2015-11-01 08:09:37 Is it possible to install wayland in alpine edge? I found the packages, but 'apk add wayland' does not do a thing. ;-) 2015-11-01 08:10:41 Definitely 'setup-xorg-base' installs only xorg, not wayland, afaig. 2015-11-01 08:22:33 I just tweaked the setup-xorg-base with all wayland packages and started a setup-wayland-base (http://hastebin.com/oxobopahiv.bash), but got "ERROR: unsatisfiable constraints:" against all wayland packages. 2015-11-01 11:36:22 what does apk info -L wayland say? 2015-11-01 12:07:49 hi all. I'm trying to get webkitgtk 2.10 on alpine. I'm having problems with `malloc_trim` (GNU extension, sigh). How do you deal with that? ifdef it out, or replace it with some brk() arcane tricks? 2015-11-01 12:46:18 ScrumpyJack: http://hastebin.com/wuliruzodi.hs 2015-11-01 18:38:44 hrnr: usually the first 2015-11-02 08:05:43 cool 2015-11-02 08:05:54 i got an email from "help desk" about my alpinelinux account 2015-11-02 08:06:01 "Your account is in the process of being upgraded to a newest 2015-11-02 08:06:01 Windows-based servers and an enhanced online email interface inline with internet infrastructure Maintenance. The new servers will provide better anti-spam and anti-virus functions, along with IMAP Support for mobile devices to enhance your usage." 2015-11-02 08:06:16 they needed my: 2015-11-02 08:06:18 * Domain\user name: 2015-11-02 08:06:18 * Password: 2015-11-02 08:07:01 i didnt know that the alpine infra was switching to windows :) 2015-11-02 08:11:16 not just switched, upgraded ;) 2015-11-02 08:17:40 ncopa: can modules.dep be split into multiple-files ? 2015-11-02 08:18:02 vkris: not that i know 2015-11-02 08:18:52 just thinking, if it could then having multiple modloop (smaller) would be easy 2015-11-02 08:19:45 but not sure if it would be secure 2015-11-02 08:22:52 is modules.dep needed afer boot ? 2015-11-02 08:23:07 vkris, yes, modprobe uses it 2015-11-02 08:23:23 when you hotplug stuff, it's needed 2015-11-02 08:24:04 so a modified modprobe and seek multiple .dep file 2015-11-02 08:24:15 can be made to seek 2015-11-02 08:24:55 that would be possible, but it'd decrease performance to some extent 2015-11-02 08:25:18 hmm 2015-11-02 08:25:21 only when we add more .modloop squashfs files 2015-11-02 08:25:31 other wise just normal 2015-11-02 08:25:50 how would you compine the multiple squashfs images? 2015-11-02 08:26:00 which means we can cut huge modloop 2015-11-02 08:26:38 we already cut several tens of megabytes with the firmware splitting 2015-11-02 08:26:43 secret, mount can do a mounting inside a readonly directory 2015-11-02 08:26:48 :-)) 2015-11-02 08:27:12 i think we dropped 30-40MB from modloop just by filtering the firmware images 2015-11-02 08:27:44 so there would be dirs like .modloop/modules/3.14.30-0-grsec/1 2015-11-02 08:27:58 preset in main modloop 2015-11-02 08:28:12 to which sub-modloop just mounts 2015-11-02 08:29:05 so if .modloop/modules/3.14.30-0-grsec/1/modules.dep present seek it 2015-11-02 08:30:06 it's complicated to do 2015-11-02 08:30:17 you'd need to patch busybox and kmod 2015-11-02 08:30:20 hmmm... just a thought 2015-11-02 08:30:39 and if the dependencies cross directory boundaries 2015-11-02 08:30:53 e.g. to load device X; you need from 1/foo/bar.ko, and 2/bar/foo.ko 2015-11-02 08:31:05 the tools would need to be able to load all indexes simultaneous 2015-11-02 08:31:12 it's doable 2015-11-02 08:31:18 but rather complicated 2015-11-02 08:31:26 especially with kmod having the .bin index files too 2015-11-02 08:31:29 yes, load all .dep file first 2015-11-02 08:31:38 ahh 2015-11-02 08:31:45 kmod does not load them fully 2015-11-02 08:31:56 instead it's doing just lookups on the .bin files 2015-11-02 08:32:44 i'm not against the idea per se 2015-11-02 08:32:56 it's just.... complicated to implement 2015-11-02 08:33:16 need to read more on how kmod works 2015-11-02 08:33:43 just thought, it would nice to have 10mb AL iso 2015-11-02 08:34:09 with the firmware split 2015-11-02 08:34:19 modloop went from 80MB -> 45MB 2015-11-02 08:34:25 yes, hardware catogorized 2015-11-02 08:34:58 ok 2015-11-02 08:51:13 or the fully generated .dep or .bin file can be there in primary modloop, only issue would be error msg when .ko is not found 2015-11-02 08:52:59 unless something verifies all files presence at boot, that would be havoc 2015-11-02 08:53:37 that's one option 2015-11-02 08:54:28 I think puppylinux dev have fould a means to merge 2 sfs files, but did not check the codes 2015-11-02 08:55:22 merge+mount 2015-11-02 08:55:28 unionfs/overlayfs is one option 2015-11-02 08:56:28 yes, but is a security issue 2015-11-02 08:57:21 hmm.. moment, live mount into primary modloop ? 2015-11-02 08:57:36 without reboot, can remount do it ? 2015-11-02 09:07:50 ? 2015-11-02 09:08:56 with overlayfs a.squashfs+b.squashfs are unionfs'd at boot 2015-11-02 09:10:00 or can i just say apk add b.squashfs and it would get attached to running/live AL ? 2015-11-02 09:11:34 that sort of goes to the area what Docker is trying to solve (to some extent) 2015-11-02 09:11:35 hope, I am making sense 2015-11-02 09:11:51 apk is not using squashfs at all 2015-11-02 09:12:55 but if I do ram install and need more kernel modules, hw specific ? 2015-11-02 09:13:27 on baremetal 2015-11-02 09:14:00 we don't currently have good solution for that. but we have had some discussions on it. 2015-11-02 09:14:18 one option is to split modules to a bit more finegrained packages, and install the ones you need 2015-11-02 09:14:28 modloop would not be needed at all then 2015-11-02 09:14:43 other option is to mount modloop+tmpfs with overlayfs 2015-11-02 09:14:53 so apk can install packages there 2015-11-02 09:16:19 without modloop it would take huge space on ram 2015-11-02 09:17:13 would just be regular .apk install 2015-11-02 09:33:20 Fedora-23 - would be comming with python3x as default 2015-11-02 09:59:29 didi read something here about webkitgtk? 2015-11-02 10:01:11 looks like it's been renamed to webkitgtk2 2015-11-02 10:02:42 anyway, could someone rebuild it in edge as it's build against ibicui18n.so.53 (from icu-libs-53) which was upgraded to icu-libs-55 2015-11-02 10:22:38 just webkitgtk2 or webkitgtk too? 2015-11-02 10:27:11 webkitgtk2 if you don't mind 2015-11-02 10:27:50 oops, i second 2015-11-02 10:27:57 s/i/1 2015-11-02 10:34:34 webkitgtk in yes, webkitgtk looks ok 2015-11-02 10:39:25 that was messy. i meant just webkitgtk2 :) 2015-11-02 10:59:15 Halloween 3.3 release \o/ 2015-11-02 10:59:31 is 3.3 spooky? :p 2015-11-02 11:03:58 a linux present in nice envelop(candy wrapper) 2015-11-02 11:04:07 i think we merged webkitgtk2 into webkitgtk 2015-11-02 11:04:20 ALPINE 2015-11-02 11:05:05 trick or alpine 2015-11-02 11:08:52 webkitgtk is in community and webkitgtk2 is in testing 2015-11-02 11:09:44 perhaps one could get knocked out? 2015-11-02 11:12:47 I rebuilt the one in testing… 2015-11-02 11:14:54 awesome thanks dude 2015-11-02 11:14:57 We AL devs are creature of habits, we still check main/ and testing/, community is at foothills of alpine, not peaky enough ! 2015-11-02 11:15:49 but yeah, looks like testing one should get removed 2015-11-02 11:17:33 also testing/davfs2 pls 2015-11-02 11:23:04 i now have webkitgtk2-2.4.9-r0 and webkitgtk2-2.4.5-r0 both appearing in apk search 2015-11-02 11:26:11 hold on to davfs2, the APKBUILDs are different 2015-11-02 11:26:36 testing has --disable-nls, maybe clandmeter would confirm 2015-11-02 11:26:48 you can trow mine away 2015-11-02 11:26:53 the one in testing. 2015-11-02 11:30:14 ScrumpyJack: I ditched the 2.4.5 one 2015-11-02 11:31:23 ta 2015-11-02 11:32:35 hmm, still in search 2015-11-02 11:33:16 even though rsync to my mirror deleted it 2015-11-02 11:36:45 arg, mirror is sending back Error 304 - Not modified to apk update 2015-11-02 11:38:55 ScrumpyJack, --force 2015-11-02 11:39:03 it'll skip the If-Updated-Since 2015-11-02 11:39:09 i think that's actually an apk bug 2015-11-02 11:39:18 i have it on my todo list to check that 2015-11-02 11:39:47 i think the problem is that when index is downloaded, it's just stored to disk (with current system timestamp) 2015-11-02 11:40:02 and If-Updated-Since is probably sent with wrong TZ 2015-11-02 11:46:24 yeah, just looking at apk_io and url. time_t and since. 2015-11-02 11:47:38 --force didn't help 2015-11-02 11:49:26 (188/191) Installing webkitgtk2 (2.4.9-r0) 2015-11-02 11:49:26 (189/191) Installing webkitgtk (2.4.9-r0) 2015-11-02 11:49:26 (190/191) Installing webkitgtk-dev (2.4.9-r0) 2015-11-02 11:49:35 messy :) 2015-11-02 11:58:34 any thoughts on raising some revenue from 'AL on Stick' crowdfunded hardware ? 2015-11-02 12:00:00 if goes well, work-from-home/parttime could become possible 2015-11-02 12:03:28 vkris, "AL on Stick" ? 2015-11-02 12:03:49 you seen https://inversepath.com/usbarmory ? 2015-11-02 12:08:36 fabled: nope, a usb based ram 2015-11-02 12:09:02 since we moving to usb v3x it would be nice 2015-11-02 12:09:20 don't need to depend on hw ram 2015-11-02 12:11:23 but the full design has more feature, ready to discuss if possible 2015-11-02 12:12:31 if would envolve some patenting too, but plan to release the patent to public domain after resonable earning or 2-4yrs 2015-11-02 12:18:28 but usbarmory, is also nice, maybe sell pre-installed AL 2015-11-02 12:57:38 usbarmory costs 100 euro/piece? WoW 2015-11-02 13:03:14 CHIP is cheaper :D 2015-11-02 14:39:20 what causes this error during at the end of a build? 2015-11-02 14:39:20 >>> vimb: Updating the cached abuild repository index... 2015-11-02 14:39:21 fatal: No names found, cannot describe anything. 2015-11-02 14:46:06 tail has started complaining about inotify 2015-11-02 14:46:08 tail: inotify cannot be used, reverting to polling: No file descriptors available 2015-11-02 14:49:16 that is busybox inotfiy 2015-11-02 14:51:15 we have inotify-tools-inc and inotify-tools 2015-11-02 14:51:30 they look the same and conflict with each other 2015-11-02 15:07:20 ScrumpyJack - inotify-tools-inc is inotify-tools patched to include an include regex (which is missing in the standard inotify-tools) 2015-11-02 15:08:13 with the ordinary version there is only an exclude regex 2015-11-02 15:10:09 inotify-tools-inc should automatically replace inotify-tools (the APKBUILD has replaces="inotify-tools") 2015-11-02 15:25:16 apk add inotify-tool works 2015-11-02 15:25:32 then apk add inotify-tool-inc -> http://tpaste.us/GEMD 2015-11-02 15:25:57 so i guess the replace isn't triggering 2015-11-02 15:58:29 Firefox 42 is available :) 2015-11-02 15:58:36 oh an hi btw :) 2015-11-02 18:49:53 is there a way to put ' conflicts="$pkg" ' in an APKBUILD ? (depends=!$pkg - did not work) 2015-11-02 18:50:57 BitL0G1c, depends="!pkg" 2015-11-02 18:51:08 why it did not work? 2015-11-02 18:51:40 you need it in quotes 2015-11-02 18:59:45 yes had it in quotes 2015-11-02 19:02:09 will try again 2015-11-02 19:28:43 ScrumpyJack: you don't have git installed 2015-11-02 19:28:56 ScrumpyJack: that "error" should be silented, will take a look tomorrow 2015-11-02 23:23:16 minor issue with 72abc71 main/busybox: split package into core and suid subpackages 2015-11-02 23:23:32 it seems to have deleted the install= line 2015-11-02 23:23:42 so the post-install hook isn't executing 2015-11-02 23:25:20 this is causing issues when building a new rootfs because chown and chmod aren't installed when the baselayout post-install hook runs 2015-11-02 23:30:09 sent in a patch 2015-11-02 23:30:26 untested but I'm p sure it's correct 2015-11-02 23:30:41 (never trust "untested but I'm p sure it's correct") 2015-11-02 23:33:42 uh is tracking latest-stable supported 2015-11-03 07:20:53 lxqt looks fairly nice 2015-11-03 07:46:27 :-) 2015-11-03 08:34:43 morning 2015-11-03 08:50:55 morning 2015-11-03 08:51:19 o/ 2015-11-03 08:51:52 BitL0G1c: why we introduce inotify-tools-inc instead of just adding regex to the inotify-tools? 2015-11-03 09:09:33 huh, when did the 3.3 packaging start happening 2015-11-03 09:09:44 is 3.3 stable even branched yet 2015-11-03 09:19:39 ammunta +1 2015-11-03 09:43:16 not branched 2015-11-03 09:43:27 we branch with v3.3.0 release 2015-11-03 09:43:48 i started the v3.3 builders last week 2015-11-03 11:10:16 ncopa: could you take a look at %833, forgot to send that patch with the xf86-input-evdev package upgrade 2015-11-03 11:10:28 should have send that in one email (sorry) 2015-11-03 11:10:37 hey :) 2015-11-03 11:10:44 ERROR: xf86-input-evdev-2.10.0-r1: trying to overwrite usr/share/X11/xorg.conf.d/10-evdev.conf owned by xorg-server-1.17.4-r0. 2015-11-03 11:10:51 i got that today during upgrading 2015-11-03 11:11:26 yeah that's what I fixed above 2015-11-03 11:11:50 I did that change when I upgrade the package but I forgot to send the patch back then 2015-11-03 11:11:55 *upgraded 2015-11-03 11:23:49 ah great :) 2015-11-03 11:23:51 thanks 2015-11-03 11:27:50 no problem, it's my fault after all 2015-11-03 13:15:22 is there a desktop dock in repos? I couldn't find one (like cairo-dock on wbar) 2015-11-03 13:50:50 hi 2015-11-03 13:51:08 nmeum: doesnt it make more sense to ship the 10-evdev.conf with xf86-input-evdev? 2015-11-03 13:53:56 I compared them and they didn't differ, but yeah in general it would make more sense to ship the file in the xf86-input-evdev package 2015-11-03 13:54:35 should I sent you a patch for that? 2015-11-03 13:58:16 i suppose you can tpaste it here 2015-11-03 13:58:28 git format-patch -1 --stdout | tpaste 2015-11-03 14:06:31 ncopa: http://sprunge.us/CMYO (simple patch of cause) 2015-11-03 14:07:02 didn't recompile xorg though because I am currently in a lecture 2015-11-03 14:08:33 ok, thanks 2015-11-03 14:08:37 i'll do that for you 2015-11-03 14:15:43 thanks 2015-11-03 15:59:02 I am moving the systemd-style net iface names to a eudev-netifnames package 2015-11-03 15:59:11 so you will have to opt-in if you want it 2015-11-03 16:06:51 so back to the old eth0 style? 2015-11-03 16:06:53 nice! 2015-11-03 16:06:58 ACTION is oldschool 2015-11-03 16:07:00 :) 2015-11-03 16:07:37 you can get newschool with apk add eudev-netifnames 2015-11-03 16:07:52 nah ... i am oldschool .. 2015-11-03 16:07:58 in case you want seem hip among your friends... :) 2015-11-03 16:08:15 for that i have ubuntu on a live stick *g* 2015-11-03 16:09:27 going home 2015-11-03 16:09:27 can you honestly still call ubuntu hip? 2015-11-03 16:09:41 see u 2015-11-03 16:11:53 chris|: amongst windows xp users ... sure 2015-11-03 17:04:46 clandmeter: do you maintain the alpine-aports mailing list? 2015-11-03 17:04:57 my patches aren't getting through 2015-11-03 17:06:41 i'm using git-format-patch and that goes through fine to other destinations, just not alpine-aports@lists.alpinelinux.org 2015-11-03 17:07:41 looking at the output git format-patch creates, there is an extra From line in the email header, although it's not a From: (missing :) 2015-11-03 17:07:51 does that confuse the mailing list software? 2015-11-03 17:08:12 or perhaps i'm marked as spam *gulp* 2015-11-04 08:01:12 ScrumpyJack: no i dont manage it. could be your smtp is blacklisted? 2015-11-04 08:19:25 I have no idea how to deal with those Go applications 2015-11-04 08:19:35 how do other distros deal with them? 2015-11-04 08:21:36 ncopa: I remember i linked you gentoo's helper 2015-11-04 08:21:54 dont remember how complete it is. 2015-11-04 08:22:27 https://groups.google.com/forum/#!topic/golang-nuts/-UG-lF_Ey-o 2015-11-04 08:22:35 but ones i mentioned Go you run away... (like me) 2015-11-04 08:25:26 https://gitweb.gentoo.org/repo/gentoo.git/tree/eclass/golang-build.eclass 2015-11-04 08:33:31 another benefit of seperate folder/aports like community/ is in event it becomes burdening/bottleneck on main/ builders they can be subdued and may even be postponed building after main release 2015-11-04 08:33:59 no 2015-11-04 08:34:01 why not have go-aports !? 2015-11-04 08:34:11 i want community to build at release time 2015-11-04 08:34:26 ok 2015-11-04 08:35:18 go introduces all new ways of building pkgs, so a different treatment 2015-11-04 08:35:52 till we figure out 2015-11-04 08:37:33 note that "6months cycle/human resource" on large community/ may fall short 2015-11-04 08:39:33 plus EU's lack for supporting opensource economic asylum !! ;-) 2015-11-04 08:40:23 the idea of community was to reduce the 2yr cycle to 6month 2015-11-04 08:41:39 yes, with given perpetual community devs hanging around 2015-11-04 08:41:51 that is a uncertainty factor 2015-11-04 08:42:23 and given economic situation, it may just creep into main/ 2015-11-04 08:44:13 gee, where can I apply for economic refuge ?! 2015-11-04 09:16:39 clandmeter: do you know who does manage it? i'll email list-owner 2015-11-04 09:37:08 if at-all, large community/ build full release can be spanned to full-EOF/2 for a release, roughly 1yr 2015-11-04 09:39:07 vkris: the plan is to only support community for the last stable release 2015-11-04 09:39:18 so 6 months 2015-11-04 09:40:34 is there no X11 calculator in the repos? 2015-11-04 09:41:06 galculator 2015-11-04 09:41:31 thanks, i was searching calc 2015-11-04 09:41:54 that's way too simple, the cool kids all use a 'g' prefix 2015-11-04 09:41:57 search %calc% :& 2015-11-04 09:42:25 even better, which means if community/ continues to build after main release, it is still all good 2015-11-04 09:43:48 which gives 2yrs for it to build (6months+1.5yrs) 2015-11-04 10:05:35 ah!, https://github.com/twitter/twemproxy/wiki, and who said wiki.a.o was incomplete 2015-11-04 10:06:17 looks like bot generated wiki 2015-11-04 10:11:08 quick apk question 2015-11-04 10:11:43 in an APKBUILD comment, what's the difference between the "Contributor:" entry and the "Maintainer:" one? 2015-11-04 10:12:39 I know what a maintainer is, but what is a "contributor"? The original author? someone else? 2015-11-04 10:18:25 Contributor is one that have contributed by improving the APKBUILD somehow afaik. 2015-11-04 10:18:45 But i think you have already figured ;) 2015-11-04 10:19:07 so, contributor to the APKBUILD, not to the software. 2015-11-04 10:19:15 yes to the apkbuild 2015-11-04 10:19:21 yes 2015-11-04 10:19:29 So I guess you can have several Contributor lines. 2015-11-04 10:19:39 yes 2015-11-04 10:19:48 and its only a comment in apkbuild 2015-11-04 10:19:53 that's a trend we had in the last few months. But yes. 2015-11-04 10:20:02 ok, thanks for the info. 2015-11-04 10:20:22 yw 2015-11-04 10:20:40 clandmeter, you around? 2015-11-04 10:21:57 make DESTDIR="$pkgdir" install 2015-11-04 10:22:09 should be make DESTDIR="$pkgdir/" install 2015-11-04 10:22:21 unless $pkgdir is slash-terminated 2015-11-04 10:23:25 the DESTDIR convention prepends paths with $DESTDIR without a slash 2015-11-04 10:23:42 $(DESTDIR)$(INSTALL_LOCATION) 2015-11-04 10:23:59 so $DESTDIR should end with a slash, always 2015-11-04 10:24:53 silly question: what if "$INSTALL_LOCATION" begins with a slash? 2015-11-04 10:25:40 then $(DESTDIR)$(INSTALL_LOCATION) has a double slash, which is resolved as a single slash, it doesn't matter 2015-11-04 10:26:39 of course if you know that $(INSTALL_LOCATION) is always absolute, then it works anyway, but is it always the case in abuild? 2015-11-04 10:27:36 sounds correct 2015-11-04 10:29:07 as long as you always install stuff into an absolute location (which makes sense for a package manager...) then there's no problem 2015-11-04 10:34:09 skarnet: so I guess I can skip sending patches for the new skarlibs/execline/s6* packages then? :) 2015-11-04 10:35:21 yup, I'll do those, once I figure out how to write an APKBUILD and use git-format-patch and git-send-email ^^' 2015-11-04 10:35:53 hint: you don't need format-patch when you send-email ;) 2015-11-04 10:36:24 I'm moving stuff from /usr/bin to /bin, are there any precautions to take as long as nothing (afaik) depends on the moved stuff? 2015-11-04 10:37:02 nah 2015-11-04 10:37:55 there might be scripts with hardoced paths? 2015-11-04 10:38:01 probably not many 2015-11-04 10:38:33 yeah, that's what I was worried about, but not many people use execline (yet!) 2015-11-04 10:39:16 better do that change sooner than later 2015-11-04 10:40:10 chris|: i think the boost update will require rebuild of many packages 2015-11-04 10:40:11 In a doc() function, if you don't use default_doc, do you have to mkdir $subpkgdir/usr/share/doc/foobar, or is it already created by default? 2015-11-04 10:40:25 s6/execline has also not been in main in any stable release jet, so little chance of people using it yet for anything other than testing 2015-11-04 10:43:24 skarnet: i think you need create it manually 2015-11-04 10:46:11 ncopa: yes, probably 2015-11-04 10:48:17 ncopa: I am not keen on the new version, I just did it while I was adding python3 support 2015-11-04 10:48:37 i think we want upgrade while at it 2015-11-04 10:52:00 fcolista: yes 2015-11-04 11:00:56 I have subpackages="$pkgname-dev $pkgname-doc" 2015-11-04 11:01:06 $pkgname-dev works fine, but not $pkgname-doc : 2015-11-04 11:01:15 >>> ERROR: skalibs-doc*: Missing /home/ska/src/aports/main/skalibs/pkg/skalibs-doc 2015-11-04 11:01:23 what am I doing wrong? 2015-11-04 11:02:05 you are missing mkdir -p "$pkgdir" in your doc() I would guess 2015-11-04 11:02:11 i think you need mkdir -p "$subpkgdir"/usr/share/doc 2015-11-04 11:02:14 in your doc() 2015-11-04 11:02:27 I called default_doc 2015-11-04 11:02:35 just like dev() calls default_dev 2015-11-04 11:02:56 and dev() is working 2015-11-04 11:03:06 if it does not find any docs as expected then it will probably not create that dir 2015-11-04 11:03:18 ah, ok. Thanks. 2015-11-04 11:03:51 (See, this is why I hate magic. Why wouldn't default_doc create the dir if it doesn't find the docs it wants?) 2015-11-04 11:06:30 yeah... i tried hard to avoid magic 2015-11-04 11:06:47 but fall for the temptation for some commonly used things like -dev and -doc 2015-11-04 11:08:23 where should I set arch to "noarch" for the doc package, if it's useful? in the doc() function? 2015-11-04 11:08:36 in practice no 2015-11-04 11:08:48 its no practicaly use currently 2015-11-04 11:09:02 so all abuilds have a warning? :) 2015-11-04 11:09:41 you set it in the doc() function 2015-11-04 11:09:46 that is a misfeature 2015-11-04 11:10:03 becuase that makes it impossible to know all arch targets from global scope 2015-11-04 11:10:16 I see. 2015-11-04 11:10:18 impossible to know the arch targets ithout actually building the package 2015-11-04 11:10:31 that also applies to depends of subpackages 2015-11-04 11:10:52 this is also main reason why i have talked about new APKBUILD format for the last year(s) 2015-11-04 11:11:20 gah, I'm just getting into apkbuild, please don't change it now XD 2015-11-04 11:11:26 :) 2015-11-04 11:11:45 i realized it will be close to impossible to change it dramatically at this point 2015-11-04 11:11:51 ok, abuild doesn't remove the usr/share/doc/* I create for the -doc package 2015-11-04 11:11:59 what am I missing here? 2015-11-04 11:12:07 i am missing lunch :) 2015-11-04 11:12:16 that's also true 2015-11-04 11:17:19 aha, the usr/ stuff is removed if I do "cd $subpkgdir", and isn't if I don't. 2015-11-04 11:17:41 Suggestion: have the script remove what's under $subpkgdir, not what's under the cwd... 2015-11-04 11:18:11 (it only does this for doc, not for dev) 2015-11-04 11:19:49 ncopa: http://kodsnack.se/120/ ;) 2015-11-04 11:53:37 the shared libs in /lib are totally devoid of symbols, even the global ones. Is that normal? 2015-11-04 11:54:22 ... how do you even dynlink without symbols? 2015-11-04 11:56:41 "strings" shows the identifiers are there... but nm -a doesn't. What. The. 2015-11-04 11:58:38 More questions... it looks like the -dev package depends on the -doc package by default. How can I prevent that? 2015-11-04 12:04:46 skarnet: i think we fixed that 2015-11-04 12:05:12 I'm using the latest abuild 2015-11-04 12:05:55 hum, i thought taht was fixed 2015-11-04 12:06:09 did you apk add docs? 2015-11-04 12:06:39 when I use "apk add skalibs-dev", it auto-installs skalibs-doc first 2015-11-04 12:07:00 not here 2015-11-04 12:07:11 (1/2) Installing skalibs (2.3.8.0-r0) 2015-11-04 12:07:11 (2/2) Installing skalibs-dev (2.3.8.0-r0) 2015-11-04 12:07:11 OK: 703 MiB in 177 packages 2015-11-04 12:07:12 because you don't have a skalibs-doc 2015-11-04 12:07:17 I just added one 2015-11-04 12:07:21 in the apkbuild 2015-11-04 12:07:32 can i see the apkbuild= 2015-11-04 12:07:33 ? 2015-11-04 12:07:42 tpaste < skalibs/APKBUILD 2015-11-04 12:07:52 gimme a sec 2015-11-04 12:09:06 what package is the tpaste binary? 2015-11-04 12:09:14 sprunge probably 2015-11-04 12:09:51 ash: getcwd: No such file or directory 2015-11-04 12:09:51 that's what "apk add sprunge" says, after the OK 2015-11-04 12:10:08 ah, nvm 2015-11-04 12:10:19 was my bad 2015-11-04 12:10:57 ncopa, what's the plan about v3.3 ? 2015-11-04 12:11:05 http://sprunge.us/iYMM 2015-11-04 12:11:13 fcolista: fix build problems asap and ship rc1 asap 2015-11-04 12:11:20 and fix the issues 2015-11-04 12:11:43 build problems re arm ? 2015-11-04 12:12:01 and x86* too 2015-11-04 12:13:52 skarnet: cannot reprocude here 2015-11-04 12:14:11 ncdev-edge-x86_64:~/aports/main/skalibs$ abuild -h 2015-11-04 12:14:11 abuild 2.25.0_rc1 2015-11-04 12:14:57 uh... my abuild is 2.22.0... 2015-11-04 12:15:14 probably from v3.2 2015-11-04 12:15:15 what repo should I use to get the latest one? 2015-11-04 12:15:18 edge 2015-11-04 12:15:30 hm 2015-11-04 12:15:32 'edge' is rolling release 2015-11-04 12:15:34 git master 2015-11-04 12:15:48 we cut release from there every 6 month and git branch 2015-11-04 12:15:50 I guess I can't mix edge and main though 2015-11-04 12:15:57 so I gotta switch all to edge 2015-11-04 12:16:02 probably yes 2015-11-04 12:16:16 due to gcc 5 c++ abi breakage 2015-11-04 12:16:38 I wanted to avoid that but eh... yolo, upgrading to edge >.> 2015-11-04 12:17:00 welcome to the world of the nuts 2015-11-04 12:17:14 i use edge for my laptop and desktop 2015-11-04 12:17:15 it means the packages I build will go to edge too, right? 2015-11-04 12:17:25 depends on where we push them 2015-11-04 12:17:25 or can I say they go to main? 2015-11-04 12:17:31 yes 2015-11-04 12:17:42 we have edge/main 2015-11-04 12:17:48 edge/community 2015-11-04 12:17:50 and edge/testing 2015-11-04 12:18:34 ah, ok, I was confused. the distinction is v3.2/edge, not main/edge 2015-11-04 12:22:06 hey 2015-11-04 12:31:27 ncopa: well, actually this is a good question 2015-11-04 12:31:39 will we keep builders running for community after 6 months? 2015-11-04 12:31:55 we'll keep builders running after 6months 2015-11-04 12:32:02 but we will probably not push so much there 2015-11-04 12:32:11 so if someone wants to maintain something, he's still able to 2015-11-04 12:32:14 yes 2015-11-04 12:32:15 sounds good 2015-11-04 12:42:23 so we are moving to gcc5 with 3.3? 2015-11-04 12:44:13 gcc 5.2.0 :D 2015-11-04 12:57:43 hmm, is it possible that we have mbedtls twice in the repostories? 2015-11-04 12:58:01 yes, and we do 2015-11-04 12:58:17 mbedtls broke API and ABI at some point 2015-11-04 12:58:26 umurmur still requires old one 2015-11-04 12:58:45 so we have mbedtls and mbedtls1 now 2015-11-04 13:00:43 hmmm 2015-11-04 13:00:53 thats not very clean .. 2015-11-04 13:05:06 because i think hiawatha meight use the wrong version now 2015-11-04 13:14:46 well, define wrong 2015-11-04 13:14:51 it uses the old mbedtls 2015-11-04 13:17:32 leo-unglaub: triggered a rebuild, should link to new one now 2015-11-04 15:13:36 could someone please rebuild mbedtls with MBEDTLS_THREADING_PTHREAD enabled 2015-11-04 15:13:48 otherwise the hiawatha webserver will not work 2015-11-04 15:15:39 thanks :) 2015-11-05 08:41:33 good morning mountain lovers 2015-11-05 08:42:36 good morning fellow climber 2015-11-05 08:49:23 rotfl 2015-11-05 09:05:49 let us get domain alpine.systems 2015-11-05 09:07:40 rather entire TLD, .alpine 2015-11-05 09:12:43 yeah, and flip all domain name, 2015-11-05 09:12:43 organisation = org.alpine 2015-11-05 09:12:43 linux = linux.alpine 2015-11-05 09:12:43 dont't the java people already do it !! 2015-11-05 09:56:33 should we also snatch up alpine.solutions just in case at some point we want to run one of those companies which sell promises and pictures of smiling people (and not much else) 2015-11-05 10:09:22 alpine.biz 2015-11-05 10:09:31 alpine.tk 2015-11-05 10:09:39 alpine.xxx for the really motivated 2015-11-05 10:21:15 I'm having an issue with apk dependencies: 2015-11-05 10:23:24 "s6" and "s6-rc-doc" are supposedly independent packages ("s6-rc-doc" comes from "s6-rc", and "s6-rc" depends on "s6", but "s6-rc-doc" doesn't depend on anything). 2015-11-05 10:23:42 However, apk insists that s6-rc-doc depends on s6. 2015-11-05 10:24:22 Aren't doc packages supposed to be independent? And anyway, if anything, s6-rc-doc should depend on s6-rc, not s6. 2015-11-05 10:25:24 I suspect there's some heuristic in some abuild script that gets messed up because of the common prefix. How would I go on debugging that? 2015-11-05 10:44:43 sounds like some issue with install_if 2015-11-05 10:50:13 s6-rc's doc() doesn't call default_doc 2015-11-05 10:50:26 is there something special I should do to correctly set dependencies? 2015-11-05 10:50:35 or should I always call default_doc ? 2015-11-05 10:51:13 (I just looked for install_if in the abuild source, it doesn't tell me much) 2015-11-05 10:59:28 do we have gdb-dbg ? cant find it https://pkgs.alpinelinux.org/packages?name=%25gdb%25&repo=all&arch=x86_64&maintainer=all 2015-11-05 11:05:15 skarnet: a subpackage can (and often should) define its own dependencies by adding a depends= stanza in the subpkg function 2015-11-05 11:06:19 skarnet: you need to have "$pkgname-doc" in subpackages 2015-11-05 11:06:35 I don't see it in aports 2015-11-05 11:25:44 barthalion: I have it. s6-rc-doc is created from s6-rc. You don't see it in aports because it's only on my machine for now. 2015-11-05 11:27:33 so if you didn't specify doc() manually, default_doc() will be used for it 2015-11-05 11:41:05 I did specify doc() manually. 2015-11-05 11:41:24 I'm moving files around in doc(); I'm not calling default_doc. 2015-11-05 11:42:17 Should I call default_doc even if I don't have anything in $pkgdir to move to $subpkgdir? 2015-11-05 11:42:48 (i.e. is there magic in default_doc that prevents the kind of problem I'm having?) 2015-11-05 11:43:11 no, I don't see the point, just add "depends=" to the top of your doc() function and that should be it 2015-11-05 11:44:22 in default_doc(), I see depends, pkgdesc, arch and install_if 2015-11-05 11:44:37 I understand the first 3, but what should I put in install_if ? 2015-11-05 11:45:28 ah, forget it, this duplicates code 2015-11-05 11:45:52 I'll just call default_doc for the variables, even if it does nothing else 2015-11-05 11:50:02 install_if is used if you need to figure out which subpackage to install when, eg. if you have a package that provides a subpackage against lua5.2 and lua5.3, you add that as a dependency in install_if, reading something like "if we have lua5.3 install subpkg_of_foo_5.3" 2015-11-05 11:51:36 so for a doc() package, there is probably nothing you could fill in there that would make much sense 2015-11-05 11:52:51 I see, thanks. It's stuff I'll probably never need, so I'll just defer to default_doc, assuming it does the right thing in the common case. 2015-11-05 11:53:34 it should do that, yes 2015-11-05 13:03:12 hey :) 2015-11-05 13:09:11 ay 2015-11-05 13:09:54 would someone with the power kindly move opensmtpd to edge/main? It would be nice to see it in 3.3 2015-11-05 13:10:20 i've been using it for a few months now quite happily 2015-11-05 13:20:00 also, could someone rebuild libgusb (libgusb-dev still depends on udev) 2015-11-05 13:30:51 today is one of those days where linux is just pissing me off .... i am the only linux guy in a meeting between mac books and i am the only one who had to work 10 minutes to get the wifi working ... also i had to reboot after suspending it because dbus was segfaulting ... 2015-11-05 13:30:57 one of those days ... 2015-11-05 13:31:08 :-/ 2015-11-05 13:31:27 its not alpines fault ... but its a linux problem in generel 2015-11-05 13:32:16 Jonathan Curran is here? 2015-11-05 13:32:23 seems to be the maintainer of opensmptd 2015-11-05 13:32:40 fcolista: ok to move libasr to main? needed by opensmtpd 2015-11-05 13:33:05 well, the other way around isn't much better. A collegue of mine is a Mac user and everytime he pull the plug of the external screen (which of course if also an Apple product), the whole machine freezes up 2015-11-05 13:33:43 you should be glad you are on the side that is actually able to do something about the problems it has ;) 2015-11-05 13:34:03 chris|: yeah, apple users have problems as well ... but it makes me angry that in 2015 we still fight with wifi issues and that stuff 2015-11-05 13:34:17 i mean wifi worked better on my first windows 2000 laptop .. 2015-11-05 13:35:17 libgusb seem to have moved to github 2015-11-05 13:37:16 ncopa: quick question, how are the namespaces for python and python3 packages defined? 2015-11-05 13:37:51 chris|: currently, not defined at all 2015-11-05 13:38:05 i realized that we didnt reach it for v3.3 2015-11-05 13:38:18 so i'll think of it post v3.3 2015-11-05 13:38:42 i think what i'd like to try is move python -> python2 and upgrade python to 3 2015-11-05 13:38:58 and rebuild all py-* to be python 3 2015-11-05 13:39:17 if there are anything that still insist on python2 then we could use py2-* 2015-11-05 13:39:38 but in general, i'd like get rid of python2 if possible 2015-11-05 13:39:49 okay, sounds good to me 2015-11-05 13:39:54 but after v3.3 2015-11-05 13:40:02 we are severily late with v3.3 2015-11-05 13:40:14 should been released last tuesday 2015-11-05 13:40:21 and we dont even have first release candidate 2015-11-05 13:40:33 didnt help that i need go look for job :-/ 2015-11-05 13:42:24 yeah, I can imagine 2015-11-05 13:42:51 if there is anything specific I can help out with, just let me know 2015-11-05 13:43:42 check bugs.a.o 2015-11-05 13:43:49 we need fix as many bugs as possible 2015-11-05 13:45:16 also try help fix the packages that currently does not build 2015-11-05 13:45:22 many of those are probaly trivial 2015-11-05 13:45:57 sure, I'll can take a look 2015-11-05 13:47:38 is there an overview of failing package builds? 2015-11-05 13:48:01 or do we work of bug reports? 2015-11-05 13:48:03 http://tpaste.us/2KMB 2015-11-05 13:48:17 in build order 2015-11-05 13:49:34 you find the build errors in http://build.alpinelinux.org/buildlogs/build-3-3-x86_64/main/ 2015-11-05 13:49:58 http://build.alpinelinux.org/buildlogs/build-3-3-x86_64/main/ack/ack-2.14-r0.log 2015-11-05 13:50:01 like that 2015-11-05 13:50:04 is trivial to fix 2015-11-05 13:50:34 just add -doc subpackage 2015-11-05 13:50:50 i just did 2015-11-05 13:50:51 :) 2015-11-05 13:51:58 libcue moved to github 2015-11-05 13:54:01 alright, I'll pick some from the list 2015-11-05 13:54:49 im on libcue 2015-11-05 13:56:28 I'm actually travelling right now, so I won't get to do a lot until later today 2015-11-05 13:57:19 I'll just send patches as usual :) 2015-11-05 14:06:01 so folks 2015-11-05 14:06:20 let's say I want to put my apks and lbu overlays on different partition 2015-11-05 14:06:35 it's what alpine_dev is for? 2015-11-05 14:07:32 not really I guess 2015-11-05 14:08:23 I was probably looking for ovl_dev then 2015-11-05 14:09:38 barthalion: currently yes, with v3.3 you should not need any of them 2015-11-05 14:09:47 how is that? 2015-11-05 14:09:58 v3.3 should work without alpine_dev and without ovl_dev 2015-11-05 14:10:17 nlplug-findfs searches all found blockdevices 2015-11-05 14:10:22 that nice 2015-11-05 14:10:40 so it shoudl work with isohybrid 2015-11-05 14:10:44 uefi 2015-11-05 14:10:52 and in the root of that ovl_dev, am I supposed to put my overlays + apks, correct? 2015-11-05 14:10:53 without needing change the syslinux.cfg 2015-11-05 14:11:04 yes 2015-11-05 14:11:18 you apks needs an empty file named .boot_repository 2015-11-05 14:12:07 tricky part is that the repo needs to be signed, so you need the signing pub key in initramfs 2015-11-05 14:12:32 thanks 2015-11-05 14:12:43 probably my last question is whether I can use f2fs or not 2015-11-05 14:13:35 currently, as long as the f2fs driver is in initramfs and you have modules=f2fs on boot line 2015-11-05 14:13:56 with v3.3, as long as libblkid can find it and f2fs kernel module is included in initramfs 2015-11-05 14:14:22 ok, I know everything know, thanks again 2015-11-05 14:14:47 I guess I'll just build a 3.3 tarball for rpi ;) 2015-11-05 14:20:24 ncopa: what's the release schedule? this or next month? 