2016-03-30 06:28:32 sounds like index is ok 2016-03-30 06:28:41 but fetching the actual file fails 2016-03-30 06:29:34 the reason it works for me is that dl-cdn will give different ip addr depening on location 2016-03-30 06:30:26 hmm okay 2016-03-30 06:30:37 can you try open http://dl-cdn.alpinelinux.org/alpine/v3.3/main/x86_64/ in a web browser 2016-03-30 06:30:53 Yeah that works for me 2016-03-30 06:31:09 can you find py-cffi there? 2016-03-30 06:31:17 Doesn't exist 2016-03-30 06:31:21 sounds like outdated index, if not for the fact that the index is fine 2016-03-30 06:31:53 its in "community" 2016-03-30 06:32:00 http://dl-cdn.alpinelinux.org/alpine/v3.3/community/x86_64/ 2016-03-30 06:32:03 ncopa 2016-03-30 06:32:17 hi oneinsect 2016-03-30 06:32:18 I get a 404 trying to open that link 2016-03-30 06:32:33 that is the problem 2016-03-30 06:32:51 hello....there is one issue with an Intel NUC board DE http://ark.intel.com/products/78576/Intel-NUC-Board-DE3815TYBE 2016-03-30 06:32:57 it comes with inbuilt 4GB emmc 2016-03-30 06:33:07 seanK_: where are you geographically? 2016-03-30 06:33:14 New York 2016-03-30 06:33:27 Minnesota ^ Same issue as SeanK 2016-03-30 06:33:30 however during installation of setup-alpine it exists saying 2016-03-30 06:33:31 I believe our droplets are in New York as well 2016-03-30 06:34:02 andyshinn: ^^^ seems like dl-cdn.alpinelinux.org/alpine/v3.3/community/x86_64 gives 404 from New York 2016-03-30 06:34:18 SeanK_: can you try without the x86_64 at the end? 2016-03-30 06:34:27 ncopa: same for here 2016-03-30 06:34:31 from even 2016-03-30 06:34:42 it didn't change since my last complaint 2016-03-30 06:35:14 I can access it without the x86_64 2016-03-30 06:35:23 what does it show 2016-03-30 06:35:31 is the community dir listed there? 2016-03-30 06:35:44 sorry, i mean is the x86_64 dir listed 2016-03-30 06:35:47 yes 2016-03-30 06:35:47 Yes 2016-03-30 06:35:57 and if you click on it you get 404 2016-03-30 06:36:05 Correct 2016-03-30 06:36:12 2 other directories that I do not get a 404 for 2016-03-30 06:36:18 x86 2016-03-30 06:36:28 and armhf do not return 404 2016-03-30 06:36:47 I'm seeing 404 from new haven (hits new york, no surprise), 200 from paris and montreal 2016-03-30 06:36:59 if that's, like, of interest to anyone 2016-03-30 06:37:34 and this does not give 404? http://fr.alpinelinux.org/alpine/v3.3/community/x86_64/ 2016-03-30 06:37:58 Correct no 404 2016-03-30 06:38:09 No it doesn't give me one either. 2016-03-30 06:38:13 ok. so its fastly that has some issue 2016-03-30 06:38:23 morning 2016-03-30 06:38:23 andyshinn: help! :) 2016-03-30 06:38:52 SeanK_: a workaround til we figure this out i to use another mirror 2016-03-30 06:39:25 ah, cdns 2016-03-30 06:39:28 Great, how can I do that? 2016-03-30 06:40:31 could setup-apkrepos and copy out one of the mirrors 2016-03-30 06:40:31 SeanK_: apk add alpine-conf && setup-apkrepos -f && apk del alpine-conf 2016-03-30 06:40:41 then edit /etc/apk/repositories 2016-03-30 06:41:47 `setup-apkrepos -f` will try connect to all the listed mirrors and pick the "fastest" one 2016-03-30 06:41:57 and add it to /etc/apk/repositories 2016-03-30 06:42:04 you can then comment out the dl-cdn 2016-03-30 06:42:12 and uncomment the community 2016-03-30 06:43:12 if I run this : apk add alpine-conf && setup-apkrepos -f && apk del alpine-conf 2016-03-30 06:43:22 I get an py-cffi error again 2016-03-30 06:43:37 Must be dependency of openrc 2016-03-30 06:43:43 no 2016-03-30 06:43:44 or alpine-conf 2016-03-30 06:43:57 it happens because the other package you tried to install is still in /etc/apk/world 2016-03-30 06:44:05 the package that depends on py-cffi 2016-03-30 06:44:12 so apk will try fix the broken install 2016-03-30 06:44:33 i think you will have to apk del package-that-failed-due-to-missing-py-cffi 2016-03-30 06:44:37 Ahh, spun up a fresh container and you are right it works 2016-03-30 06:44:49 I just deleted that container and ran it in a new one 2016-03-30 06:44:55 that works too :) 2016-03-30 06:48:29 Great that worked 2016-03-30 06:48:53 Although is there anything we can do to help? As that fix is rather clunky for a dockerfile that is otherwise pretty clean 2016-03-30 06:49:25 Worked for me as well, thanks! 