2015-09-01 06:30:58 hi there, how do pacstrap alpine from Archlinux ? want to run it in a container. :) 2015-09-01 06:34:57 hendry: LXC? there is a template in upstream lxc 2015-09-01 06:36:41 otherwise, apk add --initdb --root $sysroot --repositories-file /dev/null --keys-file /dev/null --allow-untrusted --repository http://nl.alpinelinux.org/alpine/edge/main --update-cache alpine-base 2015-09-01 06:36:44 or similar 2015-09-01 06:36:58 but it would probably be nice with an albootstrap script 2015-09-01 06:40:27 yes, was hoping to find something in Arch's AUR 2015-09-01 06:40:42 I don't apk on my system. Should I build it ? 2015-09-01 06:41:02 ncopa: tbh I'm using system-nspawn since LXC's interface freaks me out 2015-09-01 06:41:31 hendry: building apk is probably simplest 2015-09-01 06:42:02 we should have an albootstrap 2015-09-01 06:42:05 needs to be done 2015-09-01 06:42:10 not too difficult either 2015-09-01 06:43:15 ncopa: what's the git repo addr for apk ? 2015-09-01 06:43:32 http://git.alpinelinux.org/cgit/apk-tools 2015-09-01 06:43:38 git clone git://git.alpinelinux.org/apk-tools 2015-09-01 06:43:50 you can disable lua support 2015-09-01 06:44:06 does alpine use ip to setup an ip ? or what's the default dhcp client ? 2015-09-01 06:44:28 we use busybox 2015-09-01 06:44:43 busybox udhcpc 2015-09-01 06:46:01 git clone git://git.alpinelinux.org/apk-tools ; cd apk-tools ; echo LUAAPK=no > config.mk 2015-09-01 06:48:20 can't you just fetch the static apk and untar it 2015-09-01 06:48:28 you can 2015-09-01 06:49:11 how do i check it's built by yourselves ? ;) 2015-09-01 06:49:27 <[xming]> hey morning guys 2015-09-01 06:49:52 <[xming]> finally bought a 4GB memory ARM board 2015-09-01 06:50:19 well it's a question of how much you want to trust anything 2015-09-01 06:50:27 oh, ubuntu does not have libfetch 2015-09-01 06:50:58 hm 2015-09-01 06:51:03 <[xming]> now I am a bit lost on ARM h/w and distro support 2015-09-01 06:51:05 and deisabling LUA is broken 2015-09-01 06:51:38 <[xming]> uboot/kernel/... are SoC specific? 2015-09-01 06:52:02 <[xming]> if so what does alpine support? 2015-09-01 06:52:09 there is a get_static_apk here which downloads a statically built apk safely: https://github.com/lxc/lxc/blob/master/templates/lxc-alpine.in#L24 2015-09-01 07:00:05 <[xming]> is this only userland part? http://wiki.alpinelinux.org/cgi-bin/dl.cgi/v3.2/releases/armhf/alpine-uboot-3.2.3-armhf.tar.gz 2015-09-01 07:01:58 [xming]: no i think that is the entire boot image 2015-09-01 07:02:04 iirc there is a README in there 2015-09-01 07:02:24 <[xming]> so it probably won't work 100% on my board 2015-09-01 07:03:08 what board do you have? 2015-09-01 07:03:36 <[xming]> firefly with 4GB RAM and 32GB emmc, rk3288 2015-09-01 07:04:21 <[xming]> is that img armv6 or v7? 2015-09-01 07:04:50 iirc its armv6 2015-09-01 07:05:00 i think it should work with rpi 2015-09-01 07:05:14 hum 2015-09-01 07:05:21 it might be armv7 2015-09-01 07:05:23 not sure 2015-09-01 07:05:26 <[xming]> not a rPi here :D 2015-09-01 07:05:39 rpi is armv6 2015-09-01 07:05:52 but i see that we have a specific rpi image 2015-09-01 07:05:56 so it might be armv7 2015-09-01 07:07:44 <[xming]> pfff flashing needs to be connected to a PC 2015-09-01 07:18:23 damn! why lua? why can't it just be C http://s.natalian.org/2015-09-01/1441091892_1912x1036.png 2015-09-01 07:19:18 there is a bug in the makefile 2015-09-01 07:20:49 comment out LUAAPK ?= yes in src/Makefile 2015-09-01 07:20:56 hendry: what distro is it? 2015-09-01 07:21:02 do you have libfetch? 2015-09-01 07:21:06 ncopa: Arch 2015-09-01 07:21:17 sounds like it's missing #include 2015-09-01 07:21:25 that too 2015-09-01 07:21:28 maybe i should use LXC. I like systemd-nspawn, but IIUC it's just a layer on LXC 2015-09-01 07:21:43 hendry: we should fix apk-tools to build with arch/debian 2015-09-01 07:21:54 fabled: i think we might need bundle libfetch with apk-tools sources 2015-09-01 07:22:22 should we bundle then openssl too? 2015-09-01 07:22:26 no 2015-09-01 07:22:33 most distros have openssl 2015-09-01 07:22:37 and zlib 2015-09-01 07:22:38 i have lxc installed. lxc /usr/share/lxc/templates/lxc-alpine ... what lxc command line to bootstrap alp ? 2015-09-01 07:22:43 but not all has libfetch 2015-09-01 07:22:50 i dont tihnk ubuntu has libfetch 2015-09-01 07:23:09 lxc-create -t alpine -n name 2015-09-01 07:23:57 gah, hate lxc... does this mean I am not super user? http://s.natalian.org/2015-09-01/1441092221_1912x1036.png 2015-09-01 07:24:37 probably 2015-09-01 07:25:04 <[xming]> lxc can be run as user IIRC 2015-09-01 07:26:21 <[xming]> why doesn't the lxc template just wget a static apk and be done with it? 2015-09-01 07:26:33 do you have a subuid allocation configured 2015-09-01 07:26:49 ammunta: i have no idea what you just said 2015-09-01 07:27:00 ammunta: i run Archlinux 2015-09-01 07:37:57 hmm, lxc finished and i have no idea where the rootfs resides on my fs 2015-09-01 07:46:03 var/lib/lxc 2015-09-01 08:19:19 <[xming]> depends if started as user or root, to be sure look in the config for lxc.rootfs 2015-09-01 08:39:36 <[xming]> how does one install alpine on ARM? I can't find a rootfs 2015-09-01 10:04:57 interesting: OS X 10.11 switched to LibreSSL 2015-09-01 10:07:11 ncopa: have you heard from spender if he is going to throw a tantrum if we continue to distribute -stable based sources 2015-09-01 10:07:42 did he say so? 2015-09-01 10:08:06 well presumably he would considering he is already throwing a tantrum and making them non-public 2015-09-01 10:08:46 might be he is ok with it if we make sure it is not called "Grsecurity" 2015-09-01 10:08:55 he complained about 2 different issues: 2015-09-01 10:08:59 1) gpl was not respected 2015-09-01 10:09:08 2) trademark was not respected 2015-09-01 10:09:16 and it was commercial vendors 2015-09-01 10:09:25 i dont think he want go after non-commercial 2015-09-01 10:09:30 better send him an email directly ? 2015-09-01 10:09:33 yes 2015-09-01 10:09:35 we should 2015-09-01 10:09:46 i want ask both him and pipacs 2015-09-01 10:09:52 seems there's no official answer to all these questions anywhere 2015-09-01 10:09:57 however, consider that while it is true alpine itself is non-commercial, there *are* people out there using it for commercial purposes 2015-09-01 10:10:11 it is not feasible to do that with the distributions they did mention 2015-09-01 10:10:32 gentoo hradnerned and achr? 2015-09-01 10:10:35 arch* 2015-09-01 10:10:36 yeah 2015-09-01 10:10:39 they are rolling releases 2015-09-01 10:10:43 yeah pipacs hasn't been saying anything on IRC since the announce 2015-09-01 10:10:43 well, arch in particular. i guess you could with gentoo. 2015-09-01 10:10:46 yes, we cannot only do that 2015-09-01 10:11:19 i suppose we simply present the situation for spender, and our current options 2015-09-01 10:11:24 well apparently arch had a -lts grsec kernel release 2015-09-01 10:11:31 guess they'll have to drop it as well 2015-09-01 10:11:32 the way i see it is, 2015-09-01 10:11:36 our options are basically 2015-09-01 10:11:46 (i have to go to lunch. bbl) 2015-09-01 10:11:58 (1) work out some deal with spender, ideally in some sort of written construct 2015-09-01 10:13:13 (2) fork grsecurity -stable at it's current point and watch -test for changes (lots of work) 2015-09-01 10:13:26 (3) try to work with -test sources (probably a bad idea for QA reasons) 2015-09-01 10:14:26 (4) drop grsecurity 2015-09-01 10:14:46 (5) offer grsecurity as an unsupported APKBUILD where you supply the source tree yourself (after paying $200 for access) 2015-09-01 10:22:57 (6) Ask spender directly what can be done in terms of redistribution if alpine pays the sponsorship 2015-09-01 10:35:27 <[xming]> (7) install Windows 10 2015-09-01 10:35:36 <[xming]> ACTION hides 2015-09-01 11:03:46 [xming]: i installed windows 10 on my private laptop. it didnt solve the grsecurity issue in any way... :) 2015-09-01 11:03:56 but just to be sure 2015-09-01 11:04:03 i did a reinstall from scratch 2015-09-01 11:04:09 but grsecurity problem is still there 2015-09-01 11:04:38 kaniini: we can also look if we can use pax patch only 2015-09-01 11:05:43 option 2 and 3 woudl probably be combined if we go there 2015-09-01 11:05:50 but really 2015-09-01 11:06:02 we already do option 2 and 3 2015-09-01 11:06:31 we run our own grsecurity-unsupported kernel patch 2015-09-01 11:06:42 and monitor changes 2015-09-01 11:06:50 both in testing, but mostly in -stable 2015-09-01 11:21:05 [xming], #7 was the best!!! 2015-09-01 11:21:06 :D 2015-09-01 11:30:45 <[xming]> I can understand why they are doing this, but I am afraid this is going to hurt in the long run 2015-09-01 11:47:23 ncopa: I have the strong feeling that PaX patch won't be publicly available 2015-09-01 11:47:42 seems to me that the stable one isn't available already 2015-09-01 11:47:43 ? 2015-09-01 11:53:52 fabled: regarding the musl/gnupg iconv problem, maybe it would be a good idea to patch gnupg until musl upstream decides how the want to handle this issue? 2015-09-01 11:57:38 it seems to affect other packages too 2015-09-01 11:57:59 http://www.openwall.com/lists/musl/2015/09/01/3 2015-09-01 11:59:00 hmm 2015-09-01 11:59:08 perhaps we should patch it then 2015-09-01 12:44:57 hei guys 2015-09-01 12:45:02 LeSpocky: hi 2015-09-01 12:45:47 some questions regarding package build for arm 2015-09-01 12:45:51 may I asked how the raspberry builds work if you don't compile on raspberry itself? is it some generic kernel? or just userland? 2015-09-01 12:46:14 LeSpocky: afaik fabled is the best contact for that 2015-09-01 12:46:33 fabled: am i right? 2015-09-01 12:47:27 I just try to find your requirements for running and/or compiling alpine on arm 2015-09-01 12:52:01 hm, everybody seems to be busy... 2015-09-01 12:55:34 no problem, I still have some work here ;-) 2015-09-01 12:55:40 <[xming]> LeSpocky: by using cross-compiler 2015-09-01 12:56:42 [xming]: you actually do? as far as people told me, cross compiling is not possible or used for alpine at the moment? 2015-09-01 12:56:45 <[xming]> cross-compiler (and binutils) creates binaries for the TARGET system on the HOST 2015-09-01 12:57:56 <[xming]> LeSpocky: cross-complers might have issues when there are bugs in the makefiles, but generally this is how it's done when you don't want to compile on the TARGET 2015-09-01 12:58:22 <[xming]> ACTION is playing with his shinny new ARM board 2015-09-01 12:59:30 [xming]: I know them and use them at work with ptxdist and we also use them for the fli4l project, my questions we specific for alpine linux, so e.g. the precompiled downloads you provide? do you cross-compile them or not? 2015-09-01 12:59:59 <[xming]> nope, those are natively compiled 2015-09-01 13:00:19 <[xming]> on wandboards IIRC 2015-09-01 13:00:31 StarWarsFan, LeSpocky we are native building ARM images on wandboard 2015-09-01 13:01:17 okay and now is the question how building native on wandboard for raspberry pi works? do you use a generic kernel for both boards? 2015-09-01 13:01:35 is the userland the same? 2015-09-01 13:01:51 userland is identical 2015-09-01 13:02:09 wandboard runs linux-grsec; where as for rpi we build dedicated kernels from the linux-rpi aport 2015-09-01 13:02:28 the rpi kernel is built on wandboard too 2015-09-01 13:02:54 so it should be possible to run and compile alpine on other boards, e.g. odroid-xu4? 2015-09-01 13:02:58 yes 2015-09-01 13:03:08 it does not matter which board is used, as long as it's arm 2015-09-01 13:03:13 cross-compile is not supported 2015-09-01 13:03:20 except for core set of packets in limited manner 2015-09-01 13:03:27 <[xming]> fabled: is everything armv6? or v6 for rpi and v7 for the generic arm? 2015-09-01 13:03:37 userland is armv6 2015-09-01 13:03:41 linux-grsec kernel is armv7 2015-09-01 13:03:43 linux-rpi is armv6 2015-09-01 13:03:44 wandboard has only little more performance than a rpi2, so I thought if we buy a new board, we could aim at most power we can get 2015-09-01 13:04:00 <[xming]> LeSpocky: I just bought a firefly here 2015-09-01 13:04:03 yes, rpi2 might do 2015-09-01 13:04:11 though, amount of ram is relevant too 2015-09-01 13:04:16 a generic armv6 userland should run on a cortex-a15 based board 2015-09-01 13:04:35 wandboard also has SATA connector, so some of our builders hava sata disk 2015-09-01 13:04:39 <[xming]> having 4GB is a must :D 2015-09-01 13:05:25 sata is a good point, or at least the possibility to attach mass storage 2015-09-01 13:05:35 usb3 would be ok too 2015-09-01 13:05:37 or nfs 2015-09-01 13:05:41 i suppose 2015-09-01 13:07:18 <[xming]> firefly doesn't have sata, that's the only disadvantage 2015-09-01 13:07:35 <[xming]> but it has quad a17 and 4GB ram 2015-09-01 13:07:56 wandboard is 2gb only iirc 2015-09-01 13:08:23 and swap is needed to build some of the heavy weight packages 2015-09-01 13:08:34 so 4gb might be better deal than sata 2015-09-01 13:08:37 <[xming]> 2GB will be short for some pkgs, better have sata/usb3 2015-09-01 13:08:50 <[xming]> but I think 4GB is comfortable 2015-09-01 13:09:08 <[xming]> ARM h/w is very confusing 2015-09-01 13:09:24 <[xming]> especially with RK kernels 2015-09-01 13:09:43 <[xming]> not even using some conventional ways to boot 2015-09-01 13:10:25 <[xming]> swapping over nfs won't be fun :p 2015-09-01 13:11:33 <[xming]> fabled: do you install the root fs on the internal nand/emmc or on sd card? 2015-09-01 13:12:04 [xming]: I can only find a firefly board with 2GB 2015-09-01 13:13:11 <[xming]> LeSpocky: http://www.aliexpress.com/store/product/Firefly-RK3288-Quad-core-Cortex-A17-processors-Development-board/1710008_32224160718.html 2015-09-01 13:13:35 <[xming]> LeSpocky: click second icon with to the "color" 2015-09-01 13:13:46 <[xming]> s/with/right 2015-09-01 13:14:05 <[xming]> US $199.00 e.a. 2015-09-01 13:15:28 [xming], we use tmpfs run-from-ram on wandbards for the host; and do LXC containers on disk for the builders 2015-09-01 13:16:25 <[xming]> I was afraid of wearing the internal emmc 2015-09-01 13:16:52 an external uSD would also be possible 2015-09-01 13:17:03 depends on how fast this is 2015-09-01 13:17:03 yeah. i run my raspberrypis purely tmpfs. 2015-09-01 13:17:13 <[xming]> SD are not made to last :D 2015-09-01 13:17:14 depends of course on the use case if it's feasible or not 2015-09-01 13:17:35 but SDs are cheap and replacable 2015-09-01 13:17:47 <[xming]> I don't even know if they implement wear leveling on those emmc 2015-09-01 13:18:06 <[xming]> LeSpocky: gone are your data :D 2015-09-01 13:19:39 they do implement it on emmc, it is de facto a SD directly soldered onto the board 2015-09-01 13:20:01 they have to implement it, but I would probably prefare f2fs on it ;-) 2015-09-01 13:22:50 <[xming]> I just read that SD spec doesn't specify wear leveling 2015-09-01 13:22:59 <[xming]> MMC predates SD 2015-09-01 13:23:13 <[xming]> eMMC is just an evolved MMC 2015-09-01 13:23:27 <[xming]> so I am very uncertain 2015-09-01 13:27:45 do you have access to internal spec documents or just the "free" ones? AFAIR only the interface is described, not what the card does internally 2015-09-01 13:28:16 and the cards DO wear leveling, they would die much earlier with all those FAT filesystems if they wouldn't 2015-09-01 13:32:26 <[xming]> I haven't read any specs, just what google told me 2015-09-01 13:40:22 <[xming]> hmm Wear-leveling and ECC error correction and detection is not part of the SD card specification (version 2.0 SDHC) 2015-09-01 13:40:54 <[xming]> but since it's all MLC now, most cards must have that 2015-09-01 13:44:22 <[xming]> I was thinkg btrfs + compression to reduce write amplification 2015-09-01 13:46:01 there are some insights in https://electronics.stackexchange.com/questions/27619/is-it-true-that-a-sd-mmc-card-does-wear-levelling-with-its-own-controller 2015-09-01 13:46:45 so it will probably depend on the vendor, if you avoid no name cheap china stuff, you should be on the safe sid 2015-09-01 13:46:48 side 2015-09-01 13:47:17 I only buy sandisk ultra cards btw. and for rpi those guys have a list of recommended cards 2015-09-01 13:48:13 <[xming]> firefly is from china :D 2015-09-01 13:50:13 <[xming]> does uboot really needs a kernel and a initramfs partitions? Can't it read FS like grub/syslinux on x86? 2015-09-01 13:51:23 I'm not sure about eMMC but I've seen those parts only from big vendors so far (looking at a beagle bone black rigth now with an emmc ic from micron 2015-09-01 15:22:11 now i've got strange issues with go 1.5 on alpine :( 2015-09-01 15:22:34 https://gist.github.com/tboerger/0f265629fb235b468509 the exact the same command works fine on a mac and on opensuse. on alpine it bails out. 2015-09-01 15:22:45 i have even tried to add libc6-compat... nothing changes 2015-09-01 15:23:26 so my guess is that something is not really correct with the alpine go version :( 2015-09-01 15:33:33 the -race option was broken on go 1.4.2 as well (if I really correctly) 2015-09-01 15:35:45 http://git.alpinelinux.org/cgit/aports/tree/main/go/APKBUILD?id=0274802e77637110b8d1de017bd377c23505e335#n58 2015-09-01 15:37:54 I am not even the maintainer of this package but I wanted to fix the go 1.5 i386 build issues first before doing anything else with that package 2015-09-01 15:38:29 sadly I am very busy this week so I don't know if I have time to do so… 2015-09-01 15:57:09 bonsoir 2015-09-01 15:57:36 hi 2015-09-01 16:04:32 is it planned to open the alpine git repositories ? 2015-09-01 16:04:58 it's currently really hard to make a simple git clone ... 2015-09-01 16:06:51 TsT, I've just did a git pull and it is working perfectly 2015-09-01 16:11:36 alacerda I fail to clone the awall repository 2015-09-01 16:11:55 I'm planned to contribute to the Alpine Wall util ;) 2015-09-01 16:12:28 I currently working on the tarball version 2015-09-01 16:13:00 nobody likes github here ? :P 2015-09-01 16:14:46 I make emulation of lpc and stringy (in lua) and make a all-in-one standalone script ( https://github.com/tst2005/awall/tree/all-in-one/bin ) 2015-09-01 16:40:29 alacerda which URL did you use ? 2015-09-01 16:40:37 for the gil pull ... 2015-09-01 16:41:09 i clone aports repo 2015-09-01 16:41:13 let me try awall 2015-09-01 16:42:11 Tst: Fail 2015-09-01 16:42:11 http://hastebin.com/avutonuvoj.scala 2015-09-01 16:42:25 kunkku, ^^ 2015-09-01 16:42:50 ok, exactly like my try :p 2015-09-01 16:43:17 (and the tarball over cgit is empty :p ) 2015-09-01 16:45:42 TsT, look at this: http://git.alpinelinux.org/cgit/awall/tree/ 2015-09-01 16:47:15 and ? 2015-09-01 16:47:29 Bad object: LICENSE d159169d1050894d3ea3b98e1c965c4058208fe1 2015-09-01 16:47:36 Bad object: awall.lua 9a3d9cc95a9106d8a1ad5d0964a57e33f38106dd 2015-09-01 16:47:42 I download each files one by one to build my initial commit :p 2015-09-01 16:47:47 maybe that is what is breaking the repo 2015-09-01 16:47:50 awall.lua is a symlink 2015-09-01 16:48:10 LICENSE should be also one outside the repository 2015-09-01 16:48:26 cgit seems dislike symlink :p 2015-09-01 16:51:09 TsT: git clone git://git.alpinelinux.org/user/kunkku/awall 2015-09-01 16:52:13 too old?! 2015-09-01 17:47:22 strangely apk returns a bus error when /usr is full 2015-09-01 17:47:49 instead of file system full indeed :P 2015-09-01 18:04:12 ncopa: cloning the awall repo is broken 2015-09-01 18:04:41 probably has to do with the recent introduction of gitolite 2015-09-01 18:12:00 coredumb, yes, the 'disk full' error would need fixing. it's probably due to apk mmaping() the target file and read():in to mmapped area 2015-09-01 18:12:22 mmap probably succeeds, but the signal is sent when the space could not be allocated on disk 2015-09-01 18:16:46 yeah 2015-09-01 18:16:56 one cause for SIGBUS is because a just-created memory mapped file cannot be physically allocated, because the disk is full. 2015-09-01 18:24:11 fabled: ok 2015-09-01 18:24:25 i was already testing my memory sticks :P 2015-09-01 18:24:47 SIGBUS is almost never a hardware error 2015-09-01 18:25:04 now i know! 2015-09-01 18:36:42 coredumb: (6) is just the same thing as (1) 2015-09-01 18:37:51 damn what was 1 again ? :D 2015-09-01 18:42:49 coredumb: work out a deal, in writing, with spender securing our rights to use the patches 2015-09-01 18:43:04 coredumb: that includes paying him as an option of course 2015-09-01 18:45:23 alacerda yes git://git.alpinelinux.org/user/kunkku/awall is v0.x.x ... I wanna v1.2.4 :) 2015-09-01 18:52:59 kaniini: hehe ok 2015-09-02 01:20:35 if someone is interested in to create a sndio package (openbsd's audio daemon), the main developer created a mailing list for any question related to linux 2015-09-02 01:21:03 some debian devs are working in a deb package and there are a bunch of player with a sndio backend 2015-09-02 01:21:43 sndio is pretty simple compared with pulseaudio P 2015-09-02 01:22:13 http://www.sndio.org/ 2015-09-02 03:53:58 is there a Web interface where I can see the version of PHP ? 2015-09-02 03:56:53 2015-09-02 03:59:49 of the alpine package 2015-09-02 04:12:08 apk info php 2015-09-02 05:15:03 /msg NickServ VERIFY REGISTER brucey kmqwinrmhaht 2015-09-02 05:15:14 /msg NickServ VERIFY REGISTER brucey yangbo1986 2015-09-02 06:14:50 hey 2015-09-02 06:44:03 morning 2015-09-02 07:03:02 modprobe: can't change directory to '3.18.14-1-grsec': No such file or directory 2015-09-02 07:03:05 reboot? 2015-09-02 07:03:56 i only have /lib/modules/3.18.20-1-grsec 2015-09-02 07:03:58 probably 2015-09-02 07:04:08 presumable from a kernel update 2015-09-02 07:04:08 kernel upgraded 2015-09-02 07:04:53 ScrumpyJack i have had that to i think i did: apk fix 2015-09-02 07:04:53 would it help to create a symlink after a kernel upgrade until the next reboot? 2015-09-02 07:05:38 crow: apk fix didn't help i'm afraid 2015-09-02 07:06:28 dont think modules from another kernel will load 2015-09-02 07:06:50 this one did :) 2015-09-02 07:07:10 ScrumpyJack my problems were http://sprunge.us/IHAa 2015-09-02 07:07:18 seems something else, sorry :) 2015-09-02 07:08:05 crow: no problem. I'll remember that fix if i see your problem here too 2015-09-02 07:13:45 ScrumpyJack: you need have the *running* kernel to be able to modprobe 2015-09-02 07:14:04 if you did apk upgrade, and got new kernel, then you need to reboot to actually run the newer kernel 2015-09-02 07:16:59 yeah, i get that, but perhaps we could get the kernel upgrage apk to create a symlink from /lib/modules/old to /lib/modules/new (and delete at next reboot) so that modprobe works 2015-09-02 07:17:45 ScrumpyJack: if you find a nice way to implement that, then please let me know ;) 2015-09-02 07:19:57 isn't that a bit uggly ? 2015-09-02 07:20:26 modprobe may work but it may also very well not and have unexpected behaviours 2015-09-02 07:20:52 it helps when an apk upgrade has replaced your kernel but you can't schedulde a reboot 2015-09-02 07:28:10 http://linuxdocs.org/HOWTOs/Module-HOWTO/basekerncompat.html 2015-09-02 07:30:43 so the .modinfo section of the object file should prevent modules from loading. if it *does* load, then it should work 2015-09-02 07:43:27 <[xming]> some modules are kernel version dependent, symlinking won't always work 2015-09-02 07:44:26 <[xming]> maybe a better way is not to remove old modules immeadiately 2015-09-02 07:50:14 stateless: i think i might have a nasty issue related fortify 2015-09-02 07:50:32 https://github.com/minetest/minetest/issues/3124 2015-09-02 07:53:25 hm 2015-09-02 08:03:54 <[xming]> ah the new flat google logo looks like a kindergarten drawing 2015-09-02 08:04:31 <[xming]> everything has to be baby colour and childish themed these days 2015-09-02 08:10:56 <[xming]> ARM servers for 3 euro/m, with alpine image 2015-09-02 08:10:59 <[xming]> https://www.scaleway.com/pricing/ 2015-09-02 08:12:27 <[xming]> 200Mbit unmetered, this is so tempting 2015-09-02 08:12:50 <[xming]> meh and I just bought an ARM board :( 2015-09-02 08:13:12 did it come with 200mbit? 2015-09-02 08:13:13 :) 2015-09-02 08:13:39 <[xming]> it has gigabit interface 2015-09-02 08:14:10 i mean the line :) 2015-09-02 08:14:31 <[xming]> I connected a catE 5 cable to it :D 2015-09-02 08:14:34 :) 2015-09-02 08:15:01 i bet you are not allowed to do that at scaleway 2015-09-02 08:15:17 <[xming]> anyone knows them? 2015-09-02 08:17:03 <[xming]> French provider using Tiscali network 2015-09-02 08:27:30 tiscali, they used to have services in nl too 2015-09-02 08:27:38 loooong time ago 2015-09-02 08:35:08 <[xming]> and in .be too, also long time ago 2015-09-02 08:36:16 ncopa, does it build without fortify? 2015-09-02 08:37:12 stateless, GISCAN also gives errors on fortify, i wonder if we should disable fortify somehow for that 2015-09-02 08:37:13 CCLD gtk-launch 2015-09-02 08:37:13 GISCAN Gtk-3.0.gir 2015-09-02 08:37:13 /usr/include/fortify/unistd.h:41: syntax error, unexpected identifier in '# 24 "/usr/include/fortify/unistd.h" 2 3 4' at '__typeof__' 2015-09-02 08:37:13 /usr/include/fortify/unistd.h:51: syntax error, unexpected identifier in '__typeof__(getcwd) __orig_getcwd __asm__("" "getcwd"); extern __inline__ 2015-09-02 08:37:16 __attribute__((__always_inline__,__gnu_inline__,__artificial__)) char *getcwd(char *__s, size_t __l)' at '__typeof__' 2015-09-02 08:37:21 ... 2015-09-02 08:44:22 alpine-mini-3.2.3-x86.iso won't boot in qemu-system-i386 2015-09-02 08:44:41 isn't alpine-mini-3.2.3-x86.iso 32bit only? 2015-09-02 08:45:41 qemu tells me: "This kernel requires an x86_64 CPU, but only detected an i686 CPU" 2015-09-02 08:58:04 stateless: hum.. it seems like rm -r /usr/include/fortify does not make it build either 2015-09-02 08:58:13 so its probably not fortify 2015-09-02 08:59:39 that was using qemu-system-i386 2015-09-02 09:16:29 ncopa, yeah, the errors do not look fortify related 2015-09-02 09:16:40 well I wasn't sure but it kind of confirms it 2015-09-02 09:16:44 sometimes it is tricky to tell 2015-09-02 09:18:00 ncopa, have you found any bugs because of fortify yet? :P 2015-09-02 09:19:35 fabled, how is this built? any special cflags? 2015-09-02 09:19:59 this was the problem: http://tpaste.us/AjXo 2015-09-02 09:20:14 right, no idea if this is valid 2015-09-02 09:20:17 probably not 2015-09-02 09:20:38 http://stackoverflow.com/questions/5849457/using-namespace-in-c-headers 2015-09-02 09:20:48 8 2015-09-02 09:20:49 down vote 2015-09-02 09:20:49 2015-09-02 09:20:49 Item 59 in Sutter and Alexandrescu's "C++ Coding Standards: 101 Rules, Guidelines, and Best Practices": 2015-09-02 09:20:49 59. Don’t write namespace usings in a header file or before an #include. 108 2015-09-02 09:21:04 right :) 2015-09-02 09:48:29 stateless, i think the error is from GISCAN tool; i think they preprocess sources, and feed it to the introspection processor 2015-09-02 10:00:39 found the issue. stupid user problem 2015-09-02 11:48:35 has someone of you a working lua-gtk package? 2015-09-02 11:48:45 if yes, i would need one :) 2015-09-02 12:17:04 leo-unglaub: i've only seen lua-lgi 2015-09-02 12:17:15 but it appears to be broken atm 2015-09-02 12:17:36 hmm, i see 2015-09-02 13:44:16 BitL0G1c: re ovs-vswitchd 2015-09-02 13:45:14 does it help if you add "before networking" to ovs-vswitchd? 2015-09-02 14:35:42 ncopa - what I was trying to prevent in the ifupdown script is an unusable system if you install ovs but forget to configure the run levels (starting ovs-vswitchd only from within a shell script does not start the depends) 2015-09-02 14:36:38 even if you add openvswitch to /etc/modules the module does not get loaded a 2nd time from the ifupdown script 2015-09-02 14:37:25 if you start any service from a shell script it should pull in the deps 2015-09-02 14:38:13 it doesn't do it starting ovs-vswtichd in the ifupdown script 2015-09-02 14:38:19 i think it does not start the depends form the ifupdwon script because the ifupdown script is started from openrc 2015-09-02 14:38:32 how is the ifupdown script started? 