2015-11-05 14:24:50 ncopa: thanks for moving opensmtpd and libasr (that my dyslexic brain keeps calling lib-arse) 2015-11-05 14:28:11 <20WACE9TM> Hi there! Is there a binary of nagios-nrpe available someplace ?? If not is it COMPILE-able ??? :| 2015-11-05 14:28:12 ncopa, it's fine with me to move libasr to main. 2015-11-05 14:32:10 ScrumpyJack: could you please email the maintainer and ask for moving it to main? 2015-11-05 14:32:58 20WACE9TM: $ apk search nrpe 2015-11-05 14:32:58 nrpe-plugin-2.15-r1 2015-11-05 14:32:58 nrpe-2.15-r1 2015-11-05 14:33:53 <20WACE9TM> HEY ... nCopa ... !!! :) !! Are YOU OK ??? 2015-11-05 14:34:23 20WACE9TM, seems that you have some issue with caps..:) 2015-11-05 14:34:28 or shift 2015-11-05 14:34:31 <20WACE9TM> Yes ... 2015-11-05 14:34:46 ncopa: can the maintainer move it to main? 2015-11-05 14:35:13 <20WACE9TM> also I have issues with my net connection going up and down and my nick getting screwed ... :) 2015-11-05 14:35:48 or do you mean we should ask his permission? 2015-11-05 14:36:30 20WACE9TM: hi im ok. are you ok? 2015-11-05 14:36:54 ah its grharry :) 2015-11-05 14:36:58 wb! 2015-11-05 14:37:03 <20WACE9TM> yessssssssssssssss 2015-11-05 14:37:27 <20WACE9TM> I almost started to get rude to all ... :-D 2015-11-05 14:37:35 haha 2015-11-05 14:37:43 noticed :D 2015-11-05 14:37:47 <20WACE9TM> what the heck is this nick name of mine ?? 2015-11-05 14:38:07 20WACE9TM 2015-11-05 14:38:20 <20WACE9TM> yes ... where is my ... "grharry" ?? 2015-11-05 14:38:34 <20WACE9TM> =-O 2015-11-05 14:38:36 20WACE9TM (~grharry@ppp-94-65-206-145.home.otenet.gr) has joined #alpine-devel 2015-11-05 14:38:47 <20WACE9TM> :P 2015-11-05 14:38:51 <20WACE9TM> ok 2015-11-05 14:39:10 <20WACE9TM> Thanks ncopa ... you are my best friend !!! ;) 2015-11-05 14:39:25 i hope im not your only :) 2015-11-05 14:39:43 <20WACE9TM> I mean ... on alpinelinux.org ... :) 2015-11-05 14:50:41 can someone of you explain awall to me? i dont see the benefits of it to be honest 2015-11-05 14:50:50 but maybe i simply miss it because i dont udnerstand it 2015-11-05 14:58:22 it's an abstraction over iptables 2015-11-05 14:58:54 some of us feel it's more productive to write awall policies than plain iptables rules 2015-11-05 15:00:03 e.g. defining zones and variables reduces the need to repeat things 2015-11-05 15:00:22 generate iptables+ip6tables rules from a single source 2015-11-05 15:01:13 modular definitions: build higher-level policies upon basic policies 2015-11-05 15:01:54 that's useful when you have several hosts to manage that share some characteristics 2015-11-05 15:05:11 generally awall policies are less bloated and easier to read than what iptables-save spits out 2015-11-05 15:06:31 ability to put DNS names into policies 2015-11-05 15:07:40 I agree there could be better marketing material for awall :) 2015-11-05 15:12:26 kunkku: i see 2015-11-05 15:12:37 so its basically another frontend for iptables? 2015-11-05 15:12:42 kind of like nftables? 2015-11-05 15:28:26 nftables isn't only a frontend 2015-11-05 15:28:41 it's a new way of storing tables in the kernel too 2015-11-05 15:29:17 I use nftables (frontend *and* backend) and am happy with it. What I'm not happy with is the build process, but it's not worse than iptables, so... 2015-11-05 15:46:15 i've been meaning to try awall forever 2015-11-05 15:56:54 skarnet: tell me a build process that does not suck ... 2015-11-05 15:57:13 Mine. But I worked on it for months. 2015-11-05 15:57:26 And there are still things it can't do. 2015-11-05 15:57:50 Kbuild-based things aren't *too* bad either. 2015-11-05 15:59:55 I'd love a world where a Makefile is enough. Unfortunately, the rise of mobile computing ensure this worlds stays far, far away. 2015-11-05 16:00:14 ensures* 2015-11-05 16:01:01 yeah, i dream about simgle simple makefile that ist just a few lines and then simply produces a binary ... 2015-11-05 16:01:05 that would be my dream 2015-11-05 16:02:16 it's possible for very small projects that have basically no dependencies, application- or system-wise. Beyond that... not really. :/ 2015-11-05 16:18:29 dmvpn \o/ 2015-11-05 16:19:59 huh. we have ack in main? 2015-11-05 16:20:17 ack? 2015-11-05 16:22:24 https://pkgs.alpinelinux.org/package/main/x86_64/ack apparently 2015-11-05 16:53:36 got HUPed on freenode 2015-11-05 16:53:47 HUPed? 2015-11-05 16:54:20 SIGHUP 2015-11-05 16:54:26 kill -1 2015-11-05 16:54:59 or zapped, in common parlance 2015-11-05 18:11:12 would be cool to have nftables in awall 2015-11-05 18:50:42 http://build.alpinelinux.org/buildlogs/build-3-3-x86_64/main/lua-lyaml/lua-lyaml-5.1.4-r0.log wait ... what?? 2015-11-05 18:55:30 also, lua-json4 doesn't seem to split properly (and also the package name is odd) 2015-11-05 20:24:33 lol 2015-11-05 22:01:20 Is the maintainer for openjdk8 (Timo Teräs) here by chance? 2015-11-05 22:02:40 fabled is generally here during Europe work hours 2015-11-05 22:04:11 Ok thanks, now i know his user name at least :) 2015-11-05 22:04:42 :) 2015-11-06 06:34:41 (113/119) Upgrading xf86-input-evdev (2.9.2-r1 -> 2.10.0-r1) 2015-11-06 06:34:41 ERROR: xf86-input-evdev-2.10.0-r1: trying to overwrite usr/share/X11/xorg.conf.d/10-evdev.conf owned by xorg-server-1.17.3-r0. 2015-11-06 06:51:04 did it get moved 2015-11-06 08:10:43 So now grsec support is done on test kernels ? 2015-11-06 08:10:51 or is it only for edge ? 2015-11-06 08:12:35 coredumb, we've been maintaining grsec stable kernels ourselves 2015-11-06 08:12:56 we have new 3.2-stable kernel, but due to massive kernel changes it's been on-hold to get more extensive testing 2015-11-06 08:14:39 fabled: yeah I know 2015-11-06 08:14:48 so you still make the 3.18 port ? 2015-11-06 08:15:02 or just switched to 4.1 ? 2015-11-06 08:19:12 fabled: you need to upgrade xorg-server first 2015-11-06 08:30:05 ncopa: opensmtpd maintainer is happy for it go in edge/main 2015-11-06 08:40:01 where could i set the "* - nice -15" under alpinelinux? from grsecurity irc it should be set there " I think, you can solve the resource overstep by editing /etc/security/limits.conf" but we dont have /etc/security/ nor i have here local "limits.conf" somewhere on system. 2015-11-06 09:07:56 slashbeast wrote: limits.conf is pam_limits file. http://linux.die.net/man/8/pam_limits so its linux-pam package on alpinelinux https://pkgs.alpinelinux.org/contents?pkgname=linux-pam&arch=x86_64&repo=main 2015-11-06 10:42:51 tdtrask: you about? 2015-11-06 11:29:21 hey friends :) 2015-11-06 11:29:47 hi 2015-11-06 11:29:55 ho 2015-11-06 11:30:01 ACTION snores 2015-11-06 11:30:26 skarnet: are we keeping you up? 2015-11-06 11:30:57 I wish. I'm trying to learn git functions such as git-send-email. 2015-11-06 11:32:09 Reading doc is even more boring than writing it. 2015-11-06 11:32:54 patches coming? 2015-11-06 11:32:57 coredumb, we have updated 3.18.y kernel 2015-11-06 11:33:03 but not pushed to aports yet 2015-11-06 11:33:43 clandmeter: when I kick myself hard enough to finish it, yes. Prob next week though, I'll be busy this week-end. 2015-11-06 11:34:17 skarnet: hacking what? 2015-11-06 11:34:59 my stuff. To prepare for init options. 2015-11-06 11:35:41 I wish there were a feature of apk to pull from a git rev. 2015-11-06 12:01:37 ACTION wonders why our wiki has an awall howto based on shorewall. 2015-11-06 12:17:47 skarnet: i just do git format-patch HEAD^ --stdout 2015-11-06 12:18:13 that goes straight to smtp 2015-11-06 12:21:00 s/that goes/I put that/ 2015-11-06 13:20:31 we are going to do maintenance to our infra, affecting bugs www nl.a.o rsync.a.o and whatever i missed. 2015-11-06 13:33:24 ScrumpyJack: hi 2015-11-06 14:35:29 fabled is there a reason why no cacerts are installed with openjdk8 or why there is no package to provide that? 2015-11-06 14:35:49 danilobuerger, no one has taken the time to do it yet 2015-11-06 14:36:05 so a pr would be welcomed? 2015-11-06 14:36:15 i looked at it on one point; but back then the java ca store needed to be built with java tools 2015-11-06 14:36:21 and it was rather awkward 2015-11-06 14:36:27 patches are welcome 2015-11-06 14:36:39 we may have a bug about; or perhaps it was just an email request 2015-11-06 14:37:23 ok i will take a look 2015-11-06 14:38:31 are you referring to http://bugs.alpinelinux.org/issues/4128 ? 2015-11-06 14:38:41 ah, yes 2015-11-06 14:39:06 should the certificates be a separate package like arch does? 2015-11-06 14:39:19 well 2015-11-06 14:39:27 iirc, most distros ship 'update-java-ca' script or similar 2015-11-06 14:39:37 that updates the java keystore dynamically from openssl ca store 2015-11-06 14:39:50 there's also p11-kit 2015-11-06 14:39:54 that's supposed to be able to do it 2015-11-06 14:40:16 the most basic version would just need find & keytool 2015-11-06 14:40:33 + hooking it to run whenever update-ca-certificates is ran 2015-11-06 14:41:10 it needs to be able to do full sync; add added certs, remove deleted certs 2015-11-06 14:41:33 is there even a update-ca-certificates on edge? 2015-11-06 14:41:53 yes, there is. it was recently rewritten in C 2015-11-06 14:42:08 /usr/sbin/update-ca-certificates is owned by ca-certificates-20150426-r2 2015-11-06 14:42:12 ah yes :) 2015-11-06 14:42:32 the original shell version is too slow 2015-11-06 14:42:38 i rewrote it in lua a year or two ago 2015-11-06 14:42:52 but some did not like to lua dependency, so we had contributor who rewrote it in C 2015-11-06 14:43:10 http://git.alpinelinux.org/cgit/aports/tree/main/ca-certificates/update-ca.c 2015-11-06 14:43:12 yeah, first time i saw it fetching lua i was wondering too 2015-11-06 14:44:03 the speed different between lua and C version is marginal iirc 2015-11-06 14:44:11 but it's certainly less dependencies 2015-11-06 14:44:15 ok, i will look into the whole thing and how to tie it with the update-ca-certificates 2015-11-06 14:44:52 other distros generall make the update-ca-certificates run hooks as final step 2015-11-06 14:44:55 i'm happy to do that 2015-11-06 14:44:59 even write the code 2015-11-06 14:45:13 if you can ship a script that syncs the openssl ca store with java ca store 2015-11-06 14:45:56 ok 2015-11-06 14:46:11 i hope i get around to it next week 2015-11-06 14:46:24 i'll update the ticket; i'll take it, and post that we have contributor looking at it. maybe briefly explain what's needed 2015-11-06 14:46:33 ok 2015-11-06 15:52:29 ncopa: still here? 2015-11-06 15:57:56 dont think he is coming back 2015-11-06 15:58:51 https://www.alpinelinux.org/ seams to be down 2015-11-06 15:59:26 who unplugged the hamsters 2015-11-06 16:00:25 :) 2015-11-06 16:00:31 we are working on it 2015-11-06 16:00:54 me and a whole team of alpine developers 2015-11-06 16:01:23 and the ceo of alpine infra 2015-11-06 16:02:36 leo-unglaub, still down? 2015-11-06 16:03:50 nah, works again 2015-11-06 16:05:07 multiple services will/can go down 2015-11-06 16:08:11 *cough* if only someone had invented a system that monitors services and can auto restart them *cough* 2015-11-06 16:09:48 skarnet: we are willing to hire a specialist in the field. we have a 3 beers for it in the fridge. 2015-11-06 16:12:06 drink the beers, else they'll be stale for the 3.4 release ;) 2015-11-06 16:15:39 hi guys, alpine infra server are going under maintenance 2015-11-06 16:15:53 will let you know when they will be available again 2015-11-06 18:16:40 I have someone bugging me by email to package textlive (and mistaking me for fcolista???) and I don't know whyyy 2015-11-06 18:16:44 who do I punt this to 2015-11-06 18:17:04 yeah. I get mails regarding the texlive package all the time 2015-11-06 18:17:31 I'm not even sure how I got dragged into this at all though 2015-11-06 18:17:47 The people (or person?) writing those emails doesn't understand that the texlive package isn't broken but that we don't have a complete texlive suite at the moment 2015-11-06 18:18:05 some marc collin? 2015-11-06 18:18:26 yeah, but I also got an email from a different person a months ago 2015-11-06 18:19:07 I mean I need a complete texlive suite as well and I looked into packaging it, but the texlive build system is awful 2015-11-06 18:20:05 it would be more helpful if the person writing those emails would provide patches to extend the texlive package… 2015-11-06 18:22:31 yeah like 2015-11-06 18:22:46 the first time I got one I was like "I have no idea but I can take a look" 2015-11-06 18:22:52 and 2015-11-06 18:22:54 ... I did 2015-11-06 18:23:11 and you discovered that texlive is an awful piece of software just as I did? 2015-11-06 18:24:03 can we leave it at "I looked at it" 2015-11-06 18:25:33 :D 2015-11-06 18:27:49 the part I'm most confused about it why *I'm* getting these emails though 2015-11-06 18:27:56 either he's spamming everyone who uh 2015-11-06 18:28:10 has committed anything? (recently?) 2015-11-06 18:28:38 or he hasn't been reading my "I'm p sure that's no me" emails 2015-11-06 18:51:28 ammunta: I git the same mail, probably just fishing for someone to do it 2015-11-06 18:51:35 /git/got 2015-11-06 19:07:49 okay yeah I'm punting this 2015-11-06 19:18:56 hi 2015-11-06 19:19:03 im bringing up the infra 2015-11-06 19:19:10 nl.alpinelinux.org should be up now 2015-11-06 19:19:16 http at least 2015-11-06 19:19:22 please let me know what is still not working 2015-11-06 19:19:27 and i'll try fix 2015-11-06 19:27:11 works here 2015-11-06 19:55:58 patchwork seems to be still down 2015-11-06 21:27:24 btw. do I need to resend patches or will patchwork catch up on its own? 2015-11-06 21:53:29 chris|, i hope they will still popup 2015-11-06 21:53:49 i think they are in main alpinelinux smtp server, and could not foward them. 2015-11-06 21:54:13 i dont remember the complete setup anymore 2015-11-06 22:23:41 well, I gess I'll check again tomorrow 2015-11-06 22:25:39 services should be working again 2015-11-06 22:28:27 mail still seems to be a bit behind on patchwork 2015-11-06 22:50:37 is there any standard for lua packages 2015-11-06 23:03:03 if so I haven't seen it 2015-11-06 23:22:35 ncopa: btw abuilds man page symlink handling is still broken http://sprunge.us/HTBV 2015-11-07 16:41:38 clandmeter2: just fyi. patchwork does not seem to get any new patches since yesterdays downtime 2015-11-08 13:59:30 clandmeter2: tpaste is also down 2015-11-08 13:59:49 clandmeter2: so is my lovely container 2015-11-08 20:01:09 the builders run on linux-grsec not on linux-vanilla, right? 2015-11-08 20:01:22 also: does patchwork work again? 2015-11-08 20:01:54 nope 2015-11-08 20:02:33 to patchwork 2015-11-08 20:03:14 nah, that sucks 2015-11-08 20:07:30 yeah, well, it's the weekend, I'm sure it'll be back to normal tomorrow 2015-11-08 20:11:44 hopefuly, it will just pick up all the pending patches from the mail server 2015-11-08 21:34:21 maybe it will be interesting in future Kernel Self Protection Project https://lwn.net/Articles/663361/ 2015-11-08 23:04:22 wait what happened to patchwork 2015-11-09 06:44:38 wait ... docker breaks if the client is newer than the server? amazing 2015-11-09 06:45:01 that seems like the opposite of what you'd expect upgrade behaviour to look like 2015-11-09 07:01:47 any chance we could move oniguruma into main and build jq against it 2015-11-09 07:02:53 idk if anyone's willing to support it though 2015-11-09 07:16:39 morning 2015-11-09 07:16:53 chris|: did your patches show up on patchwork? 2015-11-09 07:21:05 jomat: is it fine with you if I add you as oniguruma maintainer and rebuild jq with it? 2015-11-09 07:25:50 tpaste is still down 2015-11-09 07:51:28 reminds me of the bb su securetty patch 2015-11-09 08:11:43 morning 2015-11-09 08:11:47 fcolista: you about? 2015-11-09 08:16:22 ScrumpyJack: i heard rumors about fcolista being on vacation 2015-11-09 08:17:16 ok cool. I still can't figure out why my patches to mailing list aren't getting through 2015-11-09 08:17:37 do you get any error message from mailer? 2015-11-09 08:20:09 nothing. I am subscribed, and I can issue +help email and get a responce, but emails to alpine-aports@lists.a.o don't make it 2015-11-09 08:21:07 ScrumpyJack: do you have a mail in queue you can test with? 2015-11-09 08:21:17 i can check the logs realtime when you send it 2015-11-09 08:22:29 i have set up tail -f here now so if you git send-email i'll check what the log says 2015-11-09 08:22:30 erm sorry, no ive been sending patches using old sub 2015-11-09 08:22:53 i can send a test (although not cool) 2015-11-09 08:23:12 only if you have a real patch you want send 2015-11-09 08:23:16 that not yet made it 2015-11-09 08:24:12 sending test in 10 seconds 2015-11-09 08:24:34 done 2015-11-09 08:25:05 oh no! that made it through! 2015-11-09 08:25:10 :) 2015-11-09 08:25:14 it looks so 2015-11-09 08:25:27 sorry. right, so it must be the formatting of the patches 2015-11-09 08:25:54 can I get in touch at the next patch? 2015-11-09 08:26:00 sure 2015-11-09 08:30:18 ScrumpyJack: Functions.patch got checksum error for testing/wbar 2015-11-09 08:38:16 hmm, ok 2015-11-09 08:40:41 checking 2015-11-09 08:41:02 it's ok here. must be from email 2015-11-09 08:41:28 can you delete the patch and i'll send it again (we can use it to test email problem) 2015-11-09 08:47:26 let me know when tailing the logs :) 2015-11-09 08:50:06 i am 2015-11-09 08:51:48 send 2015-11-09 08:51:50 sent 2015-11-09 08:53:30 hmm, i got an email from patchwork saying 1 patch updated (Superseded) 2015-11-09 08:53:49 but patch hasn't made it to the mailing list 2015-11-09 08:54:37 it was me who saw you had 2 patches of same patch 2015-11-09 08:54:41 in patchwork 2015-11-09 08:54:49 so i marked it as superseeded 2015-11-09 08:55:16 ok 2015-11-09 08:55:23 i know what is wrogn with patchwork 2015-11-09 08:55:39 and why sent emails are not going into patchwork 2015-11-09 08:56:43 ok, i can see the patch arrived to the mailing list, as my old list subscription email address has received it 2015-11-09 08:57:08 but my new email address hasn't received it 2015-11-09 08:58:31 here is the email i sent http://sprunge.us/ccEZ 2015-11-09 08:58:41 clandmeter: i think we have an issue 2015-11-09 09:03:47 i can see a connection from mail.a.o and a mail from alpine-aports+bounces-1580-scrumpyjack=st.ilet.to@lists.alpinelinux.org 2015-11-09 09:04:06 i think the emails are just fine 2015-11-09 09:04:27 but patches are not imported in patchwork 2015-11-09 09:04:45 and i think also emails to bugs.a.o are not imported 2015-11-09 09:06:00 whee 2015-11-09 09:08:05 clandmeter: no problem. we only needed tos tart postfix 2015-11-09 09:11:29 lol, it's mutt! it was not displaying the emails 2015-11-09 09:11:57 :) 2015-11-09 09:12:51 ncopa: morning 2015-11-09 09:12:56 im trying to fix tpaste 2015-11-09 09:13:19 its not broken, but i want a proper intid 2015-11-09 09:13:28 clandmeter: i'll go look for coffe and i'll help you 2015-11-09 09:18:12 ncopa: yes, looks like they are all there now 2015-11-09 09:18:53 ncopa: putty may not apply since you already applied the 0.65, I'll fix that later today 2015-11-09 09:23:09 ok 2015-11-09 09:23:18 i mark it as rejected? 2015-11-09 09:23:42 i marked it as not appilcable 2015-11-09 09:24:39 chris|: %928 didnt build due to missing file missing_path_wtmpx.patch 2015-11-09 09:25:29 sipsak missed an abuild checksum 2015-11-09 09:25:33 the patch was not in checksums 2015-11-09 09:25:43 thats %916 2015-11-09 09:25:57 i suppose i can clean it up myself 2015-11-09 09:27:09 ok i found the v2 for sessreq 2015-11-09 09:41:04 yeah, I thought patchwork would understand the v2 better 2015-11-09 09:56:54 whoops... i shoudl probably not have applied valerys zeromq patch 2015-11-09 09:57:06 BitL0G1c's looks better 2015-11-09 09:58:19 ncopa: after everything has been built (or not), can you generate a new failure list for me? I am assuming you have a tool for that 2015-11-09 09:58:35 ok 2015-11-09 09:58:40 chris|: one thing more 2015-11-09 09:58:44 *big* thanks 2015-11-09 09:59:00 you're welcome :) 2015-11-09 09:59:01 this is very very helpful 2015-11-09 09:59:40 just don't expect me to keep up the pace during work weeks ;) 2015-11-09 09:59:52 instead of adding more fetaures (add stuff to testing) which require attention and time 2015-11-09 10:00:00 and delays the v3.3 release 2015-11-09 10:00:09 you have done the stuff that makes v3.3 release happen 2015-11-09 10:13:15 it would be cool to get a page with build failures and we could allocate them. more of us could pitch in perhaps 2015-11-09 10:13:26 oh, patchwork works again \o/ 2015-11-09 10:20:55 yes, i think all the alpine infra is back online now 2015-11-09 10:21:12 sorry about the downtime 2015-11-09 10:21:53 ACTION <-- its his faulth 2015-11-09 10:22:12 clandmeter did a heroic job last friday to get things up but it got so late that we both got tired and started do silly mistakes 2015-11-09 10:22:32 so we decided to just go home 2015-11-09 10:22:33 i even make mistakes in typing fault 2015-11-09 10:22:39 how nice 2015-11-09 10:22:46 lol 2015-11-09 10:23:05 clandmeter: that explains a bit, doesnt it :D 2015-11-09 10:23:06 thath wathn't a mithtake, don't make thun oth people with a lithp 2015-11-09 10:23:37 lol 2015-11-09 10:24:44 ncopa: ping 10.1.34.235 2015-11-09 10:24:44 PING 10.1.34.235 (10.1.34.235): 56 data bytes 2015-11-09 10:24:44 64 bytes from 10.1.34.235: seq=2 ttl=64 time=13.077 ms 2015-11-09 10:24:44 64 bytes from 10.1.34.235: seq=0 ttl=64 time=13.077 ms 2015-11-09 10:24:44 64 bytes from 10.1.34.235: seq=1 ttl=64 time=13.062 ms 2015-11-09 10:24:52 :D 2015-11-09 10:25:04 sounds like dmvpn is working? 2015-11-09 10:25:06 yes that scaleway 2015-11-09 10:25:32 *that* is cool 2015-11-09 10:25:37 \o/ 2015-11-09 10:25:48 we bridged the two boxes 2015-11-09 10:26:00 but it makes more sence to also do it with dmvpn 2015-11-09 10:26:05 because of routing 2015-11-09 10:26:48 the 2 boxes in your serverroom? 2015-11-09 10:26:54 yes 2015-11-09 10:27:10 might make more sense to bridge them 2015-11-09 10:27:15 than do local ipsec 2015-11-09 10:27:39 yes, but then they need to be on the same subnet. 2015-11-09 10:27:47 correct 2015-11-09 10:27:57 we maybe dont want that? 2015-11-09 10:28:21 i dont know, i think i like the seperations for now. 2015-11-09 10:28:25 ok 2015-11-09 10:28:31 i'm ok with this too 2015-11-09 10:28:47 the good thing is that is simple to move it to new location 2015-11-09 10:28:55 hm, is it just me or does libxft use the wrong include path for the fontconfig headers? 2015-11-09 10:28:58 > /usr/include/X11/Xft/Xft.h:39:22: fatal error: ft2build.h: No such file or directory 2015-11-09 10:29:18 nmeum: what package? 2015-11-09 10:29:24 libxft 2015-11-09 10:29:53 builds here 2015-11-09 10:30:24 yeah it builds but the header file is wrong 2015-11-09 10:30:59 it breaks what package? 2015-11-09 10:31:31 I am currently trying to upgrade dwm, dwm needs libxft now and this is the error I get when trying to compile dwm 6.1 2015-11-09 10:31:39 ft2build.h comes from where? 2015-11-09 10:31:54 fontconfig-dev 2015-11-09 10:32:00 it should be fontconfig/f2build.h 2015-11-09 10:34:10 do you use pkg-config? 2015-11-09 10:35:15 äh I meant freetype-dev not fontconfig-dev 2015-11-09 10:35:59 nope. It's an issue with dwms build system (sorry for the noise) 2015-11-09 10:36:02 ncdev-edge-x86_64:~/aports/main/libxft$ pkg-config --cflags xft 2015-11-09 10:36:02 -I/usr/include/freetype2 -I/usr/include/libpng16 2015-11-09 10:36:04 yup 2015-11-09 10:36:20 it hsould use pkg-config --cflags xft 2015-11-09 10:37:23 clandmeter: you bridged a scaleway box to something? 2015-11-09 10:37:51 no, i think he set up dmvpn between his boxes and scaleway server 2015-11-09 10:38:07 meaning we can route between the internal subnets 2015-11-09 10:38:58 without routing via a central vpn point 2015-11-09 10:39:32 2 nodes does not make sense, but 3 or more does 2015-11-09 10:40:21 hum 2015-11-09 10:40:30 minetest and irrlicht should be moved to community 2015-11-09 10:41:02 yeah I agree 2015-11-09 10:42:45 and bsnes 2015-11-09 10:43:06 ncopa: I will try my best to support minetest/irrlicht 2015-11-09 10:43:14 s/support/maintain 2015-11-09 10:43:21 thanks 2015-11-09 10:43:31 i dont expect more than that :) 2015-11-09 10:43:37 btw, i think there is a new release 2015-11-09 10:43:53 but I am not always aware of new releases of minetest and not of alpine 2015-11-09 10:44:01 okay I will look into it 2015-11-09 10:44:08 it does not build 2015-11-09 10:44:13 dont remeber why 2015-11-09 10:44:56 gcc5? last time the build failure was due to c++0 2015-11-09 10:45:23 yeah gcc5 2015-11-09 10:45:32 ncopa: so for community I only have to care that it builds against edge? 2015-11-09 10:45:43 and last stable 2015-11-09 10:45:50 edge + last stable 2015-11-09 10:45:55 ok 2015-11-09 10:45:58 which will v3.3 2015-11-09 10:46:04 til v3.4 is released 2015-11-09 10:46:09 so I need 2 builders 2015-11-09 10:46:19 you may need that yes 2015-11-09 10:46:20 one current stable and one edge 2015-11-09 10:46:29 unless we fix some docker thing for you 2015-11-09 10:46:43 correct 2015-11-09 10:46:59 or the linux in js thing :p 2015-11-09 10:47:08 :D 2015-11-09 10:56:21 su: must be suid to work properly :/ 2015-11-09 10:56:44 blah sudo su - 2015-11-09 10:57:08 you can apk add bbsuid 2015-11-09 10:57:14 or was it busybox-suid 2015-11-09 10:57:23 are we moving mate to community? 2015-11-09 10:57:28 i think so 2015-11-09 10:57:33 would be nice 2015-11-09 10:57:41 ncopa: could you take a look at %941 and %940 (when you have time to do so) 2015-11-09 10:57:51 i think alacerda was ok with it too 2015-11-09 10:57:58 nmeum: not time today sorry 2015-11-09 10:57:58 is the idea to move all non essentials to community? and make "main" a bit like "core? 2015-11-09 10:58:13 xming: something liek that yes 2015-11-09 10:58:20 ncopa: no problem 2015-11-09 10:58:21 main is for 2 years support 2015-11-09 10:58:30 keep the main small 2015-11-09 10:58:33 community is best effort 2015-11-09 10:58:41 yes, to reduce maintenance load 2015-11-09 10:59:01 okay, pushing mate and minetest move 2015-11-09 10:59:14 so i dont spend time on fixing acient software that nobody uses anyway 2015-11-09 10:59:18 oh, you did minetest 2015-11-09 10:59:19 barthalion: i think i just did 2015-11-09 10:59:20 :) 2015-11-09 10:59:23 if some wealthy party wants extras in "main" they can pay for the maintenance :D 2015-11-09 11:02:32 what am I missing here? http://sprunge.us/XUWY 2015-11-09 11:02:36 ncopa: I'll also move gnome-* 2015-11-09 11:02:53 ncopa: what about gnumeric/abiword? 2015-11-09 11:04:07 I guess we also should move seafile 2015-11-09 11:04:10 will it be a good idea to use git submodule, to delegate community maintanance burden? 2015-11-09 11:04:15 barthalion: he just left 2015-11-09 11:04:17 no 2015-11-09 11:04:28 clandmeter: alright 2015-11-09 11:04:31 lunch? 2015-11-09 11:04:41 no for the day 2015-11-09 11:04:43 oh 2015-11-09 11:04:51 I'll start the party then 2015-11-09 11:05:00 prepare the table 2015-11-09 11:05:11 are you going to dance on it? 2015-11-09 11:05:18 xming: I have configured our gitolite to disallow certain people pushing to main 2015-11-09 11:05:37 xming: so developers have access to everything in aports, while trusted users only to testing and community 2015-11-09 11:06:02 barthalion: of course 2015-11-09 11:06:10 gitolite will reject pushes that modifies anything in main/ if you are not in the developer group 2015-11-09 11:06:17 barthalion: I meant even delegating "community" further 2015-11-09 11:06:24 not for now 2015-11-09 11:06:41 aport devs just have to maintain add/rm submodules 2015-11-09 11:06:55 submodules don't scale well 2015-11-09 11:07:06 cloning time is insane with big chunk of submodules 2015-11-09 11:07:44 not one submodule per contributor, more like one per (few) categories 2015-11-09 11:08:19 maybe one day, I have seen this model in Exherbo and maybe CRUX and it looked good to me 2015-11-09 11:08:39 it's definitely up to guys with @ in front of their nicknames 2015-11-09 11:09:32 ACTION sets operator status to barthalion 2015-11-09 11:11:04 lol 2015-11-09 11:21:14 git mv will trigger a rebuild of pkgs? 2015-11-09 11:21:23 and not fs mv? 2015-11-09 11:21:44 yes, it triggers a rebuild 2015-11-09 11:22:45 where can I see the queue? buildlogs? the buildlog dir is empty now 2015-11-09 11:23:21 http://nl.alpinelinux.org/buildlogs/ 2015-11-09 11:23:32 http://build.alpinelinux.org/buildlogs/ 2015-11-09 11:24:06 ty 2015-11-09 11:24:09 np 2015-11-09 11:28:45 lol openal-soft fails with "abuild -r" but not with "abuild" 2015-11-09 11:37:23 please pull https://github.com/alpinelinux/aports/pull/18 2015-11-09 11:46:28 so there are libressl and openssl now? 2015-11-09 11:48:38 right, so is there a nice simple list of failed builds? 2015-11-09 11:49:25 I have some time this week 2015-11-09 11:50:16 http://build.alpinelinux.org/buildlogs/ 2015-11-09 11:50:39 ScrumpyJack: I just fixed openal-soft, so don't mind it 2015-11-09 11:54:47 xming: I meant just a list of packages. could someone with shell access to that write a small shell script? I don't know what's relevant. some files there are over a year old 2015-11-09 11:55:16 ScrumpyJack: those build logs show failed pkgs 2015-11-09 11:55:41 they are all relevant to make edge working 2015-11-09 11:55:53 and 3.3 will cut from edge from some point 2015-11-09 11:56:19 so I don't quite get what else are you looking for? 2015-11-09 11:57:20 i just see a bunch of dirs and files, some called build-3-1 some build-2-6 some build-edge, some have a sum in the filename, some don't. I don't know what I'm looking at 2015-11-09 11:58:08 build-3.2* are for current stable 2015-11-09 11:58:20 build-edge* are edge 2015-11-09 11:58:33 e.g. http://build.alpinelinux.org/buildlogs/build-edge-x86_64.v3.2.0-3235-gf27f694.log 2015-11-09 11:58:59 you can see that "mate-system-monitor" failed to build for edge 2015-11-09 11:59:16 ok, so i only want to look at build-edge*.log right? 2015-11-09 11:59:41 http://build.alpinelinux.org/buildlogs/build-3-2-x86_64.v3.2.2-2-g80677ae.log 2015-11-09 11:59:51 strongswean failed for 3.2 2015-11-09 12:00:21 what about the folders? 2015-11-09 12:00:25 ScrumpyJack: I think the goal is to eliminate all failed builds 2015-11-09 12:01:00 in the dirs you find detailed build log 2015-11-09 12:01:11 ah ok, cool 2015-11-09 12:01:28 so you can do something already before you try to it locally 2015-11-09 12:12:19 changed plans 2015-11-09 12:12:21 im back 2015-11-09 12:13:27 ncopa: wb 2015-11-09 12:24:02 so, how do we stop working on the same package? 2015-11-09 12:24:09 ncopa: what plan has changed? 2015-11-09 12:24:44 taking the rest of the day off 2015-11-09 12:25:12 :) 2015-11-09 12:25:28 lunch 2015-11-09 12:25:33 speaking of mate 2015-11-09 12:25:39 we shoudl probably upgrade it to 1.12 2015-11-09 12:37:49 where is our mate maintainer? 2015-11-09 12:39:23 MIA 2015-11-09 12:41:24 clandmeter: seems like oscam is a svn snapshot 2015-11-09 12:41:31 maybe move to community? 2015-11-09 12:43:46 sure 2015-11-09 12:44:01 ill do it 2015-11-09 12:46:09 did internet just break? 2015-11-09 12:46:25 http://www.shorewall.net/ 2015-11-09 12:46:27 yeah systemd rebooted :p 2015-11-09 12:46:29 times out from here 2015-11-09 12:46:47 also this http://ipset.netfilter.org/ 2015-11-09 12:47:20 they wrongly configured their iptables :p 2015-11-09 12:47:45 -s 0.0.0.0 -j DROP 2015-11-09 12:58:45 ncopa: is it okay to leave a community pkg and its dependencies in their current versions? An only bump versions for edge? 2015-11-09 12:59:52 huh? 2015-11-09 13:00:43 yeah, once v3.3 is released then we only push to 3.3-stable when there is a specific problem we want fix 2015-11-09 13:00:55 I am working on minetest (an irrlicht) are working on 3.2 2015-11-09 13:01:27 we dont just bump version in stable tree just because there is a new version 2015-11-09 13:01:39 there are new versions of minetest and irrlicht, is it okay if I just use newer versions for edge? and leave stable as is? 2015-11-09 13:02:00 annoyingly build-edge-x86_64.v3.2.0-1830-g6fd665a.log says ERROR: audacious: Failed to build but it builds fine on my build box 2015-11-09 13:02:50 we need a better way to track failed builds and track who is fixing them 2015-11-09 13:03:32 ScrumpyJack: I think that one is too old, already fixed, check the dirs for detailed build log, it's empty there 2015-11-09 13:04:38 yeah, i figured that out when it was too late. 2015-11-09 13:05:10 you have to admit, this could be more efficient :) 2015-11-09 13:05:44 task for someone with scripting skills: 2015-11-09 13:06:00 hook on to mqtt message for build errors 2015-11-09 13:06:01 and access to the build boxes 2015-11-09 13:06:04 and generate log file 2015-11-09 13:06:11 ah yes, mqtt 2015-11-09 13:06:16 so we get a list of packages failing 2015-11-09 13:07:57 and generate a nice rss/atom 2015-11-09 13:07:57 and the publish them on http some place. mqtt topic is: "build/+/errors" 2015-11-09 13:09:01 ncopa: can anyone sub the mqtt? 2015-11-09 13:10:41 yup 2015-11-09 13:11:07 xming: mosquitto_sub -h msg.alpinelinux.org -t "build/+/errors" -v 2015-11-09 13:11:48 cool, I really need to play/learn with mqtt 2015-11-09 13:15:27 there is also an mqtt-exec package 2015-11-09 13:15:35 that can execute a given command on mqtt messages 2015-11-09 13:17:22 is there an API to redmine, so that bugs could be autoraised? 2015-11-09 13:18:30 REST API 2015-11-09 13:21:02 yes 2015-11-09 13:28:05 ok, i'll give it a go after 3.3 is released, but for now http://dev.alpinelinux.org/buildlogs/ will do 2015-11-09 13:48:00 3.3! yay! 2015-11-09 13:51:12 (i'm seeing it on my mirror - not noticed it before) 2015-11-09 14:00:02 libgweather failed, and has a log: what's wrong with http://tinyurl.com/nw9lbeb ? 2015-11-09 14:13:29 nothing that I can see :/ 2015-11-09 14:21:59 horray http://lkml.iu.edu/hypermail/linux/kernel/1511.1/00247.html 2015-11-09 14:32:56 fixing minetest https://github.com/alpinelinux/aports/pull/20 2015-11-09 14:33:29 sorry I had to close PR #19, forgot to fetch+rebase, did a merge 2015-11-09 15:02:28 05:56 < xming> blah sudo su - 2015-11-09 15:02:36 xming: sudo -i isn't sufficient? 2015-11-09 15:32:17 ncopa: mqtt is publishing http://build.alpinelinux.org/buildlogs/build-edge-armhf/.. as URL for logfile, but that's broken 2015-11-09 15:32:37 yes arm build is broken 2015-11-09 15:32:41 didnt have time to fix yet :-/ 2015-11-09 15:32:43 algitbot: old habit when sudo didn't had -i flag 2015-11-09 15:32:44 are the build logs getting written to disk for arm? 2015-11-09 15:32:46 ok 2015-11-09 15:33:02 ammunta: ^^ 2015-11-09 15:35:32 sometimes I even do sudo /bin/bash -i 2015-11-09 15:42:37 chris|: new list of failed packages: http://tpaste.us/Gxnv 2015-11-09 15:42:48 for v3.3 2015-11-09 15:43:00 hm 2015-11-09 15:43:15 we should move seafile and ccnet to commuity 2015-11-09 15:45:43 i'm killing rxmtp 2015-11-09 15:45:59 nobody with sane mind should be using it 2015-11-09 15:48:55 i have poppler upgrade + libreoffice rebuild in my queue 2015-11-09 15:53:56 ncopa: a couple of those I could not reproduce, like btrfs-progs and gphoto2 and imap. I think they just need to be retriggered, since some were just download issues 2015-11-09 15:55:31 ok 2015-11-09 15:55:48 will retrigger on next git push now 2015-11-09 16:12:39 looking at alpine package (unless it's fixed already?) 2015-11-09 16:16:11 i'm going home now 2015-11-09 16:16:16 see u 2015-11-09 16:20:21 chris|: can you cross out the ones you've looked at? 2015-11-09 16:26:58 ScrumpyJack: not from the top of my head 2015-11-09 16:27:29 but I skipped all the perl* stuff if you are interested ;) 2015-11-09 16:30:42 chris|, ScrumpyJack: i started at the end of the list. 2015-11-09 16:37:21 fabled: strongswan doesnt build 2015-11-09 16:37:31 error: redefinition of 'struct in6_addr' 2015-11-09 16:37:38 and a few others 2015-11-09 16:39:59 oooh, sounds like uapi stuff again 2015-11-09 16:40:04 everybody loves uapi 2015-11-09 17:13:14 I won't have time to look at the list for east two more hours, so please, go nuts :) 2015-11-09 17:24:02 clandmeter, see strongswan APKBUILD 2015-11-09 17:24:15 there's a custom step to remove the intree linux headers to make it build 2015-11-09 17:57:33 fabled: if i comment out "rm -r src/include/linux" it builds correctly. so should i just remove it? 2015-11-09 17:58:24 that's weird. it used to be needed there. 