2016-03-30 06:55:44 SeanK_: we need figure out why fastly gives 404 and fix it. currently it is only andyshinn that can do that (afaik) 2016-03-30 06:59:02 Perhaps this -> https://status.fastly.com/ 2016-03-30 06:59:15 Scheduled Maintenance in ATL although they claim no downtime 2016-03-30 07:10:10 btw running an alpine:edge image works fine 2016-03-30 07:21:50 could be 2016-03-30 10:56:44 ncopa, upgrade to iptables 1.6.0 broke stuff 2016-03-30 10:57:20 so:libxtables.so.11 was sover bumped 2016-03-30 11:23:17 oh 2016-03-30 11:23:28 which was the old so ver? 2016-03-30 11:24:30 to get a list of packages needs rebuild: `apk search -r --quiet --exact --origin so:libxtables.so.` 2016-03-30 12:16:47 hey :) 2016-03-30 12:16:51 how is every body? 2016-03-30 12:41:10 @ncopa/@fabled: question: Alpine provides kmod. Are the module utilities provided by busybox insufficient for what Alpine needs? 2016-03-30 12:41:33 skarnet, i believe nlplug-findfs wants to use the kmod api 2016-03-30 12:41:45 generic modprobe functionality should be enough 2016-03-30 12:42:03 depmod needs to generate the binary indexes, so kmod is needed; though i wrote patch for bb depmod to do it too 2016-03-30 12:42:24 generate the binary indexes? 2016-03-30 12:42:32 well, maybe you can answer my real question then :) 2016-03-30 12:42:43 I want to compile and install a kernel and modules (portably, not Alpine only) 2016-03-30 12:42:58 and some distros don't provide a xz-enabled kmod 2016-03-30 12:43:13 so I need to provide my own depmod in the build system 2016-03-30 12:43:26 kmod is being a huge PITA by refusing to link statically 2016-03-30 12:43:37 so I'm wondering whether busybox would do the job 2016-03-30 12:44:13 the binary indexes are needed by kmod only 2016-03-30 12:44:22 what files are those? 2016-03-30 12:44:33 modules.foobar.bin ? 2016-03-30 12:44:35 yes 2016-03-30 12:44:37 models.*.bin 2016-03-30 12:44:54 it's the same data as the file without .bin, but it has a radix the lookup table in it 2016-03-30 12:45:07 tree* 2016-03-30 12:45:17 will kmod work without those? (less efficiently of course) 2016-03-30 12:45:20 no 2016-03-30 12:45:28 will bb modprobe work? 2016-03-30 12:45:31 yes 2016-03-30 12:45:44 so why isn't the answer "fuck kmod" ? 2016-03-30 12:45:49 bb modprobe will not use them so far 2016-03-30 12:46:13 to speed up boot times, ncopa used kmod api to load the modules 2016-03-30 12:57:55 fucking rsyslog is changing a ton of libs, next version will introduce like 3 or 4 new aports :/ (due to changing in lib requirements) 2016-03-30 12:58:55 so we switch to journald? :) 2016-03-30 12:59:04 ACTION shoots ncopa 2016-03-30 12:59:10 :D 2016-03-30 12:59:19 fwiw, we *can't* 2016-03-30 13:00:13 bear with rsyslog for a few more months 2016-03-30 13:00:41 and convince dalias to add SOCK_STREAM support to syslog() 2016-03-30 13:00:51 my faith in alpine: 1. first linux distro with support for l2tpv3 2016-03-30 13:00:53 2. grsecurity 2016-03-30 13:00:56 3. NO FUCKING SYSTEMD 2016-03-30 13:01:18 skarnet: i have a hate/love relationship with rsyslog 2016-03-30 13:01:30 oh no... he said the s* word... 2016-03-30 13:01:36 i hate it but i need it :D and the only reason some new infra at work isnt alpinelinux is cause rsyslog was at 8.9 :x 2016-03-30 13:01:53 syslogd is really easy to implement as long as you have SOCK_STREAM support in libc syslog() 2016-03-30 13:02:08 but musl doesn't have it yet 2016-03-30 13:02:22 heh :D i hit a few "musl doesnt have it yet" before 2016-03-30 13:02:44 cant drop rsyslog though, its my hook to fucking get