2015-09-02 14:38:41 from the "networking" service? 2015-09-02 14:38:53 I tried "/etc/init.d/ovs-vswitchd start" 2015-09-02 14:39:04 "rc-service ovs.... start" 2015-09-02 14:39:05 from command line? 2015-09-02 14:39:31 no from /etc/network/if-pre-up.d/openvswitch 2015-09-02 14:39:53 how was /etc/network/if-pre-up.d/openvswitch called? 2015-09-02 14:40:06 did you call /etc/network/if-pre-up.d/openvswitch from command line? 2015-09-02 14:40:31 no - from a rebooted system 2015-09-02 14:40:46 i forgot to configure the runlevels & rebooted 2015-09-02 14:41:03 & was unable to login because ovsdb-server did not get started 2015-09-02 14:41:18 so when "networking" service started it called ifup 2015-09-02 14:41:24 yes 2015-09-02 14:41:46 & /etc/init.d/ovs-vswitchd start - did not start the depends 2015-09-02 14:41:58 and you want make it work even if you forgot to add ovs-vswitchd to runlevels? 2015-09-02 14:42:17 yes - the whole point of the patch is to prevent an unusable system 2015-09-02 14:42:32 ok 2015-09-02 14:42:38 sorry for being slow 2015-09-02 14:42:42 it now makes sense 2015-09-02 14:43:15 hmmm 2015-09-02 14:43:27 maybe i am to stupid or building lua gtk just sucks *g* 2015-09-02 14:43:46 I checked the openvswitch module did not get loaded twice (it didn't) 2015-09-02 14:43:55 leo-unglaub: i'm not surprised 2015-09-02 14:44:20 I tried adding rc_depends into the conf.d but it did not start the depends 2015-09-02 14:44:25 BitL0G1c: yes it should not make any harm to try load it twice 2015-09-02 14:45:10 BitL0G1c: the ovs-vswitchd service already seem to have the ovsdb-server dependency 2015-09-02 14:45:30 when i was looking for a solution it seems openrc will not load the depends if scripts are started from within a shell script 2015-09-02 14:45:51 no, the problem is nto that it from shells cript 2015-09-02 14:46:01 yes 2015-09-02 14:46:14 i think problem is that it tries to start a service already 2015-09-02 14:46:29 openrc is in the process of starting "networking" 2015-09-02 14:47:08 and while starting this service, it needs to call ifup, which calls the ifupdown script, which again calls openrc to start another service 2015-09-02 14:47:34 i think openrc does not try reoslve deps etc due to that 2015-09-02 14:48:48 I thought after seeing /lib/rc/sh/openrc-run.sh that depends just doen't get run when init scripts are started from within a shell script 2015-09-02 14:49:24 i think we start services from setup-alpine 2015-09-02 14:49:28 and i think it does work 2015-09-02 14:49:38 but idunno 2015-09-02 14:49:44 ok 2015-09-02 14:50:29 ncopa: hehe, do other lua packages suck to build as well? 2015-09-02 14:50:42 leo-unglaub: yeah... 2015-09-02 14:52:04 hehe, damn 2015-09-02 14:52:18 thats a real downside to lua 2015-09-02 14:52:24 the extensions are horribly to build 2015-09-02 15:01:22 BitL0G1c: i think somethign like this might work? http://tpaste.us/A1nZ 2015-09-02 15:01:59 i'm going home will hopefully have time too look at it tomorrow 2015-09-02 15:02:21 might be i apply your first patch and add the second checks and comment afterwards 2015-09-02 15:04:42 I was previously starting the 3 ovs scripts manually - but you didn't like possible future changes to the depends 2015-09-02 17:46:21 good evening 2015-09-03 06:57:29 ncopa, care to fix abuild manpage handling ? 2015-09-03 06:57:36 several packages that used to build, no longer do 2015-09-03 06:57:43 since uncompressed manpages in non -doc pkg are not auto-compressed 2015-09-03 06:57:50 latest found is v4l-utils 2015-09-03 06:58:01 and looking at git log the issue has been with many others too 2015-09-03 06:58:09 and just last night someone complained about same issue 2015-09-03 06:58:30 i'm ok to do -doc packages, but at least the error message should be updated, if adding $pkgname-doc is the intended fix 2015-09-03 06:59:01 nmeum: do you mind have a look at that ^^^ 2015-09-03 06:59:12 i wonder if we maybe should turn it into a warning 2015-09-03 06:59:15 also 2015-09-03 06:59:39 might be an idea to compress man pages before the package split 2015-09-03 06:59:55 for reference, it was Orochimarufan complaining about ntp pkg yesterday on #alpine-linux 2015-09-03 07:01:18 logic is wrong 2015-09-03 07:01:30 it shoudl first compress man pages 2015-09-03 07:01:33 then split 2015-09-03 07:02:07 then complain if man pages are found without -doc 2015-09-03 07:15:12 ncopa: today is one year since my last aports commit. thought I'd change that today :) 2015-09-03 07:15:26 thankfully people have kept packages where I'm listed as maintainer up-to-date 2015-09-03 07:23:58 uggedal: wow 2015-09-03 07:24:02 nice to have you back! 2015-09-03 07:24:29 we mived to gitolite since then 2015-09-03 07:24:44 ncopa: noticed :) 2015-09-03 07:24:54 you need to git remote set-url origin git@git.alpinelinux.org:aports 2015-09-03 07:24:55 I do have ssh to uggedal@git.a.o 2015-09-03 07:25:17 i think i imported your key to gitolite too 2015-09-03 07:26:11 ncopa: don't think its imported 2015-09-03 07:26:24 Next authentication method: password 2015-09-03 07:26:29 hm 2015-09-03 07:26:30 ok 2015-09-03 07:26:34 i'll fix that then 2015-09-03 07:26:39 np :) 2015-09-03 07:27:00 it was likely not imported intentionally, due to inactivity 2015-09-03 07:27:05 ;) 2015-09-03 07:27:06 yeah, sorry about that 2015-09-03 07:27:14 had to use a glibc distro for a while 2015-09-03 07:27:38 you dont need apoligize for not showing up at your unpaid job :) 2015-09-03 07:27:57 but then that distro got musl support and suddenly I'm all musl based now 2015-09-03 07:28:08 void i suppose 2015-09-03 07:28:11 yep 2015-09-03 07:28:25 but I need some stability :) 2015-09-03 07:28:31 ha! 2015-09-03 07:30:35 uggedal: try now 2015-09-03 07:32:14 ncopa: works, thanks 2015-09-03 07:33:48 so what's new in alpine world? 2015-09-03 07:35:19 we have growns 2015-09-03 07:35:23 grown* 2015-09-03 07:35:29 #alpine-linux is no 150 2015-09-03 07:35:32 now* 2015-09-03 07:35:51 we have become popular in docker world 2015-09-03 07:36:12 uggedal: where you pioneered 2015-09-03 07:36:40 no runit support unfortunally 2015-09-03 07:36:42 yet 2015-09-03 07:36:44 heh, yeah that 2015-09-03 07:36:58 i woudl like to support both openrc and runit 2015-09-03 07:36:59 I was that openrc got some kind of s6 support 2015-09-03 07:37:07 that would be great 2015-09-03 07:37:08 i saw that too 2015-09-03 07:37:25 i have not had the resources to actually add runit support 2015-09-03 07:37:31 but i think it is doable 2015-09-03 07:37:46 I actually have it in my TODO already 2015-09-03 07:37:48 to look at it 2015-09-03 07:38:10 that and something like http://void.uggedal.com/void-updates.txt 2015-09-03 07:38:12 what I really would have liked was some #!/bin/sh wrapper, similar openrc-run 2015-09-03 07:38:31 maybe #!/sbin/runit-run or similar 2015-09-03 07:39:00 and the only thing it would do would be to check if /etc/conf.d/ exists 2015-09-03 07:39:02 and source it 2015-09-03 07:39:29 that way we could have shared service "configs" 2015-09-03 07:39:43 ah, so each service in /etc/sv would have that shebang? 2015-09-03 07:39:50 that was the idea yes 2015-09-03 07:40:00 sounds like a good idea 2015-09-03 07:40:13 the hard part would be hw bringup I think 2015-09-03 07:40:21 since openrc does most of that now 2015-09-03 07:40:33 hm 2015-09-03 07:40:41 well, technically runit pid 1 could just call into openrc for that 2015-09-03 07:40:52 before it sets up the supervision tree 2015-09-03 07:41:11 woudl be kinda nice be able not needing install openrc at all 2015-09-03 07:41:28 yeah 2015-09-03 07:42:01 but i'd like avoid maintain 2 set of initscripts 2015-09-03 07:42:10 but i suppose thats not possible 2015-09-03 07:42:13 hm 2015-09-03 07:42:27 i wonder if we should have subpackages for -openrc 2015-09-03 07:42:34 where the init.d script is shipped 2015-09-03 07:42:48 i am also thinking of look into s6 2015-09-03 07:42:56 i kinda like the author of s6 :) 2015-09-03 07:43:06 and i like the s6 thinking 2015-09-03 07:43:11 much of it atleast 2015-09-03 07:43:31 re hw loading with runit 2015-09-03 07:43:48 i have an idea of netlink listener + coldplut trigger 2015-09-03 07:44:00 which i intend to implement in initramfs initially 2015-09-03 07:44:15 i need to do it for v3.3 2015-09-03 07:44:39 so i must get that done soonish 2015-09-03 07:52:18 yeah, s6 is nice. don't really like a new language (execline) though 2015-09-03 07:54:38 oh it was s6 that introduced the execline... 2015-09-03 07:54:50 yeah i didnt like that very much either 2015-09-03 07:54:58 even if the idea was kind of interesting 2015-09-03 07:55:11 is there a main/community/staging split coming? 2015-09-03 07:55:25 yes 2015-09-03 07:55:30 good 2015-09-03 07:55:32 thats true 2015-09-03 07:55:42 we need do that kinda sooinsh 2015-09-03 07:55:48 some of my packages needs to move to community :) 2015-09-03 07:56:05 ok i 'll try look at that next week 2015-09-03 07:56:24 no hurry, just saw the ml thread 2015-09-03 07:56:56 i want it before v3.3 release in nov 2015-09-03 07:57:12 in the beginning of oct we need start up the v3.3 builders 2015-09-03 07:57:18 and build world from scratch 2015-09-03 07:57:26 and focus on fixing bugs 2015-09-03 07:57:54 so all new features needs to go in during september 2015-09-03 07:58:57 what i had hoped for was: initramfs to find all block devices with known file systems (so no alpine_dev= boot option) 2015-09-03 07:59:03 I also see that the toolchain can be bootstrapped automaically now, nice 2015-09-03 07:59:05 support for isohybird 2015-09-03 07:59:41 support for efi boot 2015-09-03 08:00:08 its gonna be a busy sept :) 2015-09-03 08:00:17 heh 2015-09-03 08:00:28 i would have hoped for full python3 support 2015-09-03 08:00:45 but i cannot see how to have time for that 2015-09-03 08:01:04 as subpackages of the py-* packages? 2015-09-03 08:01:12 dunno 2015-09-03 08:01:25 i dont know how to solve it 2015-09-03 08:01:30 like the way void does it, builds both from the same source 2015-09-03 08:01:37 we could have py-* + py3-* 2015-09-03 08:01:43 yes 2015-09-03 08:01:45 that woudl be nice 2015-09-03 08:01:50 we do something similar with lua 2015-09-03 08:01:52 so 2015-09-03 08:02:13 i have been thinking of a buildflavors variable in apkbuild 2015-09-03 08:02:36 buildflavors="python2 python3" 2015-09-03 08:02:56 then build() is called for each buildflavor with the buildflavor as arg 2015-09-03 08:02:58 or similar 2015-09-03 08:03:06 same with package() 2015-09-03 08:03:33 with lua packages we have lua5.1-* lua5.2-* lua5.3-* subpackages 2015-09-03 08:04:02 the lua-* packages are empty but the proper version will be pulled in with install_ifw 2015-09-03 08:04:02 void has python-* and python3.4-* 2015-09-03 08:04:07 install_if* 2015-09-03 08:04:21 nice 2015-09-03 08:04:25 so if you apk add lua-module 2015-09-03 08:04:32 you dont get anything 2015-09-03 08:04:34 is install_if new? 2015-09-03 08:04:41 no, its old feature 2015-09-03 08:04:57 then if you apk add lua5.3 it will also pull in lua5.3-module 2015-09-03 08:05:09 if you later add lua5.2 it will also pull in lua5.2-module 2015-09-03 08:05:15 would be nice with install_if for *-doc if you have mdocml installed 2015-09-03 08:05:28 we kind of have that i think 2015-09-03 08:05:30 apk add docs 2015-09-03 08:05:34 ah nice 2015-09-03 08:05:37 and -doc gets autopulled in 2015-09-03 08:05:40 (i think) 2015-09-03 08:05:49 yup 2015-09-03 08:05:50 we do 2015-09-03 08:06:50 would be even nicer if usr/share/{doc,examples,info} went to *-doc and /usr/share/man went to *-man 2015-09-03 08:07:24 i'm afraid that would give us too many packages 2015-09-03 08:07:42 cannot really have one package per file :) 2015-09-03 08:07:50 heh 2015-09-03 08:08:50 if you have create ideas for python3, please let me know 2015-09-03 08:09:25 cool, the apk add docs actually works :) 2015-09-03 08:09:49 yeah, will first have to bring up an alpine system here and get reacquainted with it 2015-09-03 08:09:57 oh, one more new thing with abuild 2015-09-03 08:10:22 the -dev packages no longer pull in all subpackages as deps 2015-09-03 08:10:29 only what abulid think is strictly needed 2015-09-03 08:12:39 that install_if sounds kinda interesting 2015-09-03 08:12:58 pluesch: needs be used with care 2015-09-03 08:13:17 one of the deps there should always be versioned 2015-09-03 08:13:46 ncopa: so you are planning to combine buildflavors (not yet implemented) and install_if for python2/python3 packages? 2015-09-03 08:13:56 possibly 2015-09-03 08:13:59 thats what we do with lua 2015-09-03 08:14:55 that would mean that py-* packages has subpackages py2-* and py3-* with install_if 2015-09-03 08:15:02 so if you apk add py-something 2015-09-03 08:15:16 and apk add python3, it woudl also pull in py3-something 2015-09-03 08:15:30 when you apk del python3 it would remove py3-* too 2015-09-03 08:18:26 where does the init script come from in the initramfs? 2015-09-03 08:18:49 mkinitfs 2015-09-03 08:19:07 http://git.alpinelinux.org/cgit/mkinitfs/tree/initramfs-init.in 2015-09-03 08:21:06 when building an iso, the Makefile pull the init script in from there? 2015-09-03 08:22:17 from /usr/share/mkinitfs/initramfs-init 2015-09-03 08:42:03 thanks 2015-09-03 08:42:11 running qualification tests for pkgconf-1 with --version that does not pretend to be pkg-config (because 1 > 0.28) on both freebsd and alpine 2015-09-03 08:46:28 so initramfs-init is part of mkinitfs-2.8.0-r3 which gets downloaded/installed each build, so if i want my own init script, i need to repackage mkinitfs, and point apk in the iso Makefile to it? Is there a simpler way? 2015-09-03 09:23:10 ScrumpyJack: use mkinitfs -i /path/to/your-init 2015-09-03 09:24:06 i suppose we could add an MKINITFS_OPTS option so you coud set: MKINITFS_OPTS= -i /path/to/custom-init 2015-09-03 09:37:19 ncopa: I think turning it into a warning might be the best idea to fix that 2015-09-03 10:08:00 s/idea/way/ 2015-09-03 10:20:55 so, i have a custom initramfs-init, and it gets picked up at make time. where should i put files i want added to the initramfs? 2015-09-03 10:24:18 do i have to create the file in the init script with a cat "stuff" > $sysroot/file or something? 2015-09-03 11:01:34 hmm, doing that didn't work 2015-09-03 11:09:56 so, my question is, how do I add a boot script when making an alpine-iso? 2015-09-03 11:11:17 I have added a symlink in boot runlevel for local (added rc_add local boot in init script) 2015-09-03 11:12:05 now i want to add a script in /etc/local.d/script.start in the custom iso that I make 2015-09-03 11:13:44 how would i go about doing that? 2015-09-03 12:13:59 should I add a file to openrc and build the apk? 2015-09-03 20:11:51 Hello, is this a right channel to ask about questions abuild and packaging? 2015-09-03 20:21:34 squeaky_pl: yeah 2015-09-03 20:22:51 I am trying to do my first package, how can i get abuild to tell me what command exactly failed, i got that crate_apk failed 2015-09-03 20:27:01 is that the full output you get? 2015-09-03 20:29:40 afaik. abuild is generally very talkative 2015-09-03 20:31:26 pluesch, well I already lost the whole output, I deleted everything and now rebuilding from scratch also removed subpackages for now because I was suspecting it provoked the error 2015-09-03 20:31:48 i will post gist when it fails now 2015-09-03 20:32:15 ok it just happened 2015-09-03 20:33:11 https://gist.github.com/squeaky-pl/85593bfb104ce67b31f2 2015-09-03 20:33:44 so it compiled nicely and run install and then failed 2015-09-03 20:34:17 (i run abuild -r) 2015-09-03 20:34:40 I'm following http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package 2015-09-03 20:38:42 oh i found out by scanelf it links to some other packages that i didnt put in depends, could it be it? 2015-09-03 20:40:40 try that... if it doesn't solve the issue I'd suggest to run strace on it 2015-09-03 20:41:48 if that doesn't help either, you would have to wait for someone with more exp. to help you 2015-09-03 20:42:59 most activity here is during the day (CEST) 2015-09-03 20:48:11 ok, will keep that in mind, thank you 2015-09-03 20:48:57 yw 2015-09-03 20:56:06 squeaky_pl: add a set -x to your abuild 2015-09-03 20:56:28 squeaky_pl: it fails somewhere in trace_apk_deps() 2015-09-03 21:16:59 bind has two CVE's fixed in latest release btw 2015-09-03 21:17:34 my build env is not ready yet and I don't dare to push it untested 2015-09-03 21:21:20 uggedal: what do you mean with testing? just see how it builds or more? 2015-09-03 21:24:48 pluesch: that should do for a small patch like this, yes 2015-09-03 21:25:44 I can try to build it, if you send me the diff of the APKBUILD 2015-09-03 21:30:03 we'll, it was announced on oss-security 2 days ago. we could probably wait until the morning 2015-09-03 21:30:15 s/'// 2015-09-03 21:34:12 build works fine for me if I just set the version to p4 at the end 2015-09-04 04:58:04 morning 2015-09-04 06:23:52 morning 2015-09-04 06:27:34 hi _o/ 2015-09-04 06:47:10 morning 2015-09-04 07:16:38 i'm having device manager issues 2015-09-04 07:17:50 what does edge alpine come with at the moment? looks like busybox mdev is used as default. is that right? 2015-09-04 07:19:19 what is the recommended replacement? 2015-09-04 07:19:58 udev fails to start now, i get the error (udev requires /dev to be mounted devtmpfs) 2015-09-04 07:34:11 ScrumpyJack: we use mdev as default 2015-09-04 07:34:26 with v3.2 we only support udev (not eudev from edge) 2015-09-04 07:34:37 due to the devtmpfs kernel change 2015-09-04 07:34:53 you can use eudev from edge, but then you need kernel from edge too 2015-09-04 07:35:10 eudev has some weird issues still that i dont have sorted out 2015-09-04 07:52:08 morning everyone 2015-09-04 08:03:40 morning 2015-09-04 08:10:59 hm, is there a way to use specific lxc backing store by default? I have a feeling I'll forgot about adding '-B overlayfs' one day 2015-09-04 08:17:11 yo all, I'm building libcec-2.2.0 with rpi support on alpine 2.3.2 and the --as-needed abuild flag causes some libbcm_host configure tests to fail 2015-09-04 08:17:13 maybe in /etc/lxc/default.conf? 2015-09-04 08:17:26 is this expected? any ideas on how to fix it properly? 2015-09-04 08:18:20 lostd: i have no idea what it gets wrong, but it sounds like the proper fix is to fix the linker flags 2015-09-04 08:18:23 LIBS=... 2015-09-04 08:18:46 maybe just rearrange the order 2015-09-04 08:20:07 ncopa: building manually without abuild (I think that is just without --as-needed) builds fine, I'll check libs and order 2015-09-04 08:32:12 the problem seems to be related with libs being in /opt/vc/lib 2015-09-04 10:53:56 ncopa: thanks for your response. i'm in edge and using udev again. works fine 2015-09-04 11:55:56 is there any way to use flash player on chromium ? 2015-09-04 11:56:13 is the alpine build supposed to support external plugins ? 2015-09-04 13:35:11 coredumb: depends on the plugins 2015-09-04 13:36:04 seems that it requires the pepperflash stuff from the google chrome release 2015-09-04 13:36:13 let's say I won't use flash in chromium -_- 2015-09-04 13:36:57 i didnt try flash, but i did try widevine for netflix 2015-09-04 13:36:59 but didnt work 2015-09-04 14:02:34 nice to read https://clearlinux.org/features/autofdo 2015-09-04 15:57:51 coredumb: you can try flash from aports non-free and enable ppapi in chromium 2015-09-04 15:58:22 coredumb: but it was not tested 2015-09-04 15:58:38 coredumb: also they turned off ppapi in 45 chromium :( 2015-09-04 17:37:28 k0r10n: ok good to know 2015-09-04 20:11:41 hi 2015-09-04 20:12:00 can abuild also unpack src.rpms? 2015-09-04 20:13:28 and then cpio 2015-09-05 10:18:29 hi, what's the policy regarding package upgrade issues in redmine. can they be reused for new versions or not? in the past ncopa was marking them as resolved, but for some time they're getting closed, and it needs special privilege to be able to reopen already closed issue. I believe it would be less mess to have one package update issue for particular package, but it requires kind of managemental 2015-09-05 10:18:35 decision. 2015-09-05 10:51:24 The GnuPG patch (1997ebdedb85701fd11105a98de22bedf735545d) can be removed again, 3e4dc327b193eb85e43c466be9f13d3a82101f1a resolved that issue 2015-09-05 14:15:11 ncopa: could you take a look at %533? 2015-09-05 17:30:17 waaa, i am getting mad here 2015-09-05 17:30:28 building lua gtk is impossible 2015-09-05 17:30:36 the entire thing seams to be broken 2015-09-05 17:30:42 maybe is muscle or soemthing else 2015-09-06 01:50:53 huh, aports are built with -Os by default? 2015-09-06 01:53:29 ... well, it makes sense 2015-09-06 08:06:01 ammunta: yes 2015-09-06 13:52:38 will it create a circular dependency if a package has a -dev subpackage and has a runtime dependency on the -dev package of itself? 2015-09-06 13:52:53 or should I not create a -dev subpackage in that case? 2015-09-06 13:54:56 no point to split, if both need to be installed 2015-09-06 15:23:14 fabled: ok, thanks 2015-09-06 15:23:27 btw could you can revert the gnupg patch again 2015-09-06 15:23:57 yes, it can be reverted; but it does not do any harm either. 2015-09-06 15:23:59 The commit was 1997ebdedb85701fd11105a98de22bedf735545d and it was made obsolet by 3e4dc327b193eb85e43c466be9f13d3a82101f1a 2015-09-06 15:24:28 well…it doesn't harm but why would you want to keep an obsolet patch? ;) 2015-09-06 17:55:18 a took another look at the go 1.4 i386 build issue and aquired a gdb backtrace http://paste42.de/8739/ any ideas what might be the issue here? 2015-09-06 17:55:32 the grsecurity warning should be gdb and not go related, right? 2015-09-06 17:59:02 I also don't know what the problem with the MOVL instruction is. The source is adressed relative to the framepointer…so the source should be accessable and moving that to the AX register should actually work… 2015-09-06 18:41:12 ah, I can't reproduce that error under linux-vanilla so the issue is related to some grsecurity protection. Only x86 is effect here because the go setldt function isn't used on x86_64 2015-09-07 07:51:51 fcolista: f8da2e83aff9a237153754e589520dc4b58312a4 has a "non-cosistent" commit message 2015-09-07 07:56:23 i know nmeum 2015-09-07 07:56:28 figured later 2015-09-07 07:57:12 hey :) 2015-09-07 09:39:15 Hello, what is the minimum structure an APKBUILD repository needs, is just a folder like `testing` enough? 2015-09-07 09:59:19 yes 2015-09-07 12:00:28 hey ncopa, we've few patches i dunno if we can accept: %497, %533, %539, %540 2015-09-07 12:02:59 %497 no longer applies 2015-09-07 12:03:02 dunnno why 2015-09-07 12:03:30 i think %533 is ok 2015-09-07 12:03:34 does not break anything 2015-09-07 12:03:51 there's the openvswitch saga which is okish imho 2015-09-07 12:03:58 yeah 2015-09-07 12:04:10 but i think i'd like applie the firs variant 2015-09-07 12:04:20 i talked with BitL0G1c about it the other week 2015-09-07 12:04:27 so %479 only 2015-09-07 12:05:50 yes 2015-09-07 12:05:50 but 2015-09-07 12:05:56 something is duplicated 2015-09-07 12:06:03 so it needs a bit clean up 2015-09-07 12:06:09 but i'm ok to apply it 2015-09-07 12:06:17 together with the 2.4.0 update 2015-09-07 12:06:27 %512. 2015-09-07 12:06:32 ok 2015-09-07 12:06:34 btw, %497 applies here (even if there are whitespace errors) 2015-09-07 12:06:43 ok 2015-09-07 12:06:48 then maybe just push it 2015-09-07 12:07:09 thx ncopa 2015-09-07 12:17:48 ok, so just 2 left out: %539 and %540. Which are ok for me. 2015-09-07 12:37:53 ncopa: I get "FATAL: W any aports uggedal DENIED by fallthru" when pushing to aports 2015-09-07 12:40:03 what is your url set to? 2015-09-07 12:40:20 hum 2015-09-07 12:40:24 might be you are not in the group 2015-09-07 12:40:27 I'll fix that 2015-09-07 12:40:39 thanks :) 2015-09-07 12:40:53 yeah, he's not in the group 2015-09-07 12:41:41 pushed 2015-09-07 12:41:53 there, thanks 2015-09-07 15:03:25 So I wanted to have postgis in docker based on alpine linux image, thanks to wiki article it was really easy to understand basics of the packaging system. I came up with three APKBUILDs https://github.com/Rushmorefm/alpine-packages. It seems to work fine, any advices on improving my APKBUILDs? 2015-09-07 15:05:49 Also what is the friction-less distribution method. I just tar xzed resulting binary package repostiory and I am unpacking them on resulting machine passing the repo path directly to apk 2015-09-07 20:41:25 is there an easier way of keeping srcdir/pkgdir after a successfull build other than editing abuild.conf? 2015-09-07 20:43:05 maybe a flag in abuild for it would be useful. -k/keep is already taken... 2015-09-08 06:29:17 uggedal: you mean of a single build? 2015-09-08 06:33:39 maybe we can use: abuild -t ? 2015-09-08 06:33:45 keep temp files 2015-09-08 06:33:51 or -K 2015-09-08 06:38:22 hey :) 2015-09-08 06:44:42 hi 2015-09-08 06:56:13 ncopa: yeah, for a single build where I want to analyze the result. -K sounds ok 2015-09-08 07:01:16 should it also keep the dependencies installed? 2015-09-08 07:06:08 ncopa: no, I think -K should act like ERROR_CLEANUP="deps" 2015-09-08 07:07:15 you have 2 different setup of cleanup 2015-09-08 07:07:23 CLEANUP= and ERROR_CLEANUP 2015-09-08 07:07:32 one is for successful build 2015-09-08 07:07:36 yeah 2015-09-08 07:07:37 the other is for error 2015-09-08 07:08:03 so you might have configured your CLEANUP="srcdir pkgdir" 2015-09-08 07:08:12 which means that you always keep your deps 2015-09-08 07:08:21 (which makes no sense...) 2015-09-08 07:08:35 but -K, should we just always remove deps? 2015-09-08 07:08:36 saw your change, looks good 2015-09-08 07:08:47 or respect that user always want to keep them? 2015-09-08 07:08:49 no, they might be handy now that I think about it 2015-09-08 07:08:55 i think just keeping it all is best 2015-09-08 07:09:00 and makes most sense 2015-09-08 07:09:01 yep :) 2015-09-08 07:09:23 theis -K is actually a good idea 2015-09-08 07:09:32 i have wanted it myself witout realizing it :) 2015-09-08 07:10:08 I used to add a return 1 at the end of install, but this is much nicer 2015-09-08 07:10:20 yup 2015-09-08 07:19:42 morning 2015-09-08 08:14:44 i3wm 4.10.4 with two little bugfixes is here :) 2015-09-08 08:14:50 http://i3wm.org/downloads/RELEASE-NOTES-4.10.4.txt 2015-09-08 08:18:44 leo-unglaub: done 2015-09-08 08:20:11 uggedal: :) awesome! 2015-09-08 08:20:55 are there any news on the grsecurity front? 2015-09-08 08:29:01 not yet 2015-09-08 08:29:04 but i have a plan 2015-09-08 08:31:35 is there any news on the next GKH longterm kernel seeing as 3.10 will be EOL this month? 2015-09-08 08:34:33 it was supposed to be 4.1 or 4.2, I don't remember now 2015-09-08 08:34:54 4.1, I guess 2015-09-08 09:24:51 ncopa: ah, nice 2015-09-08 09:25:05 i am interrested to read about it when the plan is ready :) 2015-09-08 09:25:29 1) kidnap spender 2015-09-08 09:25:36 2) make him give us the patches 2015-09-08 09:25:45 Copyright 2015 by me 2015-09-08 09:26:00 no no no 2015-09-08 09:26:15 I'm interested as well :) 2015-09-08 09:26:16 1) kidnap his girl (or boy) friend 2015-09-08 09:26:27 2) make him give up the patches 2015-09-08 09:26:28 aw man, I hope I'll get some royalties 2015-09-08 09:26:39 it's derivative work after all 2015-09-08 09:26:57 3) Copyright barhalion (after all, he does the hard kidnampping work) 2015-09-08 09:27:01 :D 2015-09-08 09:27:01 4) ... 