2015-11-09 18:55:23 right, home now. anyone got an updated list of broken packages? 2015-11-09 19:49:24 ScrumpyJack: I only have the one that ncopa posted earlier, but it is not necessarily up to date 2015-11-09 19:51:53 well, i'm just keeping a eye on alpine-aports. i see you updated putty :) 2015-11-09 19:52:36 some of the build failures seem to be false positives, so a version bump and job done 2015-11-09 19:54:18 jup 2015-11-09 19:55:24 coova-chilli has a patch on patchwork, as halbert 2015-11-09 20:00:54 there is a *lot* of gnome stuff failing 2015-11-09 20:02:38 yeah, but I think most just needs an update to the latest version 2015-11-09 20:09:33 so it's best to look at http://patchwork.alpinelinux.org/project/aports/list/ then? 2015-11-09 20:13:13 imo yes 2015-11-09 20:27:05 god, openjdk takes ages to build 2015-11-09 20:27:51 i need a bigger build box too 2015-11-09 20:28:44 I'm building on my laptop 2015-11-09 20:29:23 what's the list? 2015-11-09 20:29:48 barthalion: http://tpaste.us/Gxnv 2015-11-09 20:30:56 those are roughly the builds that are currently failing on edge and are blocking 3.3 2015-11-09 20:31:34 yeah, I was at such parties already 2015-11-09 20:33:07 no reason to stop now :) 2015-11-09 20:37:03 I'll take some perl from the middle 2015-11-09 20:38:17 imap build fine as it is 2015-11-09 20:38:33 s/build/builds 2015-11-09 20:43:27 imap is fine, just needs to be triggered again 2015-11-09 20:43:38 it's on the list because the download failed 2015-11-09 20:44:09 yip 2015-11-09 20:44:23 some of the lua* builds should also build fine now that the lua5.*-dev packages are fixed 2015-11-09 20:44:59 I haven't looked at any perl packages 2015-11-09 20:45:50 and I am pretty sure augeas needs a look 2015-11-09 20:46:28 can you give me patchwork ids for build fixes? 2015-11-09 20:47:55 mine are 955 to 959 for now 2015-11-09 20:48:13 and 952 2015-11-09 20:51:55 954, 909 and 874 2015-11-09 20:52:59 ScrumpyJack: just for future, we usually put the version we update to in commit message 2015-11-09 20:53:03 lm-sensors didn't build because source lm-sensors.org is down. 2015-11-09 20:53:38 but it's a nitpick, so don't worry 2015-11-09 20:54:10 barthalion: no no, i thought I put the from and to versions in. did i miss one? 2015-11-09 20:54:35 gphoto2 patches have "Version update" after colon 2015-11-09 20:54:38 i mean no no, you're not nitpicking. i thought ... 2015-11-09 20:54:55 it is nitpicking, patches are good 2015-11-09 20:55:18 ah, you mean in the subject? gotchya 2015-11-09 20:55:21 yeah 2015-11-09 20:55:26 will do 2015-11-09 21:04:57 ScrumpyJack: checksums fail on your alpine patch 2015-11-09 21:05:06 ok 2015-11-09 21:05:12 I'll fix it 2015-11-09 21:05:16 what about the others? 2015-11-09 21:05:31 it's my blooming mailer 2015-11-09 21:05:41 others are good, will push them after I test chris| patches 2015-11-09 21:06:29 barthalion: if you have the time, you can also look at 961 2015-11-09 21:06:38 but it takes a while to build 2015-11-09 21:06:44 java? 2015-11-09 21:06:50 jeah 2015-11-09 21:08:03 putty never stops to amuse me 2015-11-09 21:08:12 thanks for fixing the alpine sum 2015-11-09 21:08:53 np, thank you for patches guys 2015-11-09 21:12:09 pushed… and starting java build now 2015-11-09 21:12:25 hm, halberd built here 2015-11-09 21:13:13 so did gphoto… sigh 2015-11-09 21:13:41 before the patches? 2015-11-09 21:13:55 no, after 2015-11-09 21:14:02 oh cool 2015-11-09 21:14:08 looks like some stale tarball in distfiles 2015-11-09 21:14:32 by the way, i'm getting ERROR: webkit2gtk-2.10.3-r0: package mentioned in index not found (try 'apk update') after updating 2015-11-09 21:14:55 i remember some stuff going on about webkit2gtk but i can't remember the outcome 2015-11-09 21:15:18 i think midori deps on this 2015-11-09 21:15:33 I'll trigger a rebuild tomorrow, going to bed now 2015-11-09 21:16:20 the halberd maintainer switched to github and just redirected to a non-existing tag, but he was quite willing to fix it 2015-11-09 21:16:20 my girlfriend passed on me a cold 2015-11-09 21:17:21 hot whiskey and lemon for you sir! 2015-11-09 21:18:19 whiskey? we drink tee with vodka for that here :P 2015-11-09 21:18:30 thanks and good night 2015-11-09 21:21:11 mumble is a compiler error - will dig deeper later 2015-11-09 21:26:34 barthalion: sent a v2 for openjdk, forgot to bump the pkgrel 2015-11-09 21:37:22 chris|: what TZ are you in? 2015-11-09 21:38:05 CET 2015-11-09 21:41:09 localtime 22:41 2015-11-09 21:41:55 oh so not too bad. i'm in the UK, but i'm not sure i can go much longer 2015-11-09 21:44:39 yeah, I am actually watching The Walking dead on the side ;) 2015-11-09 21:45:16 TV is off as the girls have gone to bed 2015-11-09 21:46:09 I guess I'm reaping the benefits of a single life 2015-11-09 21:49:47 mumble fixed 2015-11-09 21:50:28 show-off ;) 2015-11-09 21:51:08 wasn't hard ;) 2015-11-09 21:52:35 no-embed-qt-translation 2015-11-09 21:52:45 where did you find that? 2015-11-09 21:53:58 oh, it failed for me when I updated it to 1.2.10 2015-11-09 21:54:08 can you bump it? 2015-11-09 22:02:38 sure, building right now 2015-11-09 22:03:41 zenity is fixed, zmap is fine as is 2015-11-09 22:07:15 i'm struggling with seafile 2015-11-09 22:08:29 what should the path of include files be in an include file? relative to /usr/include? 2015-11-09 22:08:45 yes 2015-11-09 22:08:54 sent a patch for mumble 1.2.10 2015-11-09 22:10:29 darn, then include paths in libevhtp are broken 2015-11-09 22:10:58 not an uncommon problem with musl 2015-11-09 22:11:12 or glibc, depending on your point of view 2015-11-09 22:13:21 can you apk add libevhtp-dev somewhere and check 2015-11-09 22:13:33 the file /usr/include/evhtp/evhtp.h 2015-11-09 22:13:42 first line 2015-11-09 22:13:55 #include 2015-11-09 22:14:10 should it be #include 2015-11-09 22:14:40 I don't know what it's about, but theoretically yes, that's what it should be 2015-11-09 22:14:43 or should I set a search path in seafile source 2015-11-09 22:16:22 skarnet: hello kind sir. lots of s6 patches today 2015-11-09 22:16:36 wat 2015-11-09 22:17:06 did you have to patch it ? 2015-11-09 22:17:29 not me: http://patchwork.alpinelinux.org/project/aports/list/ patches 946 to 950 2015-11-09 22:18:18 how do I look for a patch by number there? 2015-11-09 22:18:57 because I have a set of patches already, I said I'd submit them this week 2015-11-09 22:19:02 mouse over, of if you are mouseless, the subject should be clear enough :) 2015-11-09 22:19:18 ah, mumuhten's patches for the new releases 2015-11-09 22:20:22 right, night folks. 2015-11-09 22:20:26 gn! 2015-11-09 22:25:47 ScrumpyJack: don't patch the includes :) 2015-11-09 22:28:26 that is a hell without return in these cases 2015-11-09 22:35:35 well, high time for me too, good night 2015-11-09 23:02:21 chris|, ScrumpyJack: you dont need to bump pkgrel on packages that need a build fix. 2015-11-09 23:02:59 what does rel stand for anyway 2015-11-09 23:03:16 its our internal version number 2015-11-09 23:03:38 I get that, but is it short for something 2015-11-09 23:04:09 package release i guess 2015-11-09 23:04:33 i didnt choose the name. 2015-11-10 06:11:50 clandmeter: good, because I don't do that ;) 2015-11-10 07:21:21 fabled: modutils patch to bb is also something to do with add multi .dep file read ? 2015-11-10 07:21:55 vkris, not really. it just supports generating the .bin files 2015-11-10 07:22:04 ok 2015-11-10 08:04:25 nmeum: did you ask uggedal if its ok with take over dmenu? 2015-11-10 08:08:13 clandmeter: I don't think I bumped the pkgrel either 2015-11-10 08:21:17 chris|: i must be blind ;) 2015-11-10 08:26:06 clandmeter: I sent in a patch that bumped a pkgrel but that had nothing to do with rebuilding, it was a legitemate bump 2015-11-10 08:26:47 you had already done that though so maybe that's the reason 2015-11-10 08:26:49 its not important, lets call it an advise. 2015-11-10 08:30:08 yikes! impressive conflict while upgrading this morning 2015-11-10 08:34:35 http://tpaste.us/3YEZ 2015-11-10 08:35:29 could someone trigger a midori rebuild? 2015-11-10 08:36:49 there is webkitgtk2 and webkit2gtk 2015-11-10 08:37:13 ScrumpyJack: mix v3.2 + edge? 2015-11-10 08:37:28 i upgraded icu in edge 2015-11-10 08:37:40 and rebuilt the whole thing 2015-11-10 08:37:40 nope, im on edge only on this desktop 2015-11-10 08:38:17 i had similar issue on my laptop the other day 2015-11-10 08:38:27 i dont remember how i managed to upgrade it 2015-11-10 08:38:42 you need find out what holds back icu-55 2015-11-10 08:38:53 webkit2gtk i think 2015-11-10 08:39:03 no 2015-11-10 08:39:09 i think its texlive 2015-11-10 08:40:46 apk del texlive made things worse :) 2015-11-10 08:40:46 i'll rebuild texlive 2015-11-10 08:41:09 |tpaste i get "413 Request Entity Too Large" lol 2015-11-10 08:41:16 ha 2015-11-10 08:41:38 can you apk del icu-dev also? 2015-11-10 08:41:56 ScrumpyJack: yes, i limited tpaste. 2015-11-10 08:42:08 sprunge can handle large requests 2015-11-10 08:42:16 icu-dev: harfbuzz-dev pango-dev gtk+2.0-dev libglade-dev 2015-11-10 08:42:36 http://sprunge.us/aYiH 2015-11-10 08:43:11 lol 2015-11-10 08:43:20 that the largest apk error ive seen 2015-11-10 08:43:22 you won a price 2015-11-10 08:43:28 yay! 2015-11-10 08:43:51 remove libreoffice? 2015-11-10 08:43:53 we can let apk bake a cake for you. 2015-11-10 08:44:19 wow 2015-11-10 08:45:06 ok apk del libreoffice and apk upgrade lets me upgrade icu-libs 2015-11-10 08:45:15 what in the fucking fuck 2015-11-10 08:45:37 clang has 110 .so's and I thought it was insane already 2015-11-10 08:45:52 how many .so's does libreoffice have? 2015-11-10 08:46:48 a little less than a zillion i think 2015-11-10 08:47:48 typical application of "because you can do something doesn't mean you should" 2015-11-10 08:50:23 ScrumpyJack: what happens if you try install libreoffice again? 2015-11-10 08:52:02 i'm still upgrading 2015-11-10 08:52:20 lol: ERROR: Failed to create usr/lib/chromium/chrome: No space left on device 2015-11-10 08:52:42 please stop torturing that poor computer 2015-11-10 08:53:08 nginx developer would not approve 2015-11-10 08:53:10 i have 25G left on / 2015-11-10 08:54:15 does apk download to tmpfs first? 2015-11-10 08:55:11 ScrumpyJack: check if you have /etc/apk/cache symlink 2015-11-10 08:57:10 i do 2015-11-10 09:00:00 then apk will cache the downloaded .apk 2015-11-10 09:00:16 try apk cache clean 2015-11-10 09:00:49 or maybe even rm /etc/apk/cache/* 2015-11-10 09:02:41 df -h /etc/apk/cache/ 2015-11-10 09:02:53 Avail 25G 2015-11-10 09:03:52 deleted cache contents 2015-11-10 09:04:23 e usr/lib/chromium/chrome 2015-11-10 09:04:35 you have that on different partition= 2015-11-10 09:05:03 same partition 2015-11-10 09:05:17 all under / 2015-11-10 09:05:34 weird 2015-11-10 09:05:45 what filesystem? btrfs? 2015-11-10 09:05:48 ext4 (so no btrfs funniness) 2015-11-10 09:06:03 heh, you ninja'd me 2015-11-10 09:06:05 that does not make any sense 2015-11-10 09:06:09 :) 2015-11-10 09:06:30 fsck.ext4 or something maybe 2015-11-10 09:06:56 well, the upgrade is doing ok so far 2015-11-10 09:13:17 can we kill gnome2 desktop? 2015-11-10 09:15:40 well, we have mate, so yes 2015-11-10 09:19:21 speaking of gnome, gnome-themes-standard conflicts with gnome-themes 2015-11-10 09:22:11 apk add webkit2gtk: ERROR: webkit2gtk-2.10.3-r0: package mentioned in index not found (try 'apk update') 2015-11-10 09:23:08 actually, that's when I had midori as well 2015-11-10 09:26:00 wow! my repo file had main testing community 2015-11-10 09:26:36 I changed it to main community testing and it works 2015-11-10 09:32:54 huh 2015-11-10 09:33:01 ok 2015-11-10 09:34:56 actually 2015-11-10 09:34:58 makes sense 2015-11-10 09:35:04 the testing repo didnt build 2015-11-10 09:43:20 what happened with gnome-icon-theme? 2015-11-10 09:43:52 it only has version 3.12 2015-11-10 09:43:56 no 3.18 2015-11-10 09:45:25 gnome-theme and gnone-theme-standard share some files 2015-11-10 09:45:45 but omg, webkit2gtk is not a happy bunny here 2015-11-10 09:45:47 http://tpaste.us/G5nB 2015-11-10 09:45:54 gnome-themes is old 2015-11-10 09:46:18 they only share three files 2015-11-10 09:47:04 oops, actually, is the index.theme they share 2015-11-10 09:47:11 and the gtkrc 2015-11-10 09:49:58 i was using gnome-themes, but gnome-themes-standard are much nicer 2015-11-10 09:52:32 guys 2015-11-10 09:52:49 we cannot move gnome-vfs to community when there are packages in main that depends on it 2015-11-10 10:30:23 ncopa: I didn't but he seems to be using OpenBSD now and doesn't seem to be interessted in alpine anymore. I can ask him though if you want me to… 2015-11-10 10:30:47 i emailed him 2015-11-10 10:32:51 I just sent him a private message 2015-11-10 10:43:54 I admire his distrohopping strength 2015-11-10 10:47:38 lol 2015-11-10 10:48:11 i bet he will be back 2015-11-10 10:50:26 i upgraded mate to 1.12 2015-11-10 11:07:31 barthalion: so do I 2015-11-10 11:07:43 aya 2015-11-10 11:20:40 AlpineJack: did something broke? 2015-11-10 11:21:10 libreoffice can't install at the moment 2015-11-10 11:21:40 and midori/webkitgtk2 doesn't play well with compton 2015-11-10 11:22:11 this is apl add libreoffice http://tpaste.us/AQrO 2015-11-10 11:28:16 ScrumpyJack, seems inkscape needs rebuild 2015-11-10 11:28:38 at least 2015-11-10 11:28:42 possibly others 2015-11-10 11:31:44 due to poppler? 2015-11-10 11:55:47 ScrumpyJack, yes 2015-11-10 11:56:35 hmm 2015-11-10 11:56:49 it's supposed to be recompiled 2015-11-10 11:57:15 inkscape was moved to community 2015-11-10 11:57:26 apparently it got deleted from main, but is not rebuilt yet 2015-11-10 11:57:30 perhaps it failed 2015-11-10 12:03:32 ncopa: uggedal replied that it would be okay if I take over dmenu 2015-11-10 12:03:51 > 11:31 Hey, since you don't seem to be working on alpine anymore: Is it okay if I take over the main/dmenu package? 2015-11-10 12:04:01 > 13:02 sure 2015-11-10 12:07:03 ok 2015-11-10 12:07:08 good 2015-11-10 12:11:23 ncopa: how much time is there usually between two Alpine releases? iow, how much time will it roughly take to whip up 3.4 after 3.3 is out? 2015-11-10 12:12:22 6 months 2015-11-10 12:12:31 we ship november and may 2015-11-10 12:13:14 so in april and october we feature freeze 2015-11-10 12:13:17 and try ship 2015-11-10 12:13:33 re s6 2015-11-10 12:13:39 i am ok to keep it in main 2015-11-10 12:15:01 hm 2015-11-10 12:15:03 new xorg 2015-11-10 12:15:15 any bojections to sneak it in for v3.3? 2015-11-10 12:15:18 objections? 2015-11-10 12:15:48 ncopa: I'm asking because I just got a contract, so I may be too busy to contribute for 4 to 8 months 2015-11-10 12:16:00 ah, congrats 2015-11-10 12:16:20 and thanks for letting us know 2015-11-10 12:16:36 the good side is that I'll be free for some time afterwards with enough money to go full-time contributor 2015-11-10 12:17:00 the bad side is that it's throwing a wrench in my s6-rc plans 2015-11-10 12:17:06 at least deferring them 2015-11-10 12:17:08 understand 2015-11-10 12:21:07 gentlefolk: can we put stuff back from community to main? i'm not sure enough thought is going into what get's moved. please contradict me, you're the guys who know about this :) (putting stuff back in main probably isn't the answer) 2015-11-10 12:21:47 ScrumpyJack: yes we can move things back if we have someone to maintain it 2015-11-10 12:21:57 what do you want back to main? 2015-11-10 12:22:14 or is it your laptop breakage? 2015-11-10 12:22:37 try apk upgrade now 2015-11-10 12:22:44 the testing repo is done building 2015-11-10 12:37:08 ncopa: minetest fixed https://github.com/alpinelinux/aports/pull/20 2015-11-10 12:37:42 nice 2015-11-10 12:37:48 i'll pull that 2015-11-10 12:37:58 neex fix a few other things first 2015-11-10 12:48:31 i think we need either update or remove the kde stuff from testing 2015-11-10 12:48:39 same with the enlightenment stuff 2015-11-10 12:51:31 does enlightenment work without systemd nowdays? 2015-11-10 12:52:14 I think so 2015-11-10 12:52:18 at least e17 did 2015-11-10 12:54:18 kde5 also works w/o systemd 2015-11-10 12:54:40 I thought that some distro even got gnome3 w/o systemd 2015-11-10 12:57:42 s/kde5/plasma5 2015-11-10 12:58:02 someone should try upgade testing kde 2015-11-10 12:59:26 ncopa: to 4.14.3 or plasma 5? 2015-11-10 12:59:41 whatever is simplest 2015-11-10 12:59:49 i suppose we want plasma5 2015-11-10 12:59:54 and get rid of kde4 2015-11-10 12:59:57 bumping from 4.13.x to 4.14.y is okayish to do 2015-11-10 13:00:13 excluding gcc5 issues 2015-11-10 13:01:08 plasma5 should be called plasma and not kde so both can coexist in the repo (not installed) 2015-11-10 13:01:26 it's really a mess ATM, migrating to plasma is a nightmare 2015-11-10 13:01:41 the good thing is we dont really have a working kde 2015-11-10 13:01:46 so we dont need to migrate 2015-11-10 13:01:48 :) 2015-11-10 13:01:51 oh 2015-11-10 13:02:02 it doesn't work? in which sense? 2015-11-10 13:02:03 its totally broken atm 2015-11-10 13:02:07 does not run 2015-11-10 13:02:15 does not install even likely 2015-11-10 13:02:30 so either purge 2015-11-10 13:02:36 or upgrade/fix 2015-11-10 13:02:42 not worth work on kde4 imho 2015-11-10 13:02:51 in that case I agree 2015-11-10 13:03:24 ncopa: you mentioned breaking deps if you move stuff to community. folks have to actually use community as a repo for the deps to work right? 2015-11-10 13:03:39 no 2015-11-10 13:03:48 nothing in main should depend on community 2015-11-10 13:04:05 that's what i was saying :) 2015-11-10 13:04:06 if you use a package that was moved to community, then you need add community repo 2015-11-10 13:04:19 i fixed a few of those issues this morning 2015-11-10 13:05:00 webkit s are in community 2015-11-10 13:05:25 anything in main that depends on webkits? 2015-11-10 13:05:28 ncopa: kde/plasma is a huge task to undertake, does Alpine wants/needs kde/plasma? 2015-11-10 13:05:43 xming: only if someone cares 2015-11-10 13:05:47 geary 2015-11-10 13:05:49 i dont care too much 2015-11-10 13:05:59 xming: so we can remove the broken stuff for now 2015-11-10 13:06:34 geary is in testing 2015-11-10 13:06:41 if you use testing you need both main and community 2015-11-10 13:07:49 can we move geary to community? works fine for me 2015-11-10 13:08:43 need to move gestting-desktop-schema too then 2015-11-10 13:09:15 can we upgrade gsetting-desktop-schema to 3.18, test it and then move? 2015-11-10 13:09:22 ok sure 2015-11-10 13:10:53 it's building now 2015-11-10 13:12:19 3.18 builds fine. do you want a patch? version bump only 2015-11-10 13:12:33 i'm on it 2015-11-10 13:12:51 ncopa is on it like a car bonnet 2015-11-10 13:13:23 ok 2015-11-10 13:13:31 if you can just confirm that nothing breaks badly 2015-11-10 13:13:36 then i'll move them to community 2015-11-10 13:14:18 in case the car bonnet thing got lost in translation -> http://tinyurl.com/pyzlbhs 2015-11-10 13:15:24 heh 2015-11-10 13:17:12 xming: was it you who used seafile? 2015-11-10 13:17:26 i've been looking at seafile 2015-11-10 13:17:37 i was never able to make it work really 2015-11-10 13:17:40 it needs pacakge splitting into client and server i think 2015-11-10 13:17:47 and a possible upgrade to 5.0.0 2015-11-10 13:17:49 i'd like to move it to community 2015-11-10 13:17:52 yeah 2015-11-10 13:19:25 ncopa: okay, 2015-11-10 13:19:47 it worked pretty well here, but totally broken after an upgrade 2015-11-10 13:20:17 seafile is so fragile 2015-11-10 13:20:54 I will take a look at 5.0 2015-11-10 13:21:27 right 2015-11-10 13:21:31 so not suitable for main 2015-11-10 13:22:52 can you rebuild ncmpcpp against updated boost? 2015-11-10 13:23:11 ok 2015-11-10 13:25:57 ScrumpyJack: pushed 2015-11-10 13:26:16 i would really prefer to try fix up main and community instead of adding more work (eg add to testing) 2015-11-10 13:26:27 moving things to community is ok 2015-11-10 13:26:29 or main 2015-11-10 13:26:52 was ncmpcpp in testing? 2015-11-10 13:28:33 yup 2015-11-10 13:28:39 main will do, it's not a problematic package 2015-11-10 13:28:54 nice to see the path completion qemu bug fixed 2015-11-10 13:34:14 new list for v3.3: http://tpaste.us/2KMw 2015-11-10 13:34:32 too much perl :) 2015-11-10 13:34:55 where is seafile? 2015-11-10 13:35:06 main/seafile iirc 2015-11-10 13:35:10 there is also ccnet 2015-11-10 13:35:17 maybe i should just move it 2015-11-10 13:35:45 i thought seafile didn't build, it was on yesterday's list 2015-11-10 13:35:54 probably some dep 2015-11-10 13:36:03 which now was resolved 2015-11-10 13:36:17 lm_sensors doesn't build because the lm sersors site is down 2015-11-10 13:36:23 so perhaps another source? 2015-11-10 13:36:37 imap built fine for me last night 2015-11-10 13:36:46 let's move seafile 2015-11-10 13:36:51 no idea what ccnet is 2015-11-10 13:37:00 a dep for seafile 2015-11-10 13:37:08 then go for it 2015-11-10 13:37:14 perl packages usually miss perl-module-build in makedepends 2015-11-10 13:38:55 should be trivial to fix then 2015-11-10 13:39:05 any objection to xorg-server-1.18? 2015-11-10 13:39:09 i think its in greyzone 2015-11-10 13:39:11 too late 2015-11-10 13:39:19 but normally not a big problem 2015-11-10 13:39:28 let's bring it in 2015-11-10 13:39:43 i'll regret it :) 2015-11-10 13:39:51 but lets do it 2015-11-10 13:43:50 libreoffice seems broken 2015-11-10 13:44:18 /usr/lib/libreoffice/program/soffice is missing 2015-11-10 13:45:16 hmm, libreoffice-common didn't get pulled in 2015-11-10 13:49:41 right, so when you install abiword in a alpine-base, it requires 65 deps 2015-11-10 13:51:20 so, how do you judge what you need to move to community if abiword goes into community? 2015-11-10 13:53:42 grep abiword main/*/APKBUILD 2015-11-10 13:53:55 if something shows up, then this also needs to move or abiword stay 2015-11-10 13:54:16 hum 2015-11-10 13:54:19 py-gtk is broke 2015-11-10 13:55:22 ok, my terrible logic was at fault 2015-11-10 14:42:50 ncopa: is there a reason easy-rsa is in testing? 2015-11-10 14:44:40 not that i know 2015-11-10 14:44:50 probably due to it got splitted out from openvpn pkg 2015-11-10 14:44:56 and i havent checked if it actually works 2015-11-10 14:48:03 nvm, i just downloaded it from github. 2015-11-10 14:49:39 hmm, it makes sence to move it to main and include it in 3.3 2015-11-10 14:49:47 or community 2015-11-10 14:49:55 yeah 2015-11-10 14:50:03 feel free to do so 2015-11-10 14:50:08 i think main is good 2015-11-10 14:52:03 hum 2015-11-10 14:52:12 xf86-video-intel has not made release since forever 2015-11-10 15:01:05 ncopa: yes, Arch switched to git 2015-11-10 15:02:25 i noticed 2015-11-10 15:03:30 seems like lastest git does not build with xorg-server 1.18 either 2015-11-10 15:14:03 ok, check / device and it's clean 2015-11-10 15:14:31 apk --verbose fix is now downloading (1/1) Installing libreoffice-common (5.0.2.2-r3) 2015-11-10 15:17:52 strace is showing 90% reads 2015-11-10 15:22:39 it looks like it downloads as .apknew.libreoffice-common-5.0.2.2-r3.5f458da7.apk 2015-11-10 15:22:43 xf86-video-nouveau release is useless 2015-11-10 15:22:51 and needs bunch of patches 2015-11-10 15:23:09 where does it try to write an apk? 2015-11-10 15:26:14 in apk cache? 2015-11-10 15:28:45 ok 2015-11-10 15:28:51 here comes xorg-server-1.18 2015-11-10 16:06:33 new build fail list: http://tpaste.us/AoOM 2015-11-10 16:06:35 i need to go 2015-11-10 16:06:36 see u 2015-11-10 16:17:55 oops, so the problem with large apk updates claiming that my disk was full was in fact mini_httpd on my alpine mirror that was dropping connections after ~80 Mb 2015-11-10 16:18:19 mini_httpd replaced with lighttpd 2015-11-10 16:28:23 ok, new xorg and nouveau! yikes! 2015-11-10 16:30:14 zmap builds here, maybe a pax issue? 2015-11-10 19:08:58 midori and ristretto seg fault on x86 2015-11-10 19:27:21 ScrumpyJack: that shows as disk full?? 2015-11-10 19:27:27 does that happen on any connection drop 2015-11-10 19:36:39 good evening gentlemen 2015-11-10 19:38:15 ncopa: do we have an updated failure list? :) 2015-11-10 20:20:00 chris| (17:06:34) (@ncopa) new build fail list: http://tpaste.us/AoOM 2015-11-10 20:20:12 thanks 2015-11-10 21:17:20 ammunta: no, just apk, so whatever apk uses. libcurl probably 2015-11-10 21:21:00 ScrumpyJack: libfetch, actually 2015-11-11 08:09:27 any git gurus about? 2015-11-11 08:09:51 if i use git format-patch HEAD^ 2015-11-11 08:11:07 git adds a From line at the top of the patch 2015-11-11 08:11:18 From 487133179e2829f8146b385170aa77cb423328b2 (with the date) 2015-11-11 08:11:40 this screws up my MTA 2015-11-11 08:12:50 I could sed it out before piping to an MTA, but can git be confed not to add that line? 2015-11-11 08:13:42 why would you send that line to MTA? 2015-11-11 08:14:31 git send-email should take that out 2015-11-11 08:35:16 barthalion: before moving anythign from main to community you need check if there are anything that depends on it 2015-11-11 08:35:25 http://build.alpinelinux.org/buildlogs/build-3-3-x86_64/main/virt-manager/virt-manager-1.2.1-r0.log 2015-11-11 08:35:58 in this case either virt-manager should move to commuinty too or gnome-icon-theme needs to stay in main 2015-11-11 08:36:25 i think i vote for moving virt-manager to community too 2015-11-11 08:36:35 anybody who use virt-manager who needs longterm support for it? 2015-11-11 08:36:41 or is latest stable good enough? 2015-11-11 08:45:32 ncopa: what are cpan makedepends? 2015-11-11 08:45:45 any magic in it? 2015-11-11 08:46:01 cpan itself? 2015-11-11 08:46:13 or hos apkbuild-cpan detects deps in perl modules? 2015-11-11 08:46:20 for instance, some perl module is missing perl-module-build 2015-11-11 08:46:37 apkbiuld-cpan pulls som json or yaml with the deps 2015-11-11 08:46:39 should i add it to makedepends or cpanmakdepends? 2015-11-11 08:47:00 ah ok, so its an automatic thing 2015-11-11 08:47:03 but i think upstream has not specified that they need perl-module-build as it used to be bundled with perl iteslf 2015-11-11 08:47:19 then ill just add it to makedepends 2015-11-11 08:47:24 yes 2015-11-11 08:47:53 it was probably part of perl-dev? 2015-11-11 08:48:55 im going to go over that list of perl modules 2015-11-11 08:50:09 not tried git send-mail 2015-11-11 08:52:40 i think it is not bundled with perl-5.22 at all 2015-11-11 08:53:38 fabled: i commented that line in strongswan and pushed it. 2015-11-11 08:54:54 hmm, git send-email requires perl, which isn't very good on a build box 2015-11-11 09:02:46 correct 2015-11-11 09:03:10 i'd love a perl-less git send-email implementation 2015-11-11 09:04:51 so would I 2015-11-11 09:08:18 if all you need is to send a patch at a time, alias git-send-patch to 2015-11-11 09:08:24 git format-patch HEAD^ --stdout| sed '1c\To: alpine-aports@alpinelinux.org | msmtp alpine-aports@alpinelinux.org 2015-11-11 09:08:39 replace msmtp with whatever you use 2015-11-11 09:13:33 ncopa: can i have -C option added to abuild as a checksum option? 2015-11-11 09:13:47 so i can do abuild -Cr 2015-11-11 09:18:11 ncopa: oops, sorry 2015-11-11 09:18:20 ncopa: let's move virt-manager, it's desktop app anyway 2015-11-11 09:18:31 i think i did 2015-11-11 09:36:08 what's the reason for leaving abiword audacious gnumeric geary geany geeqie bluefish audacity rawtherapee and the likes in main? 2015-11-11 09:40:03 no 2015-11-11 09:40:09 no reason 2015-11-11 09:40:29 :) 2015-11-11 09:40:33 i just did not have the time to check if something else depend on them so they can be moved 2015-11-11 09:40:53 well perhaps we should leave stuff in main for minimal desktop? 2015-11-11 09:40:53 geary is not in main? 2015-11-11 09:40:59 why? 2015-11-11 09:41:09 why would you like run a minimal desktop for 2 years? 2015-11-11 09:41:28 i dont mind keep things in mind as long as it does not require maintenance work 2015-11-11 09:41:36 many of those things never have any sec fixes 2015-11-11 09:41:48 so they are no problem to keep in main 2015-11-11 09:41:53 that's good then right? 2015-11-11 09:41:58 yes 2015-11-11 09:42:10 is geary in main? 2015-11-11 09:42:17 i think so, hang on 2015-11-11 09:42:40 no 2015-11-11 09:44:51 it's in testing. 2015-11-11 09:45:18 there is no email desktop client in main, nor a browser 2015-11-11 09:46:16 so we could move those things to community too then 2015-11-11 09:46:26 will community be in repo file for new iso builds? 2015-11-11 09:46:31 no point in keeping them in main 2015-11-11 09:46:38 thats a good q 2015-11-11 09:46:44 probably not 2015-11-11 09:46:57 unles we create a desktop iso of some sort 2015-11-11 09:47:42 ok, so i download the iso, in boot it up, and so apk search for desktop stuff, i'll find nothing? 2015-11-11 09:49:00 i run setup-xorg-base, apk add [some windowmanager] 2015-11-11 09:50:18 should probably make the option to add community repo from some of the setup scripts 2015-11-11 09:50:31 good idea 2015-11-11 09:50:43 in setup-repo perhaps 2015-11-11 09:51:14 setup-apkrepos evbenen 2015-11-11 09:51:46 hm 2015-11-11 09:51:56 libelf we ship is 0.8.13 2015-11-11 09:52:03 sorry, my connection is so bad, that sometimes even backspace gets dropped 2015-11-11 09:52:06 libelf arch ship is 0.164 2015-11-11 09:52:26 libelf sabotage linux ship seem to be different too 2015-11-11 09:53:11 ah 2015-11-11 09:53:16 arch linux shipf elfutils 2015-11-11 09:54:45 ok 2015-11-11 09:54:49 this is taking too much time 2015-11-11 09:55:10 any volunteers to look at making perf run on alpine? 2015-11-11 09:55:22 we need it for some kernel testing 2015-11-11 09:56:08 here is what i got so far: http://tpaste.us/2bn4 2015-11-11 09:56:17 but i think we need other libelf implementation 2015-11-11 09:56:35 or hack the makefile to not look for gnu/libc-version.h without libelf 2015-11-11 09:57:24 http://pkgs.fedoraproject.org/cgit/elfutils.git/tree/elfutils.spec 2015-11-11 09:57:44 have voidlinux got perf built ? 2015-11-11 09:58:30 ncopa: can opensmtpd go to main? 2015-11-11 09:58:51 contributor happy, and so am I :) 2015-11-11 10:02:47 sabotage linux got it built 2015-11-11 10:02:49 yes 2015-11-11 10:03:09 seems like ftp if broke from the build servers 2015-11-11 10:03:17 sigh 2015-11-11 11:02:54 ACTION slaps perl and its deps 2015-11-11 11:17:47 lldpad is driving me nuts 2015-11-11 11:18:08 so, it doesn't build, problem with undecl 2015-11-11 11:18:31 debian fixed it, but i can't find there patch 2015-11-11 11:19:09 here is the error 2015-11-11 11:19:11 http://tpaste.us/370g 2015-11-11 11:24:16 ScrumpyJack: https://bugs.debian.org/cgi-bin/bugreport.cgi?msg=10;bug=790647;filename=lldpad.debdiff;att=1 2015-11-11 11:24:32 yeah i saw that, no patch though 2015-11-11 11:25:12 huh? that link is the patch, or doesn't it work? 2015-11-11 11:25:38 that doesn't patch the issue 2015-11-11 11:26:24 have you tried -fgnu89-inline ? 2015-11-11 11:27:12 i tried GNU89, yes 2015-11-11 11:27:15 http://clang.debian.net/status.php?version=3.3&key=UNDEF_REF 2015-11-11 11:27:35 but that was mainly for clang. I think this is snice gcc 5+ 2015-11-11 11:29:48 xming: if you want to have a look, give me your public key 2015-11-11 11:32:14 ScrumpyJack: do you have the APKBUILD? 2015-11-11 11:33:18 \o/ 2015-11-11 11:33:26 perl crap fixed 2015-11-11 11:35:04 xming: I'm not there yes, still building by hand 2015-11-11 11:35:22 hang on 2015-11-11 11:37:20 http://tpaste.us/A1lE 2015-11-11 11:39:08 add #include to src/open-lldp-0.9.46/include/clif.h to get past the first error 2015-11-11 11:39:56 i do abuild deps then abuild build, fix until it works, *then* make the diff patches 2015-11-11 11:40:04 probably not very efficient 2015-11-11 11:44:36 omigosh clif.h is generated? 2015-11-11 11:49:14 ScrumpyJack: it compiles here :D 2015-11-11 11:49:38 as I said you need -fgnu89-inline 2015-11-11 11:50:01 ScrumpyJack: http://sprunge.us/NAFf 2015-11-11 11:51:55 ah, maybe i put the flag in the wrong place when I tested 2015-11-11 11:53:55 ScrumpyJack: maybe you forgot to export? 2015-11-11 11:54:27 i was testing on the command line http://tpaste.us/370g 2015-11-11 11:55:00 I put the -fgnu89-inline flag in before the -fomit-frame-po 2015-11-11 11:55:04 inter 2015-11-11 11:56:41 well there is no -fgnu89-inline in your paste, so it didn't pick that up 2015-11-11 11:56:49 need to pass that at configure stage 2015-11-11 11:57:10 that is the original line that generated the error, cleaned of my crappy tests 2015-11-11 11:57:36 xming: could you submit a patch? It's your work now :) 2015-11-11 11:58:05 ScrumpyJack: nah you do that, you still have to patch 2015-11-11 11:58:57 my part was just looking that up and showed you that was actually the correct Debian patch :D 2015-11-11 11:59:48 yeah, i missed the bottom bit of that patch 2015-11-11 12:00:12 nice spot dude 2015-11-11 12:07:57 ncopa: what is the reason for php not to be build? 2015-11-11 12:22:35 imap 2015-11-11 12:22:59 i think it fails to download from ftp 2015-11-11 12:23:11 i suspect firewall issue 2015-11-11 12:23:29 imap built for me, lemme check 2015-11-11 12:23:49 it builds for me too 2015-11-11 12:24:12 imap builds fine for me 2015-11-11 12:24:34 i'm pretty sure its fw issue in front of the build server 2015-11-11 12:25:49 cant we just move the source to dev.a.o? 2015-11-11 12:27:47 ncopa: and can you generate the list again, it should not have any perl modules anymore. 2015-11-11 12:54:30 freeswitch is seriously annoying 2015-11-11 12:54:40 they silently changed their lame fork 2015-11-11 12:54:50 same tarball previously and now: http://tpaste.us/AenO 2015-11-11 13:16:00 did someone build mumble? my git repo is all messed up 2015-11-11 13:24:31 looks like builders are building a bunch of perl-* stuff 2015-11-11 13:24:36 nice 2015-11-11 13:28:32 you spoke too quick i think http://sprunge.us/PWKG 2015-11-11 13:28:48 that's the last 20 failures of 3.3 x86_64 2015-11-11 13:29:19 nice 2015-11-11 13:49:05 oops, can someone reject my patch 979 2015-11-11 13:59:16 hey :) 2015-11-11 13:59:19 how are you? 2015-11-11 13:59:28 hi 2015-11-11 13:59:33 good thanks 2015-11-11 14:03:42 glad to hear that. i was out 5 days with 39C temperature and a headache ... 2015-11-11 14:05:42 today i read some articles about wayland 2015-11-11 14:05:54 has someone of you played around with it? 2015-11-11 14:06:05 to me it does not seam to improve stuff 2015-11-11 14:06:25 that reminds me, X11 depends on wayland at the moment 2015-11-11 14:10:38 ScrumpyJack: have you used wayland? 2015-11-11 14:24:17 never. i would like to though. 2015-11-11 16:22:13 ncopa: can we have a list before you go? 2015-11-11 16:24:31 http://tpaste.us/AOqw 2015-11-11 16:24:38 i started to work on freeswitch 2015-11-11 16:24:41 but did not finish 2015-11-11 16:24:48 need upgrade it to 1.6.2 2015-11-11 16:25:22 i think fixing gtkman will fix rawtherapee 2015-11-11 16:25:23 if you want something bigger to bite, llvm-3.7 would be nice 2015-11-11 16:25:28 probably 2015-11-11 16:25:32 llvm 3.7 2015-11-11 16:25:43 but needs fast compile machine as it is huge and slow 2015-11-11 16:25:55 but it's not in the list! :) 2015-11-11 16:25:57 patches needs to be rebased 2015-11-11 16:26:24 and we want clang statically linked against llvm libs 2015-11-11 16:26:49 can you check patch for open-lldp and see if it builds? 2015-11-11 16:27:07 checksum err 2015-11-11 16:27:11 grr 2015-11-11 16:27:11 gotta run 2015-11-11 16:27:15 see u 2015-11-11 16:27:15 and 2015-11-11 16:27:19 big thanks!!! 2015-11-11 16:27:19 ok, have a good evening 2015-11-11 16:27:28 appreciate your help 2015-11-11 16:28:12 lots of folks chipping in - i love this community 2015-11-11 16:31:16 kunkku: are you about? 2015-11-11 17:55:15 is there any tool to remove a package entry from an APKINDEX? 2015-11-11 17:59:40 ah, just regenerating the entire index using abuild index works fine 2015-11-11 18:33:18 bonsoir 2015-11-11 18:34:17 good evening 2015-11-11 18:34:58 just got home. seen this list chris| ? 