alpinelinux in production boxes at work :D its all part of my master plan 2016-03-30 13:04:10 btw, alpine is the best linux distro in the world right about now. 2016-03-30 13:04:23 ACTION cheers and waves hands in the air \o/ 2016-03-30 13:21:04 suqdiq: most "musl doesn't have it yet" occurrences are packages relying on glibcisms, and the right solution would be to fix the packages 2016-03-30 13:21:34 syslog() transport is different, and a gray area because there's no good specification 2016-03-30 13:21:53 and SOCK_STREAM transport *does* make syslogd implementation a lot easier 2016-03-30 13:22:04 but most syslogd implementations don't care, because they're monolithic 2016-03-30 13:23:00 since most people who use syslogd are using one of those monolithic implementations, musl just went with the quick approach and only supports SOCK_DGRAM 2016-03-30 13:23:27 but SOCK_STREAM support would allow for lighter syslogd implementations and a relative independence from rsyslog 2016-03-30 13:28:23 skarnet: its funny cause my last one was even a bsdism :D which conflicted me a little since i love the bsd :x anyway, it was a whole lib inside a .h file :D which musl rightfully chose not to include 2016-03-30 13:29:44 in my specific case with syslog, stream wouldnt make a different in only about 10% of the infra logging to me cause they support either dgram or RELP anyway 2016-03-30 13:30:09 sure, if you depend on rsyslog anyway, stream support wouldn't change anything for you 2016-03-30 13:30:10 anyway rsyslog is useful also for people that need to shift data into other stuff like kafka/elasticsearch/whatever and dont want to use logstash :D 2016-03-30 13:30:49 but your arguments sound good for alpine in general :) who do we have to hit with a wrench to get stream ? 2016-03-30 13:32:11 havent looked at rsyslog from a security perspective but one of the reasons i had to move beyond 8.9 was actually an overflow fixed in 8.10 :D so i guess it would be a security improvement to allow the user base not to depend on rsyslog 2016-03-30 13:32:47 Is CDN still returning 404? 2016-03-30 13:33:43 andyshinn: i dont know. it has always worked for me 2016-03-30 13:42:20 suqdiq: dalias, in #musl, but please don't hit him with a wrench, he's kinda precious. I'll handle that (with a detailed technical post, not a wrench) later. :) 2016-03-30 13:44:11 a blowjob can be a wrench 2016-03-30 13:44:19 a good techincal argument as well :D 2016-03-30 13:45:42 yeah, unless you're his wife, I'm sure he'll prefer my technical arguments. ;) 2016-03-30 13:47:01 skarnet: there was a musl donation page somewhere wasn't there ? like a kickstarter or gofundme or whatever, i remember seeing it on the twiters 2016-03-30 13:48:06 I said I'll handle that later. That means in a few months. Don't bug people with that now! 2016-03-30 13:50:23 :D 2016-03-30 13:50:38 ill go back to the cave and be silent for the next few months, laters 2016-03-30 13:51:06 for the donation page: https://www.patreon.com/musl 2016-03-30 14:12:24 is Sam Dodrill here in the channelß 2016-03-30 14:12:25 ? 2016-03-30 15:06:11 emacs will rule the world someday, https://vimeo.com/22798433 2016-03-30 15:06:20 skarnet: thanks 2016-03-30 15:08:38 emacs??? 2016-03-30 15:08:44 vim 4 ever!!!!! 2016-03-31 06:25:52 webkit2gtk fails to build here: https://github.com/WebKit/webkit/blob/master/Source/JavaScriptCore/heap/MachineStackMarker.cpp#L682 2016-03-31 09:58:39 can someone remind me how a package ends up in community? 2016-03-31 10:07:03 coredumb, it needs to have maintainer, and someone to say that it's verified to work. then it can be moved to community by asking it here/email or just sending the git patch to alpine-aports 2016-03-31 10:08:03 fabled I mean community packages are not available as binaries right ? 