2015-09-08 09:27:06 5) profit!!! 2015-09-08 09:27:38 no, wait 2015-09-08 09:27:43 i have a better idea 2015-09-08 09:27:56 1) kidnap spenders kitten/pony or dog 2015-09-08 09:27:57 ... 2015-09-08 09:27:58 :) 2015-09-08 09:28:51 gosh that's mean :( 2015-09-08 09:28:57 hm 2015-09-08 09:28:58 yea 2015-09-08 09:29:02 you're right 2015-09-08 09:29:17 better to take his gf(or bf) 2015-09-08 09:39:53 so what's the actual plan ^^ ? 2015-09-08 09:43:14 lol. coredumb what makes you think that isn't the *actual* plan? :) 2015-09-08 09:43:40 i hear he's very fond on his cat 2015-09-08 09:52:36 ScrumpyJack: :O 2015-09-08 10:08:38 ncopa: haha, as a member of the red cross i cannot comment on kidnapping ... but the engineer in me wants to write +1 ;) 2015-09-08 10:09:49 but for real, if you write spender he would propobly give alpine the patches as well 2015-09-08 10:10:08 i mean, mybe alpine becomes a "sponsor" for 1 € a year 2015-09-08 10:12:30 well, it will mean we should either keep them private or put publicly in git repositories 2015-09-08 10:12:51 the latter kinda negates entire "no -stable patches" plan of spender 2015-09-08 10:23:28 Any comments on %539 and %540? 2015-09-08 10:37:27 leo, from the grsec site, "Sponsorship begins at 200 USD/mo." 2015-09-08 10:38:02 which patches are you talikng about? 2015-09-08 10:38:19 grsec ones? 2015-09-08 10:38:42 so sftp don't work with dropbear since it references /usr/libexec/sftp-server 2015-09-08 10:38:54 crow: yes 2015-09-08 10:38:59 fix dropbear to use /usr/lib/ssh/sftp-server or fix openssh to use /usr/libexec? 2015-09-08 10:39:33 I'd prefer /usr/lib, no idea what ncopa thinks about it 2015-09-08 10:42:41 ncopa: ^ 2015-09-08 10:47:37 seems most use /usr/lib, but some defaults to /usr/libexec like git, gcc, gnupg, lxc 2015-09-08 10:47:58 maybe those should be fixed for consistency? maybe a test for usr/libexec in abuild? 2015-09-08 11:00:31 i think i prefer /usr/lib/ 2015-09-08 11:00:48 yes we can fix it for consistency 2015-09-08 11:01:05 but i dont have strong feelings for it 2015-09-08 11:05:22 re lxc 2015-09-08 11:05:34 i think we maybe should ship the alpine template with main lxc package 2015-09-08 11:17:13 anyone using zsh as user shell while building aports? 2015-09-08 11:17:51 zsh segfaults here in some configure step 2015-09-08 11:18:14 but I don't understand why its used as all since abuild has #!/bin/sh shebang 2015-09-08 11:18:20 hmm, maybe SHELL carries over 2015-09-08 11:18:39 currently using env -i HOME=$HOME /bin/sh -l for building 2015-09-08 11:32:05 some core linux kernel stuff and co Docu http://0xax.gitbooks.io/linux-insides/content/Booting/index.html 2015-09-08 11:49:26 uggedal: yeah, someone hit that already, no idea if fabled had any ideas 2015-09-08 11:49:41 I can check the logs if you don't have yours 2015-09-08 12:19:46 barthalion: thanks, I think I have logs 2015-09-08 13:01:44 hi 2015-09-08 13:02:17 o/ 2015-09-08 13:02:20 I didn't realise there was a separate channel for development, excuse for the duplication here 2015-09-08 13:02:26 is it possible to redirect from sys/poll.h to poll.h silently, without warning? the warning breaks things when building with -Werror 2015-09-08 13:02:30 http://slexy.org/view/s2YjTaBz6m 2015-09-08 13:10:47 nevermind, I realised it's an upstream behaviour inherited from musllibc & not something implemented specifically on Alpine 2015-09-08 14:57:34 ncopa: so seriously: what's your plan regarding the linux-grsec package? 2015-09-08 15:11:13 nmeum: i am writing to spender and ask for options 2015-09-08 15:11:20 ok i gotta go 2015-09-08 15:11:21 see u 2015-09-08 18:39:30 openwall is using grsec stable as well, right? How are they handling this issue? 2015-09-08 18:40:49 oh, they are not using it 2015-09-08 19:05:05 indeed 2015-09-08 19:05:13 which should indicate it's *unnecessary* 2015-09-08 19:16:30 aren't they using PaX? 2015-09-08 20:24:22 I see there are bootstrap support for bootstrapping mips in our gcc/musl packages. has anyone taken it further? 2015-09-08 20:35:43 eh, I think someone wanted to make a mips port, so we added it 2015-09-08 20:36:13 but enough time passed to make me forget who exactly 2015-09-08 21:33:34 if anyone else needs logs of this chan: http://dev.alpinelinux.org/~uggedal/alpine-devel_2013-04-18_2015-09-07.json.xz 2015-09-09 06:11:06 so, nodejs 4.0... 2015-09-09 06:12:11 hmm, "However, we have mapped a quick initial path to our first LTS version which will come in October" 2015-09-09 06:12:20 so maybe 4.0 is not LTS 2015-09-09 06:12:48 oh, "The v4.x line of releases will continue for an additional 18 months in LTS" 2015-09-09 06:13:36 it will get 12 months of security fixes after that as well 2015-09-09 06:13:51 we should probably remove testing/iojs as well 2015-09-09 06:31:05 ok 2015-09-09 07:20:05 ncopa, what do you think about %539 and %540? 2015-09-09 07:20:35 539 LGTM 2015-09-09 07:20:58 same about 540 2015-09-09 07:22:24 fcolista: looks good to me 2015-09-09 07:31:46 what are people using for testing on i686? lxc or chroot? 2015-09-09 07:35:58 I use systemd-nspawn, but I'm a minority here for obvious reasons 2015-09-09 07:36:01 I guess lxc 2015-09-09 07:38:20 uggedal: i use lxc for x86 testing 2015-09-09 07:38:24 and for stable branches 2015-09-09 07:38:43 docker might be an option too? 2015-09-09 07:38:58 thanks, will setup some lxc containers 2015-09-09 07:39:19 not that fond of docker after people started to (ab)use it at work ;) 2015-09-09 07:39:31 :) 2015-09-09 07:39:36 not surprised 2015-09-09 07:39:42 docker should be used with care 2015-09-09 07:45:53 x86-edge builder stuck? 2015-09-09 07:46:58 either stuck or busy with nodejs 2015-09-09 07:47:13 8530 8529 buildoze R 64880 0% 6 4% {cc1plus} /usr/libexec/gcc/i586-al 2015-09-09 07:47:14 18306 18304 buildoze S 8092 0% 0 0% {buildrepo} /usr/bin/lua5.2 /usr/b 2015-09-09 07:47:21 it appears to be building 2015-09-09 07:47:23 rather busy with node 2015-09-09 07:47:35 the same hw is also building kernel atm 2015-09-09 07:47:36 why not http://runc.io/ instead of docker ? 2015-09-09 07:48:47 whats the point of runc? 2015-09-09 07:48:55 they claim: can run docker images 2015-09-09 07:48:59 but so can docker... 2015-09-09 07:49:02 so why use runc? 2015-09-09 07:49:16 no central daemon 2015-09-09 07:49:20 *only* reason i can see on front page is: 2015-09-09 07:49:21 Systemd Integration 2015-09-09 07:49:21 runC does not create a daemon, so it integrates well with systemd. 2015-09-09 07:49:26 you can supervise the container directly with runit etc 2015-09-09 07:49:29 yes 2015-09-09 07:49:47 well the image point I guess is to highlight "You don't have to redo your images to use runC" 2015-09-09 07:50:13 ncopa: as uggedal says no central docker daemon 2015-09-09 07:50:14 coredumb: the question was why should i use runC in first place 2015-09-09 07:50:21 but that makes sense 2015-09-09 07:50:26 to avoid the central daemon 2015-09-09 07:51:58 coreos depends heavily on systemd and docker don't match well with it. so they created rkt. there were much drama. as a result the open container initiative and runc was started 2015-09-09 07:52:37 so runc exists due to docker not play well with systemd 2015-09-09 08:00:38 ncopa: I guess also having the central daemon and PID 1 "issues" 2015-09-09 08:00:58 but well was just pointing out the link I saw yesterday ^^ 2015-09-09 08:03:35 can anyone commit the rsync mirror for mirrors.2f30.org to the mirrors file? rsync://mirrors.2f30.org/alpine 2015-09-09 08:03:50 hadn't realized it was not present, been running rsync for some time 2015-09-09 08:15:42 stateless: something like http://tpaste.us/GVaz 2015-09-09 08:15:44 ? 2015-09-09 08:25:50 good morning 2015-09-09 08:30:52 does runc depend on systemd existing or is it just leaving the stuff that docker &c. wanted a daemon for to be done externally? 2015-09-09 08:35:11 ... what did docker &c. need a the daemon for anyway 2015-09-09 08:36:26 ammunta: http rest interface 2015-09-09 08:51:47 so I need sftp-server from openssh to use with dropbear. is it ok to add it to a subpackage (and dep to it from main package)? 2015-09-09 08:57:06 runc seems p neat 2015-09-09 08:57:13 I should play around with it a bit 2015-09-09 08:57:49 it seems in principle better than docker at least 2015-09-09 08:58:15 likely buggy as hell atm probably but hey 2015-09-09 09:04:45 ncopa, yeah add it in addition to the http mirror 2015-09-09 09:09:19 a friend just asked me if alpine has a rust package. i did not find one, or am i wrong? 2015-09-09 09:10:43 uggedal: sounds good to me 2015-09-09 12:02:20 <_mjones> I suppose this new policy withgrsecurity patches does not adversely affect Alpine? 2015-09-09 12:16:05 _mjones: it affects alpine 2015-09-09 12:16:35 if we don't get access to the stable patches, then it will create significant extra work 2015-09-09 12:16:43 ...if we want continue use grsecurity 2015-09-09 12:18:20 ncopa: ok to bump vim? 2015-09-09 12:18:29 ok 2015-09-09 12:18:48 <_mjones> @ncopa: damn. 2015-09-09 12:28:26 <_mjones> ncopa: it seems hardened gentoo and hardened arch are still accorded access as favoured projects; it would seem to me that Alpine falls into the same pro bono category, and a good citizen. 2015-09-09 12:30:10 i would think so too 2015-09-09 12:30:16 i emailed spender 2015-09-09 12:30:29 how would that work, will we need to keep the patches private? 2015-09-09 12:31:43 what i am hoping, that we get access to stable patches and changelogs 2015-09-09 12:32:24 and then we publish as little as possible 2015-09-09 12:32:31 <_mjones> @ncopa: oh good, I'm glad you're on top of that. 2015-09-09 12:32:33 we cannot completely hide it 2015-09-09 12:32:36 due to gpl 2015-09-09 12:35:11 would we not have to give out the patches on request to our users due to gpl? 2015-09-09 12:35:26 <_mjones> I think everybody knows the score with gpl, and also with upstream requirements. 2015-09-09 12:36:01 <_mjones> Red Hat are jerks who drop their kernel patches as one big diff, and only give individual patches to their contracted customers. 2015-09-09 12:36:49 <_mjones> I'd dislike having to see that happen, but if it did it wouldn't be a choice but an upstream requirement. 2015-09-09 12:37:27 <_mjones> Red Hat do it because of Oracle Linux, but it's still an unmutual move. 2015-09-09 12:43:37 uggedal: we are ofcourse required to give patches out due to gpl 2015-09-09 12:43:53 but spedner might want that we dont make them easy to find 2015-09-09 12:44:02 eg remove them from public repos 2015-09-09 12:44:10 and only give out when someone actually asks for it 2015-09-09 12:44:38 or take the redhat approach, one big diff 2015-09-09 12:44:45 to make it as useless as possible 2015-09-09 12:45:02 <_mjones> to be pedantic, you only need to give it out to someone who has received the binaries from you. 2015-09-09 12:51:25 and actually asked for sources, yes 2015-09-09 12:55:10 and you can give the sources on 8" floppies 2015-09-09 12:56:04 <_mjones> kunkku: I wouldn't want to test that in court, actually. Just the snail mail delay could be construed as obstruction in the modern world. 2015-09-09 12:57:32 <_mjones> I think the best outcome is finding a way to make grsecurity happy in general and returning to the status quo. 2015-09-09 13:06:38 <_mjones> @ncopa: what do you think is the likelihood of having UEFI boot media in time for 3.3.0? Is it known everything that needs to be done for this to ship? 2015-09-09 13:17:43 _mjones: i need fix initramfs script 2015-09-09 13:17:57 i was supposed to do that this week 2015-09-09 13:21:05 is linux-firmware needed as a dependency for linux-virtgrsec? 2015-09-09 13:44:29 <_mjones> @ncopa: oh, that sounds straightforward. I was thinking about looking into it myself, in part to improve my knowledge of booting and uefi. 2015-09-09 13:45:12 uggedal: i dont think so. you probably dont need firmware for virtual? 2015-09-09 13:45:38 _mjones: the idea is to avoid alpine_dev= boot option 2015-09-09 13:46:06 and have the initscript look at all found blockdevices 2015-09-09 13:48:51 <_mjones> @ncopa: hmmm. 2015-09-09 14:00:33 ncopa: that was what I thought to, I'll do some experiments to see if I can save 76MB :) 2015-09-09 14:02:04 ncopa: also, I noticed a few differences between grsec/virtgrsec apart from leaner hardware support in the latter 2015-09-09 14:03:18 CONFIG_MEMCG for instace. I can look them over and propose a patch to bring them more in line with eachother 2015-09-09 14:03:25 would be nice 2015-09-09 14:03:35 basically 2015-09-09 14:03:54 virtgrsec may disable all stuff that makes no sense in virtual environment 2015-09-09 14:04:14 yep 2015-09-09 14:26:46 do we need a replaces=iojs in nodejs or is it not needed since it has not been part of a stable release? 2015-09-09 14:27:17 i dont think we need it 2015-09-09 14:27:26 it is only to make apk upgrade smoother 2015-09-09 15:34:00 _mjones: gentoo and arch use the testing patches those are still available publicly 2015-09-09 15:35:49 <_mjones> nmeum: I know the testing are public, but I saw explicit accomodation of hardened gentoo and hardened arch which seemed to me to say that they would have or could have stable patches. 2015-09-09 15:36:53 "The test series, unfit in our view for production use, will however continue to be available to the public to avoid impact to the Gentoo Hardened and Arch Linux communities" 2015-09-09 15:38:16 <_mjones> nmeum: it's not overwhelmingly clear to me if the primary offense is use of ancient (3.2) kernels or noncontributing userbase, but it seems to me that the ideal outcome would be a return to the status quo. If the breach could be rectified. 2015-09-09 15:39:22 <_mjones> nmeum: as you write. I hadn't read that as ergo propter hoc. 2015-09-09 15:39:41 <_mjones> But it seems to be 'in order to avoid impact' as you say. 2015-09-09 20:29:28 I'm going to remove all manual *.la removal lines tomorrow 2015-09-10 06:51:27 barthalion: nice, make sure you also remove them from the newapkbuild template 2015-09-10 09:14:52 http://lists.alpinelinux.org/ finally got new look 2015-09-10 09:16:45 nice 2015-09-10 09:17:01 a bit of spam in alpine-infra lately 2015-09-10 09:18:15 hey :) 2015-09-10 09:24:53 -.- it finally happend 2015-09-10 09:25:03 stable grsec patches aren't available anymore 2015-09-10 09:27:45 ? 2015-09-10 09:31:06 nice! http://lists.alpinelinux.org/ 2015-09-10 09:38:32 fancy 2015-09-10 09:38:39 nmeum: good point, thanks 2015-09-10 10:17:41 i'm trying alpine out from a docker instance. how do i typically run alpine from docker so i can explore apk commands and such? 2015-09-10 10:18:04 ncopa: seems lists.a.o needs a or a 'Content-Type: text/html; charset=utf-8' http response header 2015-09-10 10:23:15 i don't get this version https://github.com/gliderlabs/docker-alpine 2015-09-10 10:23:44 i guess ppl use supervisor for managing their processes? 2015-09-10 10:24:11 hendry: docker pull alpine 2015-09-10 10:24:25 docker run -it alpine /bin/sh 2015-09-10 10:25:28 hendry: no idea, I do the monitoring on host side 2015-09-10 10:26:36 barthalion: not sure what you mean by monitoring? you mean you run processes individually ? 2015-09-10 10:28:48 I meant managing, but yes, I run one process per container 2015-09-10 10:45:17 would someone be kind enough to rebuild cdw? 2015-09-10 10:47:35 ScrumpyJack: against which library, for commit message completeness? 2015-09-10 10:48:01 libcdio and ibiso9660 2015-09-10 10:48:15 libcdio and libiso9660 even 2015-09-10 10:49:05 wee, build failure 2015-09-10 10:49:11 :( 2015-09-10 10:49:28 ah, headers 2015-09-10 10:50:41 ScrumpyJack: done, thanks 2015-09-10 10:52:07 awesome, you're a star 2015-09-10 10:55:38 ugh, 912 APKBUILDs modified 2015-09-10 10:55:44 and yet 84 left 2015-09-10 12:00:32 980 files changed, 21 insertions(+), 1102 deletions(-) 2015-09-10 12:00:34 (y) 2015-09-10 14:59:20 some packages use /usr/lib as the libexec dir some use /usr/libexec (lxc for example) what is preferred? 2015-09-10 14:59:31 s/what/which/ 2015-09-10 14:59:51 i think /usr/lib/$pkgname/ is preferred 2015-09-10 15:00:03 but no string opinion 2015-09-10 15:01:03 ok 2015-09-10 15:03:29 ncopa: I noticed that dhcpcd uses /lib while dhcpcd-dbus uses /usr/lib. I fixed that in %555 … 2015-09-10 16:09:28 hendry: runit? 2015-09-10 16:10:29 actually s6 looks like it'd play nicer with docker but eeeeh 2015-09-10 16:14:46 pretty happy with the 10.43MB FTP container I have running now 2015-09-10 16:19:50 hm? 2015-09-10 16:20:38 https://github.com/kaihendry/camftp2web 2015-09-10 16:21:49 quite proud of https://github.com/kaihendry/camftp2web/blob/master/postprocess/copy.sh ... needs polish but it works well 2015-09-10 16:22:48 hm who maintains the gliderlabs/alpine images 2015-09-10 16:23:05 it looks like it hasn't updated a while 2015-09-10 16:24:19 hendry: depending on your use-case it might be worth instead of upgrading explicitly in that Dockerfile 2015-09-10 16:24:25 rebuild the base image 2015-09-10 16:27:02 https://github.com/kaihendry/camftp2web/blob/master/Dockerfile#L1 is working out for me? 2015-09-10 16:27:47 https://github.com/kaihendry/camftp2web/blob/master/Dockerfile#L6 I meant this 2015-09-10 16:28:09 I am assuming the reason for that is that the image is like, months out of date 2015-09-10 16:34:42 ammunta: i just copied that 2015-09-10 16:36:15 oh 2015-09-10 16:36:26 https://github.com/gliderlabs/docker-alpine/tree/master/builder anyway 2015-09-10 16:36:32 it wasn't terrible difficult to use 2015-09-10 16:37:49 docker run --rm alpine-builder -esr v3.2 | docker import - docker:new 2015-09-10 16:37:56 seems to work fine 2015-09-10 16:38:10 ... wait I'm stupid 2015-09-10 16:38:20 I tagged that docker 2015-09-10 16:42:35 ammunta: we're going to bump 3.2.3 in the next couple days (maybe today) 2015-09-10 16:42:59 the official should follow shortly after that (official builds from the Glider Labs builder) 2015-09-10 16:47:52 ... when is that build supposed to update 2015-09-10 16:50:58 a couple minutes after we merge https://github.com/gliderlabs/docker-alpine/pull/67, some guys are on vacation so testing and confirmations are a little slow this week 2015-09-10 16:59:08 ... huh. 2015-09-10 16:59:35 wait is the non-updating because of failing that test or 2015-09-10 17:05:47 no, it is just in use in production, so updating existing tags is dangerous and i always prefer to have others test and sign-off on it before releasing 2015-09-10 17:11:35 ah 2015-09-10 17:11:58 well either way I have use for that builder anyway 2015-09-10 17:20:46 great, i've been thinking about separating out the builder into its own repo 2015-09-10 17:36:19 you know, the builder could be p trivially self-hosted, couldn't it 2015-09-10 17:37:22 most of the work'd just be an apk invocation 2015-09-10 17:46:36 andyshinn: btw uname -m does not work reliably 2015-09-10 17:46:55 # uname -m 2015-09-10 17:46:55 armv7l 2015-09-10 17:50:23 once I hard-coded arch=armhf that worked fine 2015-09-10 18:11:26 andyshinn: sent a pull req 2015-09-10 19:01:44 Hello. I am compiling some python package and I found out that the python3-dev package has /usr/include/python3.4m/Python.h:11:20: fatal error: limits.h: No such file or directory #include 2015-09-10 19:02:18 the file gets installed with `fortify-headers` package, should `python3-dev` depend on it? 2015-09-10 19:06:58 no 2015-09-10 19:07:14 limits.h is in musl-dev 2015-09-10 19:07:38 apk add alpine-sdk 2015-09-10 19:11:21 so python3-dev should depend on musl-dev? 2015-09-10 19:12:15 no 2015-09-10 19:12:41 alpine-sdk is needed to build alpine packages 2015-09-10 19:13:10 I am not building alpine package 2015-09-10 19:13:40 I am building a python package 2015-09-10 19:13:47 well, okay 2015-09-10 19:13:56 then you should have build-base installed 2015-09-10 19:14:14 so if python3-dev depends in its #include 2015-09-10 19:14:20 on some file that is in another package 2015-09-10 19:14:27 should it be pulled automatically? 2015-09-10 19:14:43 or the idea of apk is that it not always installs whole dependency tree? 2015-09-10 19:14:45 we expect user to have either alpine-sdk or build-base installed 2015-09-10 19:14:51 depending on the use case 2015-09-10 19:15:03 ok 2015-09-10 19:15:12 this is the only exception to resolving entire dependency tree 2015-09-10 19:15:22 for my use cases i want to have minimal install 2015-09-10 19:15:32 to have thin docker layer 2015-09-10 19:15:59 alpine is already thin 2015-09-10 19:16:15 yes, i know ;-) I am pushing it to the limits ;-) 2015-09-10 19:17:37 I was just misunderstanding the concept by assuming it always pulls full build depdendencies. So i thought that insalling `python3-dev` and `gcc` would do it 2015-09-10 19:18:13 Thanks for clearing this up 2015-09-10 19:18:51 I am having a lot of fun with alpine linux 2015-09-10 20:17:33 ncopa: comments on lua-rex update to 2.8.0 and support for 5.2/5.3: http://sprunge.us/RZFi ? 2015-09-10 20:48:20 andyshinn: tangentially, https://github.com/gliderlabs/docker-alpine/compare/master...muhmuhten:selfhosted 2015-09-10 20:50:33 andyshinn: it's definitely not suitable for actual use but it's kind of neat 2015-09-10 20:53:47 Oh that is what you meant. Actually, I have a branch I'm working on with similar changes :) 2015-09-10 20:54:09 yep 2015-09-10 20:54:23 it would buy us a few things like not needing to allow untrusted 2015-09-10 20:54:26 It is kind of scary though, because if you break the build in builder, you potentially cannot rebuild 2015-09-10 20:55:01 well like with everything else self-hosted if that breaks you use an older version to rebuild 2015-09-10 20:56:08 Yep, so I'm thinking to tag the builder something like alpine:builder, use that, then remove if it was success 2015-09-10 20:56:21 Next build would tag again based on latest 2015-09-10 20:58:51 actually, wait 2015-09-10 20:58:59 I have a better idea 2015-09-10 20:59:40 instead of opening up the key package, might as well just use the outside installation to fetch the whole base and just install that 2015-09-10 21:01:29 ... also then we never even create the cache inside 2015-09-10 21:01:59 Well, I still need it to support another PR that enables only building from certain packages to make even slimmer (no alpine-base package) 2015-09-10 21:05:51 https://github.com/gliderlabs/docker-alpine/compare/gliderlabs:79d35ac...muhmuhten:e8986da cleaned it up even more 2015-09-10 21:07:38 https://github.com/piotrminkina/docker-alpine/commit/0727c80559f4aa267b86e5cbe77028201f929e2f this one? 2015-09-10 21:16:44 Yea, it should support that 2015-09-10 21:17:41 It is a good idea, I'll comment on it a little later and maybe it can be a PR 2015-09-10 21:18:10 done 2015-09-10 21:20:23 https://github.com/gliderlabs/docker-alpine/compare/gliderlabs:79d35ac...muhmuhten:c5fd4f2 2015-09-10 21:21:02 though honestly I find the extracting alpine-base bit kind of questionable 2015-09-10 21:45:47 well, if the builder is now alpine, we can avoid that and the tzdata hack by just copying it in from the host as well 2015-09-10 21:47:36 ammunta: feel free to open that as a PR, with the idea that it will need to be rebased (and probably tweaked) after #32 is merged. but i'm happy to start a conversation on it 2015-09-10 21:52:17 andyshinn: ah, that's an idea 2015-09-10 21:52:52 mostly, because i'm working on some other things this week so i can't review right now, but if it is a PR i can come back to it and give some feedback later 2015-09-10 21:57:57 andyshinn: is there any reason we can't just leave out /etc/localtime if no -t is specified 2015-09-10 21:59:41 hmm, i did it for a reason but now i can't remember... maybe leave it out if not -t specified for now and then when i test it out i'll remember :p 2015-09-10 22:00:07 = "we'll see what goes wrong when we leave it out"? 2015-09-10 22:02:36 heh yea 2015-09-10 22:38:45 pffh found what was breaking the automated check 2015-09-10 22:38:53 embarrassing bug 2015-09-10 22:39:50 doesn't mean it *works* but 2015-09-11 06:41:18 uggedal: lua-rex looks ok to 2015-09-11 06:41:20 to me 2015-09-11 06:41:25 i didnt study deep though 2015-09-11 06:42:56 i will restart the builder containers 2015-09-11 06:43:02 and git.a.o 2015-09-11 06:43:11 can you please hold your commits for a second 2015-09-11 06:43:24 or for 5mins... 2015-09-11 06:51:16 i'm stopping the buidlers now 2015-09-11 07:09:58 ncopa: builders good to go? 2015-09-11 07:10:25 i think so 2015-09-11 07:10:28 :) 2015-09-11 07:11:22 i enabled jumbo frames 2015-09-11 07:11:57 doesn't look like good to go 2015-09-11 07:12:18 the edge builders looks fine, no? 2015-09-11 07:12:28 hm, right 2015-09-11 07:13:10 there are some kind of loop with the mqtt messaging 2015-09-11 07:13:18 it happens when they start up 2015-09-11 07:13:22 i dont know why 2015-09-11 07:13:24 hum 2015-09-11 07:13:32 uggedal: I guess you should backport f910f562 to supported branches 2015-09-11 07:13:47 seems like nfs broke 2015-09-11 07:13:56 :-( 2015-09-11 07:14:17 oh no 2015-09-11 07:14:22 totall freezy 2015-09-11 07:14:44 barthalion: don't have a stable builder to test on yet... 2015-09-11 07:15:09 but could be cherry picked for 3.2 2015-09-11 07:15:26 3.1 has 0.8 and 3.0 has 0.7 2015-09-11 07:21:36 hum 2015-09-11 07:21:42 i think i need to reboot bld1. :-( 2015-09-11 07:22:05 nfs mount makes things hang 2015-09-11 07:22:06 hard 2015-09-11 07:23:49 any idea how to recover from a hanging nfs mount? 2015-09-11 07:24:01 apparently nfs didnt like that i gave it jumbo frames 2015-09-11 07:26:09 i will have to kill the containers again:-( 2015-09-11 07:36:51 :( 2015-09-11 07:36:58 i cannot even reboot the machine 2015-09-11 07:37:00 it just hangs 2015-09-11 07:37:06 not eve reboot -f works 2015-09-11 07:37:16 its totally locked up 2015-09-11 07:37:23 i think its sync that will lock things 2015-09-11 08:00:21 actually, it shouwl work 2015-09-11 08:21:50 the box is up 2015-09-11 08:21:57 and it seems like jumboframs work 2015-09-11 08:22:08 i just need set proper nfs rsize, wsize values 2015-09-11 08:22:25 and i think there is a significant performance gain too 2015-09-11 08:22:49 iirc, it used to copy files with ~60Mb/s 2015-09-11 08:22:58 now it is around 100Mb/s 2015-09-11 10:04:57 i need fix up the firewall on bld1.a.o i think 2015-09-11 10:05:01 after lunch 2015-09-11 10:58:24 i've a strange issue with ppp interace. I've a dual-isp setup. The main one is ppp. Strangely, ppp is not adding default route (even if defaultroute is an option specified in peers/pppoe-isp). 2015-09-11 10:59:39 rather, the "second" isp which has static address assigned, works as expected. 2015-09-11 10:59:47 Now, i'm wondering two things: 2015-09-11 10:59:56 1. Can i setup "metric" on ppp ? 2015-09-11 11:00:39 2. Probably, the correct one: i've just realized that ppp comes after the second isp in /etc/network/interfaces 2015-09-11 11:01:03 so, what is picked up as first is the second isp (eth1) 2015-09-11 11:01:25 anyway, ppp does not add default route. 2015-09-11 11:33:34 hey friends :) 2015-09-11 11:33:51 ncopa: sorry about two days ago, as soon as i wanted to write you a customer called 2015-09-11 11:33:59 have you seen my text later on? 2015-09-11 11:39:04 dont remember :) 2015-09-11 11:39:47 haha 2015-09-11 11:40:04 since my post ended with a question i asume you have not seen it *g* 2015-09-11 12:10:19 ncopa: what is the plan for the new mainline kernel 4.2? 