2015-11-11 18:35:53 new list? 2015-11-11 18:36:00 this one? http://tpaste.us/AoOM 2015-11-11 18:38:22 this one http://tpaste.us/AOqw 2015-11-11 18:38:39 cool pretty short :) 2015-11-11 18:39:57 ive done aconf 2015-11-11 18:40:06 lua-xctrl just needs new checksums 2015-11-11 18:40:18 giblib the source has vanished 2015-11-11 18:40:28 same with lm_sensors 2015-11-11 18:40:43 i've done open-lldp 2015-11-11 18:41:17 that leave mariadb, open-vm-tools, the perl and onwards 2015-11-11 18:42:15 I'm on mariadb 2015-11-11 18:42:25 that's brave :) 2015-11-11 18:42:50 ah well, it compiles for ages, so it's pretty relaxing ;) 2015-11-11 18:44:24 anyone with access willing to put lm_sensors on dev.a.o? 2015-11-11 18:45:08 there must be a sensible fork in github 2015-11-11 18:47:22 there is always http://http.debian.net/debian/pool/main/l/lm-sensors/lm-sensors_3.4.0.orig.tar.bz2 2015-11-11 18:50:29 and since it will probably not see that many updates, I think just having it on dev.a.o will be sufficient 2015-11-11 18:54:43 mariadb builds fine 2015-11-11 19:00:07 cool 2015-11-11 19:00:55 gtkmm seems already fixed with r3 2015-11-11 19:01:07 don't know how mariadb ended up there 2015-11-11 19:01:47 hasn't someone done gtkmm or did you misread gtkman 2015-11-11 19:03:16 harr, got me 2015-11-11 19:09:08 ping me if you have any patches, I'll be there for a while yet 2015-11-11 19:09:44 ScrumpyJack: zmap builds fine here 2015-11-11 19:10:16 cool 2015-11-11 19:15:44 barthalion: could you take a look at 981 983 and 987 2015-11-11 19:16:24 although a checksum change of 987 without a version bump of the source is annoying 2015-11-11 19:19:43 hm, the old swatch version doesn't seem to be available anymore 2015-11-11 19:19:48 I will upgrade it to the latest one 2015-11-11 19:25:53 barthalion: %988 is a fix for main/swatch 2015-11-11 19:28:38 ScrumpyJack: aconf patch seems wrong 2015-11-11 19:28:44 it changes url, not source 2015-11-11 19:29:35 pushed the rest 2015-11-11 19:32:16 yes, that is a bit wrong. however, it builds 2015-11-11 19:32:26 but it makes no sense 2015-11-11 19:32:32 $url is not referenced anywhere 2015-11-11 19:33:07 i know. it was just that the URL was wrong 2015-11-11 19:33:29 so the build failed due to that? 2015-11-11 19:33:32 weird 2015-11-11 19:33:50 no, i don't think the build failed before 2015-11-11 19:34:07 don't know how it ended up in the list 2015-11-11 19:34:15 http://build.alpinelinux.org/buildlogs/build-3-3-x86_64/main/aconf/aconf-0.5.0-r0.log 2015-11-11 19:34:18 checksum failure 2015-11-11 19:34:31 ay! good spot 2015-11-11 19:34:53 does the patch have new sum? 2015-11-11 19:34:54 but checksums are fine 2015-11-11 19:35:03 it's just build infra being derpy 2015-11-11 19:35:16 ncopa: please remove aconf tarball from distfiles on builders 2015-11-11 19:46:20 barthalion: would you mind uploading the lmsensors tarball to dev.a.o ? 2015-11-11 19:46:56 you mean lm_sensors? 2015-11-11 19:47:06 why do we need to mirror it? 2015-11-11 19:47:24 the website is down and I don't think its comming back 2015-11-11 19:48:06 I'll switch us over to Fedora's mirror 2015-11-11 19:50:04 done 2015-11-11 19:50:53 ah, we can do that then, thanks :) 2015-11-11 19:55:40 so, what are we left with on the list? 2015-11-11 20:02:00 I'll look at freeswitch 2015-11-11 20:02:51 I cannot reproduce zmap 2015-11-11 20:11:46 zmap build fine 2015-11-11 20:11:52 builds even 2015-11-11 20:12:45 do you have grsec kernel? 2015-11-11 20:13:22 i get /usr/bin/hg not found when mercurial package is installed for rawtherapee 2015-11-11 20:13:45 you mean grsec kernel when building? 2015-11-11 20:13:51 yes 2015-11-11 20:13:57 yes i do 2015-11-11 20:14:04 ok 2015-11-11 20:14:13 i thought it could be a pax issue 2015-11-11 20:14:29 clandmeter, this looks a bit ugly: http://patchwork.alpinelinux.org/project/ 2015-11-11 20:14:35 :) 2015-11-11 20:14:57 it does indeed 2015-11-11 20:17:26 rawtherapee doesn't build since the gtkmm upgrade :( 2015-11-11 20:18:00 HappyJack: how did you get to that uri? 2015-11-11 20:18:11 just type it, or is there a link to it? 2015-11-11 20:29:13 hmm 2015-11-11 20:29:27 patchwork moved to github and have an actual release? 2015-11-11 20:30:50 https://lists.ozlabs.org/pipermail/patchwork/2015-October/001873.html so it does 2015-11-11 20:31:41 i just typed the url 2015-11-11 20:36:09 i have no experiance in django, but if i change any setting it all goes boom 2015-11-11 20:40:40 HappyJack: fixed. 2015-11-11 20:43:50 what was there? 2015-11-11 20:45:56 debug stuff 2015-11-11 20:46:04 because debug was set to true 2015-11-11 20:46:52 ah 2015-11-11 21:37:04 no U 2015-11-11 21:37:11 oops wrong buffer. 2015-11-12 08:07:41 morning 2015-11-12 08:11:19 good morning 2015-11-12 08:16:09 anyone moved to 3.3 yet 2015-11-12 08:18:16 its the same as edge? 2015-11-12 08:21:09 from 3.2 2015-11-12 08:24:49 clandmeter: yes v3.3 is currently same as edge 2015-11-12 08:24:53 different build 2015-11-12 08:24:57 so might differe a bit 2015-11-12 08:25:36 i'm on latest stable for anything that matters, and edge+testing on desktops and laptops. But servers and desk/laptops run different software (obviously), so upgrading servers to latest can sometimes lead to surprises :) I'm sure it's the same for most folk 2015-11-12 08:26:25 perhaps I better question would have been: Anyone moved their stuff that matters to 3.3 yet? :) 2015-11-12 08:31:31 yeah 2015-11-12 08:31:44 anything that matters should be moved to main or community 2015-11-12 08:49:48 ncopa: it looks like a few of those things on last night's list built fine 2015-11-12 08:50:18 namely: mariadb, zmap 2015-11-12 08:51:33 http://build.alpinelinux.org/buildlogs/build-3-3-x86_64/main/zmap/zmap-2.1.1-r0.log 2015-11-12 08:52:19 I for one cannot reproduce that error 2015-11-12 08:55:02 i borked aconf tag 2015-11-12 08:55:09 i will try clean tat up 2015-11-12 08:55:36 or more exact, will ask kunkku to clean up... 2015-11-12 09:01:42 for zmap: is it wierd that you get /bin/sh: gengetopt: not found even though you have gengetopt-dev as a dep 2015-11-12 09:02:28 I saw that with another package too 2015-11-12 09:07:54 good morning 2015-11-12 09:08:07 i see you guys have been busy updating packages :) 2015-11-12 09:08:13 84 updates in edge :) 2015-11-12 09:10:48 ScrumpyJack: its not pulling in gengetopt 2015-11-12 09:10:53 just -dev pkg 2015-11-12 09:11:14 (8/16) Installing json-c-dev (0.12-r1) 2015-11-12 09:11:14 (10/16) Installing gengetopt-doc (2.22.6-r1) 2015-11-12 09:11:14 (9/16) Installing gengetopt (2.22.6-r1) 2015-11-12 09:11:14 (11/16) Installing gengetopt-dev (2.22.6-r1) 2015-11-12 09:14:21 fixed now 2015-11-12 09:14:39 though i dont know why -dev doesnt pull in regular pkg 2015-11-12 09:24:47 it *was* pulling it on for me 2015-11-12 09:24:57 wierd 2015-11-12 09:26:33 yes thats what im saying, for us it did, but not for the builder. 2015-11-12 09:26:54 anyway, it doesnt need the dev pkg, so changing to regular one makes more sence. 2015-11-12 09:27:25 there are probably tons of pkgs that pull in -dev because dev was lazy :) 2015-11-12 09:29:19 abuild changed 2015-11-12 09:29:24 I guess you're right, I pull in -dev for most packages just to be sure 2015-11-12 09:29:25 older abuild pulled in all subpackages 2015-11-12 09:29:39 as deps for -dev 2015-11-12 09:29:42 new abuild does not 2015-11-12 09:29:53 it only pulls in what it think it needs 2015-11-12 09:30:01 eg symlink targets 2015-11-12 09:30:09 i have new abuild 2015-11-12 09:30:14 and it deos pull it in 2015-11-12 09:31:33 which verison of abuild doesn't pull in quite right? 2015-11-12 09:32:01 I have 2.25.0_rc1-r2 2015-11-12 09:32:44 current abuild does not pull in $pkg unless $pkg-dev defines a dependency on it 2015-11-12 10:04:22 why is that? 2015-11-12 10:04:44 almost always someone who wants development headers, needs also main package 2015-11-12 10:05:12 not always, there was a bug re postgresql 2015-11-12 10:05:25 docker ppl didnt wan pull in more than libpq 2015-11-12 10:05:40 so what abuild does now 2015-11-12 10:05:49 if there is a .so symlink 2015-11-12 10:05:51 in -dev 2015-11-12 10:06:10 then will it pull in whatever subpackage that has the symlink target 2015-11-12 10:06:29 so in most cases it will pull in the main package or the -libs subpackage 2015-11-12 10:06:37 which is why it works in most cases 2015-11-12 10:06:54 it will not pull in -doc like it previously did 2015-11-12 10:07:48 this also means that if a -dev package needs something in /usr/bin/ then will this dep need to be added manually 2015-11-12 10:09:05 like this: http://git.alpinelinux.org/cgit/aports/commit/?id=c0a320103ec62c20c049621f8fe92941f491ef41 2015-11-12 10:14:04 oh no /o\ 2015-11-12 10:14:13 i have made a huge mistake 2015-11-12 10:15:46 the 32 bit, x86 v3.3 build 2015-11-12 10:15:50 needs to be done from scratch 2015-11-12 10:16:04 i'm taking it offline for a while 2015-11-12 10:16:24 problem is: CHOST=i486-alpine-linux-musl 2015-11-12 10:16:33 when it should been: CHOST=i586-alpine-linux-musl 2015-11-12 10:20:12 is that why lots of x86 stuff seg faults in edge? 2015-11-12 10:21:16 mqtt shows build/build-3-3-x86 (null) 2015-11-12 10:22:42 i took it offline 2015-11-12 10:22:45 yes 2015-11-12 10:22:52 i tihnk thats why some builds fails 2015-11-12 10:22:58 mariadb for sure 2015-11-12 10:27:03 is there a problem in building for i486 when you have a 586? 2015-11-12 10:27:12 I mean, it will be less efficient, but why would it crash? 2015-11-12 10:32:23 libmysqld/../sql/mysqld.cc:4028: undefined reference to `__atomic_fetch_add_8' 2015-11-12 10:32:41 there are some missing instructions 2015-11-12 10:32:56 so some of the atomic fetch are missing 2015-11-12 10:33:02 there are workarounds 2015-11-12 10:33:13 but better just bump to i586 2015-11-12 10:33:19 we decided that right afver v3.2 2015-11-12 10:34:09 ah, some software decided to drop 486 compatibility, I see. 2015-11-12 10:46:44 hmm, the time has come. I need a alpine-iso with the whole repo on a DVD 2015-11-12 10:48:48 but edge/main is 4.1Gb 2015-11-12 10:49:02 hmm, i guess I'll pick and choose 2015-11-12 10:49:32 Use a Blu-Ray! :) 2015-11-12 10:51:00 ScrumpyJack: identify the biggest packages and suggest move them to community ;) 2015-11-12 11:49:42 ScrumpyJack: maybe you can skip -dev and -doc 2015-11-12 12:40:34 clandmeter: good idea 2015-11-12 13:00:18 wiki.a.o will be down for maintenance for aprox 1 hour 2015-11-12 13:06:39 ncopa: what kind of maintenance? 2015-11-12 13:07:11 move from pg db to mariadb 2015-11-12 13:07:27 its impossible to do that 2015-11-12 13:07:38 but seems i found someone who figured it out :) 2015-11-12 13:07:51 weird 2015-11-12 13:08:01 i emailed somebody but never got an reply afterwards 2015-11-12 13:12:53 i think there are many people emailing me that never gets any reply :) 2015-11-12 13:18:04 there is a new version of xf86-video-ati (7.6.1) don't know if you want that for 3.3, it enables DRI3 for radeons 2015-11-12 13:33:03 is mariadb better for the wiki? 2015-11-12 13:33:55 pg is no longer supported by mediawiki 2015-11-12 13:35:13 that's sad 2015-11-12 13:37:55 xming: i think we want 2015-11-12 13:38:33 does it require some configure --enable-dri3 or similar? 2015-11-12 13:38:57 checking for dri3.h... yes 2015-11-12 13:39:22 xming: you have ati? 2015-11-12 13:40:19 yes 2015-11-12 13:40:27 but on my gentoo box 2015-11-12 13:40:57 k 2015-11-12 13:42:01 wiki.a.o is back online 2015-11-12 13:42:09 i am seriously impressed 2015-11-12 13:43:03 they guy has a migrating script from pg to mariadb? nice 2015-11-12 13:43:15 i think he has a set of scripts 2015-11-12 13:43:20 and probably did some manual work too 2015-11-12 13:43:26 i dont know really 2015-11-12 13:43:34 but u think he worked on the migration for weeks... 2015-11-12 13:44:20 it's much faster now :p 2015-11-12 13:46:08 ncopa: I don't think it needs any configure script to enable dri3 2015-11-12 13:46:18 s/script/flag 2015-11-12 13:47:41 i figured 2015-11-12 13:47:50 it's looking for dri3.h 2015-11-12 13:48:03 and run time configurable 2015-11-12 14:13:35 ncopa: please ask this guy to disable caching of forms 2015-11-12 14:20:41 is it still caching forms? 2015-11-12 14:44:27 ncopa: a lot of stuff is failing to build on armhf. what's the plan there? 2015-11-12 14:54:43 ncopa: yes it still does 2015-11-12 14:54:48 try to edit a page 2015-11-12 14:54:53 and then edit it again 2015-11-12 15:06:37 ScrumpyJack: are you on edge or 3.3 now? 2015-11-12 15:10:16 no edge host has been "updated" to 3.3 yet 2015-11-12 15:12:13 Are you wondering about armhf? 2015-11-12 15:12:34 ScrumpyJack: i mean, do you have alpine on your desktop? 2015-11-12 15:13:50 yes, edge. shall i change source to 3.3? 2015-11-12 15:15:24 no i have a question 2015-11-12 15:15:34 :) ok 2015-11-12 15:15:45 can you try install xca and see if it works? 2015-11-12 15:15:54 ok desktop with edge, ok 2015-11-12 15:16:25 qt pollution , yuk 2015-11-12 15:17:36 is this in relation to the java keychain issue? 2015-11-12 15:17:45 no 2015-11-12 15:17:54 Unable to load library icui18n "Cannot load library icui18n: (Error loading shared library icui18n: No such file or directory)" 2015-11-12 15:18:04 but appart from that, it works 2015-11-12 15:18:15 ok 2015-11-12 15:18:22 thx 2015-11-12 15:18:29 do you want me to create a new db? 2015-11-12 15:18:39 you can if you want 2015-11-12 15:18:56 no thanks :) 2015-11-12 20:52:45 zomg fast 2015-11-13 08:22:53 morning 2015-11-13 08:23:52 so, are we going to jiggle the ISOs for 3.3? 2015-11-13 08:24:41 alpine-mini-3.3 becomes alpine-3.3 2015-11-13 08:25:19 good morning climbers 2015-11-13 08:26:29 ay! 2015-11-13 08:27:07 this is ugly: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=54a20552e1eae07aa240fa370a0293e006b5faed 2015-11-13 08:33:57 ncopa: how come opensmtpd hasn't made it to armhf? 2015-11-13 08:36:41 because arch="x86_64"? 2015-11-13 08:37:04 someone apparently said it only works on x86_&4 2015-11-13 08:37:43 i can check on my arm builder 2015-11-13 08:38:31 ah, sorry. i didn't look at the APKBUILD file 2015-11-13 08:38:58 clandmeter: is that arm builder a scaleway C1? 2015-11-13 08:39:03 yes 2015-11-13 08:39:45 how many do you have/use? 2015-11-13 08:40:12 builders? 2015-11-13 08:40:17 scaleways 2015-11-13 08:40:27 currently only 2 2015-11-13 08:40:32 same 2015-11-13 08:40:34 but that will go up 2015-11-13 08:40:49 fabled is working on porting to armv7 2015-11-13 08:41:25 i saw some activity yesterday 2015-11-13 08:41:52 yes, i have crosscompiler + some stuff cross compiled 2015-11-13 08:42:01 fabled: that's awesome 2015-11-13 08:42:05 but it's been few years, and i need to fix some cross compilation stuff 2015-11-13 08:42:24 libtool is a nasty beast giving problems 2015-11-13 08:42:41 clandmeter: we should try to get distcc working accross a scaleway farm :) 2015-11-13 08:45:06 shoulndt be that hard 2015-11-13 08:46:11 ScrumpyJack: im getting fts-dev (missing) 2015-11-13 08:46:21 so i guess fts is not build on arm 2015-11-13 08:46:50 i see fts-dev-1.1-r5 on arm 2015-11-13 08:47:07 ah im on stable... 2015-11-13 08:47:10 :) 2015-11-13 08:47:32 i need to convert this box to lxc 2015-11-13 08:48:00 ScrumpyJack: build on C1 isnt that bad actually, of course it can always be faster. 2015-11-13 08:48:52 ScrumpyJack: https://github.com/scaleway/kernel-tools/issues/176 2015-11-13 08:48:58 convert your scaleway to lxc? interesting, what for? 2015-11-13 08:49:22 ScrumpyJack: to build edge and stable on same box 2015-11-13 08:49:46 ah 2015-11-13 08:50:03 i just switched to that kernel 2015-11-13 08:50:10 i hope it fixes my issues. 2015-11-13 08:50:26 still wonder why not more ppl had the same issues. 2015-11-13 08:50:45 i have had 4 C1's and all of them oomed 2015-11-13 08:56:23 hey :) 2015-11-13 08:57:20 hi 2015-11-13 08:58:47 re opensmtpd: 2015-11-13 08:58:49 ../../smtpd/smtpctl.c:738:9: error: format '%ld' expects argument of type 'long int', but argument 2 has type 'unsigned int' [-Werror=format=] 2015-11-13 08:58:50 printf("struct deliver=%ld\n", sizeof (struct deliver)); 2015-11-13 08:58:50 ^ 2015-11-13 08:58:54 sigh 2015-11-13 09:00:49 thats why its 64bit only 2015-11-13 09:00:53 non-portable code 2015-11-13 09:01:31 :( 2015-11-13 09:01:55 that's simple to fix 2015-11-13 09:01:55 don't the openbsd guys have a portable version of everything? 2015-11-13 09:02:01 but yeah 2015-11-13 09:03:24 http://www.openbsd.org/images/rack2009.jpg 2015-11-13 09:06:49 perhaps Portable Version mean "non bsd" 2015-11-13 09:13:16 I think portable means non OpenBSD 2015-11-13 09:17:33 https://gitweb.gentoo.org/repo/gentoo.git/log/mail-mta/opensmtpd?showmsg=1 2015-11-13 09:17:45 32 and 64-bit 2015-11-13 09:25:07 i know fix is simple, just unexpected from openbsd guys 2015-11-13 09:27:16 ncopa: gentoo ebuild doesn't include any patches for x86 2015-11-13 09:27:40 and openbsd has x86 builds too, so I just wonder 2015-11-13 09:56:12 did we fix php deps already? 2015-11-13 09:58:47 mariadb is probably missing still 2015-11-13 09:58:50 i'll check that 2015-11-13 10:29:19 dl2 and dl3 seam offline 2015-11-13 10:31:55 works for me: http://dl-2.alpinelinux.org/alpine/v3.3/ 2015-11-13 10:32:18 dl-3 seems to have issues though 2015-11-13 10:32:19 :) 2015-11-13 11:45:27 2 seams to work again 2015-11-13 11:45:29 but 3 has issues 2015-11-13 11:49:37 can someone of you explain to me why dhcp clients suck so hard? dhcp is so easy, why are all the clients so bad? 2015-11-13 12:12:13 because most people can't code 2015-11-13 12:12:33 that's an easy answer, but a true one nonetheless 2015-11-13 12:12:42 also, busybox udhcpc is a fine client 2015-11-13 13:22:30 are there anyone here who has access to the grsecurity stable patches? if so, please ping me in private 2015-11-13 14:21:49 forums are ....@ .. <- my appauling attempt at tumbleweed 2015-11-13 15:18:34 clandmeter: you about? 2015-11-13 15:19:00 just responding to that forum post about "missing tk" 2015-11-13 15:19:33 what repo does the package search lookup, current or edge? 2015-11-13 15:19:45 it feels like it's edge 2015-11-13 15:20:39 (for exemple tk appears in main, but is that edge/main or 3.2/main?) 2015-11-13 15:21:40 (tk isn't in 3.2/main, so i guess the search tools searches edge) 2015-11-13 15:21:46 can you confirm? 2015-11-13 15:25:18 its edge 2015-11-13 15:25:29 thank you kind sir! 2015-11-13 15:25:55 maybe a note saying its edge would help some ppl 2015-11-13 15:26:08 stable does not have testing. 2015-11-13 15:29:31 ay, there was a clue? -heh 2015-11-13 15:49:31 ScrumpyJack: i need to rethink how to properly handle multiple repo's 2015-11-13 15:49:52 currently its as simple as it gets, which does not scale very well 2015-11-13 15:50:10 the more repo's, the longer it takes to regenerate the db. 2015-11-13 16:11:13 clandmeter: what's the complexity? if it's O(n) in the number of repos, that's fine 2015-11-13 16:12:08 skarnet: its not complex at all, its too simple currently, and sqlite has limitations. 2015-11-13 16:12:38 I meant the algorithmic complexity 2015-11-13 16:21:49 skarnet: in what way? 2015-11-13 16:22:04 generating the db is the main issue. 2015-11-13 16:22:18 currently it gets regenerated every 15 min 2015-11-13 16:22:22 if there are changes 2015-11-13 16:31:31 clandmeter: one for you :) http://tinyurl.com/p7otpkn 2015-11-13 16:41:41 my question was about the scaling. What are the problems you encounter when you increase the number of repos? 2015-11-13 16:42:03 how much does the regeneration cost increase? 2015-11-13 16:58:14 ncopa: I was considering to work on a bunch of patches to replace openssl with libressl after the 3.3 release, what do you think of that? 2015-11-13 17:00:23 looking at void linux it doesn't seem to be too much hassle to do that 2015-11-13 17:01:17 and if you look at the CVEs published for OpenSSL in the last year you will noticed that libressl wasn't affected by many of them 2015-11-13 19:14:11 I think there maybe a problem upgrading to the 4.1.12 kernel 2015-11-14 05:02:41 clandmeter: generate the dbs for each repo in parallel, combine the db during query 2015-11-14 07:50:04 skarnet: currently the db is updated in place, meaning a new temp db is created in the sqlite file, and filled with the data, after which it will replace the original db. but sqlite doesnt seem to like updates and queries at the same time, making access impossible for the time it takes to update. 2015-11-14 07:50:49 is the recreate-and-rename replacement viable here 2015-11-14 07:51:21 ? 2015-11-14 07:53:07 skarnet: another issue is, i suck at programming, and this project was my first attempt at writting (poor) lua code. 2015-11-14 07:53:36 you create a new sqlite db with a different name on the same device 2015-11-14 07:53:47 you write that out completely 2015-11-14 07:54:04 then, you rename(2) it over the previous db 2015-11-14 07:56:00 wait, it's lua code? 2015-11-14 07:56:49 ammunta: im not sure i catch your point 2015-11-14 07:57:28 clandmeter: that avoids making it inaccessible ever 2015-11-14 07:57:58 it's kind of a stock fs-based atomic update 2015-11-14 08:02:08 yes thats possible, but im not sure i can do that in the current setup. 2015-11-14 08:02:20 i would have to write another lua script the take care of that. 2015-11-14 08:03:10 probalby seperate the code of db generation, and the other logic. 2015-11-14 08:05:06 ah 2015-11-14 08:13:48 hm why is mdocml-apropos split out 2015-11-14 08:47:37 ... haaaa 2015-11-14 08:48:14 I just spent a bunch of time trying to figure out how to compile luajit statically 2015-11-14 08:48:51 it's 2015-11-14 09:15:32 ammuhnta (sic) is right: you want to generate the new db into another file, then atomically replace the db with rename(). 2015-11-14 09:16:01 you can write a small shell script that calls your lua code for the db generation, then performs mv -f. 2015-11-14 09:17:40 So, no interruption of service, atomic update, everyone's happy. 2015-11-14 09:21:44 well unless it was a big file and you're only updating part of it 2015-11-14 09:21:45 but 2015-11-14 09:21:49 usually that's not your issue 2015-11-14 09:22:11 because when do you only update part of a file 2015-11-14 09:24:22 skarnet: meuglement is also one of mine ... you may notice a theme here 2015-11-14 09:44:24 there's a "meuglement" user in #s6... but it's not you 2015-11-14 22:09:02 clandmeter: does lua-turbo expose headers like remote_addr? 2015-11-15 01:59:51 the openrc patch fixes /etc/hostname being empty too (which currently gives root@?) 2015-11-15 07:28:20 http://apk-tools.sourceforge.net/ is this ... ancient history 2015-11-15 08:28:30 barthalion: yes 2015-11-15 08:29:34 anyone using awall here? 2015-11-15 09:28:23 ammunta: hopefully, anything hosted on sourceforge IS ancient history 2015-11-15 11:11:08 I mean it's obviously ancient the question is whether it's history 2015-11-15 20:36:02 ncopa: why did you hardcode this? http://git.alpinelinux.org/cgit/aports/tree/main/dnssec-root/APKBUILD#n29 2015-11-16 07:10:08 nmeum: i dont remember but probabley due to the url https://data.iana.org/root-anchors/root-anchors.xml not containing any version number 2015-11-16 07:10:27 so checksum might change if downloaded later 2015-11-16 07:10:39 so i probably downloaded it and stored it in our git repo 2015-11-16 07:10:59 and then moved it into the APKBUILD to reduce number of files 2015-11-16 07:20:45 hi all 2015-11-16 07:38:57 hi 2015-11-16 07:39:07 claws-mail give me: Couldn't initialize en_US speller. 2015-11-16 07:39:32 i wonder if it due to the aspell move of dictionaries? 2015-11-16 07:39:55 we maybe need rebuild everything linked against aspell? 2015-11-16 07:55:57 i wonder where the default aspell path is takenfrom 2015-11-16 07:58:05 morning 2015-11-16 07:59:38 what builds are still broken for 3.3? 2015-11-16 08:17:20 ncopa, i know what happened with aspell 2015-11-16 08:17:26 aspell-libs got splitted 2015-11-16 08:17:33 and the main pkg is no longer pulled in by anything 2015-11-16 08:17:44 installing aspell makes claws-mail work again 2015-11-16 08:18:36 not sure who should depend on it 2015-11-16 08:18:38 maybe enchant 2015-11-16 08:37:44 if all needs it 2015-11-16 08:37:56 do we want split the lib then? 2015-11-16 08:43:00 i dunno. it contains lots of stuff: data files, and the executable 2015-11-16 08:53:12 ncopa: also: in the dnssec-root package you never actually create the root anchors dnskey as suggested on the webpackage you linked in the APKBUILD 2015-11-16 08:53:44 I might be wrong but the package doesn't seem to ship any dnssec key 2015-11-16 09:31:42 maybe I am somehow missing the purpose of this package…but it seems very incomplete to me… 2015-11-16 09:38:36 its possible 2015-11-16 09:38:44 i wonder why php no longer builds 2015-11-16 09:38:53 it fails on PSPELL 2015-11-16 09:39:04 smells like the recent aspell changes broke things for us 2015-11-16 10:10:16 giblib/src 2015-11-16 10:10:21 source url seems to be broken 2015-11-16 10:46:06 hi :) 2015-11-16 10:48:09 hi 2015-11-16 11:18:57 ammunta: making the latest skarnet.org packages, please don't update them lest it cause a conflict 2015-11-16 11:19:20 uggedal: same with s6-rc (someone please tell uggedal when he's here) 2015-11-16 11:31:34 git: 'send-email' is not a git command. See 'git --help'. 2015-11-16 11:31:44 what package should I install to get git-send-email? 2015-11-16 11:38:02 git-email 2015-11-16 11:42:46 thanks 2015-11-16 11:42:56 and here comes the whole perl truck 2015-11-16 11:48:21 I just sent my first set of patches to alpine-aports. Please review them and tell me if I did anything wrong. 2015-11-16 11:48:51 skarnet: to avoild the perl bloat, if you have a local SMTP client, you could pipe stdout of git format-patch to it 2015-11-16 11:49:37 It's a development machine, not a "production" machine, I don't mind the bloat, I just want to get the work done. But thanks. 2015-11-16 11:49:56 i use this as an alias http://tpaste.us/GVv0 2015-11-16 11:50:22 nice one. 2015-11-16 11:51:18 Alpine development machine should have as few unneeded packages on it as possible, so that abuild can trace dependancies and for ABUILD make dependancies work across build environments 2015-11-16 11:51:34 ah, I see now that I should have named the aports in the subject line. 2015-11-16 11:52:43 is there a way to make git-format-patch add those automatically, or should I always name the aport in the commit log? 2015-11-16 11:52:53 commit message* 2015-11-16 11:53:21 i don't know enough about git, but probably 2015-11-16 11:55:37 it would be easier to script some text and send it to git commit --message "your script output" 2015-11-16 11:57:16 I'll find a groove with time. I just popped my packaging cherry today, that's enough for the time being. :) 2015-11-16 12:31:24 ncopa: I fixed the dnssec-root package to follow the linked instructions more closely, could you take a look at it? %1021 and %1022 2015-11-16 12:33:25 it works as a trust-anchor-file for unbound now 2015-11-16 12:36:57 why we need the bind-tools-doc in addition to bind-doc? 2015-11-16 12:37:23 i think we dont need that 2015-11-16 12:37:31 bind-tools-doc ships documentation for the bind-tools package 2015-11-16 12:37:39 bind-doc ships documentation for the bind package 2015-11-16 12:38:16 yes, but we normally only have a single -doc 2015-11-16 12:38:51 ok, let me adjust it 2015-11-16 12:53:33 ncopa: like this? %1023 2015-11-16 12:54:24 oops fucked up 2015-11-16 12:56:09 correct the unbound fuckup in %1024 2015-11-16 12:56:12 *corrected 2015-11-16 13:14:18 you dont need the dnssec tools to run bind server? 2015-11-16 13:15:31 as far as I know you don't 2015-11-16 13:15:46 arch also moved dnssec tools to the bind-tools package 2015-11-16 13:16:44 ncopa: you merged the wrong path, now unbound is semi broken :/ 2015-11-16 13:16:48 *patch 2015-11-16 13:16:58 ugh 2015-11-16 13:16:59 sorry 2015-11-16 13:19:55 no problem, it's actually my faul 2015-11-16 13:21:03 pty.cc: In function 'gchar** __vte_pty_merge_environ(char**, gboolean)': 2015-11-16 13:21:04 pty.cc:404:75: error: missing sentinel in function call [-Werror=format=] 2015-11-16 13:21:04 g_ptr_array_add (array, g_strconcat (name, "=", value, NULL)); 2015-11-16 13:21:04 ^ 2015-11-16 13:21:11 any idea what that is about 2015-11-16 13:21:30 seems as it is complaining on g_strconcat 2015-11-16 13:21:34 but it does have NULL 2015-11-16 13:22:22 which package? 2015-11-16 13:22:32 vte3 2015-11-16 13:24:24 hm, looks wierd 2015-11-16 13:24:26 *weird 2015-11-16 13:26:02 I sent 4 patches to alpine-aports, and I received an automated email from patchwork saying 3 patches had been applied. Does it mean the 4th one failed? 2015-11-16 13:27:10 hey 2015-11-16 13:27:35 does any one know why I am receiving this message: "/dev to be a mounted devtmpfs" 2015-11-16 13:27:37 ? 2015-11-16 13:27:40 skarnet: 2/4 was not picked up by patchwork for some reason 2015-11-16 13:27:44 it was applied 2015-11-16 13:27:55 my keyboard and mouse do not work 2015-11-16 13:28:07 hi alacerda_ 2015-11-16 13:28:16 hey ncopa 2015-11-16 13:28:25 ncopa: thanks. What are the consequences of patchwork not picking it up? (I have no idea what patchwork is supposed to do.) 2015-11-16 13:28:40 patchwork is just a shiny frontend for the mailinglist 2015-11-16 13:28:51 patchwork is supposed to help us not forget patches 2015-11-16 13:29:08 patches not yet committed are listed there 2015-11-16 13:29:18 once committed they are removed from the list 2015-11-16 13:30:20 oh, so it's an automatic nagging tool? 2015-11-16 13:34:10 ncopa: here is a proper patch for unbound %1025 2015-11-16 13:38:47 wati a sec unbound-checkconf is stupid 2015-11-16 22:24:20 good evening 2015-11-17 08:50:22 ok 2015-11-17 08:50:25 good news 2015-11-17 08:50:35 i think the build log url for arm works now 2015-11-17 09:36:08 any chance if php 7.x is out before rc2 of v3.3 it can be included ? 2015-11-17 09:39:44 i dont think so 2015-11-17 09:54:25 ScrumpyJack: if you lost tpaste, apk add tpaste 2015-11-17 12:52:16 how is everything looking? 2015-11-17 12:52:33 build-wise 2015-11-17 18:17:54 ncopa: ping 2015-11-17 20:48:37 damn... all containers are broken... because s6 paths have been changed and because go moved to @community -.- 2015-11-17 21:07:47 mosez: sorry about that. But the s6 paths should *not* be hardcoded unless you're using slashpackage! 2015-11-17 21:37:28 skarnet: i'm using it within CMD of dockerfiles :) 2015-11-17 21:37:53 https://github.com/dockhippie/skydns/blob/master/Dockerfile#L29 as an example 2015-11-17 21:39:39 skarnet: but as you are here... s6 is awesome! :) 2015-11-17 21:49:47 hum... should is use s6-rc within my containers instead of plain s6? :D 2015-11-17 21:50:32 s/is/i/ 2015-11-17 22:41:13 mosez: ? that's not related to @community I think 2015-11-17 22:41:26 skarnet: uh what else are you supposed to do for shebangs then 2015-11-17 22:43:12 ammunta: og course it's related to @community. some time ago go have been part of @main, now it's moved to community. 2015-11-17 22:43:30 oh sorry I misread 2015-11-17 22:43:35 ammunta: you don't need s6 for shebangs. 2015-11-17 22:44:03 execlineb changed path 2015-11-17 22:45:00 mosez: that dockerfile shouldn't actually need the absolute path to s6-svscan anyway though? 2015-11-17 22:49:01 mosez: community move should only have affected containers building from edge without the community repo added to /etc/apk/repositories though 2015-11-17 22:49:05 which 2015-11-17 22:49:08 ... are the official images still broken 2015-11-17 22:51:23 ammunta: yes, you have to hardcode absolute paths in shebangs, and that's ugly. Shebangs breaking when you move stuff around is extremely common and extremely stupid. 2015-11-17 22:51:44 But short of implementing PATH resolution in the kernel, I don't see what we can do about that. 2015-11-17 22:52:31 mosez: thanks for the compliment. I'll take both the compliments and the criticism :) 2015-11-17 22:52:58 and the -S behaviour isn't available if using env to do PATH resolution, is it 2015-11-17 22:53:20 no, that's another limitation of the kernel shebang launcher: only one argument 2015-11-17 22:53:35 so if you use #!/usr/bin/env foobar 2015-11-17 22:53:39 you can't give options to foobar 2015-11-17 22:53:50 because your one allowed space has already been used. 2015-11-17 22:54:39 Also, using env is merely displacing the problem: /usr/bin/env? /bin/env? 2015-11-17 22:55:01 the common convention is /usr/bin/env, but it's only a convention, and will break on machines without /usr 2015-11-17 22:55:39 of course 2015-11-17 22:56:16 mosez: do you have to specify absolute paths in a Dockerfile? 2015-11-17 22:56:34 ... hm are there real systems with /usr that would actually break with a symlink /usr -> / though 2015-11-17 22:58:55 if you have no /usr at all, you can have a symlink /usr -> / but that's very, very ugly :P 2015-11-17 23:00:51 mosez: about s6-rc, it really depends on the complexity of what's inside your container. If you have several services with not-obvious dependencies and the starting order is important, you'll benefit from s6-rc. But most of the time, containers host 2 or 3 services and you can usually just yolo them in parallel. 2015-11-17 23:01:05 So I don't expect s6-rc to be useful in containers. 2015-11-17 23:04:13 the other thing is that containers can skip on a lot of the work of e.g. mounting disks and bringing up networking and keeping clocks up to date and stuff 2015-11-17 23:04:31 yup, so not many oneshots 2015-11-17 23:05:27 usually you'll have your container with your big-ass daemon to supervise and that's basically it, maybe one or two more services for health monitoring or logging or communication with another container 2015-11-17 23:05:49 but a simple supervision tree should be more than enough to manage that 2015-11-17 23:06:30 that said I have a few images with s6-rc in them, but they're also exactly the kind of mini-vm images that people go ~docker philosophy~ at 2015-11-17 23:07:00 what is the ~docker philosophy~ ? 2015-11-17 23:07:01 i never tried to use s6-svscan without a path. maybe i should have to try it 2015-11-17 23:07:50 mosez: as in CMD ["s6-svscan", path], not as in CMD ["/bin/s6-svcan"] 2015-11-17 23:07:59 mosez: check in the Docker documentation, maybe they'll say whether there's a PATH search for CMD or not at all 2015-11-17 23:08:05 although the latter ... does it in the current directory, I think 2015-11-17 23:08:12 and in most containers i just have logstash-forwarder, maybe cron and the service itself. so s6-rc should be too much 2015-11-17 23:08:20 docker does search through PATH, yes 2015-11-17 23:08:35 ... in the guest? 2015-11-17 23:08:41 yes 2015-11-17 23:08:57 nice to see Docker does things right :) 2015-11-17 23:09:01 it doesn't leak environment 2015-11-17 23:09:28 mosez: yes, if you just have 3 longruns don't bother with s6-rc 2015-11-17 23:09:36 "If you want to run your without a shell then you must express the command as a JSON array and give the full path to the executable" 2015-11-17 23:09:52 that's from the dockerfile docs. so i need an absolute path. 2015-11-17 23:09:59 they are empirically wrong 2015-11-17 23:10:18 i have updated my containers to /bin anyway 2015-11-17 23:10:24 10 bucks says they use execvp 2015-11-17 23:10:59 speaking of which 2015-11-17 23:11:33 but if you want control, you should have some /init as CMD, and your /init would set up everything according to your needs and exec into s6-svscan 2015-11-17 23:11:45 http://luaposix.github.io/luaposix/modules/posix.unistd.html#execp this is the worst case of misdocumentation wrt either not understanding what a shell is or not understanding what exec*p do 2015-11-17 23:11:53 that I have seen like 2015-11-17 23:11:54 ever 2015-11-17 23:12:15 it's the 2nd time I see you mentioning it, it really hurt you :P 2015-11-17 23:12:37 just building a test Dockerfile 2015-11-17 23:12:39 :&& I was so offended 2015-11-17 23:13:20 don't be offended by people failing to understand Unix, or you'll never stop being offended 2015-11-17 23:13:25 https://gist.github.com/muhmuhten/60673d6158c01325df23 this should do the trick 2015-11-17 23:13:50 most useful container ever 2015-11-17 23:14:24 ok, it works without an absolute path 2015-11-17 23:14:25 wait I made it more useful 2015-11-17 23:14:38 XD 2015-11-17 23:15:20 mosez: if the docker doc says you need an absolute path, then use one, because it's pure luck that it works without one :) 2015-11-17 23:15:32 :) 2015-11-17 23:15:59 but really, make a small /init script in your guest and use it as CMD entrypoint 2015-11-17 23:16:01 well pure luck and the fact that there's enough stuff that relies on it now that they'll never be able to break it 2015-11-17 23:16:23 but an /init script makes sense. just a bash script with exec s6-svscan /etc/s6 to keep everything at pid 1, right? :) 2015-11-17 23:16:30 or sh... or execline... 