2016-03-31 10:09:57 fabled: my bad I didn't know there was a new path for them! 2016-03-31 10:15:54 fabled: one other question then :) 2016-03-31 10:16:49 as I don't want to put my server on edge, how long to see a plugin in edge community to come in stable community ? 2016-03-31 10:17:59 s/plugin/package/ 2016-03-31 10:39:31 coredumb, edge is branched as new stable roughly every 6 months 2016-03-31 10:40:03 next one is due in early May 2016-03-31 10:40:28 we've had few exceptions to cherry-pick them to latest stable too but there would need to be good reasoning for that 2016-03-31 10:59:13 downloading apache httpd source is slow. maybe we should suggest that they use a faster web server for their tarballs :) 2016-03-31 11:26:27 fabled: dovecot-pigeonhole-plugin to be provided with the stable dovecot... wouldn't that be a good enough reasoning ? 2016-03-31 11:26:43 imap server without sieve is a bit mooh 2016-03-31 12:10:48 ncopa: lol 2016-03-31 12:55:08 i think i will push a rather bit bulk now 2016-03-31 12:55:27 ffmpeg, libvpx and libwebp abi breakage 2016-03-31 12:55:36 + alot rebuilds 2016-03-31 12:57:34 rebuilds of lot of heavy packages 2016-03-31 12:57:50 will likely keep arm builder busy til next week 2016-03-31 13:20:50 @ncopa: Example of Travis build for a pull request https://travis-ci.org/jirutka/alpine-aports/builds/119800445 :) 2016-03-31 13:22:43 wow 2016-03-31 13:24:38 aha 2016-03-31 13:24:49 you install a chroot 2016-03-31 13:25:01 yeah 2016-03-31 13:25:01 https://github.com/alpinelinux/aports/pull/42 2016-03-31 13:25:41 # This key will not be used anywhere, just to make abuild happy... 2016-03-31 13:25:45 it will actually be used 2016-03-31 13:25:58 if push libfoo and foo 2016-03-31 13:26:09 and foo depends on libfoo 2016-03-31 13:26:10 the built packages will not be deployed anywhere, so it doesn’t matter imho 2016-03-31 13:26:35 it may be installed as a build time dep for other package later 2016-03-31 13:27:03 With “anywhere” I mean that it will not be used anywhere outside the current Travis build 2016-03-31 13:27:04 does it use docker? 2016-03-31 13:27:12 no, chroot 2016-03-31 13:27:19 no, it doesn’t, because docker is totally unnecessary for this 2016-03-31 13:27:33 containers are said to be faster on travis 2016-03-31 13:27:37 Docker would bring extreme complexity for simple thing 2016-03-31 13:28:06 it can’t be 2016-03-31 13:28:34 i think you can have travis to spawn a container instead of a vm 2016-03-31 13:28:37 both for chroot and Docker, we can’t use Travis container-based infrastructure, it must be run on VM (sudo: required) 2016-03-31 13:29:07 https://docs.travis-ci.com/user/docker/ 2016-03-31 13:29:10 hm 2016-03-31 13:29:14 sudo: required => VM 2016-03-31 13:29:23 so docker would not bring any benefit here 2016-03-31 13:29:31 sounds like that yes 2016-03-31 13:29:54 except you don't need sudo to run on container-based infra… 2016-03-31 13:30:02 and the one without containers is certainly legacy one 2016-03-31 13:30:14 nope, you need 2016-03-31 13:30:23 because Travis doesn’t run Alpine, but Ubuntu 2016-03-31 13:30:27 that’s why I use chroot 2016-03-31 13:31:02 i wonder if we can request alpine from travis 2016-03-31 13:31:16 feature request 2016-03-31 13:31:21 sounds like reasonable request 2016-03-31 13:31:30 they won't die due to 4MB image after all 2016-03-31 13:31:35 lol 2016-03-31 13:31:46 We can try it, but currently Travis doesn’t support custom containers at all 2016-03-31 13:31:52 i think many projects may benefit 2016-03-31 13:31:53 you can’t specify what container you want to use 2016-03-31 13:32:16 i mean in addidtion to ubuntu vm, also support alpine vm 2016-03-31 13:32:21 and it’s not as easy as to add a configuration option to .travis.yml, image size doesn’t matter at all 2016-03-31 13:32:54 however, we can try to create a PR, it’d be great feature 2016-03-31 13:33:03 * PR → issue 2016-03-31 13:33:07 yup 2016-03-31 13:33:35 i mean, if you have a project you could test it against gnu libc and musl