2015-09-11 12:10:50 leo-unglaub: no, we try catch the longterm kernels 2015-09-11 12:13:44 isnt 4.2 going to be the next longterm? or am i missinformed? 2015-09-11 12:14:34 https://plus.google.com/+gregkroahhartman/posts/KYeaqXRo4y4 2015-09-11 12:14:40 seems like its gonna be 4.1 2015-09-11 12:15:33 oh sorry, then i got it wrong 2015-09-11 12:17:29 but still a valid question 2015-09-11 12:17:56 i suppose the answer is: yes, we want 4.1 kernel for alpine v3.3 2015-09-11 12:18:14 also we need someone who can make a cronjob to save the grsecurity testing patches 2015-09-11 12:18:19 and archive those for us 2015-09-11 12:19:31 thats actually something i wanted to ask you about. if the testing patches are still free, they would become stable eventually and therefor would still be able to work for us, right? 2015-09-11 12:21:53 yes 2015-09-11 12:22:19 ok, a job task: make a cron job or similar that fetches all testing patches 2015-09-11 12:22:23 so basically all spender did was to block stable patches and force users to use the testing patches that become then stable? 2015-09-11 12:22:30 yes 2015-09-11 12:22:32 but 2015-09-11 12:22:38 the stable patches are more valuable 2015-09-11 12:22:43 that is even more stupid than i thought ... 2015-09-11 12:23:03 table patches are for the stable longterm kernel they support 2015-09-11 12:23:31 they also cherry-pick sec fixes that not yet entered the given stable linux-*.y branch 2015-09-11 12:23:41 which means it contains early sec fixes 2015-09-11 12:24:01 which will likely hit next stable release (or the release after that or after that again) 2015-09-11 12:24:04 ..if ever. 2015-09-11 12:25:10 so stable grsecurity patches contains upstream sec fixes that might never be added to stable vanilla linux 2015-09-11 12:25:25 the change log gave exact commits 2015-09-11 12:25:30 hmm, and those sec fixes are not in the testing patches? 2015-09-11 12:25:47 they are often in the mainline upstream 2015-09-11 12:25:58 together with the other 1000+ commits 2015-09-11 12:26:15 hmm, damn ... so now we have to torrent the stable patches and diff them against the testing patches?? 2015-09-11 12:26:35 possible 2015-09-11 12:26:45 or we simply dont get those upstream sec fixes 2015-09-11 12:27:03 and have to wait til they show up together with stable vanilla kernel 2015-09-11 12:27:17 we will likely get them, just a bit later 2015-09-11 12:27:28 and we might miss a few 2015-09-11 12:27:31 hmm, but this would meant that staying with longterm kernels would become a problem 2015-09-11 12:27:39 <_mjones> @ncopa: did spender respond to you at all? 2015-09-11 12:27:41 not really 2015-09-11 12:27:45 _mjones: not yet 2015-09-11 12:27:59 i'll give him a few days more 2015-09-11 12:32:15 leo-unglaub: the cherry-picked upstream sec fixes is why i say that grsecurity patch is better than vanilla, even if you disable every single pax/grsecurity feature in config 2015-09-11 12:32:53 <_mjones> I didn't know about these additional fixes. Interesting. 2015-09-11 12:33:54 ncopa: but still with testing patches a current kernel would be better 2015-09-11 12:34:00 as far as i understand it 2015-09-11 12:34:22 and with linux, there are nearly zero api breaks so that it does not really matter? 2015-09-11 12:34:35 or have you ever seen huge problems between two linux versions? 2015-09-11 12:34:48 _mjones: i noticed is while porting it to 3.18 kernel and studying the changelog 2015-09-11 12:35:40 leo-unglaub: yes, i think that the pax/grsecurity functionality is good too 2015-09-11 12:36:02 so we may keep the pax/grsecurity functionality by saving the testing patches 2015-09-11 12:36:10 it means more work for us 2015-09-11 12:36:36 and we might not have the resources to port new features or fixes in grsecurity itself 2015-09-11 12:36:53 more work? ah come on, for someone as skilled as you this should not be much more work ;) 2015-09-11 12:37:04 <_mjones> More work definitely. But I suppose there might be a silver lining to this cloud, by giving an opportunity to have finer control, and better QA. Maybe. 2015-09-11 12:37:58 what are the chancses that the other distros will work together and fork it openly? 2015-09-11 12:38:21 small 2015-09-11 12:38:22 if gentoo, arch and slackware would come together, there would be a lot of man(woman)power available 2015-09-11 12:38:57 more people does not necesarily help 2015-09-11 12:39:12 i have already an open fork 2015-09-11 12:39:18 spender dislike it 2015-09-11 12:39:43 kernel patching takes time 2015-09-11 12:39:44 <_mjones> It's not clear to me what he wants to happen. 2015-09-11 12:40:02 i think what he wants to happen is more people give him money 2015-09-11 12:40:26 <_mjones> That was one of the options I imagined. 2015-09-11 12:40:40 hmm, propobly 2015-09-11 12:40:43 but why now? 2015-09-11 12:40:52 i mean he did not get payed for his work for years 2015-09-11 12:40:58 he got payed 2015-09-11 12:41:00 honestly if it was that the price wouldn't be that low ... 2015-09-11 12:41:04 he had sponsors 2015-09-11 12:41:15 _only that_ 2015-09-11 12:41:31 <_mjones> I generally think organisations should be sending money to projects a lot more. But obviously not everyone is a for-profit project. 2015-09-11 12:41:48 i suspect that some of his sponsors complained 2015-09-11 12:42:13 "Our stable series have existed solely due to the financial support of our sponsors." 2015-09-11 12:42:17 from: https://grsecurity.net/announce.php 2015-09-11 12:42:50 "We decided that it is unfair to our sponsors that the above mentioned unlawful players can get away with their activity. Therefore, two weeks from now, we will cease the public dissemination of the stable series and will make it available to sponsors only." 2015-09-11 12:43:04 so it is to protect their sponsors 2015-09-11 12:43:28 hmm, you are propobly right 2015-09-11 12:43:34 <_mjones> So then, this suggests that the current reaction is him being upset, and also an open search for additional sponsors. 2015-09-11 12:43:43 but he must know that with his desicion he is just fucking off open source projects 2015-09-11 12:44:06 <_mjones> I hadn't read it so closely to have picked that up. 2015-09-11 12:46:16 ncopa: well it's been known for ages that the choice of the stable kernel release was chosen by a/some sponsor 2015-09-11 12:47:34 yes 2015-09-11 12:48:18 if it weren't for sponsors, there's clearly no way there would be a stable release 2015-09-11 12:48:25 only testing 2015-09-11 12:48:55 seems to me it just gets back to "before there was sponsors" 2015-09-11 12:50:16 there have been stable patches as long i can remember 2015-09-11 12:51:23 have you been able to discuss with spender about community distrib sponsoring ? 2015-09-11 12:51:35 i emailed him 2015-09-11 12:51:40 but got no response so far 2015-09-11 12:51:45 k 2015-09-11 12:52:00 he probably get tons of emails 2015-09-11 12:52:06 oh yeah 2015-09-11 12:52:26 had to remind him of me once or twice ^^ 2015-09-11 12:52:55 https://twitter.com/grsecurity/status/641222600811806720 2015-09-11 12:53:25 hehe 2015-09-11 12:56:48 <_mjones> Well, that's a positive sign. 2015-09-11 12:56:58 yeah 2015-09-11 12:57:06 <_mjones> Seems he's probably getting what he wanted -- attention. 2015-09-11 12:57:20 also, there was some mentions that he didnt intend to hit opensource projects 2015-09-11 12:57:31 it was not those he were after 2015-09-11 12:57:37 at least that is how i understood things 2015-09-11 12:58:13 <_mjones> I can imagine he was probably angry when the initial decision was made. Now there's been time to consider everything. 2015-09-11 12:58:55 ncopa: yeah, thats what i understand to 2015-09-11 12:59:07 he is, at his heart, an open source guy 2015-09-11 12:59:16 so he did not aim at distros using this 2015-09-11 12:59:37 otoh, he never really liked my port of his patch 2015-09-11 13:02:54 https://forums.grsecurity.net/viewtopic.php?f=3&t=2980 2015-09-11 13:04:14 hmmm 2015-09-11 13:04:34 maybe money got to his mind ... sadly the fucking money seams to fuck people up :( 2015-09-11 13:04:57 that forum post is from 2012 2015-09-11 13:07:01 <_mjones> Everyone needs money. But spender and the team deserve proper attribution for the work they do. 2015-09-11 13:07:08 +1 2015-09-11 13:07:17 absolutely 2015-09-11 13:07:44 and grsecurity is one of those things that is underestimated imho 2015-09-11 13:08:28 <_mjones> Remember the bit about the trademark. Downstream for-profit enterprise was advertising touting the grsecurity name. 2015-09-11 13:08:52 I'm pretty sure his skillset is well paid enough at his current job :) 2015-09-11 13:09:35 https://t.co/Ug8ueRpRBN 2015-09-11 13:09:47 the entire talk is pretty goo 2015-09-11 13:09:49 good* 2015-09-11 13:10:51 +1 2015-09-11 13:32:08 ok, last apk-tools fix broke apk fetch --stdout 2015-09-11 13:34:25 no 2015-09-11 13:35:07 yes 2015-09-11 14:57:36 algitbot: hi 2015-09-11 23:02:20 is there any way to obtain more useful apk error messages? 2015-09-11 23:02:58 Even with --error I just get '1 errors; 393 packages, 2598 dirs, 35056 files, 1172 MiB' but it doesn't tell me which package caused the error 2015-09-11 23:03:02 which is midly annoying 2015-09-12 16:06:47 <_mjones> I have an armel box here running Debian 6.0.5 which is being recalcitrant. Opinions on how pointless it would be to manually bootstrap it over to Alpine by pulling a toolchain, cloning aports, and waiting days for the compile? 2015-09-12 16:07:08 <_mjones> Don't really want to crosscompile off-box if possible. 2015-09-12 16:08:04 <_mjones> This box doesn't have console or backup booting, either, so I'd really like to avoid bricking it. 2015-09-14 08:15:35 ncopa: do you have a license for http://git.alpinelinux.org/cgit/user/ncopa/upstream-monitor/ ? 2015-09-14 08:20:30 will MIT/X11 do? 2015-09-14 08:20:42 or just MIT 2015-09-14 08:20:43 yep :) 2015-09-14 08:21:16 working on extending it a bit: http://git.uggedal.com/aports-vercmp/ 2015-09-14 08:21:24 and nice project for me to learn lua 2015-09-14 08:21:25 nice 2015-09-14 08:21:29 yeah 2015-09-14 08:21:54 should I just add a ISC license with your copyright to my repo? 2015-09-14 08:22:14 send me patch and i'll apply it 2015-09-14 08:22:33 or i move it to github? 2015-09-14 08:24:04 looks like i have a bunch of uncommitted changes 2015-09-14 08:33:11 uggedal: I'm adding a license file 2015-09-14 08:33:22 ncopa: thanks 2015-09-14 08:34:07 I'll keep working on it in my repo. when its more complete we could setup a non-user git.a.o repo 2015-09-14 08:34:46 also thinking of a "override" repo, for changing url/pattern/upstream provider etc per package 2015-09-14 08:36:25 yes we want some way for a per package override 2015-09-14 08:37:26 I have a 80 line TODO here based on experience with void-updates :P 2015-09-14 08:38:36 :) 2015-09-14 08:41:22 i pushed license + some minor fixes 2015-09-14 08:42:37 ncopa: thanks :) 2015-09-14 10:53:37 could someone take a look at %563? 2015-09-14 11:17:57 good morning friends :) 2015-09-14 12:14:05 uggedal: would be using rss + git api in vercmp ? 2015-09-14 12:14:24 would you be* 2015-09-14 12:15:02 kinda thinking on it for a while, and wanting to extend nbreader 2015-09-14 12:15:10 vkris: git api? 2015-09-14 12:15:24 git has nice json api 2015-09-14 12:15:27 github 2015-09-14 12:16:13 yes, but scraping the tags page gets you the exact same info and is not under rate limiting 2015-09-14 12:17:59 any plans for web-fronted to it ? 2015-09-14 12:20:49 like eg. integrating with https://pkgs.alpinelinux.org/packages ->Version (flag outdated) 2015-09-14 12:28:22 vkris: no 2015-09-14 12:46:27 uggedal: can I clone from git.uggedal.com ? 2015-09-14 12:46:39 would give a try again 2015-09-14 12:47:57 ok git:// did not work 2015-09-14 12:51:20 vkris: fixed now 2015-09-14 12:54:59 thanks 2015-09-14 16:06:26 <_mjones> clear 2015-09-14 16:13:22 <_mjones> So last night I was checking to see if aide would be a straight port in response to a question, and I found out that it requires libgpg-error.a. I had libgpg-error installed from apk and I had the .so, but not the .a. 2015-09-14 16:14:05 <_mjones> Is that an oversight that the static lib was not built and installed, or is it part of some policy on which libs get static and which don't? 2015-09-14 16:18:14 <_mjones> I reckon that unless there's a policy or reason I don't know, that libgpg-error (and probably other things) need to be revised with --build-static or whatnot. 2015-09-14 16:20:24 <_mjones> On a related note, Alpine is the only distro I know that doesn't use gpg for package signing. I really like the reuse of the openssl routines; I hear interfacing with gpg (at least gpg 1.x) is very ugly. 2015-09-14 16:31:51 _mjones: we try to avoid static libraries iirc 2015-09-14 16:32:04 _mjones: so the best solution would be making aide use shared library 2015-09-14 16:32:10 debian might have a patch for it 2015-09-14 16:34:38 Does anyone have thoughts on this patch: http://thread.gmane.org/gmane.linux.distributions.alpine.devel/3138 ? 2015-09-14 16:35:09 Did I send it to the right list? I haven't heard any response and didn't see it get merged yet. 2015-09-14 16:35:19 <_mjones> barthalion: I don't have a problem with that, and it might be the best solution. But this is a security ware, and people might not agree. I'll see what other distros do. 2015-09-14 16:37:11 <_mjones> jlyo: interesting patch. 2015-09-14 16:39:49 <_mjones> jlyo: I'm not an expert, what specifically is the goal? 2015-09-14 16:40:12 _mjones: My use case is for an s6 based init system, having init= there makes it possible to not run busybox init. 2015-09-14 16:45:00 _mjones: Debian and Fedora have explicit policy of avoiding static libraries 2015-09-14 16:45:19 jlyo: LGTM, but this stuff is usually being merged by ncopa or fabled 2015-09-14 16:45:25 it's their turf 2015-09-14 16:45:34 <_mjones> barthalion: hey, I just knew I'd find something I'd like about Fedora, some day. ;) 2015-09-14 16:46:12 also we do the same in Arch, except we ship a static one if there is no shared counterpart 2015-09-14 16:46:30 barthalion: Thanks, it's just been sitting there for a while. I also sent some RAID and LVM fixes for mkinitfs yesterday. 2015-09-14 16:46:51 yeah, I saw these from yesterday 2015-09-14 16:47:07 clandmeter new vdr is out :) http://www.linuxtv.org/pipermail/vdr/2015-September/028872.html 2015-09-14 16:47:16 jlyo: I'll review and merge them all if ncopa don't have time to do it 2015-09-14 16:47:25 <_mjones> barthalion: this is excellent info, thanks. I just need to look now if the app will cooperate and the users accept. 2015-09-14 16:47:51 <_mjones> jlyo: anything to do with raid autodetection by chance? 2015-09-14 16:48:16 barthalion: Thanks! 2015-09-14 16:48:29 not at all :) 2015-09-14 16:48:54 uggedal: btw, did you stop using void? 2015-09-14 16:48:55 _mjones: One patch installs /sbin/mdadm and /etc/mdadm.conf in the initfs if you choose the raid feature. 2015-09-14 16:50:20 <_mjones> jlyo: my use-cases hadn't yet noticed lack of md in initrd. 2015-09-14 16:51:24 _mjones: I'm running 2 disks, with two mirorred partitions, the first is vfat for /boot, the second is encrypted LVM. 2015-09-14 16:52:45 <_mjones> you must have a use case for vfat /boot. 2015-09-14 16:52:59 <_mjones> something something uefi? 2015-09-14 16:53:16 _mjones: In anticipation of UEFI, yes :) 2015-09-14 16:56:26 It almost works in alpine, setup-disk works if you add vfat to the whitelist of supported /boot filesystems. Something in mkinitfs also tried to call ln, but it doesn't prevent it from booting. 2015-09-14 16:57:55 <_mjones> jlyo: I'm hotly anticipating uefi for a couple of my systems also. But I didn't think vfat /boot was necessary, right? It has uses, but not necessary? 2015-09-14 17:00:24 UEFI needs a UEFI system partition, which needs to be FAT. Your /boot can reside on a different partition if you use a bootloader that looks outside the EFI System Partition. 2015-09-14 17:01:08 You keep more options open for bootloaders, and simplify things, if you make your /boot your ESP. 2015-09-14 17:01:36 <_mjones> ok, that's about what I thought. 2015-09-14 17:03:10 GRUB2 UEFI can scan your disks for your /boot filesystem if it's not the ESP, but I don't think many others can. 2015-09-14 17:03:20 <_mjones> I'm not 100% sure that uefi needs an ESP on disk, though. Unless I'm dense, I'm using uefi on this desktop and I didn't make a uefi on the blank disk. Let me poke it to be sure. 2015-09-14 17:04:46 <_mjones> mmm, machine in question is grub2. I'd hate to think that grub2 could do something I'd want to do with syslinux. 2015-09-14 17:06:02 Is there an ESP on there? 2015-09-14 17:06:15 <_mjones> Microsoft was entirely too influential on UEFI. However, I'm not sure what other options were practical, given the level of functionality subsumed into UEFI. 2015-09-14 17:07:35 <_mjones> no esp on disk according to fdisk. disklabel type is dos, though. 2015-09-14 17:08:13 <_mjones> Meh, I needed to sit down and grok initramfs anyway. I'll just update myself on UEFI at the same time. 2015-09-14 17:08:29 A lot of UEFI functionality is optional in the spec, and what is implemented is usually broken. 2015-09-14 17:09:26 <_mjones> broken in the sense that it's not even working in TianoCore, or broken in the sense that most taiwanese firmware vendors are usually always broken? 2015-09-14 17:10:14 _mjones: The latter. OVMF in QEmu works pretty good. 2015-09-14 17:11:03 <_mjones> When I wanted bios to go away, I was thinking of openboot/openfirmware or Alpha SRM. Not the giant ugly beast which UEFI has become. 2015-09-14 17:13:32 <_mjones> Can't trust wintel to get it right. The light at the end of the tunnel is that a large and growing fraction of the world's amd64 machines now boot Linux, so hardware vendors should have a newfound appreciation for clean standard interfaces. 2015-09-14 17:14:46 I'm thinking about installing coreboot on my laptop, I have a (I think/hope) supported Thinkpad T60p, still needs the ATI VGA blob to work, however. 2015-09-14 17:15:30 <_mjones> I believe the procedure is well-proven, even with the ATI vgabios. 2015-09-14 17:15:41 <_mjones> well-proven on the T60 series, I meant. 2015-09-14 17:16:11 <_mjones> Earliest I have is a T420, and it's a complete teardown to flash. 2015-09-14 17:17:02 <_mjones> I've spent a dozen hours trying to shop for modern coreboot hardware. Very frustrating. The only easy answer is to buy a chromebook or chromebox. 2015-09-14 17:17:47 <_mjones> There are indications we're close to a tipping point where some new mainstream hardware ships with coreboot, but whether it happens remains to be seen. 2015-09-14 17:18:06 I was looking at the chromebooks. They look like really interesting targets to port alpine to :) 2015-09-14 17:18:54 Does the T420 have a write-lock jumper/screw like the chromebooks? 2015-09-14 17:19:49 <_mjones> I bought a wolf (2014 dell chromebook 11, amd64, 4gb optional) to run Linux, but ChromeOS worked so well, and it turns out I had enough use-case that I never did flash it. 2015-09-14 17:20:22 <_mjones> Not quite sure on the T420. It's complete disassemble to board in order to flash it. 2015-09-14 17:21:51 <_mjones> Until the 2015 Pixel 2, none of the chromebooks had more than 4gb dram. I'm uncomfortable with less than 16gb in general. Somehow chromium (irony!) manages to chew up all 16gb on occasion, not even to mention the VMs and whatever else I'm doing. 2015-09-14 17:22:35 <_mjones> I'd been planning on getting an XPS13 ever since we had a few of those in my shop, but could never bring myself to do it with only 8gb hardwired. 2015-09-14 17:22:40 I think I'd be happy with a chroot on a chromebook. I'd be more happy if the kernel came with namespaces. 2015-09-14 17:23:25 <_mjones> I'm really into recent kernels, not 'android recent' kernels. 2015-09-14 17:23:51 <_mjones> Er, hold on. That was silly. ChromeOS runs ChromiumOS kernels, not Android. 2015-09-14 17:24:40 <_mjones> Not sure what kernel they're using. I predict 3.18.x, which is pretty good, but I'm enjoying 4.x a lot at this point. 2015-09-14 17:25:48 I wish I could update my android kernel so it could support LXC. 2015-09-14 17:26:32 <_mjones> Do you know what kernel is latest for AOSP? 2015-09-14 17:26:47 The effort to DIY would probably be too imense. 2015-09-14 17:27:37 Not sure. 2015-09-14 17:30:23 <_mjones> Yes. With software, though, the results of the DIY effort can be easily replicated by others. Not so much when you build a new garden shed. ;) 2015-09-14 18:02:01 barthalion: void on desktop and alpine on servers for now 2015-09-14 18:46:36 so question for you guys, i'm working on porting ghc to armhf now (also, yeah i've been busy at work so still need to merge my x86_64 port), but it basically is best to use llvm-3.5 for codegen for that platform, i'm building 3.5.2 for current alpine but what would be the best option for llvm in this case 2015-09-14 18:47:09 sadly without going to the current HEAD it doesn't work with 3.6 2015-09-14 18:48:22 <_mjones> mitchyty: Saw your mail to the list. Not sure I have any answers or suggestions, though. 2015-09-14 18:49:16 <_mjones> Are you upstreaming this for your own purposes, i.e. production use, or general portability and support? 2015-09-14 18:49:37 <_mjones> mitchty: Are you upstreaming this for your own purposes, i.e. production use, or general portability and support? 2015-09-14 18:49:47 ghc i'm looking to get into testing at least 2015-09-14 18:50:00 i've been using the port for a few months now to be honest 2015-09-14 18:50:25 <_mjones> If it's a general platform introduction, I'd ask how you feel about packaging your HEAD for now and waiting for everything to line up for it to evolve into your stable? 2015-09-14 18:50:52 well thing is ghc HEAD is really not a stable thing 2015-09-14 18:51:09 at least in the sense of something you'd base things off of 2015-09-14 18:51:20 i can do head too later but its more, 7.10.2 is current 2015-09-14 18:51:54 for x86_64 it doesn't matter as there is a native codegen 2015-09-14 18:52:02 but it could use llvm 3.5 too technically 2015-09-14 18:52:19 oh are you the one working on the ghc porting 2015-09-14 18:52:28 heh, it works now :) 2015-09-14 18:52:31 https://github.com/mitchty/alpine-ghc/blob/master/Dockerfile 2015-09-14 18:52:44 i've got stack/cabal/ghc working on x86_64 2015-09-14 18:52:54 and working on armhf right now 2015-09-14 18:52:54 <_mjones> I'm not an authoritative source by any means, but I'm not sure how much enthusiasm there might be for packaging llvm-3.5 2015-09-14 18:53:05 well i'm working on backporting it 2015-09-14 18:53:12 curious how that would work out in this case 2015-09-14 18:53:40 <_mjones> I'm quite interested in armhf personally, but it's not a majority platform. 2015-09-14 18:53:42 if i were to add an llvm-3.5 at the very least for armhf would that be bad 2015-09-14 18:53:53 mitchty: neat 2015-09-14 18:53:58 is it aported or 2015-09-14 18:54:06 yep all apks 2015-09-14 18:54:07 sec 2015-09-14 18:54:16 just manually built or something 2015-09-14 18:54:19 https://github.com/mitchty/alpine-linux-ghc-bootstrap 2015-09-14 18:54:23 sorta 2015-09-14 18:54:29 built with that 2015-09-14 18:54:42 <_mjones> mitchty: I would personally ask about any other options than packaging llvm-3.5. 2015-09-14 18:55:27 _mjones there really isn't any for arm, the "try to use c" stuff doesn't really work, for debian they're using llvm 3.5 2015-09-14 18:55:38 "I use ubuntu+musl-cross to build a ghc cross compiler and tar+xz that up" huh, why +xz 2015-09-14 18:55:41 i can try it but this takes ages to build 2015-09-14 18:55:48 size, i'm lazy 2015-09-14 18:55:51 <_mjones> So there's a codegen for armhf but it isn't sufficient quality, or there isn't one? 2015-09-14 18:56:05 isn't a native codegen for arm assembly 2015-09-14 18:56:15 there is an old, unmaintained c backend 2015-09-14 18:56:19 that kinda works 2015-09-14 18:56:25 kinda works 2015-09-14 18:56:28 but its not really where its moving 2015-09-14 18:56:34 hard to tell its still compiling 2015-09-14 18:56:47 <_mjones> Are you going to move off of native and onto llvm for x86-64? 2015-09-14 18:57:05 don't need to but i could, its faster to compile there too 2015-09-14 18:57:49 also that page needs some updates, i've a few new tricks for the bootstrap but its not a huge difference 2015-09-14 18:57:51 <_mjones> From an outsider's POV, it could be interpreted that armhf is a second-class citizen because you maintain native codegen for a popular platform. 2015-09-14 18:58:09 _mjones that is 100% accurate actually 2015-09-14 18:58:21 snrk 2015-09-14 18:58:26 https://ghc.haskell.org/trac/ghc/wiki/Platforms 2015-09-14 18:58:29 actually 100% accurate 2015-09-14 18:58:31 its considered a tier 2 platform 2015-09-14 18:59:22 oh there's actually arm ghci now? 2015-09-14 18:59:25 since when 2015-09-14 18:59:33 only really recently 2015-09-14 18:59:34 <_mjones> MacOS? Gee, no language weenie uses OSX. Why do you even support that? 2015-09-14 18:59:36 <_mjones> ;) 2015-09-14 18:59:53 debian wheezy unstable should have it 2015-09-14 19:00:16 _mjones: excuse me 2015-09-14 19:00:23 :&& 2015-09-14 19:01:14 "Tier 3 platforms worked in the past, but probably do not work now. " 2015-09-14 19:02:42 <_mjones> mitchty: I'd bump your list post by bringing this up and seeing what is the response to this conundrum. 2015-09-14 19:03:24 <_mjones> Many of the Alpine maintainers are in the channel on european time. 2015-09-14 19:03:53 yeah its no big deal for the moment 2015-09-14 19:04:03 i'm going to get this armhf port working first 2015-09-14 19:04:49 the arm port is actually getting pretty close to "tier 1" but well, needs more users 2015-09-14 19:05:00 i figure i might as well help out 2015-09-14 19:05:20 only way to get idris working on alpine is to port ghc :) 2015-09-14 19:06:22 i suppose i could port i386 too, but i don't see much point 2015-09-14 19:09:13 <_mjones> Compiling clean on musl and without certain other quirks is an attestation of quality code. 2015-09-14 19:12:47 well it is way better than uclibc that is for sure 2015-09-14 19:12:55 that libc had to get patches 2015-09-14 19:13:10 musl, just worked out of the box for cross compiling (mostly) 2015-09-14 19:13:53 <_mjones> Did you have to fix any headers? Be honest. :) 2015-09-14 19:16:42 nope 2015-09-14 19:17:03 wait does algitbot 303 on compiling 2015-09-14 19:17:04 did have to force a few assembly things relating to pic assembly 2015-09-14 19:17:43 and for this arm port there is some old windows thing that gets compiled that i need to patch away until the upstream applies it 2015-09-14 19:17:53 mostly fiddly things 2015-09-14 19:19:00 that bootstrap repo has it all, i'm too lazy to do this manually 2015-09-14 19:19:11 if its not automated its not useful 2015-09-14 19:19:47 <_mjones> Yes, algitbot 303s on compile 2015-09-14 19:19:52 <_mjones> or compiling. 2015-09-14 19:19:59 <_mjones> compiling only. 2015-09-14 19:20:23 should have a limit to how often in a n minute window he replies with that 2015-09-14 19:20:48 <_mjones> He needs a regex upgrade too, imo. 2015-09-14 19:20:55 that or i just use the c word too much 2015-09-14 19:21:17 <_mjones> mitchty, if you're not using assembler or shell script I just don't know if I can help you. ;) 2015-09-14 19:21:46 well i do tend to write assembly every now and again, but thats for work 2015-09-14 19:22:19 and i try to stay as far away from shell as i can, half the reason for this port is so i can make fully static ghc binaries >.