2015-11-17 23:16:38 exactly 2015-11-17 23:16:47 or codep 2015-11-17 23:16:53 maybe not execline, since then you'll have shebang problems :&& 2015-11-17 23:16:59 codep? 2015-11-17 23:17:00 not anymore 2015-11-17 23:17:09 it's in /bin now and it will remain there 2015-11-17 23:17:14 :) 2015-11-17 23:17:33 well neither would /bin/s6-svscan at this point :&& 2015-11-17 23:17:39 true 2015-11-17 23:18:10 i will keep it like it is for now... 2015-11-17 23:18:18 https://gist.github.com/progrium/0ac0248f70e2adce964f 2015-11-17 23:18:39 http://cl.ly/image/340T1H0T1y3F interesting... pretty strange counting for the docker hub :P 2015-11-17 23:19:04 is this Docker hour? :) 2015-11-17 23:19:26 andyshinn: ... that's broken because as soon as you exit everything is getting KILLed 2015-11-17 23:19:38 actually, that is the point 2015-11-17 23:19:39 they don't actually get the benefit of clean termination 2015-11-17 23:20:09 I mean, unless they can clean up in the time it takes you to send the signals to everything else 2015-11-17 23:20:17 so "depending on the scheduler's mood" basically 2015-11-17 23:21:09 also, bash :&& 2015-11-17 23:21:14 oh hmm, i mis-understood the original argument, this was so that containers won't linger when one process dies - IMHO, the concern of process failing is more for the scheduler, and the container should exit if a process has failed so the scheduler can make an educated decision 2015-11-17 23:22:26 if you want to terminate the container on failed startup of a service with s6 you can do this https://github.com/dockhippie/php/blob/master/apache/rootfs/etc/s6/apache/finish#L2 2015-11-17 23:22:30 :) 2015-11-17 23:23:01 not my idea to be honest :D 2015-11-17 23:23:40 there's also a not negligible class of programs (ones that open a lockfile under /tmp with O_EXCL) which will make a codep container unsafe to restart 2015-11-17 23:23:48 or anything that writes to /tmp for that matter 2015-11-17 23:24:01 I'm ... not sure why docker doesn't mount a real tmpfs 2015-11-17 23:24:06 they can do it for /dev ... 2015-11-17 23:24:28 you can do it in your /init if you need a tmpfs 2015-11-17 23:24:39 I do do it there 2015-11-17 23:25:19 also, I don't understand your locking point: locks are released when processes are killed 2015-11-17 23:25:24 it is ugly as hell because with CAP_SYS_ADMIN dropped you can't actually mount one 2015-11-17 23:25:51 skarnet: they do the locking technique which goes "try to create O_EXCL, explode on failure, remove on exit" 2015-11-17 23:26:14 that's not a safe locking technique by any means 2015-11-17 23:26:53 and it will break horribly if the file is in persistent storage, but as you mentioned... don't do that 2015-11-17 23:27:24 also, what? docker removes some capabilities from the guest root? 2015-11-17 23:27:28 there are real programs that expect /tmp to be volatile 2015-11-17 23:27:29 yeah 2015-11-17 23:28:07 then yes, if it doesn't give you a tmpfs and doesn't give you the rights to mount your own, it's a bug 2015-11-17 23:28:49 those programs are broken. /tmp must be considered as volatile (i.e. you can't expect it to be persistent storage), but there's no obligation for it to be. 2015-11-17 23:29:05 i.e. no guarantees either way. 2015-11-17 23:32:22 I swear docker used to mount them as tmpfs before so they probably changed it for some kind of reason but I don't think I managed to dig it up last time I looked 2015-11-17 23:35:28 don't repeat it, but if you absolutely need a tmpfs, you can use a hidden, unused subdirectory of /dev. A devtmpfs is a tmpfs... 2015-11-17 23:36:07 they actually just mount a populated tmpfs at /dev 2015-11-17 23:37:18 and that's ugly, they should mount it on, say, /run, then have /dev be a symlink to /run/dev (populated) and /tmp be a symlink to /run/tmp. 2015-11-17 23:37:35 so they could provide a /dev *and* a /tmp with a unique tmpfs. 2015-11-17 23:37:56 18:25 < ammunta> it is ugly as hell because with CAP_SYS_ADMIN dropped you can't actually mount one 2015-11-17 23:38:01 what did you think I meant :&& 2015-11-17 23:38:32 I thought you couldn't get a tmpfs at all. 2015-11-17 23:38:44 ... oh 2015-11-17 23:38:46 I guess 2015-11-17 23:39:51 anyway might not have necessarily been able to do the symlink thing without like, smashing /run or /tmp dirs that already existed 2015-11-17 23:39:59 (which ... shouldn't have contained anything, but still) 2015-11-17 23:40:11 ("shouldn't") 2015-11-17 23:40:56 I said /run, they could call it /docker-tmpfs-this-directory-is-reserved-for-docker 2015-11-17 23:41:33 the point is to have a place to mount your tmpfs, and to make anything using a tmpfs a subdirectory of that place 2015-11-17 23:41:33 that's not the issue (and they already have those anyway) 2015-11-17 23:42:03 although frankly I don't think there's a real issue 2015-11-17 23:42:34 if you don't have stupid programs creating stuff in /tmp and expecting it to be very volatile, no, there's no issue 2015-11-18 00:17:42 https://github.com/docker/dockercraft 2015-11-18 09:05:50 https://github.com/ingydotnet/json-bash , seems nice 2015-11-18 09:20:20 ncopa: we dont have config.gz support in our kernels? 2015-11-18 09:23:06 we have 2015-11-18 09:23:30 modprobe configs 2015-11-18 09:23:40 oh its a module.. 2015-11-18 09:32:13 is password auth disabled by default on latest sshd? 2015-11-18 09:32:24 clandmeter, seems yes, upstream changed the default 2015-11-18 09:32:34 yes 2015-11-18 09:32:44 for root user 2015-11-18 09:32:48 along with some old keys types 2015-11-18 09:34:39 ah ok thx 2015-11-18 09:34:47 i had to enable boot password and root 2015-11-18 09:35:03 s/boot/both/ 2015-11-18 09:37:05 what's the purpose of the gdm user in our default /etc/passwd? Is it supposed to be used by gnome display manager? We don't ship gnome display manager, do we? 2015-11-18 09:38:04 relict of the past, I guess 2015-11-18 09:39:03 there is no sassc package available on alpine? :( 2015-11-18 09:39:50 i need it... i dont want to do https://github.com/drone/drone/blob/master/contrib/setup-sassc.sh :( 2015-11-18 09:43:35 you could package it according to that script 2015-11-18 09:46:33 setting up a build environment :) 2015-11-18 10:19:56 What's the purpose of the default sync user? The passwd entry implies that it is somehow used /bin/sync but /bin/sync doesn't even have a setuid bit set…I don't see why /bin/sync would need the sync user… 2015-11-18 10:54:00 hey :) 2015-11-18 13:12:53 https://gist.github.com/tboerger/415a415f99dd494a51a8 what's wrong with this package definition? why does it say autoreconf and ./configure are not found? oO 2015-11-18 13:15:46 ncopa: could you please compile mbedtls with MBEDTLS_THREADING_PTHREAD 2015-11-18 13:15:55 that way hiawatha would work again :) 2015-11-18 13:16:33 i dont get it why it can't find the packages -.- 2015-11-18 13:17:41 any packager like clandmeter, ncopa or maybe fabled got a hint for me? :) 2015-11-18 13:19:24 mosez: does src/$pkgname-$pkgver exist? 2015-11-18 13:20:18 # ls src/libsass-3.3.2 2015-11-18 13:20:19 appveyor.yml configure.ac contrib COPYING docs extconf.rb GNUmakefile.am include INSTALL LICENSE m4 Makefile Makefile.conf Readme.md res script SECURITY.md src test version.sh win 2015-11-18 13:20:34 so yes 2015-11-18 13:24:06 is abuild -r unpack build wrong? got to read docs again 2015-11-18 13:24:24 i think it is actually missing autoreconf 2015-11-18 13:24:43 libtool 2015-11-18 13:24:47 might be you need libtool 2015-11-18 13:24:52 oh 2015-11-18 13:24:56 i thinki know 2015-11-18 13:24:58 oh god... just called it wrong -.- 2015-11-18 13:25:24 ok, got the next error... ./configure: line 5201: syntax error near unexpected token `dlopen' 2015-11-18 13:25:27 ./configure: line 5201: `LT_INIT(dlopen)' 2015-11-18 13:25:46 it seems that i had tried to build this myself previously 2015-11-18 13:25:54 i had a half done apkbuild 2015-11-18 13:26:10 interesting :) 2015-11-18 13:26:35 and now it builds 2015-11-18 13:26:36 :) 2015-11-18 13:27:14 and mine is not configuring because of LT_INIT(dlopen) -.- 2015-11-18 13:29:36 and directly after getting permissions to write the build file it seems to work :D 2015-11-18 13:30:02 ncopa: so you already got build files for libsass and sassc? :) 2015-11-18 13:32:29 i started with it some time ago 2015-11-18 13:32:39 at least libsass work now 2015-11-18 13:32:59 mosez: ok to push libsass? 2015-11-18 13:34:55 ncopa: if your version works, sure 2015-11-18 13:35:24 my local try seems to work now as well... more or less -.- 2015-11-18 13:35:52 >>> ERROR: libsass-doc*: Missing /build/pkg/libsass-doc 2015-11-18 13:35:52 >>> ERROR: libsass*: prepare_subpackages failed 2015-11-18 13:35:54 :) 2015-11-18 13:36:34 it doesnt provide a doc package :D 2015-11-18 13:37:04 pushed libsass 2015-11-18 13:37:38 feel free to continue with sassc 2015-11-18 13:37:47 i'll continue fight open-vm-tools 2015-11-18 13:39:14 cool, sassc will be the next :) 2015-11-18 13:39:53 ncopa: how long does it take to be able to fetch that package through apk from testing index? :) 2015-11-18 13:40:09 build-edge-x86_64: files from v3.2.0-3704-ga4b4cf0 uploaded to testing 2015-11-18 13:40:09 build-edge-x86: files from v3.2.0-3704-ga4b4cf0 uploaded to testing 2015-11-18 13:40:13 should done now 2015-11-18 13:40:20 thats from #alpine-commits 2015-11-18 13:40:36 its available from http://rsync.alpinelinux.org/.... 2015-11-18 13:40:38 mosez: if you have nl.alpinelinux.org as your repo, instantly after builder finishes 2015-11-18 13:40:47 ah, alpine commits channel 2015-11-18 13:40:49 nl.a.o = rsync.a.o 2015-11-18 13:41:08 we will likely move master mirror to other location soonish 2015-11-18 13:42:56 barthalion: have you had issues with archlinux and ncurses-6? specifically alsamixer? 2015-11-18 13:43:01 it segfaults 2015-11-18 13:44:12 yeah, my first package contribution to alpine :D 2015-11-18 13:44:23 have to do it for couchdb as well :P 2015-11-18 13:44:26 and ceph xD 2015-11-18 13:45:40 start of couchbase: http://tpaste.us/Aa0b 2015-11-18 13:45:53 does not build iirc 2015-11-18 13:46:11 i have a bunch of ceph stuff too 2015-11-18 13:46:20 a bunch of patches 2015-11-18 13:46:30 i also started with ceph but never finished it 2015-11-18 13:50:17 alsamixer segfaults to me also 2015-11-18 13:58:52 yeah... i'm not sure if i really want to use ceph for my cluster... or if i will use gluster or even syncthing :) 2015-11-18 14:07:44 libsass-3.3.2-r0: 2015-11-18 14:07:44 masked in: @testing 2015-11-18 14:07:53 how should that be added and accepted as a dep? 2015-11-18 14:15:20 https://gist.github.com/tboerger/6254a8eafac98f394770 i don't know how to call abuild to accept dependencies from testing 2015-11-18 14:21:06 ncopa: help :) 2015-11-18 14:23:13 mosez: you can either add testing to your /etc/apk/repositores 2015-11-18 14:23:27 or you can move your working dir to testing 2015-11-18 14:24:31 i will give it a try 2015-11-18 14:24:50 testing is in repositories but with @testing 2015-11-18 14:30:56 ncopa: none I'm aware of 2015-11-18 14:32:08 ncopa: works fine with alsa 1.1.0 2015-11-18 14:32:34 ls 2015-11-18 14:32:40 ok 2015-11-18 14:32:48 $ pwd 2015-11-18 14:32:49 /build/testing/sassc 2015-11-18 14:32:58 still not working... 2015-11-18 14:33:01 clandmeter: http://build.alpinelinux.org/buildlogs/build-edge-armhf/main/fftw/fftw-3.3.4-r1.log 2015-11-18 14:33:16 and if i untag the testing repo it complains about the testing packages in my world file 2015-11-18 14:33:40 ncopa: ah its a leftover 2015-11-18 14:34:35 mosez: i suppose as a temp workaround you could move it back to your staging dir 2015-11-18 14:34:46 and ln -s ../testing/libsass . 2015-11-18 14:35:00 and then abuild -R in the sassc dir 2015-11-18 14:57:34 i just installed libsass and libsass-dev within my build env :D 2015-11-18 15:03:43 ncopa: https://gist.github.com/tboerger/6659c85dd59782e92125 anything that needs to be improved? otherwise it's working. installed it within build env and i was able to compile scss 2015-11-18 15:04:19 depends is not needed 2015-11-18 15:04:24 abuild will autodetect it 2015-11-18 15:04:41 otherwise good 2015-11-18 15:05:00 so should i email that? or can you add that directly as well? :) 2015-11-18 15:05:01 mosez: can you git format-patch -1 | sprunge 2015-11-18 15:05:13 and paste url here 2015-11-18 15:05:19 and i'll apply it 2015-11-18 15:09:51 ncopa: http://sprunge.us/DYaJ 2015-11-18 15:11:15 great 2015-11-18 15:11:16 thanks 2015-11-18 15:18:51 rnalrd: community/tg has a checksum error 2015-11-18 15:18:57 awesome 2015-11-18 15:19:04 rnalrd: one of the tarballs are generated from git master 2015-11-18 15:19:23 so if developer pushed something, it will break tomorrow 2015-11-18 15:19:28 and now it broke for v3.3 2015-11-18 15:19:36 tlparserver=master 2015-11-18 15:19:42 tl-parser.zip::https://github.com/vysheng/tl-parser/archive/$tlparserver.zip 2015-11-18 15:20:04 i think I forgot to make snapshot before pushing 2015-11-18 15:20:11 sry 2015-11-18 15:20:16 I'm on it 2015-11-18 15:21:02 i think thats the last bit for rc1 2015-11-18 15:21:54 http://cuberite.org/ awesome, i found one of my next packages xD 2015-11-18 15:28:19 mosez: could you please test the sassc package built from build server 2015-11-18 15:28:23 and test that it is okish 2015-11-18 15:28:32 and let me know 2015-11-18 15:28:42 i'd like to move this to main or community 2015-11-18 15:34:34 i will test it now with my drone build 2015-11-18 15:43:04 ncopa, pushed the snapshot version 2015-11-18 15:43:11 thanks! 2015-11-18 15:43:22 i'll wait til tomorrow if arm catches up 2015-11-18 15:43:29 need to review though, but at least it keeps you going 2015-11-18 15:43:32 and we may get rc1 tomorrow 2015-11-18 15:43:53 abuild create tarball based on package name 2015-11-18 15:44:41 so tl-parser snapshot is effectively named tg-0_gitXXXX 2015-11-18 15:44:58 which I manually renamed before upload 2015-11-18 17:26:54 nmeum: sync is probably like the shutdown &c. users 2015-11-18 17:27:05 i.e. relics of the past basically 2015-11-18 17:32:21 anyone got a test Alpine install they can try su without root password (remove root password and try su from normal user)? 2015-11-18 17:33:27 inside container, i can su to root from normal user when root has no password, wondering if this is normal behavior on regular install as well 2015-11-18 19:41:25 andyshinn: that is normal behavior 2015-11-18 20:15:58 andyshinn: that's normal for no password if with setuid su 2015-11-19 00:04:39 good evening 2015-11-19 07:32:04 andyshinn: there was a recent discussion about blank password and root 2015-11-19 07:32:08 on busybox ml 2015-11-19 07:32:10 http://lists.busybox.net/pipermail/busybox/2015-October/083392.html 2015-11-19 07:33:08 and there is a patchset for busybox to fix it properly too: http://lists.busybox.net/pipermail/busybox/2015-November/083621.html 2015-11-19 07:33:16 i inted to backport that patchset 2015-11-19 07:33:33 i have just waited for feedback on it 2015-11-19 07:34:22 andyshinn: maybe you could send an email to busybox@busybox.net (the mailing list) with your comment on this feature? 2015-11-19 07:34:30 I did here: http://lists.busybox.net/pipermail/busybox/2015-November/083629.html 2015-11-19 07:50:23 ncopa: isn't a solution to that just "don't install suid su" 2015-11-19 07:54:33 ammunta: it is 2015-11-19 07:54:49 we splitted the suid su from busybox package not too long ago 2015-11-19 12:55:54 quiet here lately 2015-11-19 13:44:00 hey :) 2015-11-19 13:52:36 hey leo-unglaub, did you get opensmtpd working the way you wanted? 2015-11-19 13:52:57 ScrumpyJack: yes, it works 2015-11-19 13:54:32 nice 2015-11-19 14:07:19 http://linux.die.net/man/3/strerror_r which version gets used on alpine? is it possible that it uses the XSI-compliant version of strerror_r? 2015-11-19 14:08:14 i'm trying to build https://github.com/cuberite/cuberite and it fails with https://gist.github.com/tboerger/41c3f691d283bae314be which is related to https://github.com/cuberite/cuberite/blob/master/src/OSSupport/Errors.cpp#L27 so it's inside of the condition for #if !defined(__APPLE__) && defined( _GNU_SOURCE) && !defined(ANDROID_NDK) 2015-11-19 14:08:52 and i have no idea which check i need to add to get into the XSI-compliant variant :(\ 2015-11-19 14:27:00 compiling this kind of tools on musl sucks -.- 2015-11-19 14:31:36 a more generic code should use if !defined(__FRUITS__) 2015-11-19 14:33:20 tst: haha, really funny... not... 2015-11-19 14:33:46 :) 2015-11-19 14:37:02 mosez: musl does not implement the broken (GNU) strerror_r 2015-11-19 14:38:38 in general, you can #if defined(__GLIBC__) && defined(_GNU_SOURCE) #endif 2015-11-19 14:39:21 #if !defined(__APPLE__) && defined( _GNU_SOURCE) && !defined(ANDROID_NDK) 2015-11-19 14:39:26 thats fundamentally broken 2015-11-19 14:39:32 it shoudl be: 2015-11-19 14:39:48 #if defined(__GLIBC__) && defined( _GNU_SOURCE) 2015-11-19 14:40:09 2015-11-19 14:40:13 #else 2015-11-19 14:40:26 2015-11-19 14:40:27 #endif 2015-11-19 14:44:38 i have removed that code for testing 2015-11-19 14:44:48 and afterwards i ran into the next error. 2015-11-19 14:44:59 now i did "CC=clang CXX=clang++ cmake . -DCROSSCOMPILE=1 -DCMAKE_BUILD_TYPE=RELEASE" and it seems to compile... we will see 2015-11-19 14:45:15 various warnings but no error so far 2015-11-19 14:50:08 awww, now it failed with OSSupport/libOSSupport.a(StackTrace.cpp.o): In function `PrintStackTrace()': 2015-11-19 14:50:11 StackTrace.cpp:(.text+0x24): undefined reference to `backtrace' 2015-11-19 14:50:11 StackTrace.cpp:(.text+0x33): undefined reference to `backtrace_symbols_fd' 2015-11-19 14:50:12 collect2: error: ld returned 1 exit status 2015-11-19 14:50:12 :( 2015-11-19 14:51:29 https://github.com/cuberite/cuberite/blob/master/src/OSSupport/StackTrace.cpp#L41-L45 -.- 2015-11-19 14:54:27 looks like i need to stop thinking about a fast minecraft server on alpine :( 2015-11-19 15:13:50 backtrace is not implemented 2015-11-19 15:13:55 i will try to build it with the same commands on debian now 2015-11-19 15:13:59 we will see if it works 2015-11-19 15:14:01 its only needed to print a stack trace if it seftaults 2015-11-19 15:14:30 ncopa: but that's the part where the compile fails 2015-11-19 15:15:10 you should probably have a configure check for backtrace 2015-11-19 15:15:19 and do: #if HAVE_BACKTRACE 2015-11-19 15:15:41 add a #else 2015-11-19 15:15:50 printf("backtrace not supported"); 2015-11-19 15:15:51 #endif 2015-11-19 15:16:26 we should have it generate a coredump instead 2015-11-19 15:16:34 and use gdb to get the backtrace 2015-11-19 15:16:55 i can't write c++ :D 2015-11-19 15:18:24 i just want to have a docker container for a fast minecraft server :D 2015-11-19 15:24:34 ncopa: I know this isn't a priority, but what format/name/content do you want to release 3.3 as? alpine-mini replaces alpine-standard? 2015-11-19 15:25:08 https://gist.github.com/tboerger/b4ec8a9234bdf13c37f3 haha, debian wheezy fails pretty hard on another part -.- 2015-11-19 15:26:34 ScrumpyJack: i think i will not change that for v3.3 2015-11-19 15:26:35 hum 2015-11-19 15:26:49 did we come up with a good name for the 'fuller' or 'offline' iso? 2015-11-19 15:27:12 a iso with 'everything' is out of question at this point 2015-11-19 15:27:27 i will not make 4G iso image 2015-11-19 15:28:19 +1 2015-11-19 15:30:08 suggestions for the 300MB iso: alpine-offline, alpine-live 2015-11-19 15:30:21 alternatively we could rename the alpine-mini to alpine-bootonly 2015-11-19 15:33:22 no, alpine-mini should become alpine 2015-11-19 15:33:43 300MB iso could be named "-full" 2015-11-19 15:33:58 but its not "full" in the sense that its not all packages 2015-11-19 15:34:16 alpine-morepkgs 2015-11-19 15:34:57 alpine-additional-pkgs 2015-11-19 15:35:13 I find full more clear than this 2015-11-19 15:35:32 I mean, it indicates to me that the non-full version is minimal, and the other one not 2015-11-19 15:35:52 we could use home / pro / enterprise naming :) 2015-11-19 15:35:58 lol 2015-11-19 15:36:10 alpine-enterprise.iso 2015-11-19 15:36:19 alpine-professional 2015-11-19 15:36:21 lol 2015-11-19 15:36:26 ultimate 2015-11-19 15:36:27 actually, that was not a bad idea 2015-11-19 15:36:44 alpine-ultimate 2015-11-19 15:37:10 core / desktop / server 2015-11-19 15:37:14 except that people will think its better and download it 2015-11-19 15:37:18 hm 2015-11-19 15:37:31 alpine-core / alpine-desktop / alpine-server 2015-11-19 15:37:39 aint bad eithe 2015-11-19 15:37:48 we have alpine-xen already 2015-11-19 15:38:05 I really think it should be just alpine for current minimal 2015-11-19 15:38:15 so 2015-11-19 15:38:24 alpine, alpine-, and alpine-xen 2015-11-19 15:38:28 makes sense to me 2015-11-19 15:39:07 alpine-peak / alpine-foothills 2015-11-19 15:39:14 ? 2015-11-19 15:39:27 or simply, alpine-more 2015-11-19 15:39:32 peak is the core, foothills includes more 2015-11-19 15:39:43 like the mountain :) 2015-11-19 15:40:00 i tend to agree with barthalion 2015-11-19 15:40:07 we want the alpine.iso to be the small one 2015-11-19 15:40:24 the question is, what do we call the bigger one? 2015-11-19 15:40:28 alpine-enterprise? 2015-11-19 15:40:36 same stuff, just more expensive 2015-11-19 15:40:50 alpine-extra ? 2015-11-19 15:41:02 sounds like you would want both isos 2015-11-19 15:41:27 alpine-expanded or alpine-extended? 2015-11-19 15:41:38 thats not bad 2015-11-19 15:41:41 alpine-extended 2015-11-19 15:42:06 alpine-enhanced 2015-11-19 15:43:02 "enhanced" sounds like you get more/better features, which is not the case 2015-11-19 15:43:13 ACTION likes alpine / alpine-xen / alpine-extended as of right now 2015-11-19 15:43:13 you'll only get more bloat 2015-11-19 15:43:22 yes 2015-11-19 15:43:29 -fat 8) 2015-11-19 15:43:32 alpine / alpine-extended / alpine-xen 2015-11-19 15:43:53 barthalion: "i can find the alpine-fat, but where do i find the alpine-ext4 and alpine-btrfs?" 2015-11-19 15:44:00 alpine-plus? 2015-11-19 15:44:06 alpine++ 2015-11-19 15:44:12 alpine# 2015-11-19 15:44:19 alpine: 2015-11-19 15:44:31 format alpine: 2015-11-19 15:44:37 i like "alpine:" as it will make fun for windows users :) 2015-11-19 15:45:06 alpine-plus, alpine-extended sounds good to me 2015-11-19 15:45:16 slightly better with alpine-extended 2015-11-19 15:45:29 alpine-superfluous 2015-11-19 15:45:55 alpine-bloated-edition 2015-11-19 15:45:58 alpine-extraordinary! 2015-11-19 15:46:19 we generally want people to pick the "alpine" variant 2015-11-19 15:46:58 ACTION has found a thesaurus, obviously 2015-11-19 15:47:11 alpine-inessential 2015-11-19 15:47:20 :) 2015-11-19 15:49:02 so far i liked the "alpine-extended" or "alpine-plus" best 2015-11-19 15:49:31 ACTION still votes for "alpine-extended" 2015-11-19 15:49:46 seems to indicate bigger but not necessarily better 2015-11-19 15:51:07 exactly 2015-11-19 16:02:37 alpine-holyshit 2015-11-19 16:04:34 alpine-dontdownloadthis 2015-11-19 16:21:09 ACTION is thinking of something witty to say about this but can't top alpine-bloated 2015-11-19 16:22:19 can we even justify an alpine-something? what's wrong with just one alpine edition? 2015-11-19 16:23:06 so there is a second iso with extra apks, how does that benefit the user? 2015-11-19 16:23:22 just hide the big version behind an extra screen except the link is subtly broken 2015-11-19 16:25:13 aline-$function works, like alpine-xen, or alpine-desktop, but alpine-large because it has extra (potentially badly chosen) packages doesn't fill a function imho 2015-11-19 16:26:18 i think we should scrap it, and keep apline.iso (which is mini) 2015-11-19 16:28:07 i need the bigger iso 2015-11-19 16:28:15 the function it serves is for run-from-RAM servers where you want some basic packages 2015-11-19 16:28:28 yup 2015-11-19 16:28:29 but, if those packages are updated, you end up needing a cache anyway 2015-11-19 16:28:41 run tmpfs installs needs them 2015-11-19 16:28:47 so, the iso is only valuable if it is up-to-date 2015-11-19 16:29:10 it also serves as fallback if the cache gets corrupt or deleted 2015-11-19 16:29:17 then you get the older version of the package 2015-11-19 16:29:22 true 2015-11-19 16:29:24 instead of boot failure 2015-11-19 16:29:41 there is definitely some value, but limited 2015-11-19 16:30:09 its a safety mechanism for tmpfs installs 2015-11-19 16:30:14 remote tmpfs installs 2015-11-19 16:30:29 goes back to the original goal for alpine - tmpfs servers 2015-11-19 16:30:30 you have the basic stuff included on the iso so you can reboot and get back connectivity 2015-11-19 16:30:33 https://github.com/cuberite/cuberite/issues/2658#issuecomment-158104988 xD 2015-11-19 16:31:33 mosez: also ask in #musl 2015-11-19 16:31:37 they are good at this kind of stuff 2015-11-19 16:41:49 i'm lost, doesn't current alpine mini boot the same way? 2015-11-19 16:43:11 you can use mini to build a run from ram 2015-11-19 16:43:29 you can pull the packages you need after that 2015-11-19 16:44:17 if you are using the bigger iso as a repos, then the contents of that repos is stuff you like to use :) others may not be able to rebuild from the iso 2015-11-19 16:47:50 if i want to use something to build/recover a host, i create a USB stick with the packages I need. That is the beauty of setup-alpine 2015-11-19 16:48:55 i think for the sake of bandwidth and clarity, alpine++ should go :) 2015-11-19 18:23:06 alpine-big 2015-11-19 18:23:27 alpine-plussized 2015-11-19 18:23:41 alpine-quarterpounderwithcheese 2015-11-19 18:49:27 alpine-basic? alpine-ram/fromram/ramfs? hm… 2015-11-19 18:52:42 alpine-kitchensink 2015-11-19 18:55:25 alpine-ram, alpine-ewe, alpine-lamb 2015-11-19 18:58:03 alpine qui tollis peccata mundi 2015-11-19 20:21:56 alpine-180% 2015-11-19 20:23:07 alpine-xxl 2015-11-19 20:23:24 (not to be confused with alpine-xxx) 2015-11-19 20:23:41 it's 100% alpine with 80% more bloat 2015-11-19 20:24:04 or name it after the user case, alpine-tmpfs 2015-11-19 20:24:10 too close to 180° 2015-11-19 20:24:26 alpine 175% then :p 2015-11-19 20:24:44 alpine-over9000 2015-11-19 20:24:59 alpine-bloated 2015-11-19 20:25:16 what does the scouter say about its bloat level 2015-11-19 20:25:32 vroooom 2015-11-19 20:25:39 fill it to a just over 700M and call it alpine-dvd 2015-11-19 20:28:04 alpine-ludicrous-sized 2015-11-19 20:28:34 if it's really bloated call it alpine-glibc :D 2015-11-19 20:29:55 how can i enable https://github.com/cuberite/cuberite/blob/master/SetFlags.cmake#L254-L257 for alpine AND freebsd? 2015-11-19 20:33:28 mosez: OR? 2015-11-19 20:33:30 https://cmake.org/cmake/help/v3.0/command/if.html 2015-11-19 20:33:41 without enableing that and adding libexecinfo@testing the build results in https://github.com/cuberite/cuberite/issues/2658#issuecomment-158184192 2015-11-19 20:34:07 ammunta: but CMAKE_SYSTEM_NAME will be Linux... and not Alpine ;) 2015-11-19 20:34:37 ... I mean, it's p rare that you need it for alpine specifically 2015-11-19 20:34:40 why do you need it 2015-11-19 20:34:44 and test for that explicitly 2015-11-19 20:36:29 as i said... build the project fails with https://github.com/cuberite/cuberite/issues/2658#issuecomment-158184192 if i don't add it just for alpine. 2015-11-19 20:57:08 so the problem is that you need libexecinfo if you're not building against glibc? 2015-11-19 20:59:13 which needs to be linked in? 2015-11-19 21:14:16 mosez: http://stackoverflow.com/questions/11976654/how-to-properly-check-for-a-function-using-cmake 2015-11-19 21:21:19 already talking with a cuberite dev... will do it different 2015-11-19 21:42:40 mosez: just curious, how fast is it? Compared to MC and how's the RAM usage? 2015-11-19 21:44:00 This looks very like minetest (c++ and lua) except that it still uses JAVA client? 2015-11-19 21:48:53 https://github.com/cuberite/cuberite/pull/2661 :D 2015-11-19 21:49:20 xming: i don't know... but i want it instead of the damn java server :D 2015-11-19 21:56:16 so you still want java client? :D 2015-11-19 21:58:50 hehe http://sdtimes.com/sd-times-blog-docker-turns-to-minecraft-for-server-ops/ 2015-11-19 22:02:36 xming: it's still required, even if i don't want to 2015-11-19 22:07:10 mosez: https://travis-ci.org/cuberite/cuberite/jobs/92145172 might want to look into what you did that failed tests 2015-11-19 22:07:50 ammunta: the test didn't fail oO something with the spawnpoint 2015-11-19 22:08:49 mosez: https://github.com/thinkofdeath/steven 2015-11-19 22:09:07 the whole build process and tests are successful 2015-11-19 22:09:14 died after [21:54:58] Did not find an acceptable spawnpoint. Generated a random spawnpoint position at {0.00, 61.00, 0.00} 2015-11-19 22:09:52 xming: seriously??????? 2015-11-19 22:09:55 awesome stuff :D 2015-11-19 22:09:58 I think that could be normal 2015-11-19 22:10:17 What doesn't work: 99% of Minecraft's features 2015-11-19 22:10:38 it's random seed and some places are not good for spawning hence the player died and it chooses a new point 2015-11-19 22:11:00 mosez: it's rendering :D the hardest part is done :D 2015-11-19 22:12:09 but a pretty cool idea 2015-11-19 22:12:15 i love golang 8) 2015-11-19 22:12:33 mosez: you can use #elif defined(__GLIBC__) 2015-11-19 22:12:46 https://github.com/tboerger/cuberite/commit/a16a9eab4fa7da5bd741e1b4dbc5315768dec6ad#diff-4efb855b470c8ac1d8b9dba0c98d11c9R40 2015-11-19 22:13:07 that avoids the arrow 2015-11-19 22:13:19 ammunta: i had that first, but disliked mixing of ifdef and if defined :D 2015-11-19 22:14:52 but anyway, it's building on alpine :D 2015-11-19 22:20:30 nice job 2015-11-19 22:31:57 i will build a docker container tomorrow :) 2015-11-20 02:15:51 #291 2015-11-20 08:57:36 since everyone seems to enjoying 'name our spin' game 2015-11-20 08:57:36 here is more, 2015-11-20 08:57:36 alpine-blueberry (bb, busybox, bareboot) 2015-11-20 08:57:36 alpine-pumpkin (plus, big) 2015-11-20 08:57:44 :-)) 2015-11-20 10:00:27 and alpine-pumkinboat (all, large, dvd) 2015-11-20 10:00:55 why not alpine-jackthepumpkinking while you're at it 2015-11-20 10:01:26 alpine-oogieboogie 2015-11-20 10:01:33 alpine-nightmare (that's a good one) 2015-11-20 10:14:17 i think i'll just do 'alpine' (current alpine-mini) and 'alpine-extended' (current alpine) 2015-11-20 10:14:40 you're too serious. :) 2015-11-20 10:15:33 :) 2015-11-20 10:16:32 morning. and happy friday! 2015-11-20 10:17:21 ncopa: you don't like my alpine-180% ? 2015-11-20 10:24:17 i got confused when you said "the function it serves is for run-from-RAM servers where you want some basic packages" 2015-11-20 10:39:23 alpine-extended works I think 2015-11-20 10:39:35 now, what packages go in alpine-extended? :) 2015-11-20 10:41:08 there are 40 squid-lang packages 2015-11-20 10:41:56 two versions of shorewall 2015-11-20 10:43:43 3 dns servers 2015-11-20 10:49:55 oogieboogie 2015-11-20 10:49:57 <3 2015-11-20 10:50:05 I think gnu tar should be added 2015-11-20 10:51:48 +1 for GNU tar, one of the 2 GNU packages that build nice and clean 2015-11-20 10:54:27 and busybox tar breaks with super long file names (last time I looked) 2015-11-20 10:54:55 screen but no tmux? *shock* :) 2015-11-20 10:59:57 asterisk, freeswitch, kamailio, there is a disproportionate amount of voice software in there 2015-11-20 11:01:27 extended means adding feature on existing, still ambiguous 2015-11-20 11:01:51 btw, I upgraded by 1st AL bare metal to 8GB ram :)) 2015-11-20 11:02:02 good points 2015-11-20 11:02:09 we dont need 3 dns servers 2015-11-20 11:02:24 i think we could pick one, nsd + unbound 2015-11-20 11:02:25 ram costs are comming down , but still not in resonable range 2015-11-20 11:02:35 we could kick out shorewall 2015-11-20 11:02:40 add gnu tar 2015-11-20 11:02:55 why not dnscache/tinydns 2015-11-20 11:02:56 add tmux, possibly kick out gnu screeen 2015-11-20 11:03:04 its slow 2015-11-20 11:03:10 lolwut 2015-11-20 11:03:20 have used dnscache/tinydns for many years 2015-11-20 11:03:24 screen is more-a-less admin defacto 2015-11-20 11:03:35 tmux is extended 2015-11-20 11:03:40 and believe me tinydns have been hard to "kill" 2015-11-20 11:03:44 tinydns just works 2015-11-20 11:03:47 can you kick all the 40 squid-lang? 2015-11-20 11:03:50 yes 2015-11-20 11:04:01 -lang + 2015-11-20 11:04:05 let's make a list 2015-11-20 11:04:08 :) 2015-11-20 11:04:08 yes 2015-11-20 11:04:13 but i am leavning now 2015-11-20 11:04:19 collab anyone? 2015-11-20 11:04:25 its friday and i am taking an early weekend 2015-11-20 11:04:32 have a nice week-end 2015-11-20 11:04:33 ncopa: good for you! 2015-11-20 11:04:56 ncopa: just a little help, what type of ram you use in your bld1 box ? 2015-11-20 11:05:19 your current 32 processor HP machine 2015-11-20 11:05:24 (dnscache is slow - due to do caching correct. has caused issues a couple of times) 2015-11-20 11:05:40 can someone spin up a collab-style instance of some sort? 2015-11-20 11:05:46 caching correct. ? 2015-11-20 11:05:47 vkris: i dont know 2015-11-20 11:06:00 ok, thanks 2015-11-20 11:06:08 skarnet: yes, it does not cache the additional records on request 2015-11-20 11:06:13 what's wrong with dnsmasq? three-in-one 2015-11-20 11:06:15 it only caches what was asked for 2015-11-20 11:06:37 does dnsmasq support proper zone files? 2015-11-20 11:06:48 so when doing resolving from root server, you need to wait til it has done all the questions 2015-11-20 11:07:20 servers normally add on 'in case you will want the authorative server in next request, then this is the answer' 2015-11-20 11:07:38 using that like unbound does will speed up things 2015-11-20 11:07:58 specially if you are on a high-latency link (satelite) 2015-11-20 11:08:18 ncopa: you still here? naff off :) 2015-11-20 11:08:20 there was also issue that cache was not big enough 2015-11-20 11:08:38 so whith 3000 workstations things got really slow 2015-11-20 11:08:49 which is why i moved to unbound 2015-11-20 11:08:52 ok 2015-11-20 11:08:58 have a nice weekend everyone! 2015-11-20 11:09:05 cache wasn't big enough... you... realize that's configurable? 2015-11-20 11:09:46 will libvirt be on the iso? 2015-11-20 11:09:49 ah well, it works for me anyway. 2015-11-20 11:11:29 clandmeter: there is no virtualization software in there at the moment is there? 2015-11-20 11:13:43 i have no idea 2015-11-20 11:13:53 i know it needs dnsmasq 2015-11-20 14:51:21 <_mjones> vkris, whenever people say things like memory and storage costs are coming down I get tense. 2015-11-20 14:51:48 <_mjones> Most of the armhf platforms have no more than 1gb ram, and the most popular one has 512mb. 2015-11-20 14:52:29 <_mjones> Moving ahead usually doesn't actually require much increase in resource use if your architecture is sound. 2015-11-20 15:13:37 ram, costs has to (it been long time), if it does not its market is being manipulated 2015-11-20 15:13:40 note that china economic has a slow-down, not all reasons, but very-likely they can make use of this as pushing more volume at less price 2015-11-20 15:14:19 rest of equation is lot of FUDs 2015-11-20 15:14:25 ;) 2015-11-20 15:21:17 <_mjones> I bricked my 128mb armel box with a new kernel; network stack comes up but no userland binds to it. Will be a pain to fix without console. 2015-11-20 15:21:44 <_mjones> Anyway, I'd like to be able to continue to run that 128mb armel box for a while longer, regardless of how many RAMs I have everywhere else. 2015-11-20 15:22:33 <_mjones> I don't run less than 16gb in any desktop, but I still like low-resource systems. 2015-11-20 15:30:18 keeping tabs on performance and optimization with low resources is healthy sign for technician/devs 2015-11-20 15:30:21 it would help in keeping bloats at bay 2015-11-20 15:37:59 healthy sign in* technician/devs ;) 2015-11-20 15:43:49 If anyone wants to have their say about what goes into alpine-extended, now is your chance: 2015-11-20 15:44:19 http://meije.alpinelinux.uk/di01hicr5x 2015-11-20 15:44:51 to edit use http://meije.alpinelinux.uk/di01hicr5x/edit and ask someone here for the credentials 2015-11-20 15:46:22 is there a list of what's in alpine (not extended) anywhere? 2015-11-20 15:51:28 yes, let me get that 2015-11-20 15:52:25 http://git.alpinelinux.org/cgit/alpine-iso/tree/alpine-mini.packages 2015-11-20 15:53:10 <_mjones> So, I'll point out that Debian has popularity-contest which measures how many people install and use each package. 