libc 2016-03-31 13:33:49 if it builds with musl libc, it'll likely build on anything :) 2016-03-31 13:34:41 well, that’s actually possible even now, you can chroot or run a docker image, but just on VM 2016-03-31 13:35:02 hey :) 2016-03-31 13:35:10 https://docs.travis-ci.com/user/multi-os/ 2016-03-31 13:35:23 I know about this 2016-03-31 13:35:23 hi leo-unglaub 2016-03-31 13:35:49 - os: linux 2016-03-31 13:35:50 dist: trusty 2016-03-31 13:35:50 sudo: required 2016-03-31 13:36:12 would be nice to be able to do: dist: alpine-edge 2016-03-31 13:39:40 anyway, Travis supports VMs and it’s unlikely to be disabled, because they don’t allow using sudo on containers (for very good reasons)… the only problem with VMs is that they boots slower, but since we don’t have dozens PRs per day, it shouldn’t be problem for as 2016-03-31 13:39:52 us 2016-03-31 13:40:13 there are circle ci too 2016-03-31 13:40:20 and probably others too 2016-03-31 13:40:38 I know, but we agreed to use Travis… 2016-03-31 13:41:05 I have no problem with using other CI, I just doesn’t have experience with others (just AppVeyor and it’s horrible, because Windows) 2016-03-31 13:41:28 was thinking if it was possible to use others without chroot 2016-03-31 13:42:13 well, but what’s problem with chroot? 2016-03-31 13:42:30 just hackish :) 2016-03-31 13:42:40 fabled: https://github.com/alpinelinux/aports/pull/42 2016-03-31 13:43:01 fabled: i think this can help us filter out bad patches 2016-03-31 13:43:49 jirutka: is it enought to add a .travis file or do we need register to travis somehow? 2016-03-31 13:44:13 you’ve proposed to use docker and simple chroot is hackish?! o.O 2016-03-31 13:44:29 you must also enable the repository on travis-ci.org 2016-03-31 13:44:56 https://travis-ci.org/profile/alpinelinux 2016-03-31 13:46:02 it's hackish, because it requires us to keep this stuff in mind and maintain our own set of scripts 2016-03-31 13:46:24 i don’t understand? 2016-03-31 13:46:26 versus travis-supported alpine image we could use that we need to request 2016-03-31 13:46:37 in any case, i think this is big a step forward 2016-03-31 13:46:43 i say we push it, enable it 2016-03-31 13:46:49 and then we improve things from there 2016-03-31 13:47:39 it’s ~30 lines https://github.com/jirutka/alpine-aports/blob/travis/.travis/install-alpine + 9 lines here https://github.com/jirutka/alpine-aports/blob/travis/.travis/common.sh#L8-L17 2016-03-31 13:47:39 the other stuff has nothing to do with chroot 2016-03-31 13:49:09 I agree that it’d be better if Travis supports Alpine, but currently it doesn’t : 2016-03-31 13:49:10 :( 2016-03-31 13:51:32 hm, there are few opened issues on travis-ci 2016-03-31 13:51:32 * https://github.com/travis-ci/travis-ci/issues/3417 2016-03-31 13:51:32 * https://github.com/travis-ci/travis-ci/issues/4825 2016-03-31 13:56:33 https://github.com/travis-ci/travis-ci/issues/4741 2016-03-31 14:00:27 i think we should file a n issue for dist: alpine 2016-03-31 14:01:17 and argue that it is nice to testbuild against musl libc in addition to gnu libc 2016-03-31 14:01:21 or similar 2016-03-31 14:01:59 yeah, I think that’s a good argument for supporting us 2016-03-31 14:05:55 hello 2016-03-31 14:06:12 hi 2016-03-31 14:06:38 do you know why this patch: http://patchwork.alpinelinux.org/patch/1812/ is not applicable? 2016-03-31 14:07:08 did I make something wrong? 2016-03-31 14:07:36 good question 2016-03-31 14:07:41 :) 2016-03-31 14:10:05 Mo0O: there are 3 different emails with autocutsel 2016-03-31 14:10:16 is it the last one you sent that i should apply? 2016-03-31 14:10:49 @ncopa, btw it’d be great to add to abuild an option to disable progress bar… or maybe it’d be better to add something like /etc/apkrc to define some options globally 2016-03-31 14:11:26 something like `apk --no-progress`? 