< 2015-09-14 19:22:42 also need to port upx as well 2015-09-14 19:22:48 my todo list is long 2015-09-14 19:49:56 go 1.5 broke docker 2015-09-14 19:50:12 feature 2015-09-14 19:50:41 beautiful 2015-09-14 19:50:45 segfaults in musl's dynlinker handling relocations 2015-09-14 19:51:04 alpine 3.2.3, broken docker included 2015-09-14 19:51:10 think I'll temporarily build it with go-bootstrap 2015-09-14 19:51:18 oh wait, it's edge 2015-09-14 19:51:20 no drama 2015-09-14 19:51:24 yeah 2015-09-14 19:51:29 uggedal: yeah, go with go-bootstrap 2015-09-14 19:51:37 I only tested the build with starting the daemon 2015-09-14 19:51:45 it's usually enough 2015-09-14 19:51:48 yep 2015-09-14 19:51:54 but go is too weird 2015-09-14 19:51:55 do something simple with client. maybe 2015-09-14 19:52:07 yeah, client is what's segfaulting 2015-09-14 19:52:08 1.4 also broke docker… 2015-09-14 19:52:53 maybe we have some other go based packages which breaks 2015-09-14 19:53:26 well, it's usually docker that breaks 2015-09-14 19:53:29 at least in Arch 2015-09-14 19:53:53 I think we have consul packaged, maybe it's complicated enough to break as well 2015-09-14 19:56:27 hub as well 2015-09-14 19:56:31 will do some tests 2015-09-14 20:11:11 <_mjones> Oh that's right, consul is in Go. 2015-09-14 20:11:26 <_mjones> libc never breaks my apps. 2015-09-14 20:11:43 <_mjones> Well, never accidentally. Pretty much never. 2015-09-14 20:19:40 uggedal: you should create a post-commit hook validating your commit messages :P 2015-09-14 20:20:17 damn 2015-09-14 20:20:44 just need to retrain my muscle memory... 2015-09-14 20:24:10 barthalion: co ~/src/aports git ci -m foo 2015-09-14 20:24:12 Missing main/testing prefix 2015-09-14 20:24:22 there :) 2015-09-14 20:39:35 consul fails to build with go 1.5 due to textrels as well. have not tested if it segfaults yet 2015-09-14 20:40:29 hub builds without textrels 2015-09-14 20:42:30 syncthing build without textrels 2015-09-14 20:45:58 godep, websocketd build without textrels as well 2015-09-14 20:46:05 gogs fails to build 2015-09-14 20:52:19 readelf -r $(which gogs) | wc -l 2015-09-14 20:52:23 281293 2015-09-14 23:09:14 so the problem seems to be docker binaries using an external linker (docker links to sqlite etc) 2015-09-14 23:10:56 docker built with go tip don't have those massive amounts of textrels 2015-09-14 23:12:00 I'll try to backport a fix to go 1.5 2015-09-15 13:57:31 hey friends :) 2015-09-15 14:21:52 ncopa: what's alpines policy on PAM, should packages be compiled with or without PAM supports 2015-09-15 14:22:01 s/supports/support?/ 2015-09-15 14:22:16 we normally don't compile with PAM support 2015-09-15 14:22:32 however, if someone asks for it we enable it 2015-09-15 14:22:52 i think someone specifically asked for pam support in ngircd 2015-09-15 14:22:58 so i enabled it 2015-09-15 15:07:15 ncopa: do you remember the options needed at kernel build time for alpine linux to run as a kvm guest? 2015-09-15 15:18:24 ScrumpyJack: linux-virt-grsec? or sth like this 2015-09-15 15:20:04 <_mjones> OpenSSH 7.0 shipped. Interesting. 2015-09-15 15:25:43 ncopa: why not disable it unconditionally? 2015-09-15 15:43:40 _mjones: did anything in particular happen 2015-09-15 16:07:46 <_mjones> ammunta: OpenSSH 7 drops some deprecated crypto and by default compiles only with SSH2 protocol, but I'm more concerned with the security fixes in it. 2015-09-15 16:08:11 <_mjones> I want to know if there are mainline backports of the fixes. Need to finish reading the announcement. 2015-09-15 16:29:40 barthalion: virtio maybe? 2015-09-15 17:20:51 ScrumpyJack: ah, you mean literal drivers 2015-09-15 17:20:55 yeah, virtio should do 2015-09-15 18:47:15 what's the correct identifier for the GPLv2? Is it GPL2 or GPL-2? 2015-09-15 18:51:21 GPL2 520 vs GPL-2 144 2015-09-15 18:52:11 How about GPLv2? 2015-09-15 18:53:46 224 2015-09-15 18:55:41 I prefer GPL2, and I think this is what we do for some time 2015-09-15 18:55:58 but it's really minor thing to complain about 2015-09-15 18:56:34 Also documented here: http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package#license 2015-09-15 19:06:44 barthalion: I am not complaining, I am just asking questions… 2015-09-15 19:12:09 nmeum: I didn't mean you, I meant people with push access 2015-09-15 20:38:11 is the license actually used for anything other than display 2015-09-15 20:48:52 no, but I dislike the fact that we have totally unconsistent license names. It makes it harder to automatically detect the license of a package 2015-09-15 20:53:29 hm do /etc/logrotate.d/acpid and /etc/udhcpd.conf really belong in with busybox 2015-09-16 03:22:32 what is gliderlabs relasonship (if any) to alpine http://alpine.gliderlabs.com/alpine/edge/main ? as soon upon https://github.com/neeravkumar/dockerfiles/blob/master/alpine-openrc/Dockerfile 2015-09-16 03:34:59 hendry: not sure what you're trying to accomplish 2015-09-16 03:35:15 a real init system does really poorly being dockerized 2015-09-16 03:43:28 well, yes, I want to use openrc instead of supervisord 2015-09-16 04:03:37 ammunta: or do you have any other suggestions ? 2015-09-16 04:28:25 hendry: openrc is fine 2015-09-16 04:28:33 it's init that's problematic 2015-09-16 04:30:48 I assumed /sbin/init was openrc .. ? 2015-09-16 04:31:23 ammunta: at this present time, how do you recommend running PHP in Alpine in a container? i.e. maintain nginx + php-fpm (I guess) 2015-09-16 04:32:05 personally? runit 2015-09-16 04:32:30 do you have an example incidentally? 2015-09-16 04:32:53 keep getting my heart broken looking at Dockerfiles on docker hub at github search. 2015-09-16 04:33:49 to enable testing, I guess ppl just echo '@testing http://nl.alpinelinux.org/alpine/edge/testing' >> /etc/apk/repositories and go from there? 2015-09-16 04:34:01 (i need php-rrd) 2015-09-16 04:36:31 https://gist.github.com/muhmuhten/986ed9bff2bb1524591b uh you can try looking at this ... it's from a few months ago and the quality is p mediocre but 2015-09-16 04:39:55 uh, this was not a good project for shell scripting 2015-09-16 04:45:00 hhmm what's /etc/service look like? I wonder how https://github.com/kaihendry/camftp2web/tree/master/supervisor.d for e.g. gets replaced 2015-09-16 04:46:13 need to see whole project, not just gist =) 2015-09-16 04:48:04 in alpine can I pin to a particular version of a package? i did notice http://wiki.alpinelinux.org/wiki/Alpine_Linux_package_management#Repository_pinning which is cool 2015-09-16 04:49:43 hendry: /etc/service has a bunch of subdirs, one for each service 2015-09-16 04:49:54 each of them usually has a script "run" 2015-09-16 04:50:05 it doesn't have to be a script though, it can just be an executable 2015-09-16 04:50:23 anyway, you have that exec into what you're trying to run 2015-09-16 04:50:29 there should be docs 2015-09-16 07:21:32 hendry: openrc does not provide /sbin/init, busybox does 2015-09-16 07:22:16 and using openrc in a docker container is a terrible idea imho :) 2015-09-16 08:04:49 uggedal: openrc works fine under docker tbh, there just isn't much it can really do 2015-09-16 09:10:15 current main/pax-utils release is at v1.1.2 2015-09-16 09:58:22 ammunta: so you're basically running entire init inside container? sounds great /s 2015-09-16 10:07:48 hey friends :) 2015-09-16 10:07:58 guten tag 2015-09-16 10:08:04 :) 2015-09-16 10:08:17 today is alpine laptop cleanup day 2015-09-16 10:08:30 there are to many things not working right now and it gets anoying over time *g* 2015-09-16 10:08:33 http://www.phoronix.com/scan.php?page=news_item&px=System-XVI-Service-Manager 2015-09-16 10:08:38 maybe interesting for alpine 2015-09-16 10:09:01 hybernate on encrypted drives, suspend on encrypted drives, wifi, decrypt of drives during boot 2015-09-16 11:07:21 abuild will now try to validate maintainers email address. This means every maintainer should add a correct working address we can contact him/her by. 2015-09-16 12:00:16 clandmeter: nice, but you regex is totally incomplete. Without having tested it one example is that email adresses containing a '+' are considered invalid by this 2015-09-16 12:02:12 validating email adresses is hard and your approach only covers the most basic cases 2015-09-16 12:05:55 uggedal: do you use runit in containers? 2015-09-16 12:06:10 hendry: I don't use containers :) 2015-09-16 12:23:53 anyone is experiencing issue with alpine 3.2 and Xorg? 2015-09-16 12:24:24 got error in relocating $driver.so 2015-09-16 12:24:35 where $driver is vesa, or fbdev, or nouveau 2015-09-16 12:25:29 vesa returns error on VBEExtendedInit: sybmol not found 2015-09-16 12:25:42 fbdev: fbdevHWProbe: symbol not found 2015-09-16 12:26:18 nouveau: glamor_set_screen_pixmap 2015-09-16 12:26:44 fcolista-weechat: :( nope 2015-09-16 12:27:17 got this issue with xserver-org @edge also 2015-09-16 12:27:25 and with mesa@edge 2015-09-16 12:28:26 ldd /usr/lib/xorg/modules/drivers/$driver.so also returns symbols not found 2015-09-16 12:28:57 looks like a compilation issue 2015-09-16 12:29:59 quite possibly 2015-09-16 12:30:32 rebuild against newer libs perhaps? 2015-09-16 12:31:00 smells like that 2015-09-16 13:24:31 ncopa, probably you already noticed that...but might be useful imho having a parser for distrowatch "latest packages" RSS, in order to know which packages have been updated. 2015-09-16 13:24:31 http://distrowatch.com/news/dwp.xml 2015-09-16 13:48:47 cool 2015-09-16 13:49:01 i tend to look at hte distrowathc list manually 2015-09-16 13:49:06 but this is a good idea 2015-09-16 13:49:17 #!/usr/bin/env python 2015-09-16 13:49:17 import feedparser 2015-09-16 13:49:17 p = feedparser.parse('http://distrowatch.com/news/dwp.xml') 2015-09-16 13:49:17 print (p['feed']['link']) 2015-09-16 13:49:17 print (p['feed']['title']) 2015-09-16 13:49:18 print (p['feed']['description']) 2015-09-16 13:49:20 print (len(p['entries'])) 2015-09-16 13:49:22 for post in p.entries: 2015-09-16 13:49:24 print (post.title + ": " + "\r") 2015-09-16 13:49:27 just playing with python feedparser 2015-09-16 13:49:36 there is a lua feedparser too 2015-09-16 13:49:40 yeah 2015-09-16 13:49:41 that i use i alpine-mksite 2015-09-16 13:50:02 ok cool...i'm in a temporary workstation where lua-feedparser is not available .) 2015-09-16 13:50:03 :) 2015-09-16 13:50:48 I need to have xorg working on alpine 3.2 or edge 2015-09-16 13:51:25 i've arch but i want to use a distribution without systemd. 2015-09-16 13:51:50 looking around, besides void linux, there's nothing that is appealing to me 2015-09-16 13:51:58 (besides alpine, obiouvsly :) 2015-09-16 14:02:24 fcolista: x86 or x86_64? 2015-09-16 14:02:33 you might need set up your xorg.conf 2015-09-16 14:02:38 x86_64 2015-09-16 14:02:50 got symbol not found 2015-09-16 14:02:57 edge or v3.2? 2015-09-16 14:02:59 both 2015-09-16 14:03:08 what xorg driver? 2015-09-16 14:03:23 got this problems with fbdev, vesa and nouveau 2015-09-16 14:03:37 all of them returned error of symbol not found (relocation problem) 2015-09-16 14:03:47 nouveau-drv.so, fbdev.so etc 2015-09-16 14:03:58 nouveau: glamor_set_screen_pixmap 2015-09-16 14:04:03 exactly 2015-09-16 14:04:48 i wonder where it is defined 2015-09-16 14:05:10 dunno. 2015-09-16 14:05:49 https://dpaste.de/cS16 2015-09-16 14:06:00 those were the error gotten 2015-09-16 14:06:28 i found it: /usr/lib/xorg/modules/libglamoregl.so:glamor_set_screen_pixmap 2015-09-16 14:06:49 who owns libglamoregl.so ? 2015-09-16 14:07:29 was someone having problems with lxc-create and backing stores? 2015-09-16 14:07:42 fcolista: xorg-server 2015-09-16 14:07:48 i think it just need to be loaded earlier 2015-09-16 14:07:58 i can't seem to create lxc contrainers with lvm backing store any more 2015-09-16 14:08:23 ncopa, you mean is a problem of loading sequence? 2015-09-16 14:08:38 thats what i think yes 2015-09-16 14:09:03 fcolista: can you paste your entire Xorg.0.log? 2015-09-16 14:09:15 sounds difficult ncopa 2015-09-16 14:09:29 but i can try 2015-09-16 14:09:36 give me 5 min 2015-09-16 14:09:55 gotta reboot this pc (i've alpine on another disk in the same pc) 2015-09-16 14:09:55 oh 2015-09-16 14:10:00 np 2015-09-16 14:10:09 5 minutes max and i'll be back 2015-09-16 14:10:09 do you have your conf 2015-09-16 14:10:13 xorg.conf 2015-09-16 14:10:19 no xorg.conf 2015-09-16 14:10:25 X -configure returns me error 2015-09-16 14:10:27 the /etc/X11/xorg.conf.d/20-modules.conf 2015-09-16 14:10:38 X -configure does not work 2015-09-16 14:10:46 oh 2015-09-16 14:11:10 check that glamoregl is listed in /etc/X11/xorg.conf.d/20-modules.conf 2015-09-16 14:11:13 i think it should be 2015-09-16 14:11:46 doing it right now 2015-09-16 14:11:46 https://dpaste.de/4jFk 2015-09-16 14:11:52 this is the xorg log 2015-09-16 14:12:13 Section "Module" 2015-09-16 14:12:13 Load "exa" 2015-09-16 14:12:13 Load "dri2" 2015-09-16 14:12:13 Load "wfb" 2015-09-16 14:12:13 Load "dri" 2015-09-16 14:12:14 Load "glx" 2015-09-16 14:12:16 Load "int10" 2015-09-16 14:12:18 Load "vbe" 2015-09-16 14:12:20 Load "fb" 2015-09-16 14:12:22 Load "shadowfb" 2015-09-16 14:12:24 Load "shadow" 2015-09-16 14:12:26 Load "vgahw" 2015-09-16 14:12:28 Load "evdev" 2015-09-16 14:12:32 Load "fbdevhw" 2015-09-16 14:12:34 Load "glamoregl" 2015-09-16 14:12:36 EndSection 2015-09-16 14:12:38 ~ 2015-09-16 14:12:40 this is the /etc/X11/xorg.conf.d/20-modules.conf file 2015-09-16 14:13:40 how do you start X? 2015-09-16 14:13:54 if i startx it hangs 2015-09-16 14:14:14 if i run only X, it does not hangs, but black screen returned 2015-09-16 14:14:36 can you log in remotely after startx? 2015-09-16 14:14:46 yes 2015-09-16 14:14:54 how does the Xorg.0.log look like after startx? 2015-09-16 14:14:56 hangs the keyboard so i cannot switch 2015-09-16 14:14:59 on terminal 2015-09-16 14:15:17 that could be because you have no input driver working 2015-09-16 14:15:19 oh 2015-09-16 14:15:22 one thing more 2015-09-16 14:15:23 i've evdev 2015-09-16 14:15:28 and mouse/keyboard 2015-09-16 14:15:35 try: apk fix kmod 2015-09-16 14:16:06 done 2015-09-16 14:16:18 i've mounted the disk and i'm in chroot 2015-09-16 14:16:21 check also that your /dev/input/* owrks 2015-09-16 14:16:27 so to verify i need to reboot 2015-09-16 14:17:08 anyway, one of the error (i'm going by memory) was related to missing /dev/dri/card0 2015-09-16 14:17:25 this appeared in xorg.log when i run only X 2015-09-16 14:17:55 maybe you need to modprobe radeon or similar? 2015-09-16 14:18:04 nvidia i mean 2015-09-16 14:18:12 modprobe nouveau 2015-09-16 14:18:18 and is loaded 2015-09-16 14:19:41 maybe you need to be in group 'video' 2015-09-16 14:19:53 i am 2015-09-16 14:20:18 in this log: https://dpaste.de/4jFk it does not look like it is loading 20-modules.conf at all 2015-09-16 14:20:25 which you dont if you do X -configure 2015-09-16 14:20:44 so i need to startx 2015-09-16 14:20:49 and check the log 2015-09-16 14:20:57 then paste it 2015-09-16 14:21:03 that would be nice yes 2015-09-16 14:21:06 ok 2015-09-16 14:21:09 give me 5 min :) 2015-09-16 14:23:54 rebooting :( 2015-09-16 14:29:54 [drm] KMS not enabled 2015-09-16 14:30:22 ncopa: http://sprunge.us/OjeQ 2015-09-16 14:30:34 this is the Xorg.log file 2015-09-16 14:31:00 ok this looks better 2015-09-16 14:31:11 should i use udev rather than mdev? 2015-09-16 14:31:17 first thing, it does load (II) Loading /usr/lib/xorg/modules/libglamoregl.so 2015-09-16 14:31:42 [ 183.311] (II) LoadModule: "nouveau" 2015-09-16 14:31:42 [ 183.311] (II) Loading /usr/lib/xorg/modules/drivers/nouveau_drv.so 2015-09-16 14:32:18 [ 183.320] (EE) [drm] KMS not enabled 2015-09-16 14:32:18 [ 183.320] (EE) open /dev/dri/card0: No such file or directory 2015-09-16 14:32:34 myabe you need unload the nouveou kernel module 2015-09-16 14:32:51 and then: modprobe nouveau modesetting=1 2015-09-16 14:33:14 done 2015-09-16 14:33:28 /dev/dri is empty 2015-09-16 14:37:53 modeset=1 2015-09-16 14:38:02 seems to be the correct option 2015-09-16 14:38:20 ok 2015-09-16 14:38:34 might want modprobe fbcon first 2015-09-16 14:40:08 ok 2015-09-16 14:40:17 ncopa: sprunge.us/EAVI 2015-09-16 14:40:21 this is dmesg 2015-09-16 14:40:35 /dev/dri still empty 2015-09-16 14:41:18 looks that this card is not supported 2015-09-16 14:41:30 this actually has kepler chipset 2015-09-16 14:43:35 grrr, updated and upgraded, still no lxc backing store :( 2015-09-16 14:44:18 well, lvm specifically 2015-09-16 14:44:56 ScrumpyJack: since when did that start to happen? 2015-09-16 14:45:13 i just pushed new lvm2 version 2015-09-16 14:45:16 might be regression 2015-09-16 14:46:32 wondering how this chipset is not supported, if arch has 2015-09-16 14:46:57 ACTION blushes 2015-09-16 14:47:21 i think it's my problem. lxc has changed the way/order it handles options i think 2015-09-16 14:47:39 thats typical :) 2015-09-16 14:52:22 back 2015-09-16 15:02:10 ncopa: http://comments.gmane.org/gmane.comp.freedesktop.xorg.nouveau/19151 2015-09-16 15:02:16 gotta wait for kernel 3.19 2015-09-16 15:02:46 taht very kind of chipset is not supported (even though keperl is supported, according with http://nouveau.freedesktop.org/wiki/CodeNames/#nve0familykepler) 2015-09-16 15:29:03 what is the alpine commit channel name? 2015-09-16 15:38:07 alpine-commits 2015-09-16 15:47:42 forgot the s 2015-09-16 16:57:43 barthalion: I don't use openrc inside containers except when I'm just throwing up one for testing and then I need like, an sshd or something 2015-09-16 16:58:25 half the default initscripts are broken anyway and anyway it's not very useful 2015-09-16 16:58:59 with runit I don't actualy run the init system, I just root it on a runsv 2015-09-16 17:33:26 hm does apk actually have a mechanism for clearing the package cache or is rm -f /var/cache/apk/* just 2015-09-16 17:36:18 ammunta: I was always doing the rm way 2015-09-16 22:06:55 https://bpaste.net/show/f6c641d9283a uhh this seems a bit like a bug 2015-09-16 22:11:09 ncopa: ↑ 2015-09-17 01:01:33 indeed, i think it is evaluating dependencies 2015-09-17 06:01:00 morning 2015-09-17 06:01:07 i dunno if its a feature or bug 2015-09-17 06:01:46 but smells like bug indeed 2015-09-17 06:51:24 hey :) 2015-09-17 07:07:43 hey 2015-09-17 07:08:09 fabled: what is the arm build hardware? 2015-09-17 07:08:23 ScrumpyJack, wandboard quad 2015-09-17 07:13:16 fabled: would these help? https://www.scaleway.com/ 2015-09-17 07:13:44 bare metal quad core ARM A7s 2015-09-17 07:13:51 (hosted) 2015-09-17 07:15:13 dl-7.alpinelinux.org seams to be down 2015-09-17 07:15:32 ScrumpyJack, possibly 2015-09-17 07:16:22 ScrumpyJack: fine print the storage is some nbd abomination, but then again compiling is usually cpu-bound right 2015-09-17 07:23:24 ncopa, %572 can be set as "accepted"? Since you applied but with some small changes 2015-09-17 07:28:19 ha 2015-09-17 07:28:30 patchwork think that i sent that patch :) 2015-09-17 07:28:31 ammunta: yeah. a little beowulf arm cluster would be cool. 2015-09-17 07:29:25 yes ncopa :) 2015-09-17 07:36:22 ScrumpyJack: scaleway is tbh p cheap as those sorts of things go but 2015-09-17 07:39:34 ncopa, i'm going to apply also %581 2015-09-17 07:39:56 do you have something in contrary? 2015-09-17 07:40:39 ammunta: but? 2015-09-17 07:41:12 fcolista: nah, do it, it makes no sense to have it absent 2015-09-17 07:41:35 i agree baetheus 2015-09-17 07:41:36 barthalion, 2015-09-17 07:41:42 thx 2015-09-17 11:08:03 running openntpd it get ntpd: unknown user _ntp 2015-09-17 11:15:27 ugh 2015-09-17 11:15:33 there was some patches about that recently 2015-09-17 11:29:35 lunch 2015-09-17 11:34:56 openbsd switched all their deamon users to have _ prefix if I recall correctly 2015-09-17 11:47:03 maybe something for us to follow? 2015-09-17 11:47:30 might break stuff 2015-09-17 11:47:38 probably not worth it 2015-09-17 11:56:03 no, but I like the idea 2015-09-17 11:56:11 me too 2015-09-17 11:56:31 so, ncurses 6.0... 2015-09-17 11:56:58 as I understand it you can either compile it with ABI compat with previous versions or compile it with new ABI 2015-09-17 11:58:47 we will compile it with new ABI 2015-09-17 11:59:08 no reason to keep the old one 2015-09-17 11:59:46 $ apk search --exact -r so:libncurses.so. 2015-09-17 11:59:46 5 --origin -q | sort -u|wc -l 2015-09-17 11:59:46 104 2015-09-17 12:00:40 there are 100+ packages which will need rebuild 2015-09-17 12:05:17 why are we using the "unstable" gnutls branch btw? 2015-09-17 12:05:26 dunno 2015-09-17 12:05:32 its latest and greatest i suppose 2015-09-17 12:05:51 i was probably not aware that it is unstable 2015-09-17 12:06:09 they break abi/api in this branch? 2015-09-17 12:06:48 not sure, see that Arch also uses "next stable" 2015-09-17 12:09:13 why do we like users with _ prefixed for daemons? 2015-09-17 12:09:40 makes it visible want is system user/daemon and what is normal user 2015-09-17 12:11:51 i've never found telling the difference to be a problem 2015-09-17 12:12:13 if *is* a differance 2015-09-17 12:12:35 if ^there^ is ... 2015-09-17 12:14:13 anyway, i should be able to run anything as any user. i should be able to run ntpd as the user drwho if i wanted 2015-09-17 12:14:29 (see what i did there?) :) 2015-09-17 12:14:57 *shrug* 2015-09-17 12:15:20 i guess some people like usernames start with capital letter too 2015-09-17 12:15:34 lowercase is not a problem 2015-09-17 12:15:51 but some might think uppercase looks better 2015-09-17 12:16:13 really? 2015-09-17 12:17:02 what i like with _ prefix is that you can very simple get a list of user accounts 2015-09-17 12:17:20 and you can get an overview which might need ssh login for exaple 2015-09-17 12:17:40 and which accounts you can safely disable ssh for 2015-09-17 12:17:42 etc 2015-09-17 12:18:02 i liked the idea with _ prefix 2015-09-17 12:18:29 but as you say 2015-09-17 12:18:33 its not a problem 2015-09-17 12:20:02 "you can very simple get a list" no you can't. there are hundreds and thousands of items of software out there that run as different users. not a single one i know creates a user prefixed with a _ 2015-09-17 12:22:46 apparently they do on openbsd 2015-09-17 12:24:22 when building the open****d software perhaps. but apache still runs as whoever i want in linux 2015-09-17 12:24:32 they shouldn't impose 2015-09-17 12:25:03 anyway. openntpd is broken. If i create the user _ntp, i then get 2015-09-17 12:25:05 Home directory of ntp needs to be /var/empty 2015-09-17 12:25:12 note, ntp without _ 2015-09-17 12:25:13 yeah 2015-09-17 12:25:18 openntpd is broken 2015-09-17 12:25:22 Please run 'usermod -d /var/empty ntp' 2015-09-17 12:25:31 again, ntp without _ 2015-09-17 12:25:50 I'm losing patienceL apk del openntpd 2015-09-17 12:27:02 i htink this broke it: http://git.alpinelinux.org/cgit/aports/commit/main/openntpd?id=8a645d5f164a82967591eb54b922fab240945f40 2015-09-17 12:27:13 the openbsd discussion centered around collisions 2015-09-17 12:28:33 makes sense 2015-09-17 12:31:30 ScrumpyJack: i pushed openntpd fix 2015-09-17 12:32:02 cool 2015-09-17 12:32:24 uggedal: users bumping into one another? how careless ;) 2015-09-17 12:32:39 (collisions) 2015-09-17 12:34:05 shame with-privsep-user isn't a runtime setting 2015-09-17 12:34:10 re ncurses 2015-09-17 12:34:22 we are not afraid of upgrade abi 2015-09-17 12:34:47 and it would probably be nice to upgrade it before v3.3 2015-09-17 12:34:55 so now is a good time to do it 2015-09-17 12:35:10 do we have scripts/helpers for major soname revbumps? 2015-09-17 12:35:38 i have some local stuff 2015-09-17 12:36:30 http://tpaste.us/GdE5 2015-09-17 12:37:07 i run that with: committhatbokeabi^ "rebuild against libfoo-6" 2015-09-17 12:37:24 then i paste in the list of files 2015-09-17 12:37:32 list of parots 2015-09-17 12:37:34 aports 2015-09-17 12:37:47 aports affected 2015-09-17 12:38:26 i find this list with something like: apk search --exact -r so:libncurses.so.5 --origin -q | sort -u 2015-09-17 12:39:21 since there are so many in this case, we might want provide a migration compat package 2015-09-17 12:39:38 rename ncurses to ncurses5 2015-09-17 12:39:56 and upgrade ncurses to 6.0 2015-09-17 12:40:13 also remove the /usr/bin/* from the compat package 2015-09-17 12:40:31 then both can be installed in parallel 2015-09-17 12:40:50 and packages will use new ncurses abi when rebuilt 2015-09-17 12:41:12 when all are rebuilt, then we can remove ncurses5 2015-09-17 12:41:35 that way we dont need rebuild all the 100+ packages in same git push 2015-09-17 12:46:15 sounds good 2015-09-17 19:11:11 hey friends 2015-09-17 19:11:18 something in xfce is broken since today 2015-09-17 19:11:34 all my xfce4 settings are gone and all icons are missing 2015-09-17 19:11:38 can someone explain this? 2015-09-17 19:12:46 hmm 2015-09-17 19:12:47 grsec: Segmentation fault occurred at (nil) in /usr/bin/xfsettingsd[xfsettingsd:2151] uid/euid:1000/1000 gid/egid:1000/1000, parent /bin/busybox[init:1] uid/euid:0/0 gid/egid:0/0 2015-09-17 19:12:51 since the latest update 2015-09-17 19:17:35 could someone please trigger a rebuild of that package so it would work again? 2015-09-17 19:19:27 well, it's a segfault, so rebuild probably won't fix it 2015-09-17 19:24:41 barthalion: its build against an old version 2015-09-17 19:37:36 leo-unglaub: of what? I don't have an env at the moment 2015-09-17 19:54:41 leo-unglaub: it was rebuilt yesterday against upower 2015-09-17 19:55:29 uggedal: since that update its broken appearently 2015-09-17 19:59:48 leo-unglaub: do you have dbus running? 2015-09-17 20:00:10 leo-unglaub: and upower? 2015-09-17 20:00:44 dbus sadly yes, upower no 2015-09-17 20:00:59 leo-unglaub: https://bugzilla.xfce.org/show_bug.cgi?id=11773 2015-09-17 20:04:53 hmmmm 2015-09-17 20:05:06 is it just me or does linux get more and more bloaded? 2015-09-17 20:05:21 dbus and now upower just to store some settings? 2015-09-17 20:05:27 that seams like fucking overkill 2015-09-17 20:06:18 leo-unglaub: linux != xfce/gnome/kde/... :P 2015-09-17 20:06:31 true, but i see a trend there 2015-09-17 20:07:02 do it still segfault with upower running? 2015-09-17 20:07:14 upower is not starting :( 2015-09-17 20:07:34 well, I don't use dbus/upower so I'm in the dark here 2015-09-17 20:07:46 what desktop environment to you use? 2015-09-17 20:10:43 none, I use herbstluftwm 2015-09-17 20:12:21 nice, thats basically like i3wm, right? 2015-09-17 20:13:58 just infinitely more customizable 2015-09-17 20:16:24 I use herbstluftwm configured to act like: http://aerosuidae.net/xoat 2015-09-17 20:17:32 that sounds interresting 2015-09-17 20:17:46 now i feel like a noob with my xfce/i3wm *g* 2015-09-17 20:19:49 sounds good 2015-09-17 20:20:18 given how I need floating windows 90% of time and only sometimes two terminals next to each other 2015-09-17 22:09:22 looks nice (I'm a awesome user) 2015-09-18 08:23:02 ncopa: what is the diff between the two columns @ http://build.alpinelinux.org 2015-09-18 08:24:01 aport total is for bootstrapping 2015-09-18 08:24:18 how many packages are totally built in repo 2015-09-18 08:24:29 there is a feature to not halt on errors 2015-09-18 08:24:56 so it means 98% of all aports are build currenly? 2015-09-18 08:25:21 it means that 98% of aports builds 2015-09-18 08:25:30 which has armhf enabled 2015-09-18 08:25:37 it means that there are a few that fails to build 2015-09-18 08:25:50 oh 2015-09-18 08:26:21 and aport build? is that the current package, or t he current outsranding packages? 