2015-11-20 15:53:24 alpine-base and network-extras are metapackages 2015-11-20 15:53:34 <_mjones> The users and audience aren't the same, but frankly I'd use their data as a powerful indicator overall. 2015-11-20 15:56:26 We could probably grab the repo http logs too see folks have been pulling. Would be nice to see the mirrors' logs as well. Perhaps a git repo of pulled logs? 2015-11-20 16:13:56 openssh in alpine-mini, not dropbear? 2015-11-20 16:18:43 <_mjones> I like dropbear, but frankly I strongly suspect that the average user will be more comforted by openssh. 2015-11-20 16:19:05 I thought Alpine didn't have average users. 2015-11-20 16:19:10 <_mjones> There are already sufficient difference from typical GNU Linux. 2015-11-20 16:19:12 Only talented ones. 2015-11-20 16:19:37 The average user would be comforted by glibc, too. And systemd. 2015-11-20 16:19:47 OK, maybe that's TOO average. 2015-11-20 16:19:52 But you get the idea. 2015-11-20 16:19:52 <_mjones> skarnet: ha! You shouldn't listen to this cool new indie band; they're just too edgy for most people. 2015-11-20 16:20:35 My point exactly. I'm not sure alpine is the place to be mainstream. 2015-11-20 16:21:08 <_mjones> Pulling repo logs is a fabulous idea, too. And reminds me of a small itch I want to scratch in that netbsd lib Alpine uses. 2015-11-20 16:22:10 I just turned my repo mirror access logs on, they were off! 2015-11-20 16:26:03 <_mjones> ScrumpyJack: in those logs, useragent shows up as libfetch, yes? 2015-11-20 16:26:10 yip 2015-11-20 16:26:28 <_mjones> Improving that was one of my projects on my list 2015-11-20 16:27:00 <_mjones> And last I looked, that lib could use a refresh but there was no obvious download source for the upgraded version, outside of source control. 2015-11-20 16:27:13 would be nice to have the useragent change with the apk command issued to distinguish between apk add and apk upgrade 2015-11-20 16:27:21 <_mjones> oh! 2015-11-20 16:27:53 <_mjones> I was going to do the apk and alpine version, but that distinction is a great idea. 2015-11-20 16:28:04 <_mjones> really good idea. 2015-11-20 16:28:22 ACTION is blushing 2015-11-20 16:28:52 <_mjones> Anybody else have ideas for the useragent? Anyone think there's a need for an anonymising switch or anything? 2015-11-20 16:30:30 you're going to do something now aren't you :) 2015-11-20 16:31:02 but we could take the weekend to chew on it 2015-11-20 16:31:56 right, gtg - have a good weekend 2015-11-20 16:32:21 <_mjones> Have a good weekend. 2015-11-20 17:51:30 I will openssh at least until dropbear's ed25519 support exists 2015-11-20 17:52:33 not sure whether that's happened yet though 2015-11-20 18:02:34 all of the squid langpacks come out to what, 500k? I don't think they merit dropping 2015-11-20 18:05:03 I don't think bash is a dep for lxc proper but most of the template scripts are bash scripts? 2015-11-20 18:26:23 <_mjones> I dislike bash dependencies. It was always a silly early Linux decision to use bash instead of ash. Ash has been around since '89. 2015-11-20 18:32:06 has the gnu stuff always built with not bash 2015-11-20 18:38:08 ammunta: for squid-lang it's not about size of course 2015-11-20 18:47:53 <_mjones> libfetch isn't on the package list even though it's used by apk 2015-11-20 20:01:19 what's the purpose of "update_config_sub || return 1" ? If update_config_sub fails, compilation got interrupted. 2015-11-20 20:02:24 wait, what /is/ acf 2015-11-20 20:03:09 Alpine Configuration Framework 2015-11-20 20:03:39 http://wiki.alpinelinux.org/wiki/Alpine_Configuration_Framework_Design 2015-11-20 20:05:25 umh 2015-11-20 20:05:27 wait 2015-11-20 20:05:31 ok, got it 2015-11-20 20:06:04 config.sub should be patched in order to support alpine-arm-musl 2015-11-20 20:06:45 so if is not supported, i should put on arch x86 and x86_64 rather than all 2015-11-20 20:48:50 _mjones: was thinking about libfetch and anonymisation. an option in setup-alpine that asks "can we collect stats?" 2015-11-20 20:49:22 and if yes, create the file /etc/apk/stats 2015-11-20 20:50:06 <_mjones> stats of package installs? 2015-11-20 20:50:22 <_mjones> Also, ScrumpyJack, can you put a title on the spreadsheet to clarify the purpose? 2015-11-20 20:50:45 <_mjones> I think 'package review for inclusion in new base install image and expanded install image' is accurate, right? 2015-11-20 20:50:53 yes, and if /etc/apk/stats exists, useragents can change to reflect apk add or apk update etc 2015-11-20 20:51:45 <_mjones> I think the functionality is intrusive enough to warrant another package, even if it's a metapackage. 2015-11-20 20:51:59 _mjones: you can put a title on the speadsheet :) (I don't have a browser here) 2015-11-20 20:52:19 right, back to party :) 2015-11-20 20:52:53 <_mjones> Was just thinking that the useragent could reflect the base command issues 'Alpine 3.4.1 armhf; apk 2.4.4 (apk update)" or something. 2015-11-20 20:57:48 probably no good reason to distinguish the command issued 2015-11-20 20:57:54 cf. -U vs update 2015-11-20 20:58:06 that seems weirdly invasive 2015-11-20 20:59:16 although 2015-11-20 20:59:26 don't the server logs show *what* got requested anyway 2015-11-20 20:59:36 that's probably a lot of information by itself 2015-11-20 21:09:30 <_mjones> ScrumpyJack brainstormed that and I thought it was interesting from a data point of view, so you could distinguish initial installs for purposes of popularity monitoring. 2015-11-20 21:10:11 <_mjones> Just dividing everything into add || update would be interesting. 2015-11-20 21:17:11 <_mjones> er, I've been meaning to say 'add vs. upgrade', not update. 2015-11-20 21:18:05 <_mjones> Yes, I think this spreadsheet badly needs an intro that clarifies from which media the proposed drops are from. Maybe another column. 2015-11-20 21:23:29 <_mjones> And also, I should be looking at the graphviz of dependencies before I think further about removals. 2015-11-21 01:23:37 wait is there no lua sha2 things 2015-11-21 22:18:26 http://www.uclibc-ng.org/ not like it makes any difference now 2015-11-22 01:28:18 hm, is anybody actually using pulseaudio on alpine? The dependencies are broken since the udev-dev package doesn't exist 2015-11-22 01:48:38 so...spim doesn't seem to build on the x86 builder, how can retrieve the buildlog again? 2015-11-22 02:02:08 oops got it 2015-11-22 12:39:55 Hi, just saw https://grsecurity.net/announce.php and wondering whether it affects the alpine's future work, too? 2015-11-22 12:43:17 It categorically says that the future public release will be unfit for production use. "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." 2015-11-22 12:43:40 Any say about the direction grsecurity+pax is taking since September 15, 2015?! 2015-11-22 17:11:44 hey there, I am looking for a relieable way to contact someone from the alpine staff 2015-11-22 17:12:30 'cause I do have a question about some of the firefox patches 2015-11-22 17:14:05 actually about the one from the git at aports/community/firefox/fix-netwerk.patch 2015-11-22 17:16:01 I thought it's a typo 2015-11-22 17:16:06 but it's not 2015-11-22 17:16:14 stefson: anyway, what is your question? 2015-11-22 17:16:48 if there is any bug report upstream, as there is no description in the header of the patch 2015-11-22 17:17:02 I don't think so 2015-11-22 17:17:11 according to git log, we took it from musl-pkgsrc-patches 2015-11-22 17:20:36 hmm, I'm using the checked out git repo on my harddisk normaly. so please tell me, at which point do I find the git log? 2015-11-22 17:21:37 2df82e3d3d783c51aaba69c49ddb00a9f25ab138 2015-11-22 17:23:00 the git log, not the specific one at 2df...etc 2015-11-22 17:23:29 'git log' command? 2015-11-22 17:25:10 ah, ok 2015-11-22 17:25:34 it's there since 2 years 2015-11-22 17:25:39 yeah 2015-11-22 17:27:05 so, how may I find out if upstream knows about this problem? 2015-11-22 17:27:41 that patch is in use since 2013, so I just bet it doesn't 2015-11-22 17:28:28 hm 2015-11-22 17:28:58 are you using alpine for your desktop? 2015-11-22 17:29:12 no 2015-11-22 17:29:22 but there are people who do here 2015-11-22 17:29:22 k 2015-11-22 17:29:31 I prefer Arch on my laptop 2015-11-22 17:30:53 it is not on i686 by chance? 2015-11-22 17:31:10 i686 and x86_64 2015-11-22 17:31:50 because I am experiencing strange breakages on i686 2015-11-22 17:32:07 firefox has missing texts and buttons in the gui 2015-11-22 17:32:16 same goes for xfce4-menues 2015-11-22 17:32:32 on Alpine? you probably need gnome-icon-something 2015-11-22 17:32:56 no, on gentoo 2015-11-22 17:33:42 the patches from your git have been really helpfull actually, but in this case I am stuck 2015-11-22 17:34:06 anything interesting in console when you start firefox? 2015-11-22 17:35:06 there is a silent segfault, if you would like to call it like that, which appears in dmesg 2015-11-22 17:35:44 I'll upload the log if you would like to see it 2015-11-22 17:36:42 sure, I can take a look, but I can't promise anything 2015-11-22 17:36:52 thats ok 2015-11-22 17:36:56 https://bpaste.net/show/74598b448300 2015-11-22 17:37:02 it is very minimal information anyway 2015-11-22 17:38:34 yeah, quite 2015-11-22 17:38:41 :) 2015-11-22 17:39:18 works like a charm on amd64 2015-11-22 17:40:37 and for you on i686 with alpine, there are no such kind of glitches, or however you would like to call them? 2015-11-22 17:40:44 is firefox usable for you? 2015-11-22 17:41:14 yeah, looks fine 2015-11-22 17:41:31 there is also a bug report, but not with any usefull information http://bugs.alpinelinux.org/issues/4248 2015-11-22 17:42:58 I use alpine on my notebook 2015-11-22 17:43:12 i686 as well? 2015-11-22 17:43:15 no 2015-11-22 17:43:28 x86_64 2015-11-22 17:45:27 so you don't have any issues like missing textes in guis and so on? 2015-11-22 17:46:43 by the way @ barthalion , which desktop env. are you using? gnome, xfce, kde? 2015-11-22 17:49:03 out of those three, alpine only has xfce afaik 2015-11-22 17:52:56 ah, okay, didn't knew that 2015-11-22 19:20:00 I use awesome as window manager 2015-11-22 19:20:33 And usually don't have issues like missing fonts 2015-11-23 08:03:08 morning climbers 2015-11-23 08:08:28 morning all 2015-11-23 08:08:48 do you know where is sys_siglist funcion implemented in musl? 2015-11-23 08:08:56 *function 2015-11-23 08:09:11 is it in musl 2015-11-23 08:10:15 ammunta, ok. 2015-11-23 08:10:22 I don't see it at all 2015-11-23 08:10:29 me too 2015-11-23 08:13:05 but see also src/string/strsignal.c 2015-11-23 08:14:33 ammunta, path? 2015-11-23 08:14:44 couldn't find it 2015-11-23 08:15:30 fcolista: from top of the repo? 2015-11-23 08:16:05 i couldn't find src/string/strsignal.c 2015-11-23 08:16:15 what's the path starting from the / ? 2015-11-23 08:16:38 find that name in the entire disk does not return anything. 2015-11-23 08:16:47 Probably i'm missing some headers 2015-11-23 08:17:16 i've musl-dev, linux-headers and bsd-compat-headers 2015-11-23 08:18:03 ? 2015-11-23 08:18:03 oh 2015-11-23 08:18:09 I'm looking at the musl source 2015-11-23 08:18:11 not just the headers 2015-11-23 08:18:33 ah 2015-11-23 08:18:41 http://git.musl-libc.org/cgit/musl/tree/src/string/strsignal.c it's this file 2015-11-23 08:18:45 no i'm looking for the header files 2015-11-23 08:18:53 i'm working on build openvas 2015-11-23 08:19:04 /home/fcolista/aports/testing/openvas-manager/src/openvas-manager-6.0.6/src/manage_sql.c:12712:10: error: 'sys_siglist' undeclared (first use in this function) 2015-11-23 08:19:10 that's the issue 2015-11-23 08:19:38 ah-a 2015-11-23 08:19:38 https://github.com/alpinelinux/aports/blob/master/testing/kbuild/sys_siglist.patch 2015-11-23 08:19:50 looks we need a workaround 2015-11-23 08:20:07 just a different name 2015-11-23 08:20:22 and the way is called 2015-11-23 08:20:23 ok 2015-11-23 08:20:37 hm 2015-11-23 08:20:51 is string reps of signals something that never got standardized 2015-11-23 09:48:46 Is there someone working with xpra 0.16.0beta version for alpinelinux? I tried to compile, but without success. The available version reports two errors: 1. Unable to find rencode module and 2. --sync-xvfb parameter not found (both are available in Ubuntu 14.04 fyi). Any inputs? 2015-11-23 10:04:27 where did you get xpra 0.16.0beta? 2015-11-23 10:23:37 I suppose trunk is 0.16beta, they have deb for it https://www.xpra.org/beta/jessie/main/binary-amd64/ 2015-11-23 10:24:16 sorry, i mean where did you get it in the alpine repos 2015-11-23 10:24:21 I don't see it in edge 2015-11-23 10:24:32 on 0.15 2015-11-23 10:29:10 I think he's trying to build 0.16beta 2015-11-23 10:32:10 bloody rockchip, arm is not so difficult to understand, but rockchip made it all so confusing, made merunning in circles 2015-11-23 10:32:32 they have their own partitioning scheme, that was confusing me 2015-11-23 10:37:01 ScrumyJ1ck: I found at xpra.org site. http://xpra.org/beta/ 2015-11-23 10:38:19 ScrumpyJack: I move lensfun to community. ok? 2015-11-23 10:38:20 xming: thanks for pointing it out to ScrumpyJack before I did. :-) 2015-11-23 10:40:17 I tried to follow https://www.xpra.org/trac/ticket/820, but neither I could install py-gtk nor py-gobject as it reports unmet dependenices in edge. 2015-11-23 10:58:12 ncopa: yes, that's fine. doesn't rawtherapee dep on lensfun too? 2015-11-23 10:59:27 i think i moved rawtherapee 2015-11-23 10:59:32 or disabled it 2015-11-23 10:59:46 yeah 2015-11-23 10:59:52 moved it to unmaintained 2015-11-23 10:59:54 it does not build 2015-11-23 11:01:26 ah sigc++ 2015-11-23 11:01:32 needs c++11 2015-11-23 11:40:31 ncopa: since grsecurity has gone closed and the pubic release will be "not production ready" according to grsecurity.net site, how will it affect the future of grsecurity in alpinelinux? 2015-11-23 11:57:22 From what I read here: https://www.xpra.org/trac/ticket/820 xpra of any version is not going to work in alpine linux due to toolchain issue. I am getting "could not determine ld path for x86_64, please file an xpra bug" 2015-11-23 11:57:24 :( 2015-11-23 12:03:23 zenny1: grsecurity is no longer open source, and yes, it affects alpine linux 2015-11-23 12:03:27 it means more work for us 2015-11-23 12:07:39 ncopa: however they stated that they will release their "not-production-ready" patches for arch and gentoo reportedly here: https://grsecurity.net/announce.php 2015-11-23 12:09:07 i knkow 2015-11-23 12:11:43 we based our 4.1.y kernel on that 2015-11-23 12:35:49 ncopa: did you mean 4.1.y kernel is based on the "not production ready" patch? Isn't that too insecure? Just curious! 2015-11-23 12:36:44 yes, its based on the "not produciton ready" patch 2015-11-23 12:36:58 the option is to drop grsecurity 2015-11-23 12:37:12 so either we use that or we dont use grsecurity at all 2015-11-23 12:37:41 i still believe that the grsecurity option is the most "secure" option of those 2 2015-11-23 13:11:46 how can i tell CMake to check if an .h exists? 2015-11-23 13:14:08 The equivalent of AC_CHECK_HEADERS([execinfo.h]) 2015-11-23 13:14:40 ok figured 2015-11-23 13:39:24 hey :) 2015-11-23 13:49:07 ncopa: did you just get rawtherapee to build? what changed as it builds fine for 3.2 2015-11-23 13:56:46 sigc++ 2.5.2 2015-11-23 13:56:53 requires c++11 2015-11-23 14:51:08 is that "yes, it builds now" or "go fix the APKBUILD"? :) 2015-11-23 15:01:45 it is, "yes it builds now" 2015-11-23 15:01:51 but it does not build on arm 2015-11-23 15:02:00 question on linux block devices 2015-11-23 15:02:44 i suppose the only correct way to compare 2 block deives to see if they are the same is to compare major/minor? 2015-11-23 15:03:14 it is re #4863 2015-11-23 15:03:33 ncopa-desktop:~/Sync/Projects/mkinitfs$ ls -l /dev/mapper/vg0-test 2015-11-23 15:03:33 brw------- 1 root root 253, 5 Nov 23 15:50 /dev/mapper/vg0-test 2015-11-23 15:03:33 ncopa-desktop:~/Sync/Projects/mkinitfs$ ls -l /dev/dm-5 2015-11-23 15:03:33 brw-rw---- 1 root disk 253, 5 Nov 23 15:54 /dev/dm-5 2015-11-23 15:03:51 i think those 2 are the same device 2015-11-23 15:05:02 that is a great question ncopa 2015-11-23 15:05:12 can#t you blkid them? 2015-11-23 15:05:26 i suppose i can 2015-11-23 15:05:47 but i wonder if that is the correct thing to do 2015-11-23 15:05:52 i dont think it is 2015-11-23 15:05:57 what does blkid read if the devices are i. e. filled with 0? 2015-11-23 15:05:59 hmm, good question. i am not sure 2015-11-23 15:06:02 ncopa-desktop:~/Sync/Projects/mkinitfs$ ls -l /dev/vg0/test 2015-11-23 15:06:02 lrwxrwxrwx 1 root root 20 Nov 23 15:46 /dev/vg0/test -> /dev/mapper/vg0-test 2015-11-23 15:06:21 so if cryptroot=/dev/vg0/test 2015-11-23 15:06:46 lvm will create /dev/vg0/test symlink pointing to /dev/mapper/vg0-test 2015-11-23 15:06:52 which is a block device 2015-11-23 15:07:37 the coldplug devtmpfs will create /dev/dm-5 2015-11-23 15:07:46 which i think is the same thing 2015-11-23 15:08:03 I think that the major and minor numbers identify a device explicit 2015-11-23 15:08:09 i think so too 2015-11-23 15:08:29 As the kernel doesn't recognize it by the name but by these numbers 2015-11-23 15:08:43 thats how i understand it too 2015-11-23 15:08:44 yes, that's how it works 2015-11-23 15:08:58 the name doesn't matter, the major and minor do 2015-11-23 15:09:06 so then the only correct way to wait for a blockdevice to turn up is to compare major:minor 2015-11-23 15:09:11 in contrast blkid reads some infos from the filesystem 2015-11-23 15:09:40 interresting :) learned something new again :) 2015-11-23 15:10:00 the correct way to wait for a block device to turn up is to watch the netlink 2015-11-23 15:10:13 skarnet: thats what i am doing 2015-11-23 15:10:27 but user specify: wait for /dev/vg0/something 2015-11-23 15:10:36 which is a symlink 2015-11-23 15:10:36 hmm 2015-11-23 15:10:42 then you need a conversion table 2015-11-23 15:10:54 i think i need stat(3) 2015-11-23 15:11:05 and compare major/minor numbers 2015-11-23 15:11:30 what do you do if the name expected by the user turns up but the numbers mismatch? 2015-11-23 15:12:38 i think the logic is: compare /dev/ with 2015-11-23 15:12:48 if equal, then return found 2015-11-23 15:13:04 if not, then stat() 2015-11-23 15:13:09 and compare major/minor 2015-11-23 15:13:18 ugh 2015-11-23 15:13:42 see #4863 2015-11-23 15:13:52 thats what i am trying to solve 2015-11-23 15:13:53 if /dev/console is a link to /dev/ppp and I want to look for /dev/console, I don't want the program to tell me "it's ok, you have /dev/console" 2015-11-23 15:13:58 I want it to fail loudly 2015-11-23 15:14:17 stat() doesn't give major/minor#, does it? 2015-11-23 15:14:28 what gives major/minor? 2015-11-23 15:14:35 major() and minor() 2015-11-23 15:14:46 that's defined... somewhere in the linux headers 2015-11-23 15:14:49 yes 2015-11-23 15:15:03 man makedev(3) 2015-11-23 15:16:01 at least for glibc 2015-11-23 15:17:27 how do i get the dev_t? 2015-11-23 15:17:45 seems to be an uint64_t 2015-11-23 15:18:14 i think stat and rdev 2015-11-23 15:18:26 st_rdev 2015-11-23 15:20:43 dev and ino are in stat 2015-11-23 15:21:11 [XSI][Option Start] 2015-11-23 15:21:11 dev_t st_rdev Device ID (if file is character or block special). 2015-11-23 15:21:35 st_dev doesn't work for specials? ugh. 2015-11-23 15:22:04 it tells what mounted device file is on 2015-11-23 15:22:13 makes sense to me 2015-11-23 15:22:38 oh, rdev is the device representing the special. OK, got it. 2015-11-23 15:23:21 and you call major() and minor() on the st_rdev, right? 2015-11-23 15:23:27 yes 2015-11-23 15:23:32 but i wonder if i have to do that 2015-11-23 15:23:35 i suppose i do 2015-11-23 15:23:58 unless you can find an official document that says no other information is encoded in the st_rdev 2015-11-23 15:24:06 but I doubt you will 2015-11-23 15:24:29 i think using major() and minor() is the correct way to do it 2015-11-23 15:24:51 If you want to be robust against a change in the internal st_rdev semantics, yes, it is 2015-11-23 15:27:54 makedev(major,minor) 2015-11-23 15:27:58 is the opposite 2015-11-23 15:28:07 gives a dev_t 2015-11-23 15:28:29 the question is whether anything else can appear in a dev_t 2015-11-23 15:29:01 if a dev_t (as used in st_rdev) is only made of major and minor and nothing else, then you can just compare the st_rdev 2015-11-23 15:29:20 otherwise, extract the major and minor and compare them 2015-11-23 15:29:38 that's what I would do, because it costs peanuts 2015-11-23 15:32:25 is atoi(major_str_from_kernel_uevent) safe? 2015-11-23 15:35:35 if you trust the kernel, yes 2015-11-23 15:35:48 thats the question 2015-11-23 15:35:51 can i trust the kernel 2015-11-23 15:36:09 if you don't, there's not much point in running a system :) 2015-11-23 15:36:19 exactly 2015-11-23 16:35:05 https://pkgs.alpinelinux.org/contents?filename=%25xvfb%25&path=&pkgname=&repo=all&arch=x86_64 2015-11-23 16:35:18 doenst look like any package has the wrapper script 2015-11-23 16:39:00 oops, wrong channel... 2015-11-23 17:48:03 fabled: BTW: many packages do expect JAVA_HOME to be set when using java 2015-11-23 17:48:05 maven for instance 2015-11-23 17:48:23 I think it would seriously be a good idea to ship a /etc/profile.d file in OpenJDK{7,8} 2015-11-23 17:48:45 and when doing so you might as well adjust PATH to include _all_ tools in your $PATH instead of using symlinks 2015-11-23 17:48:48 (just a suggestion) 2015-11-23 18:15:45 btw... is there an api for https://pkgs.alpinelinux.org/? 2015-11-23 18:28:07 nope 2015-11-23 18:28:17 we should really host the source for that on git.alpinelnux.org 2015-11-23 18:28:22 always forget the github url for that project 2015-11-23 18:28:38 https://github.com/clandmeter/aports-turbo 2015-11-23 18:29:07 clandmeter: is there any reason why it isn't hosted on git.alpinelinux.org? 2015-11-23 18:29:21 it doesn't even have a license ;_; 2015-11-23 18:32:22 you agree to give up your kidney if you use this code 2015-11-23 18:32:47 also your firstborn 2015-11-23 18:35:19 nah, it really sucks if software doesn't have a license 2015-11-23 18:39:06 I'm sure it's something like WTFPL but you poke clandmeter about it anyway 2015-11-23 19:16:13 it's probably clandmeter never got around to thinking about it 2015-11-24 06:05:17 nmeum, patches are welcome for the java stuff 2015-11-24 07:05:40 morning 2015-11-24 07:05:54 fabled: i have a weird issue with nlplug-findfs and cryptsetup 2015-11-24 07:06:23 basically, the dm-X device uevent never shows up 2015-11-24 07:06:33 but the /dev/mapper/foo is created 2015-11-24 07:07:04 increasing timeout does not help 2015-11-24 07:18:54 ncopa, strange 2015-11-24 07:19:09 very strange 2015-11-24 07:19:10 maybe some kmod alias or softdep is missing? 2015-11-24 07:19:18 no 2015-11-24 07:19:22 the device shows up 2015-11-24 07:19:29 udevadm monitor gets the event 2015-11-24 07:19:53 KERNEL[1121185.235013] add /devices/virtual/bdi/253:4 (bdi) 2015-11-24 07:19:54 KERNEL[1121185.235044] change /devices/virtual/block/dm-5 (block) 2015-11-24 07:19:54 KERNEL[1121185.235065] add /devices/virtual/block/dm-4 (block) 2015-11-24 07:19:54 KERNEL[1121185.235883] change /devices/virtual/block/dm-4 (block) 2015-11-24 07:20:24 the dm-5 is the lvm device that is encrypted 2015-11-24 07:20:36 the dm-4 is the cryptdm 2015-11-24 07:21:27 if i revert the async fork patch, then it works again 2015-11-24 07:22:42 adding more debug, but it looks like it is doing everything correct 2015-11-24 07:32:42 nlplug-findfs returns error also with lvm on boot. Since lvm binary is not yet installed, returns error with /bin/lvm (not found) 2015-11-24 07:33:27 fcolista: that is if root is not on lvm but you have lvm devices for other 2015-11-24 07:33:35 its a known issue 2015-11-24 07:33:41 and i have ideas how to fix it 2015-11-24 07:33:42 right 2015-11-24 07:33:55 trying to fix this other thing first 2015-11-24 07:34:03 if you have it on fstab, then is not mounted 2015-11-24 07:34:35 nlplug-findfs does not really use fstab 2015-11-24 07:34:44 localmount should do it 2015-11-24 07:34:50 but you need start the lvm service too 2015-11-24 07:35:03 yeah, localmount fails because lvm is not yet started 2015-11-24 07:35:45 fabled: i think i am on something now 2015-11-24 07:35:56 the event *is* arriving and read 2015-11-24 07:36:05 must be the parsing of the message that goes wrong 2015-11-24 07:56:45 do we have a chance to have gcc-mingw32 | gcc-mingw64 on alpine ? 2015-11-24 07:57:10 patches are welcome... 2015-11-24 07:57:29 will probably not happen til after v3.3 2015-11-24 07:59:54 nmeum: i switched it to github because it has bugs/pr's 2015-11-24 08:00:41 ok 2015-11-24 08:00:50 i'm working on openvas porting 2015-11-24 08:01:00 the only missing package so far is openvas-smb 2015-11-24 08:01:17 that needs gcc-minigw[32|64] 2015-11-24 08:02:30 minigw? :D 2015-11-24 08:04:11 s/minigw/mingw 2015-11-24 08:04:42 yeah took me 2 sec to realize that 2015-11-24 08:04:57 xming, just a typo ;) 2015-11-24 08:05:50 I actaully like that typo :D 2015-11-24 08:06:24 mingw looks too much like my first name 2015-11-24 08:06:56 lol 2015-11-24 08:07:05 wmingw 2015-11-24 08:07:10 *xmingw 2015-11-24 08:07:53 I am still struggling with my arm board, but I think I am getting to understand this weird rockchip stuff, now I am waiting for a ttl cable so I have serial debug 2015-11-24 08:08:08 ok 2015-11-24 08:08:09 fcolista: mingw are actually the first 5 lettres of my first name 2015-11-24 08:08:17 i think this is an udev race or something 2015-11-24 08:08:27 udev/blkid race 2015-11-24 08:17:12 morning 2015-11-24 08:45:34 Hi people 2015-11-24 08:46:16 hi 2015-11-24 08:46:17 what's the most lightweight graphical environment you'd recommend, as in, a lightweight WM? (ideally that would work with Alpine, ofc) 2015-11-24 08:46:31 what's the policy naming for python3 packages? 2015-11-24 08:46:48 py3-$name 2015-11-24 08:46:49 fcolista: we dont have any policy for it yet 2015-11-24 08:46:58 umh 2015-11-24 08:47:16 i was hoping to do py-* = python3 and those few who needs python2 would do py2-* 2015-11-24 08:47:27 py-setuptools does not support python3 so far. 2015-11-24 08:47:36 Package that we have i mean 2015-11-24 08:47:46 nobody cared enough for python3 before the v3.3 feature freeze 2015-11-24 08:47:58 but we have tons of packages depending to py-setuptools 2015-11-24 08:48:00 so now we have to wait til after v3.3 release 2015-11-24 08:48:08 wait til after v3.3 2015-11-24 08:48:13 so having py-setuptools with python3, would break. 2015-11-24 08:48:16 yes, i guess so 2015-11-24 08:48:38 skarnet: i use xfce, but i dont consider it lightweight 2015-11-24 08:48:49 skarnet: I have used fluxbox in the past, which is pretty light 2015-11-24 08:49:07 then there is things like evilwm, which is too simplistic imo 2015-11-24 08:49:29 skarnet: i suppose it depends on how "weird" you want go 2015-11-24 08:49:53 fabled: i know whats going on with cryptsetup 2015-11-24 08:50:13 kernel triggers uevent corectly 2015-11-24 08:50:26 but the /dev/dm-4 does not exist 2015-11-24 08:50:35 nor does /dev/mapper/testcrypt 2015-11-24 08:51:00 no device node exists at the time the uevent is triggered 2015-11-24 08:51:05 so blkid fails 2015-11-24 08:51:13 ncopa: as long as it allows me to move windows around, it's good enough for me :) 2015-11-24 08:52:12 so the parallelisation of uevents is not good? 2015-11-24 08:52:38 i think it is 2015-11-24 08:52:44 i mean it is needed 2015-11-24 08:53:08 but i dont know if it is cryptsetup itself that rm and mknod 2015-11-24 08:53:10 or udev 2015-11-24 08:53:13 or something else 2015-11-24 08:53:30 the device node appears shortly after 2015-11-24 08:53:43 i dont know hos devtmpfs works 2015-11-24 08:53:57 if it makes sure that device node exists before it triggers the event 2015-11-24 08:54:20 or if event may be triggered before device node exists 2015-11-24 08:54:57 there would be no way to use it at all if the event could be triggered before the device node is created 2015-11-24 08:55:04 so you can just assume the device exists 2015-11-24 08:55:13 thats what i thought too 2015-11-24 08:55:25 but, since you have the major/minor with the event 2015-11-24 08:55:29 the thing is 2015-11-24 08:55:34 then you could mknod it 2015-11-24 08:55:39 devtmpfs will only create a minimal node 2015-11-24 08:55:46 but i thought that devtmpfs would do it for me 2015-11-24 08:55:47 a basic hardcoded name 2015-11-24 08:56:01 all the symlinks and other bells and whistles are done by udevd 2015-11-24 08:56:35 i suppose it might be udev or udisks or something that is deleting/moving things around 2015-11-24 08:56:39 yes 2015-11-24 08:56:47 or cryptsetup itself 2015-11-24 08:57:01 devtmpfs only does the skeleton things 2015-11-24 08:57:15 if you have anything fancy, it's done in userspace 2015-11-24 09:04:06 i wonder if it is udev or if it is cryptsetup itself that does it 2015-11-24 09:04:49 if its udev, then it will probably work 2015-11-24 09:04:59 but i want test it before i push... 2015-11-24 09:11:21 morning all 2015-11-24 09:18:36 are we in a package freeze at the moment for 3.3? or could someone look at 913 as a candidate for edge/main. been using wbar for a while without problem, and it would be the nice addition for desktop environments 2015-11-24 09:22:06 ok ,this cryptsetup business scares me 2015-11-24 09:22:26 fabled: I send you patches regarding the java issue you rejected them back then 2015-11-24 09:22:41 i have no idea what is moving around the /dev/dm-* node 2015-11-24 09:22:51 i *think* it is udev and not cryptsetup 2015-11-24 09:22:59 in which case it should work 2015-11-24 09:23:04 but i cannot prove it 2015-11-24 09:23:15 in which case things may break for people if i push 2015-11-24 09:23:22 nmeum, iirc, i asked if it was really required. you now proved it is. i think the patch should just go to java-common and let it determine what JAVA_HOME to set 2015-11-24 09:24:04 alright 2015-11-24 09:25:17 sounds like you're a bit busy with other stuff :) 2015-11-24 09:36:16 fabled: but why not deprecate java-common and change PATH instead if you need to install a profile file none the less? 2015-11-24 09:51:31 jomat: i think i have a fix for #4863 2015-11-24 09:51:40 but i have problems testing it 2015-11-24 09:51:58 my tests in my real hw fails 2015-11-24 09:52:04 but in qemu it appears to work okis 2015-11-24 09:52:21 okish 2015-11-24 09:52:59 jomat: do you think you could help me test it? 2015-11-24 09:59:04 nmeum, so which profile takes precedence when you have java7, java8 and gcc-java installed? 2015-11-24 10:28:24 fabled: hm, good point 2015-11-24 10:42:37 that's the sole point of java-common - to be the place which decides the default. of course user can override in ~/.profile if needed 2015-11-24 10:52:53 right, but iirc java-common is missing some symlink, maybe we should change that? 2015-11-24 10:54:03 we also need fix the java cert store 2015-11-24 10:54:14 maybe generate it with p11-kit? 2015-11-24 11:33:23 anyone know how to set working directory in an openrc initd file? 2015-11-24 12:01:04 cd ? 2015-11-24 12:04:33 yes, but i excepted it to be a option. 2015-11-24 12:15:42 i think you can start_pre() { cd $thepath; } 2015-11-24 12:15:43 or similar 2015-11-24 12:15:54 ah 2015-11-24 12:15:54 no 2015-11-24 12:15:57 -d , -chdir path 2015-11-24 12:15:57 chdir to this directory before starting the daemon. 2015-11-24 12:16:02 http://www.calculate-linux.org/main/en/start-stop-daemon 2015-11-24 12:16:06 ok 2015-11-24 12:16:39 i verified the kernel sources 2015-11-24 12:17:07 kernel will create devtmpfs entry before uevent 2015-11-24 12:17:34 so some userspace is doing funky 2015-11-24 12:18:32 now i wonder if it is udev or cryptsetup 2015-11-24 12:19:35 i bet it is udev 2015-11-24 12:20:42 always blame poettering 2015-11-24 12:21:37 this is the reason that there is a need for systemd, create new things to hide old mess 2015-11-24 12:22:23 sorry I am grumpy today 2015-11-24 12:47:41 ncopa: ah good catch :) 2015-11-24 12:47:57 i did start_pre, but this looks better. 2015-11-24 13:03:46 arg, can't spin up a scaleway any more 2015-11-24 13:04:45 I'm looking for a desktop card/cal dav client in Alpine. Does anyone use one? 2015-11-24 13:05:07 i am also interested in one if you find 2015-11-24 13:05:23 has anyone attempted to build thunderbird? 2015-11-24 13:05:31 not for a long time 2015-11-24 13:14:54 fabled: i have a nasty synchronisation issue with cryptsetup 2015-11-24 13:15:04 it is very nasty 2015-11-24 13:15:18 basically 2015-11-24 13:15:40 chromium + archon and andy caldav android client? 2015-11-24 13:15:47 s/any/andy 2015-11-24 13:16:04 while waiting for password input to cryptsetup, we need process incoming uevents (for plugging keyboard driver) 2015-11-24 13:16:15 however, once password is entered 2015-11-24 13:16:25 archon? 2015-11-24 13:16:45 we kind of need pause the uevent handling til cryptsetup process has exited 2015-11-24 13:17:27 ScrumpyJack: running android apps in chrome/chromium 2015-11-24 13:18:08 ScrumpyJack: https://github.com/vladikoff/chromeos-apk/blob/master/archon.md 2015-11-24 13:19:01 there is korganizer but alpine doesn't have working kde 2015-11-24 13:22:04 fabled: https://dpaste.de/VasX 2015-11-24 13:22:34 http://tpaste.us/GBJo 2015-11-24 13:22:56 i added --verbose and --debug to cryptsetup call 2015-11-24 13:32:00 seems like libdm has some udev synchronization code 2015-11-24 13:36:42 ncopa: Yes, I use luks in lvm on real hardware, I can help testing 2015-11-24 13:37:03 how many cpu cores? 2015-11-24 13:37:09 i think i have a synchronization issue 2015-11-24 13:37:16 i can reproduce it 2015-11-24 13:37:17 4 2015-11-24 13:37:27 i think things may work by random 2015-11-24 13:37:42 the problem is: 2015-11-24 13:37:51 cryptsetup is executed 2015-11-24 13:38:18 uevents needs to be handled in parallel (so keyboard driver is loaded in case it is not already) 2015-11-24 13:38:44 then cryptsetup will create the /dev/dm-X device 2015-11-24 13:38:47 and reload it 2015-11-24 13:39:01 and then finally will it create /dev/mapper/$name 2015-11-24 13:40:08 but when uevents are handled in parallel 2015-11-24 13:40:22 then the uevent for the /dev/dm-X device will arrive 2015-11-24 13:40:31 before the /dev/mapper/$name is created 2015-11-24 13:40:51 Oh 2015-11-24 13:41:15 also, during reload, the /dev/dm-X will temporary disappear 2015-11-24 13:42:09 i see that cryptsetup has som udev synchronization code 2015-11-24 13:44:19 i think what i want is that cryptsetup tell nlplug-findfs to take a break til it exits after password was entered 2015-11-24 13:47:15 1 2015-11-24 13:47:18 oops, sorry 2015-11-24 14:01:09 /dev/dm-X in this case X != 0 2015-11-24 14:01:36 As X = 0 is the lvm-device 2015-11-24 14:01:52 ACTION is just waking up and still a bit slow :-) 2015-11-24 14:02:08 yes, it may be dm-1 dm-2 dm-3 etc 2015-11-24 14:31:53 Oh, running "sudo nlplug-findfs" just switched my X11 keyboard layout back to english… that was unexpected :-) 2015-11-24 14:32:21 no, it will trigger coldplugging 2015-11-24 14:32:42 it resets keybard to english on my lappy too 2015-11-24 14:33:40 jomat: in case you are testing: http://tpaste.us/2b14 2015-11-24 14:33:50 not yets pushed git commits 2015-11-24 14:33:57 i think it will solve your lvm issue 2015-11-24 14:44:10 fabled: i think we will have to use cryptsetup-libs 2015-11-24 14:44:15 https://gitlab.com/cryptsetup/cryptsetup/wikis/API/index.html 2015-11-24 14:44:27 does not seem too complicated 2015-11-24 14:45:44 crypt_init(); crypt_load(); crypt_activate_by_passphrase(); 2015-11-24 14:47:24 Ohno, the upgrade deleted my /lib/modules/`uname -r` again m) 2015-11-24 14:47:38 huh? 2015-11-24 14:47:42 yeah 2015-11-24 14:47:50 upgrading kernel means you need reboot 2015-11-24 14:48:26 yes, and of course i forgot to mount /boot and now I don't have ext2.ko anymore :-) 2015-11-24 14:48:55 aw 2015-11-24 14:49:23 i think you can try rename the /lib/modules/$new 2015-11-24 14:49:38 the recompile was trivial 2015-11-24 14:49:42 so old modules will likely work 2015-11-24 14:49:59 Mhh, unsure, I'll try to insmod it… 2015-11-24 14:50:23 ahh, worked :-) 2015-11-24 14:50:28 :) 2015-11-24 14:50:47 btw: you don't need to rename it, "insmod /lib/modules/4.1.12-2-grsec/kernel/fs/ext2/ext2.ko" just works, too :-) 2015-11-24 14:51:12 yeah 2015-11-24 14:51:52 (if the dependencies are satisfied, ofc) 2015-11-24 14:52:26 ok this cryptsetup issue was a bit painful 2015-11-24 14:52:41 i have 2 different options on how to solve it 2015-11-24 14:52:53 and a hack 2015-11-24 14:54:01 the hack is that we can add a loop/sleep when block device appears when cryptsetup is executing 2015-11-24 14:54:48 do { retry; sleep } until (found || timeout); 2015-11-24 14:55:17 ugly so i dont consider that an option really 2015-11-24 14:55:46 other option is to store the pid of cryptsetup 2015-11-24 14:56:48 and when cryptsetup has exited, do an extra check for the device(s) we are looking for 2015-11-24 14:57:01 that would probably work but is semi ugly 2015-11-24 14:57:21 and the last option is to use libcryptsetup 2015-11-24 14:58:47 storing the cryptsetup pid would introduce a loop, too, wouldn't it? Just replacing s/found/cryptsetup exited/ 2015-11-24 14:59:18 or sth. like that 2015-11-24 15:01:41 no 2015-11-24 15:02:00 we catch any proces exiting so we can reap it 2015-11-24 15:02:26 http://git.alpinelinux.org/cgit/mkinitfs/tree/nlplug-findfs.c?h=nlplug-findfs#n868 2015-11-24 15:02:41 pid = waitpid(...) 