2016-03-31 14:11:28 the workaround I’ve used works, but it’s more a hack 2016-03-31 14:11:39 yeah 2016-03-31 14:11:51 --no-progress Disable progress bar even for TTYs 2016-03-31 14:11:57 the problem is that abuild calls apk, but it’s not possible to specify what options it should pass to it 2016-03-31 14:12:16 got you 2016-03-31 14:12:17 abuild command doesn’t have that option 2016-03-31 14:12:26 ncopa: yes it's the last one, sorry for the repetion, I've failed while configuring git send-email, my bad 2016-03-31 14:12:31 I’m doing this https://github.com/jirutka/alpine-aports/blob/travis/.travis/setup-alpine#L273 2016-03-31 14:12:31 https://github.com/jirutka/alpine-aports/blob/travis/.travis/abuild-apk 2016-03-31 14:12:38 all 3 patch are the same 2016-03-31 14:12:57 just ignore the to others 2016-03-31 14:13:00 so sorry 2016-03-31 14:14:11 thanks ncopa 2016-03-31 14:21:19 Mo0O: pushed. thanks! 2016-03-31 14:22:08 (16:11:51) ncopa: --no-progress Disable progress bar even for TTYs 2016-03-31 14:22:08 i dont think thats correct ... when i look it up i get the following: 2016-03-31 14:22:08 --no-progress The progress of the alpine rust package 2016-03-31 14:22:12 hahahahahahaha 2016-03-31 14:22:16 sorry 2016-03-31 14:22:22 could not resist *g* 2016-03-31 14:22:23 :D 2016-03-31 14:22:46 ncopa: how can we push a community package into stable ? 2016-03-31 14:22:48 :) 2016-03-31 14:22:55 XD 2016-03-31 14:23:10 wait 2016-03-31 14:23:19 an edge community package into stable 2016-03-31 14:23:21 :D 2016-03-31 14:23:25 coredumb: which community package do you want? 2016-03-31 14:23:28 Ah, similar question! when can be a testing package pushed into stable? :) 2016-03-31 14:23:40 dovecot-pigeonhole-plugin 2016-03-31 14:23:43 (or community) 2016-03-31 14:25:19 coredumb: the normal process would be to file a bug on bugs.a.o and give a good reason to why it should be backported to stable 2016-03-31 14:26:30 with target 3.3.x 2016-03-31 14:26:40 because imap without sieve support is not cool at all 2016-03-31 14:26:57 what is sieve used for? 2016-03-31 14:27:08 server side filtering 2016-03-31 14:27:28 are there any clients that supports it? 2016-03-31 14:27:51 there is some if you allow external connections 2016-03-31 14:27:57 but that's the beauty of it 2016-03-31 14:28:06 you can write your script by hand 2016-03-31 14:28:27 where/how are those scripts stored? 2016-03-31 14:28:30 and have your filtering done all the time without worrying with which client you connect to imap 2016-03-31 14:28:40 in your maildir 2016-03-31 14:28:52 you need ssh access for that? 2016-03-31 14:29:02 or can you upload your script via email client? 2016-03-31 14:29:33 to push a file yeah, if your client support sieve you connect directly to the sieve port and write your rules 2016-03-31 14:30:15 http://sieve.info/clients 2016-03-31 14:30:23 just googled that quickly 2016-03-31 14:31:48 sieve sounds interesting 2016-03-31 14:32:12 a must have :) 2016-03-31 14:32:22 like server-side rules in m$ exchange 2016-03-31 14:33:07 suqdiq: exactly like that 2016-03-31 14:33:33 coredumb: can you please file a request on bugs.a.o? 2016-03-31 14:33:45 its just so i have documentation on who requested it 2016-03-31 14:34:02 so in 1 year ahead, if i wonder, why did i backport that? 2016-03-31 14:34:11 yep 2016-03-31 14:34:14 then i can ref to a specific request 2016-03-31 14:34:17 let me do it 2016-03-31 14:39:07 well this evening... I just migrated my email server and have no client connected nor bugs.a.o account :D 2016-03-31 14:40:20 :) 2016-03-31 15:11:38 ACTION is using dovecot-pigeonhole-plugin 2016-03-31 15:16:30 ncopa: regarding that python 2/3 discussion we had a few weeks ago: is there any plan/ETA on when python3 library apks will appear _at all_ (regardless of a big shift to py3k)? 2016-03-31 15:17:36 it seems to me that since there is none yet, nobody wants to be the first to start adding one. or maybe there are project agreements I don't know about? 