2015-09-18 08:27:05 that is the progress of the current run 2015-09-18 08:27:22 the packages it will try to build for this run 2015-09-18 08:27:47 if it builds in "continue on errors" it will not try rebuild it til the APKBUILD changes 2015-09-18 08:28:12 it will also not try build packages which depends on packages that previously failed 2015-09-18 08:28:36 "aport built" will tell you the progress of the packages it wil actually try to build 2015-09-18 08:28:49 in this run 2015-09-18 08:29:03 if you git push 5 packages while it is building 2015-09-18 08:29:15 then it will restart when it is done with that build 2015-09-18 08:29:23 and try build those 5 2015-09-18 08:29:47 the aport total column should probably be removed 2015-09-18 08:30:18 and we should probably have some other way to display the statistics 2015-09-18 08:32:47 maybe call the the column different, something like aports health status :) 2015-09-18 08:37:51 fabled: is there a logfile for arm builder? 2015-09-18 08:45:58 morning folks. could someone rebuild gnutls-utils? it's broken in edge 2015-09-18 08:47:44 for the error: ldd /usr/bin/certtool 2015-09-18 08:48:20 ScrumpyJack: I might have broken it, so will look at it 2015-09-18 08:49:39 ScrumpyJack: what is the error? 2015-09-18 08:49:43 libgnutls.so.30 => /usr/lib/libgnutls.so.30 2015-09-18 08:49:50 same soname 2015-09-18 08:51:11 Error relocating /usr/bin/certtool: gnutls_pkcs7_verify_direct: symbol not found 2015-09-18 08:51:28 and then some more 2015-09-18 08:52:28 ScrumpyJack: can you give me the exact command you're using? 2015-09-18 08:52:33 ncopa: thanks for the ntp patch yesterday. apk add openntpd still gives me the same error. how does apk tell if the package has been updated? 2015-09-18 08:53:04 apk version openntpd 2015-09-18 08:53:14 r0 -> r1? 2015-09-18 08:53:49 yes 2015-09-18 08:54:25 ScrumpyJack: no relocation errors in certtool here 2015-09-18 08:54:39 apk version gnutls ? 2015-09-18 08:55:21 clandmeter, arm builder logs are not available on public ip currently; ncopa is working on having the builders upload their logs 2015-09-18 08:55:34 uggedal: 3.4.5-r0 2015-09-18 08:55:36 i have an idea how to do it 2015-09-18 08:55:44 just dont have time to complete that project 2015-09-18 08:57:31 :) 2015-09-18 08:57:41 story of our life 2015-09-18 08:57:50 ScrumpyJack: strange, works here 2015-09-18 08:58:08 ScrumpyJack: do you mix stable repo with edge? 2015-09-18 08:58:33 ScrumpyJack: ah, apk info gnutls-tools ? 2015-09-18 08:58:34 ncopa: yes, i think that's the problem 2015-09-18 08:58:46 for openntpd i stii get http://tpaste.us/24E1 2015-09-18 08:58:47 s/info/version 2015-09-18 08:59:04 mixing the repos will not work due to gcc libstdc++ ABI breakage 2015-09-18 09:01:08 i pushed libtirpc update 2015-09-18 09:01:19 the reverted the auth_des stuff 2015-09-18 09:01:30 for gnutls, i have this http://tpaste.us/3Mbk 2015-09-18 09:01:44 however, i made it build, and nfs-utils builds against it 2015-09-18 09:01:47 i think it is good 2015-09-18 09:01:59 would be nice if someone could test if nfs still works 2015-09-18 09:02:12 ncopa: for openntpd box, it's edge, not repo mix 2015-09-18 09:02:48 openntpd r1 gives me this http://tpaste.us/24E1 2015-09-18 09:02:49 ScrumpyJack: yeah, that will not work :) 2015-09-18 09:03:23 uggedal: i should update gnutls to edge then 2015-09-18 09:04:44 ScrumpyJack: try apk del openntpd, then remove ntp user and group from system (remove lines from /etc/passwd, /etc/group and /etc/shadow) 2015-09-18 09:04:49 and then try apk add it again 2015-09-18 09:04:53 :) ok 2015-09-18 09:05:02 alternatively you could try do what error message suggest 2015-09-18 09:05:13 usermod -d /var/empty ntp 2015-09-18 09:06:07 deleting the user by hand worked 2015-09-18 09:09:48 arg, crashed on one box 2015-09-18 09:14:27 woah! now my lighttpd is running as ntp! 2015-09-18 09:15:31 ok, it's lighttpd in a LXC. the host has mapped the new ntp user to that uid lol 2015-09-18 09:21:56 hum 2015-09-18 09:22:06 any zfs users here? 2015-09-18 09:22:22 it seems like they have removed the gentoo/openrc specific init script 2015-09-18 09:22:26 openrc thinks openntpd-5.7_p4-r1 has crashed after service start, even though ntp is running 2015-09-18 09:22:44 does it create pid file? 2015-09-18 09:23:01 probably not :) 2015-09-18 09:27:15 i don't know much about openrc. does it create a pid file in /var/run for each service? 2015-09-18 10:06:18 i pushed a zfs update 2015-09-18 10:06:28 it would be nice if someone could test that the init.d scripts works as expected 2015-09-18 10:06:39 earlier there was /etc/init.d/zfs 2015-09-18 10:06:47 now there are 4 differenet inits scripts 2015-09-18 10:07:01 apparently, 2 needs to be in boot runlevel and 2 in default runlevel 2015-09-18 10:07:54 does alpine support ZFS on root? 2015-09-18 10:31:22 taskd also seems broken. that is from edge: http://tpaste.us/GEOJ 2015-09-18 10:40:48 ScrumpyJack: no, works fine on clean edge env 2015-09-18 10:41:17 ScrumpyJack: what you're seeing is the libstdc++ ABI bump with gcc 5.x 2015-09-18 10:43:24 ScrumpyJack: apk version libstdc++ 2015-09-18 10:43:26 libstdc++-5.2.0-r0 = 5.2.0-r0 2015-09-18 10:54:47 uggedal: perhaps we can move taskd to main then? 2015-09-18 10:55:11 hmm, that won't fix it. i'm on 3.2 2015-09-18 10:56:14 ok, perhaps i should bit the bullet and put my really really important box that needs to be up all the time on edge :) 2015-09-18 10:59:23 ok, so i am on 3.2 and i have edge/main and edge/testing pinned. If i want to unpin edge and stay on 3.2, how can i remove and (downgrade where applicable) all the stuff i got from edge? 2015-09-18 11:24:53 remove it from repositories and apk upgrade -U -a 2015-09-18 12:03:00 ok, here goes :) 2015-09-18 12:03:53 i get a lot of there: WARNING: The repository tag for world dependency 'davfs2-doc@testing' does not exist 2015-09-18 12:06:35 ERROR: Not continuing with upgrade due to missing repository tags. Use --force to override. 2015-09-18 12:27:33 we dont ship testing repos with stable releases 2015-09-18 12:46:38 ncopa: can you please look at http://bugs.alpinelinux.org/issues/4645 2015-09-18 12:46:50 not sure how to handle the versioning of lua-json4 2015-09-18 12:48:09 cant you just update it to 1.0.0? 2015-09-18 12:48:35 using the github tarball 2015-09-18 12:50:11 and change the alpine version to what? 2015-09-18 12:50:17 (it's already 1.0.0) 2015-09-18 12:51:00 1.0.0.1? 2015-09-18 12:56:12 should i --force the upgrade? 2015-09-18 13:11:35 or should i apk del each package that is in testing 2015-09-21 07:43:29 morniong 2015-09-21 07:43:36 we need be careful with golang apps 2015-09-21 07:43:47 go is not maintained for long time 2015-09-21 07:44:09 so we cannot provide support for those for 2 years 2015-09-21 07:44:29 so anything golang can not go to "main" repo 2015-09-21 07:51:02 ncopa do we have "community" ? 2015-09-21 07:51:27 crow: not yet 2015-09-21 07:51:37 but i expect it to show up soonish 2015-09-21 07:51:41 before v3.3 2015-09-21 07:51:46 we need it before v3.3 2015-09-21 07:51:58 so hopefully within next 2 weeks or so 2015-09-21 07:52:09 well if we have community then these packages should be moved there. 2015-09-21 07:52:14 yup 2015-09-21 07:52:17 thats the idea 2015-09-21 07:53:23 why going with golang if not maintained? 2015-09-21 08:27:27 anyone experinced issues with creating x86 lxc containers? all triggers seems to fail when creating the rootfs (ERROR: busybox-1.23.2-r9.trigger: script exited with error 1) 2015-09-21 08:28:44 uggedal, possible happens if /proc is not mounted on target container when running apk 2015-09-21 08:29:02 musl needs /proc for certain operations glibc/uclibc did not 2015-09-21 08:29:09 e.g. realpath() 2015-09-21 08:29:42 works for x86_64 though 2015-09-21 08:30:35 and /proc is mounted 2015-09-21 08:30:51 is host and container same arch ? 2015-09-21 08:31:15 ncopa: those go CVE's needs rebuilds of go packages :( 2015-09-21 08:32:03 they do? 2015-09-21 08:32:04 arch 2015-09-21 08:32:07 argh! 2015-09-21 08:33:48 go only dynlinks against libc, the runtime and libs are statically compiled 2015-09-21 08:34:21 well, it only links against libc if it uses c code through cgo 2015-09-21 08:34:40 otherwise it handles syscalls directly 2015-09-21 08:39:34 iM not gonna maintain go 2015-09-21 08:40:20 go is not suitable for anything that is not under fastmoving development 2015-09-21 08:40:21 :) 2015-09-21 08:40:46 so if you are planning to do a longtime running stable service, do go for go ;) 2015-09-21 08:49:16 ncopa, what about %583 ? 2015-09-21 08:49:17 :) 2015-09-21 08:49:46 sure 2015-09-21 08:49:50 but dont move it to main 2015-09-21 08:50:07 and dont put me as maintainer 2015-09-21 08:50:19 lol, no, i'm not accostumed to do that :) 2015-09-21 08:51:14 sounds like a nightmare mantains this packages. 2015-09-21 08:51:40 It would end up in a bunch of old packages that nobody's want to deal with, imho 2015-09-21 08:52:07 so, community repo (like aur) would be a correct place where to move them 2015-09-21 08:53:32 I look forward to community/ :) 2015-09-21 08:55:37 oh 2015-09-21 08:55:44 looks that we don't have go for x86 2015-09-21 08:56:15 that might be resolved if we enable it, can't recall what the problem with x86 was 2015-09-21 09:00:53 probably pie or pic issue 2015-09-21 09:03:21 would community solve this? ls -1 testing | wc -l 2015-09-21 09:03:22 1508 2015-09-21 09:03:38 that is, could some packages from testing be moved to community? 2015-09-21 09:03:59 and the rest which were not touched recently be moved to unmaintained? 2015-09-21 09:09:23 yes 2015-09-21 09:09:31 we even talked about rename testing to "staging" 2015-09-21 09:09:51 to make clear that packages are not supposed to stay there 2015-09-21 09:10:01 and maybe autopurge after a certain time 2015-09-21 09:11:25 sounds good 2015-09-21 09:17:58 the main problem seems to be non-committers contributing to testing and never reporting back wether it works and can be moved to main (and community in the future) 2015-09-21 09:28:03 maybe we can add a TTL field to pkgs.a.o when packages are in a specific repo (like testing) 2015-09-21 09:42:27 i wonder if we shoudl just create the community repo now 2015-09-21 09:42:33 just add or move a package there 2015-09-21 09:42:38 and then confiugre the builders 2015-09-21 10:06:53 speaking for which, i've been using this lot for months now without issues http://tpaste.us/AvJW can they be moved to edge/main? 2015-09-21 10:12:48 ScrumpyJack: will you be the maintainer? 2015-09-21 10:24:41 yes, ok 2015-09-21 10:25:09 i have changed my email address so perhaps you could do a sed for me? :) 2015-09-21 11:24:37 ScrumpyJack: sure, pm me your maintainer line. 2015-09-21 11:29:59 hey 2015-09-21 11:53:14 hmm, how is the apk add used to create the lxc rootfs supposed to work? the busybox trigger for instance calls busybox --install 2015-09-21 11:53:28 when host is x86_64 and target is x86 2015-09-21 12:35:11 it does nto work? 2015-09-21 12:35:37 uggedal: lxc-create will call apk add --root $rootfs 2015-09-21 12:35:54 which will run the busybox --install trigger in chroot 2015-09-21 12:35:58 so it shoudl just work 2015-09-21 12:36:09 hum, is somebody interested in creating an package for https://bip.milkypond.org/? :) 2015-09-21 12:36:41 or is there already some kind of irc proxy/bouncer available on alpine? 2015-09-21 12:38:07 mosez: i have been looking for something similar myself, without not really finding anything 2015-09-21 12:38:11 bip look snice 2015-09-21 12:41:47 yeah i thought that as well :) 2015-09-21 12:48:18 mosez ncopa whats wrong with psyBNC or shrugBNC 2015-09-21 12:54:16 ncopa: ah, did not know --root runs in chroot 2015-09-21 13:25:18 hmm, I get: execve("var/cache/misc/busybox-1.23.2-r9.trigger", ["trigger", "/sbin", "/bin", "/usr/sbin", "/usr/bin"], [/* 1 var */]) = -1 ENOEXEC (Exec format error) 2015-09-21 13:25:33 with apk add -U --initdb --root /root/test --repository http://nl. 2015-09-21 13:25:38 alpinelinux.org/alpine/edge/main --allow-untrusted --arch x86 alpine-base 2015-09-21 13:36:35 crow: https://pkgs.alpinelinux.org/packages?name=%25bnc%25&repo=all&arch=x86_64&maintainer=all it just doesnt exist as well :) 2015-09-21 13:38:29 last update 2013: http://sourceforge.net/projects/psybnc/ and last update jan 2015 http://www.shroudbnc.info/redmine/projects/shroudbnc/repository 2015-09-21 14:00:49 ah, linux-virtgrsec has CONFIG_IA32_EMULATION disabled 2015-09-21 16:12:57 mosez ok :) well if ncopa is interested in bip soon we could see that APKBUILD :)= 2015-09-21 18:52:08 good evening 2015-09-21 18:52:10 :) 2015-09-21 18:52:32 i think i tracked down the xfsettings bug from a few days ago 2015-09-21 18:52:48 dbus is crashing and after that xfconfig is gone as well 2015-09-21 19:04:35 is anybody able to build main/openssh? CVE-2015-5600.patch doesn't apply here 2015-09-21 19:55:39 nmeum: applies and builds here 2015-09-21 20:31:49 after installing grsec latest iso and rebooting my network is not detected http://sprunge.us/Yejc eth0 ...ip: ioctl 0x8913 failed: No such device . i was able to use iso and install alpine to disk... weird . maybe because of this: * Loading modules ...modprobe: can't change directory to '3.18.20': No such file or directory ? 2015-09-21 20:41:30 i have only /lib/modules/3.18.21-0-grsec (just normal installation to disk) using old stuff written in this wiki http://wiki.alpinelinux.org/wiki/Installing_on_GPT_LVM just without creating vg0 but deleting old lv and creating new.. 2015-09-21 21:31:27 also trying to rebuild initramfs with mkinitfs result in an error http://sprunge.us/cLFH 2015-09-22 06:42:13 morning, 2015-09-22 06:53:05 morning 2015-09-22 06:57:17 ScrumpyJack you have had some problems with modules few days ago to? but for you it was edge and lxc right? (i am not sure if you see last chats from yesterday here on channel) 2015-09-22 06:58:29 just reading 2015-09-22 07:03:09 ok 2015-09-22 07:03:27 i am not sure what i could do wrong. 2015-09-22 07:04:09 iso sha256sum was correct when i copied it to usb, installer usb worked to, network was up and running. but after reboot to new system i get that msg 2015-09-22 07:04:38 hmm, wierd 2015-09-22 07:05:11 to temp fix it you could create a symlink from 3.18.20 to 3.18.21 2015-09-22 07:05:39 but that's sellotape 2015-09-22 07:06:47 was this a fresh install or an upgrade with iso as source? 2015-09-22 07:10:19 fresh install 2015-09-22 07:10:31 sys install to lvm lv partition 2015-09-22 07:13:18 just downloading 3.2.3 now 2015-09-22 07:14:04 ScrumpyJack thank you, i created /lib/modules/3.18.20 to /lib/modules/3.18.21-0-grsec, but now i see Loading modules ...modprobe: can`t load module af_packet (kernel/net/packet/af_packet.ko): invalid module format 2015-09-22 07:15:13 i can try new install, maybe is just this installation borked. but i did it like always. did this many times on this apu4c and lvm2.. 2015-09-22 07:18:41 what does uname -r say? 2015-09-22 07:19:19 3.18.20 which is weird, where is 3.18.21-0-grsec 2015-09-22 07:19:26 maybe i should delete whole vg 2015-09-22 07:20:36 but in /boot/ there are vmlinuz-grsec and initfamfs-grsec no signt ov vmlinuz or initramfs.. 2015-09-22 07:20:50 extlinux.conf line is also correct. 2015-09-22 07:22:50 if you have installed from the grsec patched kernel iso, vmlinuz-grsec and initfamfs-grsec is correct 2015-09-22 07:23:34 ScrumpyJack yes i did, i used alpine-3.2.3-x86_64.iso 2015-09-22 07:28:53 did you apk upgrade after an install? 2015-09-22 07:32:00 ScrumpyJack no i did not, just added these three packages lvm2 e2fsprogs syslinux, 2015-09-22 07:32:08 ScrumpyJack for you it works? 2015-09-22 07:47:43 ncopa: been working on getting linux-virtgrsec in sync with linux-grsec. would it be best to leave pkgrel at 0 and let it build next time there is a new upstream version do avoid rebuilding all out-of-tree kernel modules? 2015-09-22 07:48:01 s/do/to 2015-09-22 07:48:57 might be an idea 2015-09-22 07:49:09 but i dont mind that we just rebuild them too 2015-09-22 07:49:45 if you send me a git-am'able patch (sptring/tpaste) i can apply it here and rebuild the 3rd parties 2015-09-22 07:49:56 sprunge* 2015-09-22 07:50:17 i am working on initramfs program to find root= 2015-09-22 07:50:33 i will have it trigger mdadm and lvm too 2015-09-22 07:50:42 ncopa any idea what would be problem in my case? 2015-09-22 07:51:42 crow: what happened? 2015-09-22 07:51:47 ncopa: cool, will do. need to do some boot tests first. i had removed some modules mkinitfs expected which I'll have to bring back 2015-09-22 07:52:34 hid-cherry.ko and fusion/* 2015-09-22 07:52:56 hid-cherry probably dont make sense in linux-virtgrsec 2015-09-22 07:54:19 no, but its just easier to leave it in :) 2015-09-22 07:54:46 mkinitfs did not fail, but annoying with the error messages 2015-09-22 07:55:24 crow: interesting, the initramfs contains modules for 3.18.20-1 but the overlay is 3.18.21-0 2015-09-22 07:56:44 that should work, but for you it seems like your module path is still the initramfs path, and not the new overlay. i don't know how that can happen 2015-09-22 08:15:59 ncopa sorry for for pasting these three lines 2015-09-22 08:16:00 22:31:49) (crow) after installing grsec latest iso and rebooting my network is not detected http://sprunge.us/Yejc eth0 ...ip: ioctl 0x8913 failed: No such device . i was able to use iso and install alpine to disk... weird . maybe because of this: * Loading modules ...modprobe: can't change directory to '3.18.20': No such file or directory ? 2015-09-22 08:16:02 (22:41:30) (crow) i have only /lib/modules/3.18.21-0-grsec (just normal installation to disk) using old stuff written in this wiki http://wiki.alpinelinux.org/wiki/Installing_on_GPT_LVM just without creating vg0 but deleting old lv and creating new.. 2015-09-22 08:16:10 (23:31:27) (crow) also trying to rebuild initramfs with mkinitfs result in an error http://sprunge.us/cLFH 2015-09-22 08:17:17 ok 2015-09-22 08:17:29 so you are running old kernel but installed new kernel modules 2015-09-22 08:17:53 ncopa well this is new installation, i am not sure where from is old kernel, 2015-09-22 08:18:12 ncopa ScrumpyJack) crow: interesting, the initramfs contains modules for 3.18.20-1 but the overlay is 3.18.21-0 2015-09-22 08:18:51 can you mount your sysroot manually? 2015-09-22 08:18:58 ncopa i used alpine-3.2.3-x86_64.iso and did install it from usb stick, like always, i am on lvm partitions. i can do clean everthing, vg to and do again new install 2015-09-22 08:19:09 ncopa i am logged in, just not network 2015-09-22 08:19:27 now=no 2015-09-22 08:19:28 i think you need boot using usb again 2015-09-22 08:19:47 then you mount /dev/ /mnt 2015-09-22 08:19:50 and /mnt/boot 2015-09-22 08:19:57 ok just sec 2015-09-22 08:19:59 so you mount up your root and boot 2015-09-22 08:20:24 also, make sure you have working network from your boot usb 2015-09-22 08:20:38 (setup-alpine -q , select no disk) 2015-09-22 08:21:00 then you do: apk fix --root /mnt linux-grsec 2015-09-22 08:23:36 \o/ 2015-09-22 08:23:36 ncopa, the initramfs on alpine-3.2.3-x86_64.iso contains modules for 3.18.20-1 but the overlay is 3.18.21-0 2015-09-22 08:23:52 my initramfs program can start and find mdadm devices 2015-09-22 08:23:56 could that be a problem 2015-09-22 08:24:05 ./nlplug: Found LABEL=testroot (/dev/md3) 2015-09-22 08:24:05 ./nlplug: FOUND LABEL=testroot 2015-09-22 08:24:05 ./nlplug: modaliases: 118, forks: 0, events: 638, total bufsize: 249819 2015-09-22 08:24:08 the overlay? 2015-09-22 08:24:20 the real root 2015-09-22 08:24:30 it should not be a problem 2015-09-22 08:24:45 but it seems like you booted using the kernel from iso 2015-09-22 08:25:07 trying to use the rootfs with mismatching kernel 2015-09-22 08:25:47 ncopa well is not that the same case i have. ScrumpyJack downloaded iso to test and see if the error i get is same 2015-09-22 08:26:04 so by mounting up the rootfs and $rootfs/boot 2015-09-22 08:26:29 and then apk fix --root $rootfs linux-grsec we try make sure that installed kernel and initramfs correspond 2015-09-22 08:26:42 make sure that you boot from your disk and not from the iso 2015-09-22 08:31:49 ncopa for my case you wrote to boot usb again, i did, installed lv2m efprogrs and syslinux, activated vgchange -ay ; mountet rootfs to /mnt (i dont have seperate boot partition), i used setup-alpine -q (setup just locales, and network), there was no question regarding disk), while doing apk fix --root /mnt linux-grsec i get thishttp://sprunge.us/HXEf 2015-09-22 08:33:25 udhcpc failed to get a DHCP lease 2015-09-22 08:33:28 you dont have network 2015-09-22 08:33:42 make network work first 2015-09-22 08:33:48 in the live system 2015-09-22 08:36:54 ncopa its just from other interfaces, network works fine (after setup-alpine -q) 2015-09-22 08:37:02 ok 2015-09-22 08:37:11 this device have 3 eth interfaces and 1x wif 2015-09-22 08:37:11 so you can ping google.com 2015-09-22 08:37:14 sure 2015-09-22 08:37:44 i installev lvm2 ef2progs and syslinux 2015-09-22 08:37:48 in that paste 2015-09-22 08:37:51 ok 2015-09-22 08:37:58 and your /boot? 2015-09-22 08:38:01 where is it? 2015-09-22 08:38:05 is it also on lvm? 2015-09-22 08:38:10 on same rootfs lvm, 2015-09-22 08:38:18 ok, there is your problem 2015-09-22 08:38:26 syslinux cannot boot from lvm 2015-09-22 08:38:35 http://sprunge.us/BXdc 2015-09-22 08:38:37 you need a non-lvm boot partition 2015-09-22 08:38:47 where the kernel is 2015-09-22 08:38:48 ncopa see this on botton http://wiki.alpinelinux.org/wiki/Installing_on_GPT_LVM 2015-09-22 08:39:36 ok hmm 2015-09-22 08:39:43 /boot is device /dev/sda2 2015-09-22 08:40:05 which is a raw partition, not lvm 2015-09-22 08:40:08 i think i f.. it there 2015-09-22 08:40:21 it may be ext4 2015-09-22 08:40:24 ok doing everything again :) 2015-09-22 08:41:11 thank you seems last night was to late and i missed to mount this partition :( 2015-09-22 08:41:19 vanilla kernel was there before so its loading it from there. 2015-09-22 08:41:39 ncopa one think i have to complainis selecting mirrors process. here it takes 10min ... http://sprunge.us/QKOe 2015-09-22 08:41:45 can we have timeout after 10sec.. 2015-09-22 08:42:24 you see there 137 (i think seconds) is just to much to wait for it. i dont know how much second is needed for ERROR: 2015-09-22 08:48:21 yeah, 10 sec or so is probably good 2015-09-22 08:50:06 should i create feature request? 2015-09-22 10:09:43 who maintains the mailing lists? 2015-09-22 10:13:10 ncopa: http://sprunge.us/eUIJ 2015-09-22 10:14:05 install size down from 50790400 to 38637568 2015-09-22 10:15:16 and more in sync with linux-grsec apart from drivers (cgroups, ia32, zram, etc) 2015-09-22 10:30:47 uggedal: http://sprunge.us/eUIJ <- 'Remove features present in linux-grsec' 2015-09-22 10:30:50 can some of features not be used in virt environment, like passing USB devices ? 2015-09-22 10:37:27 vkris: some of them can, but if you're doing something special like that you should be using the linux-grsec kernel 2015-09-22 10:38:20 vkris: linux-virtgrsec was tuned for server usage before my changes, so I don't think specialized drivers etc fit 2015-09-22 10:40:07 ok 2015-09-22 10:42:00 feature 'xz ramdisk compression' seems nice, was is it in kernel stagging ? 2015-09-22 10:42:23 or use zram 2015-09-22 10:43:15 vkris: linux-grsec don't have xz ramdisk support, so why should linux-virtgrsec? 2015-09-22 10:44:05 and since mkinitfs uses gzip it won't be used 2015-09-22 10:45:18 I think, ncopa was having a discussion on using xz, a while ago 2015-09-22 10:47:37 '#alpine-linux-2014-04.log:2014-04-24' 2015-09-22 10:51:42 yeah, but then it should be enabled for all kernels and changed in mkinitfs 2015-09-22 10:52:01 nice 2015-09-22 10:52:09 i wonder about the sound support 2015-09-22 10:52:19 are you sure we dont want sound support for virtual? 2015-09-22 11:01:34 ncopa: my thinking was that this is for virtual servers (since the it uses 100hz vs 300hz, deadline vs cfq scheduler, no forced preemption vs voluntary preemption) 2015-09-22 11:02:59 for a virtual desktop I think you want linux-grsec anyways (and it has virtio drivers etc) 2015-09-22 11:04:57 if its intended for virtual desktops we can add back sound (and maybe agp/dri also) 2015-09-22 11:06:37 ok 2015-09-22 11:06:45 makes sense 2015-09-22 11:23:21 ncopa thank you for pointing me to this /boot partition, even i wrote that wiki i missed that part last time. ScrumpyJack the problem was that i did not mount the /boot partition which is needed for syslinux. 2015-09-22 11:39:49 ncopa: what do you use for tracking new CVEs (thinking of integrating it into aports-vercmp)? 2015-09-22 11:40:45 we have a volunteer who follows mailing lists etc and file bugs for us 2015-09-22 11:41:22 ah, nice 2015-09-22 11:41:46 but i struggle to keep up with fixing things :-/ 2015-09-22 11:43:05 starting to get decent coverage in new upstream versions: http://sprunge.us/EcOj 2015-09-22 13:37:17 could someone take a look at %589? 2015-09-22 14:00:49 nmeum: nmeum it's already in /usr/sbin 2015-09-22 14:01:31 nope 2015-09-22 14:01:42 http://sprunge.us/ahIZ 2015-09-22 14:03:31 lol, i was doing a which sendmail on my opensmtpd box and i got the busybox one back first 2015-09-22 14:04:02 yep, that's why I created that patch 2015-09-22 14:11:52 nmeum: applied 2015-09-22 14:11:53 thanks 2015-09-22 14:12:16 cool! my "nlplug" coldplug from initramfs appears to work 2015-09-22 14:13:15 it succesfully inits mdadm and lvm2 2015-09-22 14:13:20 and switch root 2015-09-22 14:13:28 i need probably deal with cryptsetup 2015-09-22 16:08:38 hmm, i think smtpd sendmail shouldn't overwrite busybox link 2015-09-22 16:09:53 bah, never mind 2015-09-22 16:11:02 maybe there should be a seperate package for sendmail in smtpd as it's already provided by busybox. that way users know they are explicitly going to get sendmail 2015-09-22 16:12:15 like procps for exemple 2015-09-22 16:15:35 could someone move taskd to edge/main? work fine for me 2015-09-22 16:16:54 (it is already there - i should get some sleep) 2015-09-22 19:37:45 there was default_kernel_opts="nomodeset quiet" with vanilla kernel, but with grsec it isnt there. is this needed anymore? 2015-09-22 19:39:26 the patch seems still there http://git.alpinelinux.org/cgit/aports/tree/main/alpine-conf/0001-setup-disk-pass-nomodeset-boot-option.patch?id=545977c1fde1dadc99987aedc2912d93243c28a9 but i dont get it what this if checks 2015-09-23 09:51:08 Firefox 41 stable is available :) 2015-09-23 09:57:18 stable ? lie ! :p 2015-09-23 09:58:10 lawl 2015-09-23 09:59:12 remember the version that drop usefull extention ? :) 2015-09-23 09:59:26 dont worry, we have chromium :p 2015-09-23 09:59:44 clandmeter the %30 browser ? :p 2015-09-23 10:00:17 google's blackhole 2015-09-23 11:24:44 that sandy aport needs to: 1) use $pkgver in $source 2) override/extend default CFLAGS/LDFLAGS (default CFLAGS is -O0) 2015-09-23 13:06:57 ok nlplug-findfs works 2015-09-23 13:07:21 nplug-findfs will coldplug til given is found 2015-09-23 13:07:59 it will automatically load kernel modules for found hardware til given device is found 2015-09-23 13:08:49 i am not sure how to deal with nfs and those 2015-09-23 13:09:10 ok 2015-09-23 13:09:35 now i wonder, how do we make it mount and look for apkovl and .boot_repository? 2015-09-23 13:11:29 the idea was to do something like: nlplug-findfs -a '*/*/.boot_repository' -a '*.apkovl.tar.gz*' 2015-09-23 13:12:14 when if has found all the globs it looks for it will exit 2015-09-23 13:13:57 hm 2015-09-23 13:26:08 do we want mount $UUID as: 2015-09-23 13:26:14 /media/$UUID 2015-09-23 13:26:15 or 2015-09-23 13:26:21 /media/UUID=$UUID 2015-09-23 13:26:27 we currently do the latter 2015-09-23 13:26:35 do we want change it to the former? 