2015-11-24 15:03:16 if (pid == pid_of_cryptsetup) found |= searchdev(....); 2015-11-24 15:03:30 ah, sure 2015-11-24 15:10:06 rnalrd: what do you think about %1043 2015-11-24 15:10:24 i have a rebased patch of it if you are interested in splitting nmap 2015-11-24 15:10:28 imho, it makes sense 2015-11-24 15:10:48 jomat: but i think the best solution is to use libcryptsetup 2015-11-24 15:13:01 ncopa: dito 2015-11-24 15:13:33 I'm soon ready to reboot :-) 2015-11-24 15:14:31 it may work, it may break, depending on timing 2015-11-24 16:19:34 ncopa: nlplug-findfs from mkinitfs 2.8.0_git20151024-r2 has these patches http://tpaste.us/2b14 applied? 2015-11-24 16:20:09 no 2015-11-24 16:20:14 Ah :-) 2015-11-24 16:21:12 at least now I know that 4.1.12-2-grsec still boots :-) 2015-11-24 16:21:25 yeah :) 2015-11-24 16:21:52 ncopa: any idea when could we expect first rc with test isos and images? 2015-11-24 16:22:02 this week 2015-11-24 16:22:35 while working on jomats issue i discovered a worse problem with cryptsetup 2015-11-24 16:22:46 i have a couple of ideas how to solve it 2015-11-24 16:23:37 i suppose we could ship a rc1 with cryptsetup semi broken 2015-11-24 16:24:13 there is one more thing i want fix in the initramfs: relocation of mounts so they match a given fstab 2015-11-24 16:24:26 but i dont think that will be too hard to do 2015-11-24 16:25:05 the nlplug-findfs/cryptsetup synchronization issue was worse 2015-11-24 16:26:40 patch applied, brb, reboot :-) 2015-11-24 16:32:20 ncopa: hm, I guess I can live without both 2015-11-24 16:32:52 ncopa: confirmed. Now boots from /dev/dm-0, /dev/mapper/root-alpine and /dev/root/alpine (symlink to /dev/mapper/root-alpine) 2015-11-24 16:32:57 I think I'll build rpi image today 2015-11-24 17:24:58 wut, I need arm to build rpi image 2015-11-24 17:55:42 why wouldn't you 2015-11-24 22:50:59 ping anyone 2015-11-24 22:51:33 this must be a FAQ. Alpine Edge, everything by default. busybox vi doesn't print accented characters (prints two dots instead). 2015-11-24 22:52:01 I checked in main/busybox: CONFIG_FEATURE_VI_8BIT=y as expected 2015-11-24 22:52:26 environment: CHARSET=UTF-8, shouldn't even be necessary with musl 2015-11-24 22:52:43 so, what am I doing wrong, and how can I get accented characters properly printed? 2015-11-25 01:45:29 skarnet: I always assumed that this was a bug in busybox 2015-11-25 05:53:13 nmeum: nope, it works on a bb I compile myself. 2015-11-25 05:56:44 do you have the spare time to fiddle with options until you find what goes wrong :&& 2015-11-25 05:58:07 nope, sorry - this time I'm a dumb user, not a hacker 2015-11-25 09:53:26 \o/ 2015-11-25 09:54:50 guess what, my nlplug-findfs solution seems to work 2015-11-25 09:57:16 and it seems to work pretty well too 2015-11-25 10:20:26 noticed in AL v3.1 that curl pulls in lua5.2, hope its ok ? 2015-11-25 10:21:11 yes, due to update-certificates 2015-11-25 10:21:17 we made an implementation in C 2015-11-25 10:21:23 so with v3.3 it will no longer be needed 2015-11-25 10:23:11 ok, thanks 2015-11-25 10:27:00 jomat: would you like to test another patch for initramfs? 2015-11-25 10:27:05 http://tpaste.us/Apan 2015-11-25 10:28:13 fabled: if/when you have time, i would like a code review of http://tpaste.us/Apan (on top of current git nlplug-findfs branch) 2015-11-25 10:28:20 i think it works as expected 2015-11-25 10:29:04 I also have this: http://tpaste.us/AWxx 2015-11-25 10:29:12 but that one i dont think i will push 2015-11-25 11:02:34 oh nice! seems like soundtouch is the only pkg in main that does not build on arm now 2015-11-25 11:02:59 in community: 2015-11-25 11:03:01 build-3-3-armhf:~/aports/community$ ls -d */src 2015-11-25 11:03:01 firefox/src webkit2gtk/src 2015-11-25 11:04:17 firefox was internal error 2015-11-25 11:04:23 probably due to out of memory 2015-11-25 11:07:00 webkit2gtk has issue with asm 2015-11-25 11:07:05 ok lunch 2015-11-25 11:15:08 skarnet: oh, that's interessting 2015-11-25 11:16:03 skarnet: would you mind sharing the config you used to compile it yourself? Then I might be able to figure out which options are missing or incorrectly set in our config… 2015-11-25 11:27:49 morning 2015-11-25 11:28:23 where do we expect to find .so symlinks to libs .so.version? 2015-11-25 11:28:41 in pkg-dev or in pkg-libs? 2015-11-25 11:29:15 the symlink from /usr/lib/libsqlite3.so to /usr/lib/libsqlite3.so.0 is in sqlite-dev 2015-11-25 11:29:59 so you can't access libsqlite3.so without sqlite-dev 2015-11-25 11:31:21 (csync2 is looking for libsqlite3.so) 2015-11-25 11:31:58 maybe csync2 should be looking for libsqlite3.so.0 2015-11-25 12:03:36 what is csync2 doing? 2015-11-25 12:03:40 dlopen? 2015-11-25 12:11:03 just thinking if it would be better for main/youtube-dl to be in community/ 2015-11-25 12:12:35 hey :) 2015-11-25 12:12:48 hi :) 2015-11-25 12:16:28 leo-unglaub: you use cryptsetup, right? 2015-11-25 12:16:49 ncopa: yes i do 2015-11-25 12:16:56 on all drives 2015-11-25 12:17:09 you have lvm on cryptsetup too? 2015-11-25 12:17:11 or mdadm? 2015-11-25 12:17:24 or cryptsetup on mdadm? 2015-11-25 12:17:44 lvm no because its to slow, but mdadm with cryptsetup on it 2015-11-25 12:18:00 i am about to push a new mkinitfs, with updated nlplug-findfs 2015-11-25 12:18:23 great, i am goingto test it as soon as the build is done 2015-11-25 12:19:02 reminds me 2015-11-25 12:19:08 there was one more issue with lvm 2015-11-25 12:19:24 prints error message from initramfs 2015-11-25 12:19:52 there are a lot of issues with cryptsetup+mdadm and booting in general 2015-11-25 12:37:39 ncopa: pushed? 2015-11-25 12:37:44 not yet 2015-11-25 12:42:53 leo-unglaub: pushed! 2015-11-25 12:43:46 nmeum: it wasn't on Alpine, it was on a manually built system, but I'll dump the .config somewhere when I get the time 2015-11-25 12:48:07 ncopa: updating 2015-11-25 12:56:39 ncopa: was it on a musl based system? 2015-11-25 12:56:43 ncopa: okay, it boots but with a lot of error messages 2015-11-25 13:25:28 ncopa: strace csync2 shows: open("/usr/lib/libsqlite3.so", O_RDONLY|O_CLOEXEC) = 4 2015-11-25 14:03:41 hmm, i also have problems with cryptsetup initrd. just created one, and the initrd does not contain cryptsetup, even though mkinitfs.conf contains cryptsetup as feature. 2015-11-25 14:16:45 qdeyXHmyEe5v: what do you have in your /etc/mkinitfs/features.d/cryptsetup.files? 2015-11-25 14:17:13 also, with todays push, it will now use libcryptsetup and not the /sbin/cryptsetup binary 2015-11-25 14:20:31 files contains /sbin/cryptsetup 2015-11-25 14:20:45 but initrd didnt include it? 2015-11-25 14:21:06 i figured out that mkinitfs somehow calls the file in the end /boot/grsec.gz and not /boot/initramfs-grsec 2015-11-25 14:21:19 is it v3.2 or edge? 2015-11-25 14:21:56 this is an old 3.0 stick from which i install... /o\ 2015-11-25 14:22:13 that explains :) 2015-11-25 14:23:29 now i have an emergency shell, where manually i can open the partition, and mount it. but somehow cryptroot= is not used correctly. 2015-11-25 14:25:16 i dont know if cryptsetup was supported in alpine 3.0 2015-11-25 14:30:12 it was 2015-11-25 14:30:34 i used it back then for other systems. i even hacked the mkinitfs script to support detached headers. 2015-11-25 14:30:40 detached luks headers 2015-11-25 14:35:37 hmmm, now i booted into it, and did a setup-xorg-base and i get udev requires /dev to be a mounted devtmpfs please reconfigure your system. what exactly do i have to do? 2015-11-25 14:36:09 (booted into it == manually opened the luks root in the emergency shell, mounted it on sysroot, and exited) 2015-11-25 14:36:13 sounds like you mix some stable with edge 2015-11-25 14:36:59 edge's udev (eudev) will not work with older kernels 2015-11-25 14:37:09 you need new kernel for that, with devtmpfs 2015-11-25 14:37:20 i have now 4.1.12-2-grsec running here. 2015-11-25 14:37:44 and you have custom initrd? 2015-11-25 14:37:51 not really 2015-11-25 14:38:08 only one line changed for the cryptsetup using an additional --header param. 2015-11-25 14:38:25 hmmm. maybe i redo the mkinitfs using this system. 2015-11-25 14:38:28 i'll try that 2015-11-25 14:38:46 you need this: http://git.alpinelinux.org/cgit/mkinitfs/commit/?id=ac397ebafd39a277ca94cd37b70ad15d1c1e3d46 2015-11-25 14:40:41 but there have been alot of changes in initramfs recently 2015-11-25 14:40:59 we dont have support for cryptsetup with detached headers currently 2015-11-25 14:41:18 damn. 2015-11-25 14:41:21 oki 2015-11-25 14:41:26 how does that work? 2015-11-25 14:41:29 the --header thingy? 2015-11-25 14:41:48 do you need special boot option for that? 2015-11-25 14:41:51 instead of having the luks metadata at the beginning of the partition, you have it somewhere else in a file 2015-11-25 14:42:17 so i added a --header=/luks-header to the initramfs invocation of cryptsetup 2015-11-25 14:42:34 where /luks-header is the configurable location of the header 2015-11-25 14:42:38 got you 2015-11-25 14:42:59 how big is this header file? 2015-11-25 14:43:10 4M /o\ 2015-11-25 14:43:19 because of some stupid luks limitation 2015-11-25 14:43:23 the useful data is much less 2015-11-25 14:43:45 and you include this in the initramfs image? 2015-11-25 14:44:14 yes. using mkinitfs/features/cryptsetup.files i added this to be included 2015-11-25 14:44:38 interesting 2015-11-25 14:45:00 how about this 2015-11-25 14:45:11 we have a magic file 2015-11-25 14:45:32 lets say /etc/cryptsetup/luks-header 2015-11-25 14:45:40 if it exists it is automatically used 2015-11-25 14:45:55 i like this 2015-11-25 14:46:07 if not, we assume header meta data to be at the beginning 2015-11-25 14:46:13 alternatively 2015-11-25 14:46:21 we could support separate device 2015-11-25 14:46:26 eg, sd card reader 2015-11-25 14:46:42 indeed. that's the point 2015-11-25 14:46:52 the boot partition is now on an external usb stick. with this header 2015-11-25 14:47:33 intereseting 2015-11-25 14:47:44 so you have the entire boot partition on usb 2015-11-25 14:47:50 yes 2015-11-25 14:48:02 and a headerless luks partition on the hd 2015-11-25 14:48:13 without the usb no booting, no evil maiding. 2015-11-25 14:48:21 nmeum: http://skarnet.org/tmp/bbconfig.txt 2015-11-25 14:48:25 (at least unless not doing bios evilmaiding) 2015-11-25 14:48:37 thats kinda cool 2015-11-25 14:48:58 skarnet: thanks, did you built it on a musl based system back then? 2015-11-25 14:49:34 there's way too much metadata on the luks header. 2015-11-25 14:50:36 i wonder it it would make sense to have a separate boot partition on disk, and have a luksheader= boot option 2015-11-25 14:50:53 but probably better with the entire boot on usb 2015-11-25 14:51:51 in theory we could make initramfs finnd the boot partition and find the luks header there 2015-11-25 14:52:11 but it is probably simpler with a magic header file that gets copied to initramfs 2015-11-25 14:52:32 nmeum: yes, I believe so. With an Aboriginal Linux toolchain, hacked to link stuff against musl. 2015-11-25 14:53:17 i dunno, i liek to be explicit with my keymaterial, automatic location of it sounds scary, but dunno yet why 2015-11-25 14:53:58 i mean 2015-11-25 14:54:05 nmeum: I can't *swear* it, because my first builds were uClibc 32-bit, and I switched to musl 64-bit shortly after, but I believe I did rebuild bb. If you cannot reproduce, it means I'm wrong. 2015-11-25 14:54:26 qdeyXHmyEe5v: at boot option you'd need a luksheader=/path/to/header 2015-11-25 14:55:02 at boot nlplug-findfd would mount detected devices and look for /path/to/header 2015-11-25 14:55:46 once cryptroot and /path/to/header was found, it would init the cryptroot 2015-11-25 14:57:01 hmmm, would be very convenient. i guess i'd like that. 2015-11-25 14:57:32 the good thing is that it would be faster to load the initramfs as its 4MB smaller 2015-11-25 14:57:53 and creating the initramfs would be faster too 2015-11-25 15:00:45 true 2015-11-25 15:01:05 but magic luks-header file is faster to implement 2015-11-25 15:01:19 does that mean in my case i have one boot device, in your case you have a stick and an sd card for the luks header? 2015-11-25 15:01:54 it means you have the luks header file on your boot device 2015-11-25 15:02:11 aah. yes. like i do now. ok. gotit 2015-11-25 15:02:19 actually 2015-11-25 15:02:30 you could have it on separate too 2015-11-25 15:02:36 lets say sd card 2015-11-25 15:02:41 and boot partition 2015-11-25 15:02:50 bot partition on internal disk 2015-11-25 15:03:03 hmmm, indeed. both scenarios work 2015-11-25 15:03:14 the luks header could be on any removable device 2015-11-25 15:03:37 as long as it corresponds to the boot option 2015-11-25 15:03:50 hmmm 2015-11-25 15:05:19 it would take some time to implement 2015-11-25 15:05:29 i would need refactor nlplug-findfs 2015-11-25 15:05:51 however, a magic header file should be relatively quick to implement 2015-11-25 15:07:59 great, i added the patch to my init script, and now the /dev is mounted as it should be. 2015-11-25 15:08:18 i guess now i could say i have a custom init script... 2015-11-25 15:09:30 can this wait for v3.4 release? 2015-11-25 15:09:37 sure 2015-11-25 15:09:44 i can hack on it myself in the mean time 2015-11-25 15:09:45 i'm about to do v3.3 and are not too keen on add new features at thsi stage 2015-11-25 15:10:05 no worries. i'm happy to hack on this without official support 2015-11-25 15:10:34 i think it is interesting ideas though 2015-11-25 15:10:37 been doing this already for quite some time 2015-11-25 15:10:45 if you have the header file on external media 2015-11-25 15:11:10 then you can make machine completely unbootable byt removing 2015-11-25 15:11:22 sdcard slot 2015-11-25 15:11:26 for example 2015-11-25 15:11:35 yep 2015-11-25 15:11:47 just remove the sdcard and laptop is "uninstalled" 2015-11-25 15:11:57 well, there's a default windows on that laptop, which is like a decoy... 2015-11-25 15:12:06 :) 2015-11-25 15:12:27 booting only with bios choose bootmedia menu 2015-11-25 15:12:45 just a second "uninitialized" partition 2015-11-25 15:17:57 back 2015-11-25 15:18:29 ncopa: did you get the email screenshot? 2015-11-25 15:19:44 leo-unglaub: yup 2015-11-25 15:19:53 i dont think its related the new initramfs image 2015-11-25 15:19:58 i dont know why it happens 2015-11-25 15:20:11 apparently it finds the root partition and manage to boot it 2015-11-25 15:20:23 but the other partitions are not found for some reason 2015-11-25 15:20:43 maybe cryptsetup init.d script broke? 2015-11-25 15:23:55 hmmm 2015-11-25 15:24:17 the strange thing is that it worked before so i am not sure what changed 2015-11-25 15:24:24 it decrypts my first drive 2015-11-25 15:24:36 but not my secound drive and then it fails to mount stuff 2015-11-25 15:26:53 at what time is the second drive decrypted? 2015-11-25 15:26:57 from initfs too? 2015-11-25 15:27:05 or from openrc 2015-11-25 15:27:10 its in /etc/crypttab 2015-11-25 15:27:15 but its never executed 2015-11-25 15:27:35 openrc got updated 2015-11-25 15:27:50 and cryptsetup was updated 11 nov 2015-11-25 15:56:20 whee 2015-11-25 15:56:21 \o/ 2015-11-25 15:56:25 isohybrid works 2015-11-25 15:56:27 kind of 2015-11-25 15:57:18 yes 2015-11-25 15:57:22 it works \o/ 2015-11-25 15:58:14 \o/ 2015-11-25 16:02:02 i need clean up the initramfs init script and fix the modloop script 2015-11-25 16:27:11 ncopa: I'll help you clean up all this for 3.4 if I get the time :) 2015-11-25 16:27:37 what is the difference between the Raspberry Pi iso image and the Generic ARM iso image? 2015-11-25 16:28:04 I want to install Alpine on a qemu-system-arm, what image should I use? 2015-11-25 16:28:29 (maybe I should ask those kind of questions in #alpine-user...) 2015-11-25 16:47:30 ok, those aren't images, those are tarballs. Meh. 2015-11-25 17:41:41 Looks like nobody has written a tutorial about running Alpine in a qemu-system-arm. 2015-11-25 17:41:57 I mean, installing. 2015-11-25 17:42:35 If someone has pointers, I'll gladly take them. Otherwise, I'll explore, and gather materials for such a tutorial... but I'd rather not have to do it. 2015-11-25 17:45:18 <_mjones> skarnet: good question. I was doing some work with armel not long ago, and soon will be working with armhf, but that's all hardware and not qemu. 2015-11-25 17:45:45 <_mjones> I suppose one needs a devicetree that's supported by the kernel. 2015-11-25 17:46:23 <_mjones> So my informed guess is that if you want to use the rasp-pi install media, you'll need to configure qemu to look just like RPi. I bet a tutorial is a quick bing away. 2015-11-25 17:46:29 I could use a big hammer and boot Aboriginal on the arm qemu, then build Alpine from source on it... 2015-11-25 17:47:25 <_mjones> I was in the process of doing that when I bricked my armel hardware, ha. It's fixable but painful, so I put that project on hold for a bit. 2015-11-25 17:47:50 <_mjones> I'm not terribly well informed on devicetree stuff. 2015-11-25 17:48:55 I'm not either. But the real hardware won't look one bit like a RPi, and I just want some place to compile my stuff to arm while I'm waiting for the hw. 2015-11-25 17:48:57 <_mjones> Planning on trying to get mainline kernel up on an Amsong SoC, and I'd like to make an alpine package for said kernel for distro. 2015-11-25 17:49:12 <_mjones> oh, you didn't say that before. 2015-11-25 17:49:38 I asked what the difference was between "Rpi" and "Generic ARM" 2015-11-25 17:49:47 <_mjones> If you have real hardware then the RPi qemu won't help a lot, unless you do one in preparation for doing a totally different one. 2015-11-25 17:50:01 <_mjones> The thing is I don' 2015-11-25 17:50:07 <_mjones> t know what 'generic arm' means. 2015-11-25 17:50:36 <_mjones> There aren't many way to have a generic distribution of armhf Linux because of the devicetree and hardware-specific stuff. 2015-11-25 17:50:48 I will have real hw at *some* point, depending on other people's reactivity and I have limited trust in that one. So I want to work on a qemu in the meantime. 2015-11-25 17:50:52 <_mjones> Let me look at the aports and maybe I will spot something. 2015-11-25 17:50:56 <_mjones> what hw? 2015-11-25 17:51:02 <_mjones> ACTION checks aports for generic arm. 2015-11-25 17:51:10 some AMLogic thingy. 2015-11-25 17:52:03 <_mjones> I was planning on s805 on odroid-c1+, but I've been very lazy about ordering hardware, and then another project on arm64 came up. 2015-11-25 17:52:39 the choice of SoC doesn't depend on me, and I'm not experienced enough with hw to ask to be a part of the decision process in that one 2015-11-25 17:52:40 <_mjones> uh, have to merge aports changes to my local... 2015-11-25 17:52:58 <_mjones> So the hw decision isn't totally done yet. 2015-11-25 17:54:51 AIUI, they want to try things with 2 or 3 hw manufacturers before settling on one. 2015-11-25 17:54:58 <_mjones> I don't see an aport for generic arm kernel. Unexpected. 2015-11-25 17:56:49 <_mjones> grsec armhf, that's it. 2015-11-25 17:57:32 I don't know what the "hf" means in armhf 2015-11-25 17:57:45 <_mjones> hardfloat. 2015-11-25 17:57:50 ... of course. 2015-11-25 17:57:50 <_mjones> means ARMv7. 2015-11-25 17:58:01 <_mjones> armel is armv6 I think. 2015-11-25 17:58:08 <_mjones> armv8 is 64bit 'arm64'. 2015-11-25 17:58:12 <_mjones> armhf means armv7. 2015-11-25 17:58:17 I know it will be armhf then. 2015-11-25 17:58:28 <_mjones> If it's modern and 32bit it will be armhf. 2015-11-25 17:59:14 <_mjones> There's a patch in-tree for imx6 so that's supported by generic. I'm looking at the kernel config now to see what else. 2015-11-25 18:00:30 <_mjones> OK, it's long but you want to look in main/linux-grsec/config-grsec.armhf for the different hw support. 2015-11-25 18:00:43 <_mjones> I see TI OMAP. 2015-11-25 18:00:47 skarnet: I am in the process of installing alpine on a ARM device 2015-11-25 18:01:16 xming: it would be awesome if you were documenting the steps you take :) 2015-11-25 18:01:26 from what I've gathered, you need alpine's uboot 2015-11-25 18:02:03 <_mjones> You'll need a uboot specific to the hardware. 2015-11-25 18:02:04 my device is really a weirdo I need more recent mainline kernel then 4.1.x 2015-11-25 18:02:11 <_mjones> xming, thanks for bringing that up. 2015-11-25 18:02:34 _mjones: Alpine needs a forked uboot to support tarballs, that's what fabled told me 2015-11-25 18:02:46 <_mjones> s805 needs 4.4 and I'm not even sure if everything is merged into that. Meson platform. :-/ 2015-11-25 18:03:00 I'm not worried about booting the hw right now, I'm just trying to run a qemu! 2015-11-25 18:03:06 <_mjones> xming: ah, I didn 2015-11-25 18:03:08 <_mjones> 't know that. 2015-11-25 18:03:21 <_mjones> booting qemu, booting hardware, same thing. 2015-11-25 18:03:45 <_mjones> The uboot image will need to match the emulated hw, in my experience. 2015-11-25 18:03:46 in general, ARM boot SPL then uboot then extlinux.conf style config 2015-11-25 18:03:51 except you can pick and choose what you emulate, so if there's something simple, I'll take it 2015-11-25 18:04:01 <_mjones> That's why I recommended emulated RPi, because Alpine already has a target and distro-media. 2015-11-25 18:04:03 with mainline uboot you can use DOS or GPT partitions 2015-11-25 18:04:15 boot partition need boot flag set 2015-11-25 18:04:35 after that, it's more or less like x86 2015-11-25 18:04:51 <_mjones> RPi isn't a very good SoC, but it's a good least-common-denominator target that everyone knows and understands. 2015-11-25 18:05:18 <_mjones> xming: good stuff. skarnet is right about needing docs though. 2015-11-25 18:05:20 one exception is that you provide FDDIR line in extlinux.conf for the kernel to find DT 2015-11-25 18:05:32 <_mjones> A fork of arch supports armhf, maybe there's docs to borrow. 2015-11-25 18:05:56 _mjones: yes, it took me sometime just to gather the above info 2015-11-25 18:06:16 <_mjones> Just putting your last few comments in a wiki article would be a good start. 2015-11-25 18:06:17 currently it doesn't boot and I am blind, still waiting for my USB TTL cable 2015-11-25 18:06:49 <_mjones> I bricked my armel because the hw lacks console. Not doing that again soon. 2015-11-25 18:07:12 no console on the hw? not even a serial port? 2015-11-25 18:07:12 also when partitioning, make sure that the first partition starts at 1MB (2048 block/sector) 2015-11-25 18:07:18 so you have place for the uboot 2015-11-25 18:07:38 skarnet: there is, but it's 3v ttl and I only have 5v rs232 here 2015-11-25 18:07:57 <_mjones> skarnet: it was sold as a nas. Buffalo linkstation mini. Nice little hw with twin sata ports. 2015-11-25 18:07:58 oh sorry you were talking to _mjones 2015-11-25 18:08:21 oh, I'll take any experience :) 2015-11-25 18:08:56 I have this board http://en.t-firefly.com/en/ 2015-11-25 18:09:06 _mjones: how do you even install stuff on that... 2015-11-25 18:09:07 the 4GM RAM version :D 2015-11-25 18:10:42 I will document when I get at least the kernel booting, right now I don't even sure if I am doing things correctly 2015-11-25 18:10:54 <_mjones> skarnet: complex bootstrap from the initial vendor code. I bootstrapped to Debian 7.x. Was concerned about systemd on a 128mb ram box, considering source-porting Alpine over. 2015-11-25 18:11:08 _mjones: doesn't it have loader/recovery mode 2015-11-25 18:11:27 <_mjones> skarnet: built a mainline kernel and the kernel comes up but not userland, so bricked. I'll have to tear it apart to move back to old kernel by mounting drives from another machine. 2015-11-25 18:11:41 _mjones: I mean, without a console, how can you work? blind? 2015-11-25 18:11:51 <_mjones> xming: there's a tftp recovery but only works if the drives are zeroed. :P 2015-11-25 18:11:59 <_mjones> skarnet: yes, blind. 2015-11-25 18:12:00 also most Linux for ARM info are Android specific, and they are made for ROMs and not DOS/GPT partitions 2015-11-25 18:12:04 ugh. 2015-11-25 18:12:24 ACTION just doesn't want to work blind, ever. Too risky. 2015-11-25 18:12:46 <_mjones> skarnet: I'll fix it eventually. But for now I'm moving to odroid armhf hardware with sdcards and console. 2015-11-25 18:12:52 blind is like doing remote Linux install/upgrade w/o iKVM :D 2015-11-25 18:12:57 <_mjones> armhf and 1gb. 2015-11-25 18:13:11 odroid is nice 2015-11-25 18:13:18 but only 1/2 GB 2015-11-25 18:13:29 <_mjones> no, 1gb on c1/c1+. 2015-11-25 18:13:36 xming: reminds me of my favourite interview question ever 2015-11-25 18:13:52 <_mjones> I haven't done 512mb since 1998. 2015-11-25 18:13:55 "You have an unmanned datacenter on the moon, and you need to do a kernel upgrade." 2015-11-25 18:14:36 <_mjones> skarnet: I'll need a mirror here for testing. ;) 2015-11-25 18:14:39 take a shuttle and see you in 2 weeks :D 2015-11-25 18:15:12 starting with a local canary is a good idea :) 2015-11-25 18:15:14 <_mjones> radhard chips are ridiculously expensive and export-controlled defense items. 2015-11-25 18:15:32 <_mjones> skarnet: you're not in states, correct? 2015-11-25 18:16:18 Correct. EU, where law isn't quite as brain-damaged as NA 2015-11-25 18:16:25 (slowly getting there though) 2015-11-25 18:17:38 <_mjones> Somewhat differently brain-damaged. 2015-11-25 18:17:53 well UK is 2015-11-25 18:17:57 <_mjones> But with a population that is less wary of government power in general. 2015-11-25 18:18:04 luckily I am not in UK 2015-11-25 18:18:40 most parts of West europe are getting very close 2015-11-25 18:18:49 which saddens me 2015-11-25 18:19:10 <_mjones> I used to have a minor sideline in exporting computer hardware, and years ago you couldn't export machines past a certain speed because they could be used to sim nuclear tests and FEA pits. 2015-11-25 18:20:09 <_mjones> There are things you can buy over the counter here in new york with which you can't take in your luggage to moscow or you'd be breaking the law. They arrested a woman for that a couple of years ago. 2015-11-25 18:20:44 or: how politics get in the way of tech and QoL 2015-11-25 18:21:14 <_mjones> In my experience, europeans and commonwealth citizens tend to be naive about how governments will enforce the laws they want. 2015-11-25 18:22:04 _mjones: still slightly better than the NA system, vote for the lesser of 2 evils :D 2015-11-25 18:22:32 <_mjones> example of import bust: http://jalopnik.com/why-are-the-feds-obsessed-with-seizing-these-peoples-ol-1672381729 2015-11-25 18:23:10 <_mjones> xming: possibly. But people sometimes have misconceptions about NA/American situation too. What bothers you most? 2015-11-25 18:23:30 _mjones: Pax Americana 2015-11-25 18:24:11 <_mjones> oh, interventionism. I assumed you meant respective domestic regulatory regime. 2015-11-25 18:25:21 right, the whole earth is domestic 2015-11-25 18:26:18 <_mjones> It would be ultimately embarassing to have a clinton and a bush running against each other in next year's American presidential election. 2015-11-25 18:26:45 does it matter? Obama is worse in my POV 2015-11-25 18:26:52 <_mjones> Looks like one of them won't be, but the D party is insanely determined to make the other one their candidate. 2015-11-25 18:27:39 <_mjones> xming: funny how nobody had a problem with obama until his second term. :-/ 2015-11-25 18:28:42 _mjones: nobody might be not accurate, I never liked him in the first place, I'd vote for Clinton if I were an American 2015-11-25 18:29:20 He was too obivious as a con man 2015-11-25 18:29:34 <_mjones> Clinton is essentially an incumbent career politician with a $2,000,000,000 election fund. Wouldn't vote for her if my life depended on it. 2015-11-25 18:29:57 <_mjones> Obama was juiced in to the Chicago political machine, and on the board of the Joyce foundation. No thanks. 2015-11-25 18:30:10 if you want to continue thise nice chat, maybe we can do this in private, we are going way off-topic 2015-11-25 18:30:17 amazing 2015-11-25 18:30:35 <_mjones> I don't want to be off-topic. 2015-11-25 18:30:50 so let's talk about ARM :D 2015-11-25 18:31:03 amurricquese politics is broken as hell next topic 2015-11-25 18:31:26 <_mjones> Who has wiki access? xming's information can be the nucleus of an article. 2015-11-25 18:31:40 so is most other politics but ah well 2015-11-25 18:32:14 <_mjones> The last person you want to give power is the person who wants power. That's why I go by the ietf motto: 2015-11-25 18:32:36 <_mjones> "We reject kings, presidents and voting. We respect rough consensus and running code." 2015-11-25 18:33:30 _mjones: register and you have access :D I have acces, I am just a bit reluctent to put these info on the wiki right now, before I have verified the results 2015-11-25 18:34:05 <_mjones> xming: I respect that but sometimes you have to treat it as nascent code. Even if it's ugly, it's a start. 2015-11-25 18:34:46 maybe I will do it on a "General ARM" page marked was WIP 2015-11-25 18:35:10 <_mjones> sounds good to me. 2015-11-25 18:35:30 <_mjones> There needs to be a qemu-arm howto eventually too. 2015-11-25 18:35:50 but still I only have one ARM board (rockchip) and I don't know if those info apply to other SoCs, I think they are but ... 2015-11-25 18:36:34 <_mjones> I'm guilty too; I have a stack of small changes I haven't submitted to appropriate upstreams, and a list of todo projects besides. 2015-11-25 18:36:45 <_mjones> xming: which board exactly? 2015-11-25 18:37:12 firefly rk3288, 4GB RAM/32 eMMC 2015-11-25 18:37:28 it's going to reaplce my x86 home server 2015-11-25 18:37:33 <_mjones> ah. Lots of dram you got there. 2015-11-25 18:37:40 probably saving me like 50~100W/h 2015-11-25 18:38:01 _mjones: yes the most affordable ARM board with that much RAM 2015-11-25 18:38:15 <_mjones> Heh. Good idea. I just inherited a pair of big dell rackmount servers with twin Xeons each at 80w tdp. 2015-11-25 18:38:55 that firefly uses max. 25W 2015-11-25 18:39:02 <_mjones> I was doing arm work in part for power savings. And heat savings. 2015-11-25 18:39:13 oh and gigabit ethernet and ac wireless 2015-11-25 18:39:17 <_mjones> 25w for the board all together I'm sure. 2015-11-25 18:39:26 although wlan doesn't work in mainline kernel yet 2015-11-25 18:39:37 _mjones: yes 2015-11-25 18:39:41 <_mjones> Anything with less than gigabit I reject, and normally don't deal with anything less than 1gb. 2015-11-25 18:40:19 <_mjones> I wonder if netbooting these is easier for development work. 2015-11-25 18:40:33 only drawback is that it doens't have SATA port, but most SATA ports are just usb2-SATA bridge so external HDD is the same 2015-11-25 18:41:04 yes mainline there is a uboot branch which supports tftp boot 2015-11-25 18:41:07 <_mjones> depends on the SoC. I think allwinner a20 isn't usb-sata. 2015-11-25 18:41:28 on that fireffly board that is 2015-11-25 18:41:47 <_mjones> these boards should have enough resources for ipxe if it wasn't hardware dependent. https boot is nice. 2015-11-25 18:42:08 it's uboot implementation and uboot space 2015-11-25 18:42:12 <_mjones> tftp is limited to 64k packets, if that. 2015-11-25 18:42:19 for rk3288 the uboot is limited to 32KB 2015-11-25 18:42:40 so one cannot stuff too much stuff in it 2015-11-26 07:58:46 seems like chat.freenode.net are having issues? 2015-11-26 07:59:51 seems so 2015-11-26 08:00:06 noooo, what is giving you that impression ? :P 2015-11-26 09:26:31 bah texlive is messed up 2015-11-26 09:35:00 morning 2015-11-26 09:39:19 hey :) 2015-11-26 09:39:24 how is everyone? 2015-11-26 09:40:54 looking forward to Christmas :) 2015-11-26 09:44:28 hehe 2015-11-26 09:49:16 ncopa: hey 2015-11-26 09:49:22 the boot issue still exists 2015-11-26 09:49:32 i now removed the line from fstab and now it boots 2015-11-26 09:49:38 the bug is somewhere in dmcrypt 2015-11-26 09:49:49 because it always just decrypts the first line from /etc/crypttab 2015-11-26 09:49:54 all other lines are ignored 2015-11-26 09:51:09 here is my crypttab: http://paste.debian.net/plainh/12f1b174 2015-11-26 09:52:50 my attempt to clarify things on ARM http://wiki.alpinelinux.org/wiki/Alpine_on_ARM 2015-11-26 09:53:00 skarnet: ^ 2015-11-26 10:01:33 I maybe have found a bug with the qemu package (qemu-kvm to be more specific). Everything seems to work fine in a vm till it gets load (my assumption). Eventually it crashes. Problam is that there is like only one line from qemu-system-86_64 and it seems to be not very helpful. I've also created an strace now. However it's size is 10GB in total. 2015-11-26 10:03:08 The error message is this: qemu-system-x86_64: Guest moved used index from 11773 to 11776 2015-11-26 10:03:35 I'll need to cut the strace and upload it somewhere. I don't think uploading the whole 10GB thing will help. 2015-11-26 10:04:39 To be sure it's not my configuration, I'll try testing it with another host OS and the exact same settings. 2015-11-26 10:05:18 Does anyone have any more ideas here? What I can try and stuff. 2015-11-26 10:05:30 is qemu running on alpine host? 2015-11-26 10:05:40 ncopa: alpine host and alpine vm 2015-11-26 10:05:42 so yes 2015-11-26 10:05:44 both 3.2 2015-11-26 10:06:20 I've already tried with linux and linux-vanilla on the host. Doesn't make any difference. 2015-11-26 10:07:13 So it should not be related to grsecurity at least. 2015-11-26 10:07:30 do you use vnc? 2015-11-26 10:08:02 sounds similar to this: https://lists.nongnu.org/archive/html/qemu-devel/2010-09/msg01145.html 2015-11-26 10:08:26 how much memory do you give the guest? 2015-11-26 10:08:34 do you use --enable-kvm ? 2015-11-26 10:08:34 no. the qemu-system-x86_64 line has -nographics at the end. I use serial console. Have only used spice for installing the vm. 2015-11-26 10:08:46 yes i use --enable-kvm 2015-11-26 10:09:12 I use qemu on x86_64 all the time 2015-11-26 10:09:19 but i am using edge 2015-11-26 10:09:21 there are some similar reports according to google, it could be a qemu bug 2015-11-26 10:09:27 s/nographics/nographic/ 2015-11-26 10:09:38 pluesch: can you paste the command line how you start qemu-kvm? 2015-11-26 10:09:45 good idea 2015-11-26 10:10:18 http://sprunge.us/ZERM 2015-11-26 10:10:34 $5 rPi Zero, imaging a beowulf cluster of ... 2015-11-26 10:11:20 normally it's of course without the strace 2015-11-26 10:11:54 pretty vanilla, have you tried w/o smp? 2015-11-26 10:12:01 xming: not yet 2015-11-26 10:12:07 but i can, sure :) 2015-11-26 10:12:24 there is really no point to have smp and thread, unless you want to test those things 2015-11-26 10:12:50 -smp cores=6 is the same 2015-11-26 10:13:00 okay so I'll do it like that: 1) try it without smp and thread 2) try it on edge 3) try it on different os with same host hardware 2015-11-26 10:13:04 does that sound okay? 2015-11-26 10:13:08 yes 2015-11-26 10:13:15 pluesch: and try -smp cores=3 2015-11-26 10:13:18 pluesch: and try -smp cores=6 2015-11-26 10:13:39 it will take me a while for doing all this and i need to go now. will report back once I have more information. 2015-11-26 10:14:04 https://bugzilla.redhat.com/show_bug.cgi?id=962413 2015-11-26 10:14:08 it usually also takes like 1-2 hours till the vm crashes 2015-11-26 10:14:38 I think that "Guest moved used index ..." is printed because the guest isn't there any more 2015-11-26 10:14:42 is there some way you can stress the guest to produce crash earlier 2015-11-26 10:14:48 that's not the cause 2015-11-26 10:15:15 also, what hw is it? 2015-11-26 10:15:22 intel or amd? 2015-11-26 10:15:25 intel 2015-11-26 10:15:54 btw there is no helpful output in dmesg on both vm (as long is i can look at it) and on host 2015-11-26 10:18:05 smells like qemu bug 2015-11-26 10:18:09 pluesch: your host has more than 6 cores and 12GB memory? 