2016-03-31 15:22:27 hello 2016-03-31 15:22:32 I think that we need to figure out how to make packages to support both versions. It’d be IMO quite silly to have py2 and py3 variant of each Python package. 2016-03-31 15:22:44 anyone can suggest a client for playing webradio over slow networks? 2016-03-31 15:23:17 mpv goes silent incase the network is out for too long 2016-03-31 15:23:22 same for mpg123 2016-03-31 15:41:33 oneinsect: there's not much you can do about that except increase the buffer size 2016-03-31 15:41:42 don't those tools allow you to choose the buffer size? 2016-03-31 15:41:44 darn 2016-03-31 15:41:50 yes they do 2016-03-31 15:42:04 i specify a buffer but sometimes the buffer runs out 2016-03-31 15:42:14 and then after a minute or two 2016-03-31 15:42:27 when the internet comes back on (3G) they are just silent 2016-03-31 15:42:29 dont play 2016-03-31 15:42:34 the players 2016-03-31 15:42:36 ah 2016-03-31 15:42:52 so they fail permanently on a temporary error 2016-03-31 15:42:54 that is bad 2016-03-31 15:43:11 any ideas? 2016-03-31 15:43:15 on mpv-player 2016-03-31 15:43:18 or mpg123 2016-03-31 15:43:24 or anyother? 2016-03-31 15:43:24 but at the same time, an audio stream with interruptions is generally not worth playing 2016-03-31 15:43:44 i am trying to build a continuous internet radio player 2016-03-31 15:43:46 command line 2016-03-31 15:43:55 i just want it to continue playing 2016-03-31 15:44:23 may be its fine when it has not internet but then when it has it can connect and play 2016-03-31 15:44:27 any ideas? 2016-03-31 15:44:29 but what if there's no data to play ? do you want it to wait and resume when there's data? 2016-03-31 15:44:34 yessss 2016-03-31 15:44:40 can any player do that? 2016-03-31 15:45:00 The Sampled Symphony, by oneinsect 2016-03-31 15:45:09 lol 2016-03-31 15:45:35 that is what will happen 2016-03-31 15:46:49 i understand 2016-03-31 15:46:58 but isnt there any way to deal with such problems 2016-03-31 15:46:59 ? 2016-03-31 15:50:32 if your network bandwidth is smaller than what the webradio server sends, ultimately no 2016-03-31 15:50:48 cant i use something like 2016-03-31 15:50:48 http://pira.cz/show.asp?art=silence 2016-03-31 15:50:51 the only thing you can do is make a giant buffer to delay the problem as much as possible 2016-03-31 15:51:26 you certainly could 2016-03-31 15:51:27 If the stream have a hick-up, I let "Pira CZ Silence Detector" kick the m3u/pls file once again after 4 seconds of silence. I schedule the silence detector only to react when I am running streams... 2016-03-31 15:51:40 sounds good 2016-03-31 15:51:43 something like that for linux? 2016-03-31 15:51:47 thats for windows? 2016-03-31 15:52:02 sorry, I'm not the person to ask for about application software ^^' 2016-03-31 15:52:11 to ask* 2016-03-31 15:52:20 (geez my English gets worse every day) 2016-03-31 15:53:03 lol 2016-03-31 15:53:10 never mind 2016-03-31 15:53:15 i will dig around 2016-03-31 15:53:21 thanks for the help skarnet: 2016-03-31 15:53:31 i just wanted to know what is possible 2016-03-31 15:55:53 jirutka: well, that would work for pure-Python packages only, I guess? 2016-03-31 15:56:11 and only if they support both 2 and 3. 2016-03-31 15:56:49 C-based ones are bound to the header files of a certain interpreter version (if I'm not mistaken) 2016-03-31 15:57:51 so stuff like PIL/Pillow, lxml etc. still is very version-specific. 2016-03-31 16:00:29 from a practical point of view: with some tools I maintain I hesitate to drop Python 2 support altogether because I cannot provide a Alpine-based Dockerfile that builds an image from pre-compiled packages because the slightly nasty (i.e. C-based ones) are not available for Python 3 in Alpine. 2016-03-31 16:01:48 so I basically suggest to add python3-pillow, python3-lxml that explicitly depend on 'python3'. also it's not that those need to be dropped later one, but will become more important in the future. 