2015-09-23 13:27:01 or do we want move it to /run/media/ or similar? 2015-09-23 14:01:16 ncopa: i have no preference 2015-09-23 14:01:42 ./media already exists, so it can stay in there 2015-09-23 14:03:52 the idea is that nlplug-findfs should be able to find the boot repository and/or apkovl 2015-09-23 14:04:24 or the device specified by root=... boot option 2015-09-23 14:05:14 i wonder how to tell nlplug-findfs to find boot repository and apkovl 2015-09-23 14:05:19 and how to return it 2015-09-23 14:05:37 ideally i'd like to find both in same run 2015-09-23 14:05:43 so i dont need to re-run it 2015-09-23 14:06:03 but then, how should it output what it finds? 2015-09-23 14:06:57 maybe something like: -b /tmp/repositories 2015-09-23 14:07:45 which will leave devices with boot_repository mounted and write the boot repositories found to the specified file 2015-09-23 14:08:30 i think that might make sense 2015-09-23 14:08:48 we need make it find many boot repos 2015-09-23 14:09:00 it needs support multiple boot repos 2015-09-23 14:49:36 hmmm 2015-09-23 14:49:54 if you can just add one mountpoint just use and overlay filesystem 2015-09-23 14:50:04 i know its a bad solution, but it should work 2015-09-23 15:50:42 uggedal: git://github.com/asciimoo/vutr 2015-09-23 16:25:16 vkris: thanks 2015-09-23 20:13:05 is there any particular reason why openssh hasn't been upgraded to 7.1 yet? 2015-09-23 22:03:22 created a script for creating lxc containers for stable releases and building in them 2015-09-23 22:04:03 cd $APORTSDIR 2015-09-23 22:04:07 acontain create ; acontain start ; acontain prepare 2015-09-23 22:04:10 cd main/ncdu 2015-09-23 22:04:17 acontain build -r 2015-09-23 22:04:52 with git co 3.2-stable it will build ncdu for x86/x86_64 on 3.2 2015-09-23 22:05:03 with git checkout master it will build edge 2015-09-23 22:19:25 uggedal_: would you mind sharing that script? 2015-09-23 22:22:24 sure, but its not finished yet. it might wipe your / and such 2015-09-23 22:22:54 nmeum: http://git.uggedal.com/dotfiles/tree/.local/bin/acontain 2015-09-23 22:23:08 thanks 2015-09-24 03:05:33 http://www.audi.jp/rice/ 2015-09-24 08:03:06 ncopa, i'm trying to build dsniff for alpine. I've an issue related to rpc/rpc.h: it is not found. I figured that this is an ancient implementation, and libitrpc-dev contains itrpc/rpc/rpc.h. So i patched the includes to the right path. 2015-09-24 08:03:20 Now i've this issue: rpc/types.h cannot be found 2015-09-24 08:03:30 but this is referenced by libitrpc 2015-09-24 08:03:37 probably i'm hitting this: https://cygwin.com/ml/cygwin/2015-05/msg00246.html 2015-09-24 08:09:31 fcolista: the proper way to do it is to add $(pkg-config --cflags libtirpc) or simiar to CFLAGS 2015-09-24 08:10:04 interesting 2015-09-24 08:10:09 $ pkg-config libtirpc --cflags 2015-09-24 08:10:09 -I/usr/include/tirpc 2015-09-24 08:10:49 and pkg-config --libs libtirpc to LIBS 2015-09-24 08:11:18 i need to add this to the APKBUILD or patch the Makefile? 2015-09-24 08:11:23 What's the best approach? 2015-09-24 08:12:42 probably patch the makefile 2015-09-24 08:12:45 or configure script 2015-09-24 08:13:00 that way you can send the patch back to dsniff developers 2015-09-24 08:20:31 currently firefox is build with --disable-crashreporter, would anyone have a problem with setting it to enabled? 2015-09-24 08:25:23 ncopa: maybe this is a candidate for abuild repo: http://git.uggedal.com/dotfiles/tree/.local/bin/acontain ? 2015-09-24 08:32:18 nice 2015-09-24 08:32:45 or we could have it as a separate project 2015-09-24 08:34:12 ncopa: http://sprunge.us/LXah 2015-09-24 08:35:17 yes that should work 2015-09-24 08:35:30 ok, thx 2015-09-24 08:35:35 i'm gonna try 2015-09-24 08:35:45 or you could do: LIBTIRPC_CFLAGS := $(shell pkg-config --cflags libtirpc) 2015-09-24 08:35:58 and then CFLAGS += $(LIBTIRPC_CFLAGS) 2015-09-24 08:36:18 then will you not spawn pkg-config for every compiled file 2015-09-24 08:36:28 yeah 2015-09-24 08:36:29 good point 2015-09-24 08:36:38 also, it might be that not all programs needs link against libtirpc 2015-09-24 08:37:02 LIBTIRPC_LIBS := $(pkg-config --libs libtirpc) 2015-09-24 08:37:56 looks like it would be more consistent to do: 2015-09-24 08:38:20 TIRPCINC := $(pkg-config --cflags libtirpc) 2015-09-24 08:40:23 TIRPCINC := $(shell pkg-config --cflags libtirpc) 2015-09-24 08:40:23 CFLAGS = @CFLAGS@ -DDSNIFF_LIBDIR=\"$(libdir)/\" 2015-09-24 08:40:23 CFLAGS += $(TIRPCINC) 2015-09-24 08:42:19 LIBTIRPC_LIBS := $(pkg-config --libs libtirpc) 2015-09-24 08:42:19 LIBS = @LIBS@ -L$(srcdir) -lmissing 2015-09-24 08:42:19 LIBS += $(LIBTIRPC_LIBS) 2015-09-24 08:42:34 sorry, TIRPCINC := $(pkg-config --cflags libtirpc) 2015-09-24 08:44:09 INCS = -I. $(NIDSINC) $(PCAPINC) $(LNETINC) $(DBINC) $(SSLINC) $(X11INC) \ 2015-09-24 08:44:10 -I$(srcdir)/missing 2015-09-24 08:44:26 looks like it would be more appropiate to add TIRPCINC to INCS 2015-09-24 08:44:38 INCS = ..... $(TIRPCINC) 2015-09-24 08:44:57 INCS is what they sue to append -I cflags 2015-09-24 08:45:37 oh, ok 2015-09-24 08:46:36 that's valid both for cflags that libs ? 2015-09-24 08:47:48 INCS = include files 2015-09-24 08:47:51 libs are librarys 2015-09-24 08:48:01 includes are #include in source files 2015-09-24 08:48:04 compile time 2015-09-24 08:48:08 gcc -c 2015-09-24 08:48:28 link time is when you link all the compiled .o files + needed libs 2015-09-24 08:48:45 gcc -o program files.o -llib1 -llib2 2015-09-24 08:49:00 gcc -o program files.o $LIBS 2015-09-24 08:49:16 i have not studied the relevant Makefile 2015-09-24 08:49:34 but you need make gcc -c .... append the output of pkg-config --cflags libtirpc 2015-09-24 08:49:40 and the link operation 2015-09-24 08:49:51 what is pkg-config supposed to do? 2015-09-24 08:49:58 need to append output of pkg-config --libs libtirpc 2015-09-24 08:50:30 pkg-config is a helper tool used when compiling applications and libraries. It helps you insert the correct compiler options on the command line so an application can use gcc -o test test.cpkg-config --libs --cflags glib-2.0 for instance, rather than hard-coding values on where to find glib (or other libraries). 2015-09-24 08:51:04 so that would not be needed: 2015-09-24 08:51:08 LDFLAGS="-lresolv -lglib-2.0 -lgthread-2.0 $LDFLAGS" ./configure ... 2015-09-24 08:51:56 you normally do the detection of -lresolv -lglib-2.0 in the configure script 2015-09-24 08:52:38 pkg-config let you store the needed flags for a library 2015-09-24 08:52:44 so if you want use libtirpc 2015-09-24 08:53:05 you dont need to know what -I/pathc/to/libtirpc/headers you need 2015-09-24 08:53:15 you dont need to know the exactpath where the headers are installed 2015-09-24 08:53:38 instead you ask pkg-config, what cflags do i need to use libtirpc? 2015-09-24 08:54:05 pkg-config knows that libtirpc was installed in /usr/include/something or /usr/local/something 2015-09-24 08:54:15 how pkg-config is able to find this info? 2015-09-24 08:54:28 each libary provides a .pc file 2015-09-24 08:54:37 which are stored in /usr/lib/pkgconfig 2015-09-24 08:55:22 many/most libraries provides a .pc file 2015-09-24 08:55:28 ah-a 2015-09-24 08:55:32 we ship them with -dev packages 2015-09-24 13:18:54 ncopa i created #4655 so it doesn't get forgotten 2015-09-24 13:25:24 crow: thanks. i would have forgotten it 2015-09-24 13:35:15 ncopa you saw that what i posted yesterday about lm_sensors? 2015-09-24 13:37:14 ncopa: go(-bootstrap) build just fine on x86 2015-09-24 13:37:20 not sure what was going on before 2015-09-24 13:37:33 it does? 2015-09-24 13:37:53 on linux-grsec it didn't build ~a week ago 2015-09-24 13:38:09 ah 2015-09-24 13:38:19 my build has vanilla 2015-09-24 13:38:26 yeah, it builds fine on vanilla 2015-09-24 13:38:26 builder* 2015-09-24 13:38:55 some x86 assembler file didn't work properly due to a grsecurity protection 2015-09-24 13:39:37 ncopa should i report that about lm_sensors to ? 2015-09-24 13:39:46 this is what i wrote :) 2015-09-24 13:39:49 I don't use x86 so I wasn't interested enough to fix this or to report it upstream 2015-09-24 13:39:50 sorry its big i know 2015-09-24 15:24:36 could someone take a look at %595? 2015-09-24 15:50:15 why does the default group file shipped by alpine-base-layout contain somemany groups that are not needed by default like postfix or apache? 2015-09-24 16:35:41 what does abuild depend on if I want to use it on not alpine 2015-09-24 16:43:41 nmeum: I suspect it'd be a breaking change to fix that 2015-09-24 16:44:45 no, it wouldn't be a breaking change, you could simply add .pre-install files to postfix and apache 2015-09-24 16:45:09 and apk info -R shows you which packages abuild depends on 2015-09-24 16:47:45 then you'd have to upgrade postfix and apache at the same time as the baselayout, surely? 2015-09-24 16:48:28 I guess you could just push out the postfix and apache upgrades earlier 2015-09-24 16:48:51 oh huh since when is s6 in edge main 2015-09-24 16:49:54 since july 19 2015-09-24 20:39:33 ncopa: could you consider applying my proposed patches tomorrow? Especially %598, %597 and %595 as well as %584 are important to me 2015-09-25 07:26:45 i'm getting an error on boot. modprobe invalid switch -h 2015-09-25 07:27:00 i can't see anything runnign modprobe with -h 2015-09-25 08:55:49 ScrumpyJack: perhaps you have something with -h in /etc/modules or similar 2015-09-25 08:59:36 nope :( 2015-09-25 12:32:26 /etc/init.d/mdadm-raid contains modprobe -k which isn't in busybox modprobe 2015-09-25 12:34:30 [ -f /proc/mdstat ] || modprobe -k md > /dev/null 2>&1 2015-09-25 12:34:33 what's that about? 2015-09-25 12:56:23 that seems to load the mdadm module 2015-09-25 13:16:43 i get modprobe: unrecognized option: k 2015-09-25 13:18:19 ScrumpyJack: add 'md' to /etc/modules 2015-09-25 13:19:43 just removing that '-k' flag form that line should also work 2015-09-25 13:23:46 yeah 2015-09-25 13:23:49 we should do that 2015-09-25 13:24:24 does anyone know what it originally did? 2015-09-25 13:24:29 ncopa: i'm trying to track down a "modprobe invalid option see -h for details" error on boot 2015-09-25 13:24:37 i dont think kmod supports it 2015-09-25 13:26:55 where did that mdadm-raid service even came from? 2015-09-25 13:27:14 it doesn't seem to be importet from gentoo 2015-09-25 13:27:22 *imported 2015-09-25 13:28:50 mdadm-3.3.4-r0 contains: etc/init.d/mdadm-raid 2015-09-25 13:36:32 I am a aware that it belongs to the mdadm package, the question is who wrote that service and why? Because usually mdadm should start raid devices from mdadm.conf automatically (iirc) 2015-09-25 13:51:33 BTW: could someone take a look at %597 and %598? 2015-09-25 14:45:52 ncopa: did you got a reply from spender re grsec patches? 2015-09-25 14:46:32 pluesch: no 2015-09-25 15:08:45 ncopa: have you sent the mail to contact@ or spender@? 2015-09-25 15:09:01 to spender@ 2015-09-25 15:09:39 sned it again to contact@ ... they have changed the addresses. I have got a reply from contact@ within a day. 2015-09-25 15:10:06 they have also changed it on the page: https://grsecurity.net/contribute.php 2015-09-25 15:13:58 I have waited around 2 weeks for a reply from spender@, sent it again to contact@, got reply after about a day. Requested a quote for a private individual. I have a quote for 200 usd per month now. Don't know yet if I will do it or not. It actually depends if Alpine gets access to stable patches again... 2015-09-25 15:15:45 pluesch: thanks for the tip. let spender know the ^^^ 2015-09-25 15:16:07 i resent to contact 2015-09-25 15:16:16 big thanks 2015-09-25 15:16:20 and have a nice weekend 2015-09-25 15:16:32 ty 2015-09-25 15:16:36 you too 2015-09-25 15:17:47 if there's a known fact that the patches will be redistributed the quote won't stay at 200$ 2015-09-25 15:19:59 it is very strange imho~ 2015-09-25 15:20:06 GPL allows you to distribute 2015-09-25 15:20:32 I can't find anything on http://grsecurity.net/agree/agreement.php that should prevent it. 2015-09-25 15:21:18 can someone read through it maybe... my english isn't very good so maybe I have just missed a thing 2015-09-25 15:21:21 pluesch: oh it's not prevented per se 2015-09-25 15:22:18 If spender knows you'll be redistributing the patch - cause you have to comply with the GPL - the entry price is higher 2015-09-25 15:22:44 he may even base the price on the number of "redistributions" you plan on doing 2015-09-25 15:23:05 coredumb: so I should be able to "buy" it and give it to alpine, if they won't give it to alpine for free... 2015-09-25 15:23:15 they haven't asked me anything about redistribution... 2015-09-25 15:23:33 now I have no proof but this what I understood from my quote :) 2015-09-25 15:23:52 what quote did you got? 2015-09-25 15:24:29 I have no actual quote atm. just email conversation. 2015-09-25 15:24:52 300/mo + renegociation when we start - if we do actually it's still not sure - distribuing the kernel to our customers 2015-09-25 15:25:01 tbh 2015-09-25 15:26:30 and will have to sign an agreement that if I make the the patch publicly available they'd drop my sponsorship and refund me 2015-09-25 15:27:03 okay... I'll get that too I guess -.- 2015-09-25 15:27:38 meaning that it's still not that simple to redistribute the patch to a distribution 2015-09-25 15:28:27 yeah. I can understand that now... 2015-09-25 15:28:58 and as Alpine is not using the stable kernel anyway ... I feel that grsec gonna be quite reluctant to find any agreement 2015-09-25 15:29:12 ^ pure speculations on my side 2015-09-25 15:29:29 sometimes alpine uses stable branch directly from grsec iirc. 2015-09-25 15:29:38 nope we dont for a long time 2015-09-25 15:29:46 3.14? 2015-09-25 15:29:53 3.18 2015-09-25 15:30:04 yeah. but 3.14 was directly from grsec, or not? 2015-09-25 15:30:09 no 2015-09-25 15:30:12 oh. 2015-09-25 15:30:13 we have our own modifications 2015-09-25 15:30:35 okay... got that wrong then. sorry about that. 2015-09-25 15:30:45 as coredumb said, it is unlikely that spender will work out a deal. but it is good to investigate all options first. 2015-09-25 15:31:07 spender does not give a damn about gpl compliance, it is about revenue 2015-09-25 15:31:36 I guess... but why has he waited so long to do it? 2015-09-25 15:31:44 dunno, you would have to ask them 2015-09-25 15:31:56 maybe they want to work on grsec fulltime instead of have a day job 2015-09-25 15:33:01 @kaniini | spender does not give a damn about gpl compliance, it is about revenue < actually I think a bit of both... As he totally agree that as a sponsor you have to comply to it, but in return you have to pay :D 2015-09-25 15:33:21 coredumb: it is really about revenue tbh 2015-09-25 15:33:37 so most likely we (alpine) will just continue to maintain our own branch based on the test branch from grsec with modifications. 2015-09-25 15:33:45 maybe 2015-09-25 15:33:45 kaniini: honestly don't you think that he's already fully working on grsec ? 2015-09-25 15:33:51 coredumb: i know he is not 2015-09-25 15:34:01 I bet his employer already pays him to do so 2015-09-25 15:34:06 coredumb: they don't 2015-09-25 15:34:07 at least at some levels 2015-09-25 15:34:14 really ? 2015-09-25 15:34:21 coredumb: they pay him to work on cuckoo 2015-09-25 15:34:25 (accuvant) 2015-09-25 15:35:34 coredumb: https://github.com/brad-accuvant 2015-09-25 15:35:45 interesting. 2015-09-25 15:35:52 ok 2015-09-25 15:36:05 yep so the github profile :P 2015-09-25 15:36:09 i think 2015-09-25 15:36:15 optiv is about to implode 2015-09-25 15:36:19 so he is trying to get revenue 2015-09-25 15:36:26 so he can work on grsecurity full time instead 2015-09-25 15:36:54 and sure, they probably do pay for him to do grsecurity some of the time, but most of his work appears to be related to malware analysis 2015-09-25 15:37:32 oh well i do not really care either way 2015-09-25 15:37:33 ok 2015-09-25 15:37:38 we will figure it out 2015-09-25 15:38:02 I'm thb pretty happy if he gets some paying customers so he can work fulltime on grsec. 2015-09-25 15:38:10 bc grsec is more important to me :) 2015-09-25 15:38:18 tbh* 2015-09-25 15:38:40 if you were using alpine because it ships grsec and only for that reason, you completely missed the message 2015-09-25 15:39:05 not really. I'm using alpine for various reasons. 2015-09-25 15:39:11 https://lkml.org/lkml/2015/9/13/195 during all this, this is getting the cold shoulder from kernel devs :( 2015-09-25 15:41:23 one primary reason for me is because it's not using systemd. another is it's small size. grsec is a reason too, that's true. but it's not the only one. 2015-09-25 19:20:31 > 17:26 < coredumb> and will have to sign an agreement that if I make the the patch publicly available they'd drop my sponsorship and refund me 2015-09-25 19:20:36 That's probably a violation of the GPL 2015-09-25 19:21:30 I don't see how 2015-09-25 19:22:21 he can do whetever he wants with the patches he's already gotten, but won't get access to new ones 2015-09-25 19:22:25 well the GPL explicitly allows you to redistribute copies „so you can help your neighbours” 2015-09-25 19:23:09 ah, ok then 2015-09-25 19:23:14 yeah, he can redistribute. but then he wont get access to newer copies 2015-09-25 19:23:40 yeah, I got that 2015-09-25 19:23:42 IANAL an all that 2015-09-25 19:46:24 so each time spender learns you are redistributing, you need to assume a new identity to become sponsor again 2015-09-25 19:47:01 kunkku: yeah gonna be tough for the whole year : 2015-09-25 19:47:02 :D 2015-09-25 19:47:46 fortunately we have a few users 2015-09-25 19:49:12 ^^ 2015-09-25 20:19:10 or just drop it... 2015-09-25 20:19:26 "<@dalias> pax is crap" 2015-09-25 20:23:45 well, this is something we also considered 2015-09-26 06:04:11 ? 2015-09-26 13:16:40 Hello people, i was just wandering if theres any information that you can point me out on how to compile a kernel for alpine 2015-09-26 13:17:39 there's no information in the wiki but, package creation and few more things 2015-09-26 13:18:15 any pointer will be good, thanks! 2015-09-26 13:18:42 argami: I'd try to compile the linux-grsec package 2015-09-26 13:18:59 in aports/main/linux-grsec 2015-09-26 13:19:05 abuild -r 2015-09-26 13:19:05 ok 2015-09-26 13:19:20 but never tested it myself 2015-09-26 13:20:40 and you need to apk add alpine-sdk 2015-09-26 13:21:07 a friend of mine for a project ask me to build a 4.x kernel fro alpine (he already did it) but he want me to find out the dificulty of the task doing it my self 2015-09-26 13:21:26 thats few great pointers 2015-09-26 13:21:29 thanks 2015-09-26 13:23:03 you're welcome 2015-09-26 13:23:05 good luck :-) 2015-09-26 13:23:19 thanks i think im going to need it 2015-09-26 14:41:34 jomat: its already building at least the 3.x version later i will start to fight with the 4.x 2015-09-26 14:41:36 thanks!! 2015-09-27 03:31:50 Hello again people, does any one has receive this error when compiling the kernel? 2015-09-27 03:32:16 home/argami/aports/main/linux-grsec/src/linux-3.18/Makefile:683: *** Your gcc installation does not support plugins. If the necessary headers for plugin support are missing, they should be installed. On Debian, apt-get install gcc--plugin-dev. If you choose to ignore this error and lessen the improvements provided by this patch, re-run make with the DISABLE_PAX_PLUGINS=y argument.. Stop. 2015-09-27 03:54:08 apk add gcc-plugin-dev 2015-09-27 04:01:46 hey thanks for the answer but it doesn't work for me 2015-09-27 04:01:57 ERROR: unsatisfiable constraints: 2015-09-27 04:01:57 gcc-plugin-dev (missing): 2015-09-27 04:01:58 required by: world[gcc-plugin-dev] 2015-09-27 04:03:14 i also look into the apk list and theres not any by version either 2015-09-27 04:22:28 apk search gcc | grep dev 2015-09-27 04:22:31 its something like that 2015-09-27 04:23:43 i did it and no result im looking now into the mirrors directly to know if i have any problem with my apl 2015-09-27 04:23:44 apk 2015-09-27 04:30:15 thanks anyway 2015-09-27 09:18:06 argami: setup development on your alpinelinux installation http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package 2015-09-27 16:00:22 Hello again!, i was finally able to fix all my issues and now i have a fully functional alpine in xhyve 2015-09-27 16:01:39 just a curiosity, anybody knows how long it takes the kernel to be build in alpine? i just realize that is a little slow so just wandering if anybody knows 2015-09-27 16:08:35 that depends on your hardware 2015-09-27 16:09:47 I believe I found a fucked up bug in apk-tools. I managed to create a package which contains files that are not installed by apk-tools even though they are present in the .apk file if you unpack it manually 2015-09-27 16:10:10 i give 2G and 2 cpu to the virtual from 1,6 GHz Intel Core i5 2015-09-27 16:10:27 i just want some posible comparation point 2015-09-27 16:10:58 in the mean time im preparing a install of alpine in my server (16GB and 8cpu) to build there 2015-09-27 16:45:19 why does apk-tools ship its own tar implementation instead of using libtar? 2015-09-27 17:24:09 argami seems you was quit irc i wrote this for you: argami: setup development on your alpinelinux installation http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package 2015-09-27 17:24:24 crow: thanks man 2015-09-27 17:24:34 i already did all that 2015-09-27 17:24:51 its working 2015-09-27 17:24:57 slow but working 2015-09-27 19:28:32 argami_ :) is the host slow or? compiling kernel with everything on is ofcourse long process 2015-09-27 19:30:46 crow: is slow on my virtual machine in xhyve but because my computer is slow, but in the server i configured it with 8GB ram and 2cpus and its was a lot faster 2015-09-27 19:31:03 i already was able to compile the 3.X 2015-09-27 19:31:12 now im working on the 4.X test 2015-09-27 19:31:16 argami_ may i ask why recompiling? or you need some feature not included 2015-09-27 19:31:23 hahaha 2015-09-27 19:31:59 we need kernel 4.X 2015-09-27 19:32:20 some partition ciphering that we need 2015-09-27 19:32:45 not really sure which one now because im just updating my self in how to work with alpine 2015-09-27 19:38:50 :) 2015-09-27 19:39:48 what im fearing most is what is going to happen with the grsec patches issue 2015-09-27 19:40:30 im building 4.X now with the testing patch but do you know if theres already any option? 2015-09-27 19:48:21 argami_: for 4.X there is only the testing patch 2015-09-27 19:48:29 there is no stable patch for 4.X as of now 2015-09-27 19:49:25 pluesch: well my question is more on a long time, how alpine is goin to be able to keep giving same level of sec with out those patches 2015-09-27 19:51:37 argami_: alpine already maintains an own branch for 3.18 based on the testing branch 2015-09-27 19:51:50 there is no information if alpine will get access to stable 2015-09-27 19:51:58 probably it will not happen 2015-09-27 19:54:01 that's what i tought 2015-09-28 00:20:15 guys crow pluesh nmeum thanks for the help i already have my 4.1.7 working!! 2015-09-28 08:36:29 ncopa: have been thinking about alternative init/supervision a bit. maybe we should first start with moving openrc init/conf files into subpackages like: http://sprunge.us/LTfj ? 2015-09-28 08:51:32 uggedal: yes 2015-09-28 08:51:39 i think that is a good start 2015-09-28 08:51:57 install_if="docs $pkgname=$pkgver-r$pkgrel" 2015-09-28 08:52:02 shoudl prbably be different 2015-09-28 08:52:20 install_if="openrc $pkgname......" 2015-09-28 08:52:51 but then we have some packages that has $pkgname-server subpackage 2015-09-28 08:52:57 then we need change the install_if 2015-09-28 08:53:06 ncopa: yeah :P 2015-09-28 08:53:27 i dont know if we want introduce this for v3.3 2015-09-28 08:54:21 when is the freeze? 2015-09-28 08:55:45 will community/staging make it for 3.3? 2015-09-28 09:11:10 i was thinking around 1 october 2015-09-28 09:11:23 community/staging is supposed to be before v3.3 yes 2015-09-28 09:11:28 when you say "alternative init/supervision", you're excluding systemd right? 2015-09-28 09:12:27 it's impossible to build systemd on musl 2015-09-28 09:12:40 so it excludes itself without our help 2015-09-28 09:12:54 barthalion: i did not know that 2015-09-28 09:15:00 there *was* uselessd 2015-09-28 09:15:09 well still is.. 2015-09-28 09:15:44 unmaintained 2015-09-28 09:16:22 same guy started writing his own init, because we need more NIH 2015-09-28 09:23:20 barthalion: regarding http://git.alpinelinux.org/cgit/aports/commit/?id=b4a13e7314bfe25e63b2f1ae5014d5429c17b539 2015-09-28 09:23:41 current newapkbuild still has it 2015-09-28 09:24:00 I think I pushed it to git there 2015-09-28 09:24:05 probably need to patch abuild in aports/main 2015-09-28 09:24:10 hm, true 2015-09-28 09:24:12 will do 2015-09-28 09:54:04 there was also an article about new init system here http://www.phoronix.com/scan.php?page=news_item&px=System-XVI-Service-Manager 2015-09-28 10:04:13 yeah, I meant this one 2015-09-28 12:03:27 ncopa: could you take a look at the pending patches in patchwork? 2015-09-28 12:36:05 ncopa: can we add this: http://sprunge.us/hNiH so that I can get abump to work with my container setup? 2015-09-28 12:38:47 uggedal: ok 2015-09-28 12:40:19 nmeum: wow patchwork is exploding 2015-09-28 12:41:00 nmeum's fault 2015-09-28 12:41:10 no 2015-09-28 12:41:12 many of those are superseded 2015-09-28 12:41:13 its my fault 2015-09-28 12:41:23 I'm joking guys :( 2015-09-28 12:41:40 i should have given nmeum git push access 2015-09-28 12:41:52 but wanted to wait til we got community repo up 2015-09-28 12:41:53 ok 2015-09-28 12:41:58 well, I asked you if I can do it 2 weeks ago 2015-09-28 12:42:03 I guess it went to /dev/null 2015-09-28 12:42:16 ncopa: yeah, just wait until the community repo is up 2015-09-28 12:42:33 which leads to: i need get it up *now* :) 2015-09-28 12:42:53 first candidate package to move to community? 2015-09-28 12:43:03 what is not superseded? 2015-09-28 12:43:23 also, does it mean you're going to rename testing to staging *now*? :P 2015-09-28 12:43:48 barthalion: probably not *now* 2015-09-28 12:43:56 but likely soonish 2015-09-28 12:43:57 maybe you should merge those patches first, since they won't apply anymore if you rename testing/ 2015-09-28 12:44:07 yeah 2015-09-28 12:44:21 better get community up first 2015-09-28 12:46:20 nmeum: do you have login to patchwork.a.o? 2015-09-28 12:47:48 no, I don't is it possible to close your own patches if you have one? 2015-09-28 12:47:48 nmeum: could you please create a patchwork login? then you can mark your patches superseeded yourself 2015-09-28 12:47:56 ah, alright will do so 2015-09-28 12:48:37 ncopa: it also means we're going to need new mailing list, or re-use alpine-aports for that 2015-09-28 12:48:43 and another patchwork project 2015-09-28 12:49:09 and configure all edge build servers to also build community repo 2015-09-28 12:49:18 i think we can reuse alpine-aports 2015-09-28 12:49:42 and reuse the patchwork project 2015-09-28 12:49:47 its just a subdir 2015-09-28 12:50:06 will not be a separate git repo 2015-09-28 12:50:45 hm 2015-09-28 12:50:56 we are going to use vref in gitolite? 2015-09-28 12:51:21 because I don't see other way to separate privileges to various directories without separate git repo 2015-09-28 12:52:39 barthalion: yes, i was thinking using gitolite to handle permissions in subdirs 2015-09-28 12:52:45 and branches 2015-09-28 12:52:58 branches are easy 2015-09-28 12:53:09 and subdirs? 2015-09-28 12:53:10 I'm not so sure about entire directories 2015-09-28 12:55:18 okay, doable 2015-09-28 12:55:23 aka. RTFM 2015-09-28 12:55:25 http://gitolite.com/gitolite/vref.html#NAME 2015-09-28 12:55:32 http://stackoverflow.com/questions/8047416/gitolite-or-gitosis-permission-on-directory-inside-of-the-repository 2015-09-28 12:58:31 so we can just do '- VREF/NAME/main = @trusted_contributors' or smth 2015-09-28 12:59:16 maybe we can try that right away 2015-09-28 12:59:25 and git nmeum access to testing 2015-09-28 12:59:29 give* 2015-09-28 12:59:32 also we might need https://github.com/sitaramc/gitolite/blob/master/src/VREF/refex-expr 2015-09-28 13:00:02 or maybe not 2015-09-28 13:00:06 no, I don't think so 2015-09-28 13:00:24 barthalion: do you think you test that out with nmeum? 2015-09-28 13:00:26 we allow regular devs to push everywhere and just disallow trusted users to touch anything in main 2015-09-28 13:00:33 sounds good 2015-09-28 13:00:47 ncopa: just tell me if you need my ssh public key 2015-09-28 13:00:59 nmeum: yeah, we do, send it to me or ncopa 2015-09-28 13:01:14 nmeum: send it to barthalion 2015-09-28 13:01:44 barthalion: verify that the ssh key you get has access to the email address published in the commits 2015-09-28 13:01:51 barthalion: see the pm I send you 2015-09-28 13:01:58 so it is not an AI bot 2015-09-28 13:02:02 :) 2015-09-28 13:03:03 i mean, anyone could create a nick on irc and claim to be the sören tempel that fills up our patchqueue 2015-09-28 13:07:05 :p 2015-09-28 13:14:56 hi everyone 2015-09-28 13:15:39 hi 2015-09-28 13:16:43 hi 2015-09-28 13:18:49 I'm trying to package libgit2 for alpine and I followed the wiki page http://wiki.alpinelinux.org/wiki/Creating_an_Alpine_package 2015-09-28 13:19:15 but when I abuild -r, shouldn't the final package be present somewhere ? 2015-09-28 13:19:48 it certainly should 2015-09-28 13:20:13 it could land in some directory in your $HOME, I don't remember its name because I'm a pineapple 2015-09-28 13:20:17 packages maybe? 2015-09-28 13:20:44 yeah just found it in $HOME/packages/testing/x86_64 2015-09-28 13:20:46 thanks :) 2015-09-28 13:35:28 i have another man page confict: 2015-09-28 13:35:31 ERROR: coreutils-doc-8.24-r0: trying to overwrite usr/share/man/man1/groups.1.gz owned by shadow-doc-4.2.1-r3 2015-09-28 13:35:57 ncopa: so everything is fine except it doesn't work 2015-09-28 13:36:14 I'm going to need ssh access to gitolite user, I need to debug it and look at logs 2015-09-28 13:36:26 oooh 2015-09-28 13:36:28 wait 2015-09-28 13:36:31 I'm stupid 2015-09-28 13:42:17 ncopa, does pingu add routing tables automatically? 2015-09-28 13:42:29 i'm referring to /etc/iproute2/rt_tables 2015-09-28 13:42:44 on pingu there's the directive route-table 2015-09-28 13:43:06 so i suppose pingu adds directly the route-table a runtime, and there's no need to specify it on rt_tables 2015-09-28 13:43:09 am i correct? 2015-09-28 13:53:25 fcolista: correct 2015-09-28 13:53:44 ok. Thanks ncopa. 2015-09-28 13:54:02 the only thing /etc/iproute2/rt_tables is used to be able to use names for routing tables instead of numbers 2015-09-28 13:54:21 but pingu does not use them 2015-09-28 13:54:25 ip rule show 2015-09-28 13:54:26 etc 2015-09-28 13:54:30 ok, good. 2015-09-28 13:54:41 will display number if its not in rt_tables 2015-09-28 13:55:03 so youcan specify route-table in pingu.conf and give it a name in rt_tables 2015-09-28 13:55:29 but pingu will not touch or use rt_tables 2015-09-28 13:55:46 it's clear now...thx! 2015-09-28 14:12:55 i am about to push openssh 7..1 2015-09-28 14:12:57 7.1 2015-09-28 14:13:16 it seems that it disables rsa pubkyes by default 2015-09-28 14:13:52 any objection to push openssh 7.1? 2015-09-28 14:14:57 it disables dsa, not rsa if i reacall correctly :) 2015-09-28 14:15:00 +1 with me 2015-09-28 14:15:45 oh i had dsa key then 2015-09-28 14:16:05 ok i push 2015-09-28 14:30:31 ncopa: just tested openssh 7.1 2015-09-28 14:30:34 works well 2015-09-28 14:31:56 good thanks 2015-09-28 15:57:26 hmmmm, since the openssh update i noticed something very strange 2015-09-28 15:57:43 when i ssh to a server and idle there it works fine an no connection loss 2015-09-28 15:58:12 but when i do a tail -f on the remote server and the file does not change, the connection breaks up and the ssh connection is gone 2015-09-28 21:31:48 oh boy ... i am talking with the rust people for 3 hours now about getting rustc to work on alpine 2015-09-28 21:31:53 this seams like an impossible task 2015-09-28 21:32:00 even if they say its "supported" 2015-09-28 21:35:32 https://github.com/rust-lang/rust/issues/28667 2015-09-28 21:40:58 Oh nice https://twitter.com/LeoUnglaub/status/643182531588792320 \o/ 2015-09-28 21:41:32 Oh, -echan 2015-09-28 21:45:38 jomat: yeah, i am trying to push alpine really hard into the red cross eco system 2015-09-28 21:45:59 it will take some years, but i am confidend i will get rid of that windows crap 2015-09-28 21:47:40 cool :-) your tweet was just mentioned in the Chaos Salzburg channel :-) 2015-09-28 21:54:47 well, greetings to the chaos salzburg channel from the metalab in vienna :) i am going to bed now 2015-09-28 21:54:50 see you guys later 2015-09-29 06:07:31 hello, what is the best way to submit small patches to the alpine packaging tools (apk-tools + abuild)? 2015-09-29 06:08:16 there are a few places with missing header files that just so happen to work on musl, but do not on other systems (such as in abuild-fetch.c, PATH_MAX is used without #include ), which makes it harder to target alpine in a chroot or such 2015-09-29 06:08:41 should I file a bug in redmine, or just send patches to the ML? 2015-09-29 07:09:22 awilfox: in the meantime you could use the static apk-tools version 2015-09-29 07:13:31 awilfox: if you have the time, you could do both. file a bug report and propose the solution in your patch :) 2015-09-29 07:17:20 ScrumpyJack: okay, will do. just wasn't sure which one (if either) was more preferred by the community. thanks 2015-09-29 07:17:34 uggedal: doesn't help abuild too much :) 2015-09-29 07:38:50 ncopa: will the builders handle missing dependencies (go on x86) or do I need to add arch='x86_64 armhf' to all go packages? 2015-09-29 07:44:20 the really strange thing is that it go/x86 no longer builds on 3.2 2015-09-29 07:44:47 maybe the musl upgrades is at fault 2015-09-29 08:13:49 does anyone see the reason why i get this oscam segfault? http://pastebin.com/pUVUv6zt (note its oscam just renamed to oscamcacheex) 2015-09-29 08:26:09 uggedal: the builders will not autodetect missing deps due to arch 2015-09-29 08:26:21 however, i think the go on x86 issue is fixable 2015-09-29 08:26:37 it is strange that it suggenly appeared 2015-09-29 08:26:42 suddenly* 2015-09-29 08:26:45 in v2.3 branch 2015-09-29 08:26:48 v3.2* 2015-09-29 08:27:00 but as i understand, it does work on vanilla kernel 2015-09-29 08:27:12 so i think we need paxmark somthing 2015-09-29 08:29:04 crow: it fails in fclose() 2015-09-29 08:29:20 i would check if it maybe close same file double 2015-09-29 08:29:44 or if there are some use of thread unsafe calls in threads 2015-09-29 08:30:02 but it is difficult to know without debug symbols 2015-09-29 08:39:30 ncopa: found the source of the go x86 failure. it was grsec kernel changes: https://github.com/alpinelinux/linux-stable-grsec/commit/1134b46895b28613a080b371fe70e4b440a10eb3#diff-80716dbec75df699671148c08bb3eb80R25 2015-09-29 08:43:18 thats good news 2015-09-29 08:43:25 ncopa well oscam debug symbols are there, seems i am missing the musl debug symbols? "Reading symbols from /usr/bin/oscamcacheex...done." 2015-09-29 08:44:49 uggedal: seems like we can fix it with a config then 2015-09-29 08:45:01 ncopa: sysctl -w kernel.modify_ldt=1 2015-09-29 08:45:19 makes it build here under grsec (from edge) in x86 lxc 2015-09-29 08:45:44 are the builders running under lxc or native? 2015-09-29 08:45:57 uggedal: lxc 2015-09-29 08:46:42 ok i have one more issue i need to deal with and then i'll try set that on the builders 2015-09-29 08:47:33 good 2015-09-29 08:47:59 I'll add a comment to the APKBUILD 2015-09-29 08:58:33 ncopa: just let me know when the builders are set and I'll push 2015-09-29 09:05:52 hi there 2015-09-29 09:53:18 uggedal: builders now have kernel.modify_ldt=1 2015-09-29 10:10:02 ncopa: thanks :) 2015-09-29 10:30:43 ncopa do you know maybe how to get musl symbols (as you wrote that is difficult to know without debug symbols) oscam one are there . 2015-09-29 10:34:41 crow: apk add musl-dbg ? 2015-09-29 10:35:56 uggedal thank you, i am hiding myself because of not knowing this from begining. 2015-09-29 10:40:54 ncopa now there are a lot of more info: http://pastebin.com/Ddkyi2kX 2015-09-29 10:54:07 seems it's calling fclose(0) 2015-09-29 10:55:08 glibc seems to check for that 2015-09-29 10:55:11 posix does not require it 2015-09-29 10:55:20 so musl crashes, because application is invoking UB 2015-09-29 10:55:43 sounds to me that oscam should be fixed 2015-09-29 11:09:55 fabled ok 2015-09-29 11:10:02 ill report it then on "other" side. 2015-09-29 11:11:20 we are happy to apply patches early, though 2015-09-29 11:17:53 fabled if i would only know how. 2015-09-29 11:20:08 sounds like flush_config_file (f=0x0, conf_filename=conf_filename@entry=0x4862241aba8 "oscam.srvid2") at oscam-conf.c:527 is at fault 2015-09-29 11:22:06 maybe it's permission issue 2015-09-29 11:22:12 it's trying to write oscam.srvid2 2015-09-29 11:22:15 can you strace ? 2015-09-29 11:22:20 ah i see let me check, 2015-09-29 11:22:32 the error path is still incorrect in code 2015-09-29 11:22:50 you have an example how to strace? 2015-09-29 11:25:45 just start the command as you would, but prefix it with strace 2015-09-29 11:26:03 you need to have done earlier: apk add strace 2015-09-29 11:26:05 so it's installed 2015-09-29 11:30:23 ok here is the strace http://pastebin.com/8pnV8C3b 2015-09-29 11:31:08 it may be differen with the oscam user (but it is /bin/false/ and i am not sure how to run it without init.d skript with this user 2015-09-29 11:37:20 can you run: strace -f ... 2015-09-29 11:37:25 to follow forks 2015-09-29 11:48:34 ncopa: what's your opinion on having a default -{bash,fish,zsh}-completion package in abuild? 2015-09-29 12:11:30 im a bit sceptical 2015-09-29 12:11:56 adding many hidden magic features may make it end up as gentoo 2015-09-29 12:12:03 with lots of magic stuff 2015-09-29 12:12:17 but you need to know a full API to actually understand whats going on 2015-09-29 12:12:36 do we have many *-completion candidates? 2015-09-29 12:13:28 i'd prefer not add default-*-completion unless there are many packages that it helps 2015-09-29 12:13:36 -doc and -dev are pretty many 2015-09-29 12:23:46 I totally get your point, so far there are not that many packages so you are probably right it isn't worth the complexity increase so far 2015-09-29 12:23:58 ncopa if i add -f it will never stop.. 2015-09-29 12:30:00 finally i got dsniff packaged 2015-09-29 12:30:02 \0/ 2015-09-29 12:33:07 umh. dsniff does not build on x86 2015-09-29 12:33:27 it's related to libnids 2015-09-29 12:40:53 ncopa, what can be the issue? 2015-09-29 12:40:53 /usr/lib/libnids.so: undefined reference to `ip_fast_csum' 2015-09-29 12:40:53 /usr/lib/libnids.so: undefined reference to `ip_compute_csum' 2015-09-29 12:40:53 /usr/lib/libnids.so: undefined reference to `my_tcp_check' 2015-09-29 12:40:53 /usr/lib/libnids.so: undefined reference to `my_udp_check' 2015-09-29 12:40:54 collect2: error: ld returned 1 exit status 2015-09-29 12:40:59 only for x86 2015-09-29 12:49:57 sounds like somethign was not built on x86 2015-09-29 13:10:40 fabled ncopa segfault fixed upstream (not by me) http://www.streamboard.tv/oscam/changeset/11019/ but this version we do not have in main. i was just testing newer version 2015-09-29 13:41:59 ncopa: do you think that it is a good idea to generate SSL certs and keys in post-install files? 2015-09-29 13:54:06 crow: should i push that fix to edge? 2015-09-29 13:57:33 nmeum: i think that might be an idea 2015-09-29 13:58:00 in any case, ssl certs in .apk are a bad idea 2015-09-29 13:58:53 ncopa: I am currently working on merging %533 and creates openssl certs without a config meaning it uses the default value for everything which I absolutly dislike 2015-09-29 13:59:01 sorry, I meant %593 2015-09-29 14:00:05 I decided to remove the post-install file completly and fixed a bunch of typos etc, if you don't have any objections then I would push this 2015-09-29 14:03:39 ok 2015-09-29 14:04:38 yes i agree 2015-09-29 14:06:35 BTW: don't want to annoy you but it would be nice if I could get additional privileges in patchwork in order to be able to mark patches as accepted/rejected, etc 2015-09-29 14:07:25 ncopa no leave it as is, ill check and send PR 2015-09-29 14:08:12 crow: i already pushed the patch 2015-09-29 14:08:34 nmeum: i thought i already gave you 2015-09-29 14:08:38 i'll do it in a min 2015-09-29 14:10:24 ncopa ok 2015-09-29 14:11:41 ncopa maybe to disable automaticly account registration on wiki.. http://wiki.alpinelinux.org/w/index.php?title=Special:RecentChanges&days=30&from= 2015-09-29 14:11:55 see older post. only to enable user "manuell let say on irc or such". 2015-09-29 15:08:32 took me few years to notice that there is no /opt 2015-09-29 15:08:37 for purpose I guess? 2015-09-29 15:52:45 fabled: I have a very strange issue with apk-tools where I managed to create a package which contains a file if you untar it but the file is not installed…any idea what might be causing this? 2015-09-29 15:53:05 does it give error? 2015-09-29 15:53:11 nope 2015-09-29 15:53:27 hidden files (beginning with dot) might get special treatment too depending on certain things 2015-09-29 15:53:49 yeah I looked at the code, but it is not a hidden file 2015-09-29 15:56:23 if you are interessted in it the following APKBUILD creates this package http://sprunge.us/VVPT 2015-09-29 15:56:44 not sure if anyone else can reproduce, but I can 2015-09-29 18:50:58 nmeum: what was that issue (seeing as you pushed with only options="!strip" added) 2015-09-29 18:52:11 I don't know what the issue is but if you increase the size of the binary apk-tools seems to install the binary properly 2015-09-29 18:52:40 and go binaries shouldn't be stripped none the less 2015-09-29 18:53:13 yeah, know that, but pretty strage about the file size 2015-09-29 18:54:02 could you reproduce the bug with the APKBUILD linked above? 2015-09-29 18:54:57 nmeum: testing now 2015-09-29 18:55:19 on x86_64 I guess? 2015-09-29 18:55:22 yep 2015-09-29 18:58:03 nmeum: seems to work here: http://sprunge.us/TZai 2015-09-29 18:58:24 I used the APKBUILD in testing and removed the !strip option 2015-09-29 18:58:31 hmhm 2015-09-29 19:00:04 apk -V 2015-09-29 19:00:06 apk-tools 2.6.4, compiled for x86_64 2015-09-29 19:00:30 apk-tools-2.6.4-r1 2015-09-29 19:04:31 which abuild version? 2015-09-29 19:06:43 nmeum: abuild-2.24.0-r7 2015-09-29 19:08:24 nmeum: this was on a minimal build container, only having alpine-base, abuild, build-base and patch installed 2015-09-29 19:40:55 ncopa: could you review this libfetch upgrade and resulting apk-tooks rebuild in the morning? compile/apk-add/update/upgrade tested ok here on x86/x86_64: http://sprunge.us/fgSX 2015-09-29 19:55:43 Arch never was upstream of libfetch 2015-09-29 19:56:06 we just took what was in cvs 2015-09-29 20:03:49 uggedal: where were you when I had to package libfetch myself :) 2015-09-29 20:04:56 barthalion: yeah, by upstream I meant source= :P 2015-09-30 00:43:46 hi 2015-09-30 06:38:19 morning 2015-09-30 06:38:47 uggedal: i have been thinking that we maybe should bundle libfetch sources in the apk-tools tarball 2015-09-30 06:39:12 because there are no libfetch packages for debian and other distros 2015-09-30 06:39:25 so it is difficult to build apk-tools anywhere else than alpine 2015-09-30 06:39:35 we also link to libfetch static 2015-09-30 06:40:21 so the benefit of having it as separate package is not that big 2015-09-30 06:40:46 it also looks like apk-tools is the only package that actually use it 2015-09-30 06:41:09 ncopa, i'm going to apply %591 2015-09-30 06:41:28 fcolista: ack 2015-09-30 06:41:32 thanks 2015-09-30 06:42:28 fabled: what do you think about bundle libfetch sources with apk-tools? 2015-09-30 06:42:43 i guess that'd be ok if wanted 2015-09-30 06:43:00 remends me: mdoc uses $PAGER, and with the default 'more' 2015-09-30 06:43:03 which does not work 2015-09-30 06:43:12 $ man ls 2015-09-30 06:43:12 more: -s: No such file or directory 2015-09-30 06:44:43 we have committer who send new aports without maintainer nor contributor. I don't care about contributor, but what about maintainer? 2015-09-30 06:44:53 If is testing we leave it as is? 2015-09-30 06:45:32 ncopa: sounds good 2015-09-30 06:46:31 fcolista: i'm ok to push things to testing without maintainer, but we should not move that to main unless someone are willing to maintain it 2015-09-30 06:46:52 wow 2015-09-30 06:47:02 in my buildenv without maintainer i got error 2015-09-30 06:47:18 package not available for that arch 2015-09-30 06:47:22 :-o 2015-09-30 06:48:04 without maintainer AND contributor 2015-09-30 07:04:07 excellent. I've upgraded the buildenv and i'm no longer able to do commits...due to openssh 7 2015-09-30 07:04:09 :-/ 2015-09-30 07:07:44 :) 2015-09-30 07:08:46 ncopa, i think i'm going to send you some new keys... 2015-09-30 08:07:54 community repo is alive 2015-09-30 08:09:05 \o/ 2015-09-30 08:09:42 now lets move majority of packages to there 2015-09-30 08:09:47 :) 2015-09-30 08:26:05 uggedal: i saw you discussed with skarnet yesterday about s6 2015-09-30 08:27:02 ncopa: yes 2015-09-30 08:27:09 he has this new s6-rc thing 2015-09-30 08:27:14 i know 2015-09-30 08:27:29 2015-09-30 08:27:30 packaging it now for testing to take a look at it 2015-09-30 08:27:35 ok i gotta go 2015-09-30 08:27:37 bbl 2015-09-30 08:57:04 back 2015-09-30 08:57:15 so s6 2015-09-30 08:57:21 ncopa: moved go and related packages to community 2015-09-30 08:57:28 great! 2015-09-30 08:57:29 thanks 2015-09-30 08:58:02 from what i have read about s6, it fits pretty well into the "alpine thinking" 2015-09-30 08:58:03 small 2015-09-30 08:58:05 simple 2015-09-30 08:58:07 secure 2015-09-30 08:58:10 yes 2015-09-30 08:58:17 i also like the supervision thingy 2015-09-30 08:58:22 vs pid files 2015-09-30 08:58:29 s6-rc with deps is ~2MB 2015-09-30 08:58:36 hm 2015-09-30 08:58:41 openrc is like 600k? 2015-09-30 08:58:54 might be smaller 2015-09-30 08:59:03 OK: 155 MiB in 52 packages 2015-09-30 08:59:08 OK: 157 MiB in 56 packages 2015-09-30 08:59:17 but above 1MB 2015-09-30 08:59:24 thats a bit big 2015-09-30 09:00:50 1644kb 2015-09-30 09:01:05 skalibs/execline/s6/s6-rc 2015-09-30 09:01:09 the only thing i have been a bit sceptic to is that sharnet is "weird" 2015-09-30 09:01:16 yeah 2015-09-30 09:01:31 even though the implementation is minimal I also find s6-rc a bit complex 2015-09-30 09:01:58 but many of the ideas behind are good i think 2015-09-30 09:02:30 i am normally not afraid of thinking different 2015-09-30 09:02:41 thats why we use musl libc and busybox instead of the traditional stuff 2015-09-30 09:02:48 if we want minimal we can do as sabotage, enable runit in busybox and run runsvdir from inittab with respawn 2015-09-30 09:03:02 but s6 is a bit too weird 2015-09-30 09:03:04 so it scares me 2015-09-30 09:03:34 then thing im not 100% happy with in runit is that it is poll vs push 2015-09-30 09:03:59 as in dependencies? 2015-09-30 09:04:15 yes 2015-09-30 09:04:18 as i understand it does 2015-09-30 09:04:21 start some service 2015-09-30 09:04:28 yeah, that can be a pain 2015-09-30 09:04:31 if not all deps are ok, then sellp for a while and retry 2015-09-30 09:04:35 sleep* 2015-09-30 09:04:55 for instance, under runit my transmission-daemon fails to start at boot 2015-09-30 09:05:01 since net is not up 2015-09-30 09:05:05 i kinda like the way systemd does it 2015-09-30 09:05:08 and it does not seem to recover 2015-09-30 09:05:32 when a resource becomes available it triggers start of the things that depends on it 2015-09-30 09:05:46 so when net comes up, everything that depends on net starts 2015-09-30 09:05:56 which is "push" approach 2015-09-30 09:06:17 yeah 2015-09-30 09:06:45 s6-rc has that 2015-09-30 09:06:55 thats what i think i remember 2015-09-30 09:07:15 and it has oneshots too, which must be hacked with: inittables-restore; pause 2015-09-30 09:07:18 in runit 2015-09-30 09:07:42 in theory those shouldnt be handled by runit 2015-09-30 09:08:16 i have kind of been hoping for that someone step up and create a systemd replacement 2015-09-30 09:08:16 well, you want a down version for some (like alsactl) 2015-09-30 09:08:35 https://github.com/ServiceManager/ServiceManager 2015-09-30 09:08:53 skarnet is one of the peoples i think could be able to do so 2015-09-30 09:08:55 he is clever 2015-09-30 09:09:10 http://homepage.ntlworld.com/jonathan.deboynepollard/Softwares/nosh.html (c++..) 2015-09-30 09:10:20 I think the best approach would start to make openrc optional 2015-09-30 09:10:25 and start to experiment a bit 2015-09-30 09:10:27 yes 2015-09-30 09:10:40 we cannot just swap it out in one shot 2015-09-30 09:10:59 so making openrc optional is a good start 2015-09-30 09:11:20 we don't even have to do that at first, could just change init= cmdline at first 2015-09-30 09:12:12 s6-rc would need triggers to recompile the service bundle when you install a apk 2015-09-30 09:13:19 and people would have to create their own bundle to list the services they want to start 2015-09-30 09:13:22 like https://github.com/skarnet/s6-rc/tree/master/examples/source/ok-all 2015-09-30 09:13:26 ok, you mee recompile the service? 2015-09-30 09:13:48 yes, the service tree is compiled to some kind of binary format 2015-09-30 09:14:31 doing that by triggers is ok 2015-09-30 09:14:39 unless it takes ages to do the recompile 2015-09-30 09:15:13 but enabling services is not as accessible as in init.d/runit 2015-09-30 09:15:36 lunch, bbl 2015-09-30 10:48:38 maybe firefox should be moved to community as well? seeing as the LTS releases are maintained shorter than alpine releases? 2015-09-30 11:02:28 yup 2015-09-30 11:02:30 it should 2015-09-30 11:03:32 i'm moving it 2015-09-30 11:03:38 same with chromium 2015-09-30 12:13:44 ncopa: regarding community msg on ml, wouldn't it be better to have it on the wiki? 2015-09-30 12:22:40 clandmeter: that too 2015-09-30 12:33:42 what about the 'whonix'guy: "... would you like that your project also has the ability to care for anonymize ?" could someone grasp his intention/idea? 2015-09-30 12:34:16 you mean TNT BOM BOM? 2015-09-30 12:34:21 yup 2015-09-30 12:34:24 :) 2015-09-30 12:34:26 hehe 2015-09-30 12:34:51 http://www.jornaldototonho.com.br/wp-content/uploads/2014/03/BOMBOM-500x311.jpg 2015-09-30 12:35:07 this is bom bom here in Brazil ^ 2015-09-30 12:35:26 bom bom is tasty! 2015-09-30 12:35:31 really good :) 2015-09-30 12:35:33 kkk 2015-09-30 12:40:20 ncopa: did you grant nmeum access to community repo? 2015-09-30 12:42:21 ah, you didn't need to 2015-09-30 12:48:47 i think he has access to all except main/ 2015-09-30 12:54:29 are there any other important things we need before v3.3 feature freeze? 2015-09-30 12:54:45 uggedal btw, we will have a "soft" feature freeze in oct 2015-09-30 12:55:07 but will will not branch 3.3-stable til we do the actual release, in nov 2015-09-30 12:55:23 so during october i woudl like to focus on bugfixing only 2015-09-30 12:55:35 for v3.3 2015-09-30 12:57:32 ncopa, nlplug work? and i'd like to look at making root be squashfs+overlayfs+tmpfs mount -- so it's easy to create firmwares for embedded devices 2015-09-30 13:10:29 ncopa: ok :) 2015-09-30 13:10:53 btw, what about linux 4.1? 2015-09-30 13:15:03 3.18 has projected EOL jan 2017 (could be a problem for 3.2 also...) 2015-09-30 13:16:44 shouldnt here gzip be installed on default as dep: http://sprunge.us/jPgK 2015-09-30 13:17:32 i see in lxc APKBUILD have depends="bash gzip" 2015-09-30 13:18:48 crow: which repo do you use? 2015-09-30 13:19:04 lxc-checkconfig won't work on grsec kernels btw 2015-09-30 13:19:05 stable doesnt have gzip yet 2015-09-30 13:19:19 no /proc/config.gz 2015-09-30 13:20:11 bash and gzip dep should be moved to subpkgs which needs it. 2015-09-30 13:22:16 clandmeter i am on main 2015-09-30 13:22:37 uggedal well if i add gzip and redo it again it show nice output at least /second part of paste 2015-09-30 13:22:43 crow: edge or stable? 2015-09-30 13:23:10 clandmeter stable: http://nl.alpinelinux.org/alpine/v3.2/main 2015-09-30 13:23:18 i am on 3.2.3 2015-09-30 13:24:05 http://git.alpinelinux.org/cgit/aports/tree/main/lxc/APKBUILD?h=3.2-stable 2015-09-30 13:24:35 clandmeter ah yes i was checking trunk APKGBUILD 2015-09-30 13:24:53 well can the gzip be added to stable? 2015-09-30 13:48:19 everything is possible :) 2015-09-30 13:49:55 crow: might just be linux-virtgrsec kernel which is missing /proc/config.gz 2015-09-30 13:50:03 ncopa: so its safe to add community to pkgs.a.o? 2015-09-30 13:51:05 uggedal or that, but for me addind gzip it solves the problem. 2015-09-30 14:30:07 ncopa: I still don't have permisson to mark patches as accepted/rejected in patchwork 2015-09-30 14:41:43 nmeum: well, there is no need to do that manually if you push the patch as it is 2015-09-30 14:42:41 sure, but for instance I would like to reject %630 since the author readded his post-install script which I explitcitly removed 2015-09-30 14:43:26 reject is not enough, send also the mail to alpine-aports 2015-09-30 14:44:11 yeah, I am writing that email currently 2015-09-30 14:44:35 I don't think I have any special permissions on patchwork, clandmeter can help you 2015-09-30 14:46:55 uggedal: you can modprobe configs 2015-09-30 14:47:04 yes we want linux 4.1 2015-09-30 14:47:12 i need to do that this week :-/ 2015-09-30 14:48:29 clandmeter: yes you can safely add community to pkgs.a.o 2015-09-30 14:50:51 ncopa: i already did :) 2015-09-30 14:52:58 nmeum: did you try logout login? 2015-09-30 14:54:19 i have igven you all permssions in patchwork 2015-09-30 15:15:54 ncopa: I'm not sure if you saw my question in the other channel, but where would I look in the repo to find the scripts that generate the bootable system 2015-09-30 15:16:26 or are there steps on rebuilding a bootable iso with pre-installed packages (not using LBU/apk cache to install them live). 2015-09-30 15:30:49 a18cf0d3343c487f4a6a262f0038f21d95c35b66 just broke git-send-email http://sprunge.us/TQMW could somebody with access to main/ please revert it? 2015-09-30 17:28:57 ++ $message") or die $smtp->message; 2015-09-30 17:29:00 here's the issue 2015-09-30 17:29:43 it looks like it should be $smtp->datasend("$message") or die $smtp->message; 2015-09-30 17:31:49 or should just be gone 2015-09-30 18:41:52 if we're going to have %631 can we also http://sprunge.us/JGag 2015-09-30 18:42:13 huh, the ' probably shouldn't be happening here 2015-09-30 18:42:31 ... or the " 2015-09-30 23:02:45 nmeum: did %638 fix send-email, it'd be embarrassing if it didn't 2015-09-30 23:31:56 ammunta: didn't test it yet, will do so tomorrow 2015-09-30 23:38:13 ammunta: btw are you the guy who suggested the patch? did you suggest it upstream? 2015-09-30 23:38:37 I'm not 2015-09-30 23:38:59 I just noticed that the patch didn't actually match the source 2015-09-30 23:42:01 tbh I'm not even sure where upstream would be to send it to 2015-09-30 23:45:35 if anything it should probably go up-upstream to perl