2016-03-31 16:02:09 Y0Gi: Gentoo supports both versions nicely 2016-03-31 16:02:49 I guess they provide packages with binaries compiled against both versions? 2016-03-31 16:02:53 however, it basically just installs the package twice (if you want both versions) 2016-03-31 16:03:20 well, that probably works, but seems to contradict the idea of an as-small-as-possible (~) distribution 2016-03-31 16:03:27 yeah 2016-03-31 16:03:55 quick question, if a package that is already in main/ suddenly requires 2 new aports to be created, will these aports go into testing first or directly into main ? since package wont compile without new libs 2016-03-31 16:04:14 Alpine solves these issues with subpackages, but I really don’t like idea of having lxml and lxml-py3 :/ 2016-03-31 16:04:37 jirutka: because of the naming? 2016-03-31 16:04:56 yeah… 2016-03-31 16:05:14 maybe lxml-py2, lxml-py3 and both `provides=lxml` 2016-03-31 16:05:26 but still… it looks hackish to me eyes 2016-03-31 16:05:41 * me → my 2016-03-31 16:05:55 since 2/3 are incompatible at times, I think it's the cleanest way. 2016-03-31 16:06:35 on a related note: I have the feeling that using 'py' instead of 'python' in package names might do a disservice. e.g. when searching for packages. 2016-03-31 16:06:50 I don’t consider it a clear solution, but perhaps the less evil that is currently possible 2016-03-31 16:07:15 well, it's consistent with the current state, so yeah. 2016-03-31 16:08:02 ah, that’s another problem, the current convention is to _prefix_ python packages, so python-lxml-py2, python-lxml-py3, python-lxml… uh 2016-03-31 16:09:00 or we can start prefixing python 2 packages with python2, violate convention of suffixes for subpackages, and name them python2-lxml, python3-lxml, and virtual python-lxml 2016-03-31 16:09:23 then again, as long as I can finally install those binaries as APKs instead of collecting (and removing them afterwards) all development tools and headers, and wasting time with compilation, I'd happily turn a blind eye to the package names. 2016-03-31 16:10:46 "python2-lxml, python3-lxml, and virtual python-lxml" <- those closely match my expectations. but with a Debian background, I'm likely biased. 2016-03-31 16:11:28 @ncopa, what do you think about this? 2016-03-31 16:12:11 in this case, at least questioning (and bending or adjusting) the conventions *might* be worthwile. 2016-03-31 16:14:10 btw if it was just up to me, I’d consider python2 as deprecated, make python3 as default and python2 packages name with suffix legacy :P 2016-03-31 16:14:27 jirutka: no py suffixes 2016-03-31 16:14:43 patch abuild to support ^py3- subpackages 2016-03-31 16:15:05 ? 2016-03-31 16:17:00 huge amount of python packages requires to run setup.py with destdir set, so this should be default package function for py2 & py3 (sub)packages 2016-03-31 16:17:15 wrt virtual python-lxml, look at how lua packaging is done now 2016-03-31 16:17:55 ah, I should look how lua packages are handled 2016-03-31 20:25:34 or we could simply not support py2 2016-03-31 20:49:04 hey :) 2016-03-31 20:56:23 python 2 is end of life very soon 2016-03-31 23:26:15 Adran: python 2 reaches its end of life in 2020 (iirc) that's not pretty soonish in my book 2016-03-31 23:26:32 4 years 2016-03-31 23:28:42 fcolista: your package upgrade of perl-package-deprecation manager (a20bf2885f9dd96b686f6be725cdf450387932cc) added a dependency which is not in main/ currently. I will revert that commit for now to unbreak the builders 2016-03-31 23:30:14 I also considered moving that dependencyc (perl-namespace-autoclean) to main/ but it depends on perl-moose which is in testing/ currently and has a shitload of testing/ dependencies and a circular dependency on perl-package-deprecation 2016-03-31 23:36:46 s/a20bf2885f9dd96b686f6be725cdf450387932cc/6562cd444773e9c8d6e75d9bba7fb315d19f6a86/