2015-11-26 10:18:32 ncopa: +1 2015-11-26 10:20:09 intel e5 2620 v3 is the host cpu 2015-11-26 10:20:19 memory it has 32gb in total 2015-11-26 10:20:39 the qemu vm is the only thing running on the host 2015-11-26 10:20:59 trying without smp at all right now 2015-11-26 10:21:24 would be nice to find a way to provode the error quicker than 1-2 hours 2015-11-26 10:21:43 -.- yeah .... 2015-11-26 10:21:49 can we DDOS you VM? :D 2015-11-26 10:23:02 too bad the rPi Zero doesn't have any NIC 2015-11-26 10:24:05 I'm running redis and python (celery) on it that creates a lot of files and does a lot of network connection. To be more specific I'm archiving 4chan boards with it. 2015-11-26 10:24:50 The script runs in many other places too. Some also KVM virtual machines. So it shouldn't by my script crashing the vm. 2015-11-26 10:24:51 wait, you don't use virtio-io for the block device? 2015-11-26 10:25:24 if you have tons of IO bettter use that 2015-11-26 10:25:44 -drive file=${DISKIMG},if=virtio,format=raw 2015-11-26 10:25:49 I do... I think. 2015-11-26 10:27:39 need to go now. Thanks for all your help. I'll report back when I'm home again (tonight). 2015-11-26 10:28:01 ah right missed that line 2015-11-26 10:28:21 i have a windows 7 vm on my work desktop in qemu 2015-11-26 10:28:32 using virt-manager 2015-11-26 10:28:37 works perfect for me 2015-11-26 10:28:47 disk is raw lvm image 2015-11-26 10:29:32 i also test boot all kernel in qemu with ncurses output on dev server 2015-11-26 10:29:38 but those never run longer than 5 mins 2015-11-26 10:29:47 my homw server has a bunch of VMs w/o graphics 2015-11-26 10:30:24 was planning to migrate to lxc, now changed of plan, migrate to ARM + lxc 2015-11-26 10:30:31 :) 2015-11-26 10:30:34 I'm using qemu-kvm all the time. This is my first time with a alpine host. 2015-11-26 10:31:06 I don't see the need of VM when the host and all the guests are private 2015-11-26 10:31:39 when alpine is both host and guest i use lxc 2015-11-26 10:31:56 (unless i want test/devel kernel or boot scripts) 2015-11-26 10:32:09 migrating to ARM will save me 50~100W/h 2015-11-26 10:32:15 I like to crypt my vms with LUKS :D 2015-11-26 10:32:44 ncopa: maybe you'll know that. What's the point of the Generic ArM alpine tar.gz on the download page? Say I have a qemu for armhf and I want to install Alpine into it, what should I do? 2015-11-26 10:33:07 pluesch: when you are doing anything near 4chan, you would want a lot :D 2015-11-26 10:33:12 I'd like to avoid going all the way to Raspberry Pi emulation if possible. 2015-11-26 10:34:03 xming: does your current PC have moving parts? if it does, start with removing them, that's the most power you'll gain :P 2015-11-26 10:34:07 skarnet: iirc, i formated a mmc as fat (wandboard required it) 2015-11-26 10:34:15 then i untarred the arm tarball 2015-11-26 10:34:27 found the proper uboot image 2015-11-26 10:34:32 or similar 2015-11-26 10:34:45 mmc? 2015-11-26 10:34:48 skarnet: I have a power hungry AMD CPU in it, and a power hungry AMD GPU 2015-11-26 10:35:06 skarnet: internal storage 2015-11-26 10:35:06 wand2:~# ls /media/mmcblk0p1/ 2015-11-26 10:35:06 apks cache wand2.apkovl.tar.gz 2015-11-26 10:35:06 boot extlinux 2015-11-26 10:35:06 boot.old u-boot 2015-11-26 10:35:37 skarnet: there is a readme.txt in u-boot 2015-11-26 10:36:17 http://sprunge.us/YJFj 2015-11-26 10:37:28 ncopa: that's all fine and dandy but that doesn't tell me how to boot the qemu. 2015-11-26 10:37:28 first partition was created a bit in 2015-11-26 10:38:02 the point is, I'm not (yet) tied to a hw architecture. 2015-11-26 10:38:15 skarnet: try dd uboot to the block device 2015-11-26 10:38:18 I just want the simplest possible thing to boot a qemu-system-arm on. 2015-11-26 10:38:59 hmm qemu doesn't need uboot apparently 2015-11-26 10:40:15 skarnet: what happens if you: qemu-system-arm -kernel boot/vmlinuz-grsec -initrd boot/initramfs-grsec 2015-11-26 10:40:36 let me try 2015-11-26 10:40:45 with the kernel shipped in tarball 2015-11-26 10:41:05 this is actually interesting 2015-11-26 10:41:12 skarnet: just use qemu's direct kernel boot 2015-11-26 10:41:14 i'd like figure out how to do that myself 2015-11-26 10:41:34 No machine specified, and there is no default. 2015-11-26 10:41:34 Use -machine help to list supported machines! 2015-11-26 10:41:44 boot kernel + initrd from the command line, use a block device as rootfs 2015-11-26 10:42:09 if it boots you into initramfs then it is a good start 2015-11-26 10:42:15 so I'll pick a random machine in the list... 2015-11-26 10:43:05 see if qemu has a device mathcing http://wiki.alpinelinux.org/wiki/Alpine_on_ARM 2015-11-26 10:43:27 ACTION compiles qemu with arm target 2015-11-26 10:44:32 "-machine virt" seems to be what I want 2015-11-26 10:44:49 ... now THIS is fun 2015-11-26 10:44:58 running for some time... 2015-11-26 10:45:05 VNC server running on `127.0.0.1:5900' 2015-11-26 10:45:05 qemu: fatal: Trying to execute code outside RAM or ROM at 0x08000000 2015-11-26 10:45:32 then a big register dump and they are almost all 0 2015-11-26 10:46:18 ncopa: is there a sign that will definitely tell me I'm in the initramfs? 2015-11-26 10:46:52 for arm, i dont know 2015-11-26 10:47:07 maybe you could add kernel opt for serial console? 2015-11-26 10:47:44 I thought the console was by default the terminal I run qemu in 2015-11-26 10:47:58 and run qemu with -nographic 2015-11-26 10:49:00 it's running, but I'm not seeing anything 2015-11-26 10:49:10 qemu-system-arm -nographic -machine virt -kernel ./bo 2015-11-26 10:49:10 ot/vmlinuz-grsec -initrd ./boot/initramfs-grsec 2015-11-26 10:49:26 upside: it hasn't crashed yet 2015-11-26 10:50:48 downside: it's eating a full CPU 2015-11-26 10:51:05 ... so it's probably stuck in a loop somewhere and doesn't work 2015-11-26 10:51:47 and ^C doesn't stop it, I had to kill it by hand. Wheee. 2015-11-26 10:51:50 I wonder about -machine virt 2015-11-26 10:52:10 well I'll try something else. It's fun, and I'm actually being paid for it. 2015-11-26 10:52:13 if that's for running kvm arm guests on arm host 2015-11-26 10:57:31 what annoys me is that Ubuntu can do it with debootstrap 2015-11-26 10:58:16 ah, I have an idea. Hold on. 2015-11-26 11:08:31 Uncompressing Linux... done, booting the kernel. 2015-11-26 11:14:03 i am soooo glad i am running alpine linux on my servers ... my gateway is currently under heavy attack but i dont really worry about it ... 2015-11-26 11:14:09 alpine linux stays strong :) 2015-11-26 11:15:14 that's the power of knowing what you're doing with software :) 2015-11-26 11:16:00 yes :) 2015-11-26 11:16:21 i am just sitting back and watching tail -f on my messages log 2015-11-26 11:16:38 its fun to see what people try to get into ssh 2015-11-26 11:16:57 most of the attacks are sooo old ... you would need a 5 year old suse system to be affected ... 2015-11-26 11:17:07 ssh1 attacks? really?? 2015-11-26 11:17:26 you'd be surprised... those attacks do work on some industrial systems >.> 2015-11-26 11:19:23 yeah ... 2015-11-26 11:19:41 austrian train services had to shut down last week because there windows 3.1 system crashed!!!!!! 2015-11-26 11:26:39 ncopa: i really like the alpine package manager 2015-11-26 11:27:02 but i have one issue with it, the fact that i cannot inspect all files of a package without installing it 2015-11-26 11:29:47 the alpine ARM doesn't boot with qemu 2015-11-26 11:34:03 qemu-system-arm -M vexpress-a15 -kernel tt/boot/vmlinuz-grsec -initrd tt/boot/initramfs-grsec -dtb tt/boot/dtbs/vexpress-v2p-ca15_a7.dtb -serial stdio 2015-11-26 12:01:48 anyone found that tail -f stops working after a few days? 2015-11-26 12:02:00 leo-unglaub: that is the price for having a light index 2015-11-26 12:02:16 if we included the entire file lists in the index then it would grow *huge* 2015-11-26 12:02:26 there is a database on network though 2015-11-26 12:02:29 it's probably logrotate lol - ignore me, i need food 2015-11-26 12:02:31 pkgs.alpinelinux.org 2015-11-26 12:03:42 ncopa: a packages still not building? 2015-11-26 12:03:56 ? 2015-11-26 12:04:05 webkit2gtk 2015-11-26 12:04:07 on arm 2015-11-26 12:04:08 s/a/any? 2015-11-26 12:04:14 asm issues 2015-11-26 12:04:22 arm asm isues 2015-11-26 12:04:34 we can check what fedora and others do with those 2015-11-26 12:04:36 i don't think I can help there 2015-11-26 12:04:49 also midori seems broken 2015-11-26 12:04:55 and alsamixer segfaults 2015-11-26 12:05:02 i am currently working on fixing modloop 2015-11-26 12:05:10 i think rc1 will be out later today 2015-11-26 12:05:18 looks promising 2015-11-26 12:10:49 yeah, self compiled kernel boots 2015-11-26 12:10:57 yes, midori is all over the place. i thought it was my compositor, but it behaves strangly with just X running 2015-11-26 12:11:38 now I need to find what's config makes the diffrence 2015-11-26 12:25:03 i think its webkit 2015-11-26 12:25:40 xming: self-compiled kernel as in you cross-compile a kernel yourself for armhf? 2015-11-26 12:25:51 instead of using the provided vmlinuz? 2015-11-26 12:26:28 k0r10n: yes 2015-11-26 12:26:34 skarnet: yes 2015-11-26 12:27:33 whee \o/ 2015-11-26 12:27:37 modloop works 2015-11-26 12:27:43 i think we may have a winner 2015-11-26 12:28:09 now, if i have done things correctly, booting isohybrid as disk should just work 2015-11-26 12:28:31 nice 2015-11-26 12:29:25 Booting from Hard Disk... 2015-11-26 12:29:38 \o/ 2015-11-26 12:29:42 modloop works 2015-11-26 12:29:47 isohybrid works 2015-11-26 12:30:10 \o/ 2015-11-26 12:30:18 localhost:~# /etc/init.d/modloop stop 2015-11-26 12:30:18 * WARNING: you are stopping a sysinit service 2015-11-26 12:30:18 * Unmounting /.modloop ... [ ok ] 2015-11-26 12:30:18 * Unmounting /media/sda ... [ ok ] 2015-11-26 12:30:35 it does not depend on alpine_dev on boot prompt 2015-11-26 12:31:23 xming: could you please post your kernel .config? I'd like to try it 2015-11-26 12:38:03 skarnet: ARCH=arm CROSS_COMPILE=armv7a-hardfloat-linux-gnueabi- make -j12 vexpress_defconfig 2015-11-26 12:39:03 ncopa: I am trying to apply grsec/pax patches to kernel 4.1.12 but they both fail 2015-11-26 12:40:39 xming: thanks, but it's the vexpress_defconfig I'm interested in. Even better if you have a versatilepb_defconfig :) 2015-11-26 12:40:55 skarnet: it's default config 2015-11-26 12:41:03 skarnet: included in the kernel source 2015-11-26 12:41:11 ah, ok. thanks. 2015-11-26 12:41:38 where did you get the toolchain though? is there an Alpine package? 2015-11-26 12:41:48 skarnet: ls arch/arm/configs/ 2015-11-26 12:42:00 skarnet: I am using gentoo on my desktop 2015-11-26 12:50:36 ncopa: did you read the backlog? 2015-11-26 12:50:42 about the mdcrypt boot issue? 2015-11-26 12:51:22 its crypttab or /etc/init.d/dmcrypt issue? 2015-11-26 12:51:39 i havent looked closer on it 2015-11-26 12:51:58 currently i'm focusing on initramfs and modloop 2015-11-26 12:52:01 so i can ship rc1 2015-11-26 12:55:12 ok 2015-11-26 12:55:21 modloop init.d script can find the modloop now 2015-11-26 12:55:22 how strange that abuild can patch them cleanly 2015-11-26 12:56:01 but in theory, you know the path to modloop image when generating the boot media 2015-11-26 12:56:14 so we should proably require that it is added as boot option 2015-11-26 12:56:46 we need specify kernel image and initrd image, so we should probably also specify the modloop image too 2015-11-26 13:21:41 this also means that we can get rid of isolinux.cfg 2015-11-26 13:21:43 weet 2015-11-26 13:21:46 sweet 2015-11-26 13:30:36 clandmeter: do we still support unionfs modloop? 2015-11-26 13:31:10 introduced here: http://git.alpinelinux.org/cgit/aports/commit/?id=16dbbaafe3bf7a3b9516e61b2cfe447eb7e56a86 2015-11-26 13:51:35 anything we should remove? http://git.alpinelinux.org/cgit/alpine-iso/tree/alpine-extended.packages 2015-11-26 13:51:45 i'd like to remove samba 2015-11-26 13:54:33 do we drop *asterisk* but keep freeswitch? 2015-11-26 13:54:50 or we drop both... 2015-11-26 13:55:16 keep freeswitch please 2015-11-26 13:55:31 by drop do you mean community? 2015-11-26 13:55:40 or git rm? 2015-11-26 13:55:48 no, drop from alpine-extended iso image 2015-11-26 13:55:56 anything we should include? 2015-11-26 13:56:25 oh then nvm, you can exclude both if it were up to me 2015-11-26 15:35:38 please hold your commits for a sec 2015-11-26 15:35:47 i'm tagging 3.3.0_rc1 2015-11-26 15:45:36 \o/ 2015-11-26 15:48:15 http://nl.alpinelinux.org/alpine/v3.3/releases/x86_64/ 2015-11-26 15:48:25 i had a minor bug in release script 2015-11-26 15:50:26 sweet 2015-11-26 15:54:19 ok you can commit again 2015-11-26 16:26:16 nice 2015-11-26 17:00:46 finally got a kernel to show something starting with the config from alpine 2015-11-26 17:00:55 then I got this 2015-11-26 17:01:03 grsec: Segmentation fault occurred at 0000001c in /init[init:1] uid/euid:0/0 gid/egid:0/0, parent /[swapper/0:0] uid/euid:0/0 gid/egid:0/0 2015-11-26 17:21:41 xming: I'm following in your tracks 2015-11-26 17:21:58 I finally managed to boot my Linux kernel under qemu-system-arm! woot. 2015-11-26 17:22:02 ncopa: \o/ 2015-11-26 17:22:14 algitbot: woot 2015-11-26 17:22:24 nice to know :) 2015-11-26 17:22:37 skarnet: try the raspberry pi zero next? 2015-11-26 17:23:05 no, I first need to make a rootfs and do something useful. Currently I have a booting kernel, but a crash when it looks for a rootfs :) 2015-11-26 17:23:43 ncopa: re packages: http://meije.alpinelinux.uk/di01hicr5x/view 2015-11-26 17:24:03 use http://meije.alpinelinux.uk/di01hicr5x/edit if you want to comment 2015-11-26 17:32:45 skarnet: with vexpress_defconfig or alpine's config? 2015-11-26 17:33:43 with versatilebp_defconfig, I'm emulating a versatilebp 2015-11-26 17:34:05 pb* 2015-11-26 17:34:26 boot with debian rootfs, use apktoolstatic, chroot and be done 2015-11-26 17:34:47 I am trying with alpine's .config 2015-11-26 17:35:00 The last time I needed anything Debian to get things done was 10 years ago and it's not changing today 2015-11-26 17:35:47 I'm the guy who builds rootfses, not the guy who uses them 2015-11-26 17:36:34 interesting bit of Linux trivia here: 2015-11-26 17:36:42 ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000004 2015-11-26 17:36:50 (and several minutes later...) 2015-11-26 17:36:57 random: nonblocking pool is initialized 2015-11-26 17:37:12 self healing kernel yeah 2015-11-26 17:37:24 so a kernel panic may actually leave several threads running... 2015-11-26 17:41:05 woot 2015-11-26 17:41:08 w00t 2015-11-26 17:41:23 algitbot: you are 1337 2015-11-26 17:41:48 1337 algitbot 2015-11-26 17:41:56 algitbot 1337 2015-11-26 17:42:02 :D 2015-11-26 17:42:03 algitbot: you are not 1337 2015-11-26 17:42:18 algitbot: you are a bot 2015-11-26 17:42:34 reverse-engineering is hard :P 2015-11-26 17:43:18 algitbot !say me? thanks! :) 2015-11-26 17:43:22 :p 2015-11-26 18:44:32 knows somebody if steamcmd is runnable on alpine? :) 2015-11-26 18:50:47 xming, ncopa: I'm back .... so far my vm runs stable without smp and threads.... 2015-11-26 18:52:02 it wasn't under that much load tho. will try to get it a bit more loaded now 2015-11-26 19:12:24 regression, http://c-ares.haxx.se/mail/c-ares-archive-2014-11/0003.shtml ? 2015-11-26 19:13:33 ok, [w]doze related 2015-11-26 19:17:05 hum... i don't want to use a different os for a steam game server container :( 2015-11-26 19:19:18 oh god.... this damn steamcmd tool is only 32bit oO 2015-11-26 19:21:56 mosez: isn't steamcmd opensource? 2015-11-26 19:31:00 hum... anyway it requires glibc... 2015-11-26 19:42:49 looks like i have to build the container on another distro -.- 2015-11-26 19:45:17 # ./steamcmd.sh +login anonymous +force_install_dir testing +app_update 211820 +quit 2015-11-26 19:45:20 ./steamcmd.sh: line 29: /root/linux32/steamcmd: No such file or directory 2015-11-26 19:45:25 # ll /root/linux32/steamcmd 2015-11-26 19:45:26 -rwxrw-r-- 1 root root 5.4M Feb 5 2013 /root/linux32/steamcmd* 2015-11-26 19:45:29 lol 2015-11-26 19:45:59 maybe related to 64bit <=> 32bit -.- 2015-11-26 19:55:39 welcome to highly available freenode 2015-11-26 19:56:48 is there a way to install 32bit libs and tools on an 64bit alpine? 2015-11-26 22:19:16 xming, ncopa_, ncopa: guest crashed again... without smp and thrads 2015-11-26 22:19:24 -.- 2015-11-26 22:19:51 upgrading to edge now 2015-11-26 23:09:20 ncopa_: I am trying to improve the abuild man page hardlink compression stuff, wouldn't it be better way easier to simply replace hardlinks with symlinks? 2015-11-26 23:25:55 nmeum: not sure what it's about exactly, but symlinks take inodes whereas hardlinks don't, it's the main reason why people use hardlinks these days 2015-11-26 23:26:22 I know the difference between hardlinks and symlinks 2015-11-26 23:27:50 I'm sure you do, and I should definitely read about what you're trying to do before jumping in 2015-11-26 23:27:52 skarnet: the problem is that we automatically compress man pages with gzip. gzip complains if a file has > 1 links and doesn't run in that case, thus you need to remove all hardlinks to that file and update them later before compressing the man pages 2015-11-26 23:28:42 the relevant code is here: http://git.alpinelinux.org/cgit/abuild/tree/abuild.in#n1407 2015-11-26 23:29:01 I'd say the problem is "a man page as a hardlink" in the first place XD 2015-11-26 23:29:05 has* 2015-11-26 23:29:30 the problem with that code is that it assumes that inodes are numbered in a decsinding way which isn't the case on fancy 'tree-based' filesystems like btrfs or zfs 2015-11-26 23:30:40 honestly, I question the utility of checking for hard links among man pages 2015-11-26 23:31:11 if a package does that, patch it so it makes symlinks instead 2015-11-26 23:31:24 that's an option as well 2015-11-26 23:31:32 but it would be more work 2015-11-26 23:31:41 a man page has no business being hard linked 2015-11-26 23:31:52 have you benchmarked how many times this code is called? 2015-11-26 23:32:01 nope 2015-11-26 23:32:11 but I know quite a few packages which use hardlinks 2015-11-26 23:32:15 for man pages 2015-11-26 23:32:23 bind, nvi, mdocml (iirc), … 2015-11-26 23:32:39 write a cron job to mail insults to the developers 2015-11-26 23:33:43 :) 2015-11-26 23:34:45 I also have an abuild patch which removes all man page hardlinks and replaces them with symlinks http://sprunge.us/DVca 2015-11-26 23:35:06 didn't test this a lot so far, but it seems to work… 2015-11-26 23:35:13 even though doing it that way sucks as well 2015-11-26 23:35:16 it doesn't make the same assumption about inode order? 2015-11-26 23:35:20 nope 2015-11-26 23:35:29 then it's less sucky :P 2015-11-26 23:35:53 yeah, it's less sucky but it still sucks 2015-11-26 23:35:56 hardlinks in general suck 2015-11-26 23:36:10 (until some software does something really evil that DEPENDS on some file being a hard link to the man page) 2015-11-26 23:36:31 (in which case my answer would be: nuke the package from the aports) 2015-11-26 23:37:01 exactly 2015-11-26 23:37:49 I will this my patchset a bit more and send it to the -devel list later on… 2015-11-26 23:37:53 s/this/test/ 2015-11-26 23:52:44 ncopa_: I just sent a few abuild patches to -devel, would be nice if you could take a look at them before the weekend 2015-11-27 06:19:42 xming, ncopa_, ncopa: vm also crashes if host is on edge 2015-11-27 07:34:51 do we have an example where the subpackage in the APKBUILD, rather than only _move_ files, it needs to download a file with source and then compile it? 2015-11-27 07:47:28 pluesch: still no idea when it crashes? Ubnder CPU load, IO load? 2015-11-27 07:49:20 fcolista: just put the flie in the source, it will get unpacked, in you sub() cd; confingure;make 2015-11-27 08:24:28 morning 2015-11-27 08:25:49 xming: currently at work already again. I'll leave the vm online for today with no load at all and see if it crashes again. 2015-11-27 08:26:48 pluesch: can't you use something to minitor it? 2015-11-27 08:27:02 so you will know exactly when it crash and the condiftion it was in 2015-11-27 08:28:11 ScrumpyJ1ck: morning, are you familiar with how alpine boots with initrd, modloop, ...? 2015-11-27 08:28:12 xming: yeah, would need to find "something".... 2015-11-27 08:28:12 what exactly should I measure? 2015-11-27 08:29:31 pluesch: load avg, cpu and io load, maybe even network load 2015-11-27 08:29:32 well I don't think it's CPU, but could still be virt-block/net 2015-11-27 08:30:37 pluesch: nagios/monit/zabbix, or even just a cron job gathering those output and email yourself 2015-11-27 08:31:38 sysstat, iostat, ... 2015-11-27 08:33:33 I am having issues that the kernel cannot find root when trying to boot ARM with Qemu 2015-11-27 08:33:59 xming: I'll try :) Thank you for helping me. 2015-11-27 08:34:01 with this "console=ttyAMA0 modules=loop,squashfs,sd-mod,usb-storage alpine_dev=mmcblk0p1 modloop=modloop-grsec" 2015-11-27 08:36:11 is there a way to get shell before pivot root? 2015-11-27 09:05:38 xming: so nice of you to discover problems before I get to it 2015-11-27 09:06:16 if you have trouble mounting a rootfs, I'll try bypassing the initrd entirely 2015-11-27 09:06:28 initramfs* 2015-11-27 09:07:13 I don't know what's in the initramfs-grsec thingy and I hate debugging initramfs, so let's see if I can make the kernel find a rootfs directly 2015-11-27 13:03:40 skarnet: w/o initrd it's easy, it even boot nicely with a debian initramfs, I want to get Alpine's initramfs-grsec+modloop-grsec working 2015-11-27 13:04:08 I am having issues that the kernel cannot find root when trying to boot ARM with Qemu 2015-11-27 13:04:16 is there a way to get shell before pivot root? 2015-11-27 13:45:35 ncopa / clandmeter / nangel: I'm not sure who is responsible for dl-3 mirror, but it is not properly syncing 2015-11-27 13:45:59 specifically, I updated acf-weblog on alpine 2.7 yesterday, and it does not yet have the new package 2015-11-27 13:58:20 Does the default Alpine kernel support TAP? 2015-11-27 13:58:33 # tunctl -t tap0 2015-11-27 13:58:33 tunctl: can't open '/dev/net/tun': No such file or directory 2015-11-27 13:58:49 or is it just some udev thing I should do manually? 2015-11-27 14:01:41 modprobe? 2015-11-27 14:02:01 it does, you need to modprobe 2015-11-27 14:02:09 ok, what's the module name? XD 2015-11-27 14:02:19 tun, iirc 2015-11-27 14:02:39 or tap 2015-11-27 14:02:50 tun works apparently, thanks 2015-11-27 14:02:55 yw 2015-11-27 14:03:13 the kernel autoloads some modules when you try to open a nonexistent device 2015-11-27 14:03:24 where is this configured? I'd like to autoload tun, if possible 2015-11-27 14:04:14 you need to specify it into /etc/modules 2015-11-27 14:04:43 AIUI, /etc/modules is autoload all the time, not on demand 2015-11-27 14:04:56 but nvm 2015-11-27 14:05:10 not sure if network scripts autoload tun if specified into interfaces 2015-11-27 14:06:59 i'm not aware of Alpine kernel autoloading needed modules on-demand 2015-11-27 14:16:23 clandmeter, you around? 2015-11-27 16:06:23 fcolista: i am 2015-11-27 16:11:53 ncopa / clandmeter / nangel: dl-3 mirror is now properly syncing 2.7 packages (thanks to nangel) 2015-11-27 17:09:40 ncopa: based on the BusyBox mailing list thread, it looks like BusyBox is more against patching su to ignore password-less root, you think a dummy unusable password hash like "!" is next best hack? 2015-11-27 18:26:39 http://git.alpinelinux.org/cgit/aports/tree/main/alpine-baselayout/APKBUILD#n45 is this so that ISO image can login as root? 2015-11-27 18:28:08 i wonder if it could be in another package further downstream, that isn't dependent of alpine-base, but still in ISO so that we can build LXC and Docker images without the hack 2015-11-27 18:58:13 andyshinn: I think I prefer the approach of not shipping suid su 2015-11-27 19:00:09 the only thing swaying me towards disabling password, is that other images also already do this 2015-11-27 19:00:24 but afaik it is indeed so that can log in as root from the iso 2015-11-27 19:02:11 I slightly suspect busybox login is even less likely to take a patch for this 2015-11-27 19:02:25 the problem right now, is that disabling suid means doing it on /bin/bbsuid, which may affect other suid bins 2015-11-27 19:02:34 the scope of testing is larger 2015-11-27 19:03:03 well busybox-suid is already moved into a separate package 2015-11-27 19:03:30 would it be too much to expect that if you're installing busybox-suid that you have already managed to set a password 2015-11-27 19:03:43 sorry that line came out weird because I was distracted 2015-11-27 19:04:32 hmm, while i was researching i didn't see any separate package for the bbsuid bin 2015-11-27 19:04:51 when were you researching? it's ... relatively recent I think 2015-11-27 19:05:15 https://pkgs.alpinelinux.org/contents?pkgname=busybox-suid&arch=x86_64&repo=main 2015-11-27 19:06:16 since ~a month ago to the day 2015-11-27 19:07:20 no, very recently... maybe i didn't catch that, so the best solution is to just remove busybox-suid package maybe 2015-11-27 19:07:57 in my test system though i don't have busybox-suid, but the bbsuid program is still there... 2015-11-27 19:14:40 right now, alpine-base depends on busybox-suid, i wonder if a reasonable solution is to remove this dependency from alpine-base and move further down the line (or just have it manual addition) 2015-11-27 19:19:00 andyshinn: fwiw I dropped alpine-base for manually selecting packages in my docker images 2015-11-27 19:20:18 looks like I currently have pkg=alpine-baselayout,alpine-keys,apk-tools,musl-utils 2015-11-27 19:43:42 hmm, i guess libc-utils == musl-utils? 2015-11-27 19:44:38 i think there is a PR where i mentioned we will try to remove alpine-base and just do baselayout + other essential in images starting with 3.3 2015-11-27 19:48:41 https://github.com/gliderlabs/docker-alpine/issues/103 found it 2015-11-27 19:49:18 ah yeah that's where -p came from 2015-11-27 19:49:26 yea, in #32 is where it started 2015-11-28 09:41:23 xming, ncopa: i have an atop log now for my virtual machine crash problem. it doesn't show anything strange at the end where the crash happend 2015-11-28 09:41:43 will try a different os for the host now 2015-11-28 09:42:26 pluesch: so it wasn't under heavy load? 2015-11-28 09:47:52 it was under load but not under heavy load 2015-11-28 09:48:11 both i/o and cpu 2015-11-28 09:48:47 network should be the same all the time too... but I guess I can test this with some network stress test scripts 2015-11-28 09:50:27 pluesch: try bonnie or similar for IO stress test, and run some CPU stress test 2015-11-28 09:51:23 pluesch: what kind of network IO is it? Are you just send http get/post and getting replies? 2015-11-28 09:52:52 exactly 2015-11-28 09:52:58 via https to be more specific 2015-11-28 09:58:01 okay... we need some stress tools packaged for alpine :3 2015-11-28 10:35:28 pluesch: you want to be able to reproduce this realiably 2015-11-28 10:55:16 xming: stress tested cpu, memory and io so far with "stress".... nothing helped 2015-11-28 10:55:39 i hate those kind of bugs ^^ 2015-11-28 10:55:55 every stresstest was running for 10min 2015-11-28 10:57:38 pluesch: just something obvious, you don't have anything running that would shutdown the guest? 2015-11-28 10:58:46 pluesch: are you running the VM as root or as normal user? 2015-11-28 10:58:51 root 2015-11-28 10:59:09 I don't think I have running something that shutdowns it 2015-11-28 10:59:26 the script is the exact same that I'm using on other nodes 2015-11-28 10:59:35 and on the host? 2015-11-28 10:59:47 somehing like killall qemu :D 2015-11-28 11:00:44 the host is alpine edge with nothing but qemu and currently atop installed and running 2015-11-28 11:01:13 iperf from host to vm also did not cause any crashes 2015-11-28 11:12:56 pluesch: you are running qemu from the command line right? Not using something like libvirt? 2015-11-28 11:30:09 xming: yes 2015-11-28 11:30:27 inside a tmux 2015-11-28 20:58:52 i have no idea how to debug this further 2015-11-28 21:07:30 what I'm going to try is upgrading the guest to edge too 2015-11-29 07:52:53 <_mjones> Aw, Alpine aren't enabling /proc/config.gz? Debian's reasons for refusing to do so are awfully spurious. 2015-11-29 09:30:16 there is a module for that 2015-11-29 09:30:21 it's not autoloaded 2015-11-29 14:51:47 ncopa: Sorry, was busy these days… do you still need something tested out? 2015-11-30 08:21:15 pluesch: do you have a simple way for me to try reproduce? 2015-11-30 08:44:18 chris|: i think i applied wrong libxml patch for 2.7-stable 2015-11-30 08:44:26 i think i applied the first 2015-11-30 08:44:32 before realizing that there was a v2 2015-11-30 08:44:36 what was the difference? 2015-11-30 10:10:45 ncopa: no, I don't found a way to make it easy reproducable. sometime it took like +8 hrs for the vm to crash. sometime it was almost instantly as I started my scripts. 2015-11-30 10:11:13 pluesch: what do i need to test it? 2015-11-30 10:11:32 I've tried find a way with "stress" to make it easy reproducable. but stress testing with "stress" didn't seem to trigger the "bug". 2015-11-30 10:11:43 understand 2015-11-30 10:11:50 if i create an alpine guest 2015-11-30 10:11:53 and just boot it 2015-11-30 10:11:56 and let it idle 2015-11-30 10:11:59 will it trigger the issue? 2015-11-30 10:12:05 ncopa: a lot of space and my scripts + redis and some python deps 2015-11-30 10:12:08 ncopa: sadly not... 2015-11-30 10:12:53 I've switched host to voidlinux now, and the guest seem to work stable. I'll get another node soon. If you want I can install Alpine there again for some more testing. 2015-11-30 10:13:52 void linux and musl? 2015-11-30 10:13:53 or glibc 2015-11-30 10:14:00 what version of qemu? 2015-11-30 10:14:34 glibc, afaik. I'm not so familiar with it yet but I've used the .iso without "musl" in it. 2015-11-30 10:15:10 and vanilla kernel? 2015-11-30 10:15:18 version=2.4.1 2015-11-30 10:15:42 doesn't seem to be kernel related. I've tested both (vanilla and grsec) with alpine. 2015-11-30 10:15:46 ok 2015-11-30 10:15:50 ah yes, you said 2015-11-30 10:16:11 and usually if grsec is to blame, it will write something to dmesg 2015-11-30 10:16:15 true 2015-11-30 10:16:29 so the qemu that had the issue was 2.4.0.1, in alpine v3.2? 2015-11-30 10:16:52 i wonder if alpine 3.3.0_rc1 has the issue 2015-11-30 10:17:29 I tested it on v3.2 and edge. So it shouldn't be the qemu version. 2015-11-30 10:17:34 On edge it's currently pkgver=2.4.1 2015-11-30 10:17:53 ok 2015-11-30 10:18:01 so it happens with qemu-2.4.1 too 2015-11-30 10:18:18 yep, at least with alpine as host. 2015-11-30 10:18:35 really strange... so it can only be something related to musl? 2015-11-30 10:19:05 seems like that 2015-11-30 10:19:35 I can tary voidlinux with musl on the other node once I have it. 2015-11-30 10:19:44 would be good to make the musl devs aware on the issue 2015-11-30 10:19:51 that would be nice 2015-11-30 10:20:20 if it happens with voidlinux qemu with musl, then it may indicate that it is related musl 2015-11-30 10:20:38 we could also compare the configure/build flags 2015-11-30 10:25:43 ncopa: the v1 has an APKBUILD.(orig|rej) file included... :( 2015-11-30 10:26:39 ok 2015-11-30 10:26:41 sorry about that 2015-11-30 10:27:17 so git rm *.orig *.rej should do it 2015-11-30 10:27:22 yes 2015-11-30 10:27:31 besides that, the patches should be identical 2015-11-30 10:29:17 anyone familar with voidlinux here? do they really package all packages for both musl and glibc? It doesn't seem to be this way when I look at their repository. (http://repo.voidlinux.eu/current/) ... guess I should better ask this in their # maybe. 2015-11-30 10:29:40 pluesch: try in #xbps 2015-11-30 10:37:11 thank you. I have found it out I guess. They have completely different urls for musl and glibc packages. I'll report back once I have tested it with musl on void. 2015-11-30 10:50:36 chris|: i just want to say: big thank you 2015-11-30 10:50:43 for the *-stable patches 2015-11-30 10:50:47 appreciate 2015-11-30 10:50:51 sure thing 2015-11-30 10:51:27 chris|: are you registered to bugs.a.o? 2015-11-30 10:54:12 not sure actually :D 2015-11-30 10:54:16 so probably not 2015-11-30 10:54:18 seems not 2015-11-30 10:54:37 could you please do? 2015-11-30 10:54:53 then you could help resolve bugs too 2015-11-30 10:55:12 normally it is enough to add a line in commit message: "fixes " 2015-11-30 11:08:09 done 2015-11-30 11:18:04 ncopa: firefox is broken on x86. has anyone mentioned it? 2015-11-30 11:18:57 pluesch, ncopa: seabios might make the difference too 2015-11-30 11:19:18 (in 3.3) 2015-11-30 11:19:34 seabios oO 2015-11-30 11:20:31 yeah, IIRC there were bugs (dissapearing guests) with some versions of seabios, abeit not recent version 2015-11-30 11:20:49 still need to the with the same versions of seabios to be sure 2015-11-30 12:22:29 ScrumpyJack: x86 firefox has been broken since 3.2 or maybe even 3.1 2015-11-30 12:22:32 i dont know why 2015-11-30 12:24:08 ncopa: regarding my abuild patches: I didn't want those in for 3.3 in the first play, just merge them (if you want to) after 3.3 is released 2015-11-30 12:24:14 s/play/place/ 2015-11-30 12:24:16 ok 2015-11-30 12:24:19 agree 2015-11-30 12:24:31 it might make sense to fix btrfs/zfs though 2015-11-30 12:24:36 since its broken 2015-11-30 12:25:37 that's what the patch does…but doing that while retaining the hardlinks is actually kind of hard 2015-11-30 12:27:07 https://github.com/gentoo/portage/blob/master/bin/ebuild-helpers/ecompressdir#L209-L211 2015-11-30 12:27:10 ^ gentoos approach 2015-11-30 12:30:33 hm 2015-11-30 12:30:43 so we could just break the hardlinks i suppose 2015-11-30 12:30:46 ala gentoo 2015-11-30 12:31:55 that's pretty dirty, imho but yeah we could try that 2015-11-30 12:34:04 iirc we have only one package that acutally use hardlinks 2015-11-30 12:34:12 and it is tcl 2015-11-30 12:41:29 what are your thoughts on the contents on alpine-extended? 2015-11-30 12:42:44 several people have made changes to http://meije.alpinelinux.uk/di01hicr5x 2015-11-30 12:43:04 (the list of proposed additions/removals 2015-11-30 12:46:49 i have removed acf-* 2015-11-30 12:47:00 asterisk-* 2015-11-30 12:47:31 kamailio-* 2015-11-30 12:48:12 i added lxc 2015-11-30 12:48:22 and gnu ta 2015-11-30 12:48:24 gnu tar 2015-11-30 12:48:34 i added btrfs-tools 2015-11-30 12:48:39 btrfs-progs 2015-11-30 12:49:03 i removed postgresql too 2015-11-30 12:49:18 and i added nfs-utils 2015-11-30 12:57:10 i quite liked the idea of having acf in the iso, it's alpine linux specific, and being in the iso might make people notice it more 2015-11-30 13:01:45 ncopa: no, we have more packages that use hardlinks 2015-11-30 13:01:49 bind for example 2015-11-30 13:01:57 and mdocml (iirc) 2015-11-30 13:02:48 but yeah there are not many packages that use hardlinks for man pages that's why I suggested replacing them with symlinks in the first place 2015-11-30 13:55:53 ugh -_- 2015-11-30 13:56:20 just realized my "make strip" didn't work with cross-compilation, which means stuff will break when you build for arm. 2015-11-30 13:56:38 I'll fix that some time this week... 2015-11-30 14:00:31 is valery kartel here? 2015-11-30 14:00:35 the abuild patch does not apply 2015-11-30 14:35:01 i think lxc might be broken :( 2015-11-30 14:35:49 lxc-1.1.5-r1 2015-11-30 14:39:46 what happens to buildlog? 2015-11-30 14:45:17 ok, my bad. I apk upgraded without rebooting 2015-11-30 14:50:19 why do you need to reboot for an lxc upgrade? 2015-11-30 14:50:25 sounds…wrong 2015-11-30 14:50:42 a new kernel got thrown in with the upgrade 2015-11-30 14:56:27 what is loading the ipv6 module? 2015-11-30 15:54:51 sigh 2015-11-30 15:55:00 i think we need think of QA after v3.3 2015-11-30 15:55:19 i really really really want avoid bad packaging 2015-11-30 15:55:39 i spent half day with sec fixes for nsrp and nss 2015-11-30 15:55:44 they should be simple 2015-11-30 15:55:49 they are relatively simple 2015-11-30 15:55:58 but sqlite -dev had wrong depends 2015-11-30 15:56:10 so i had to clean up someone garbage 2015-11-30 15:56:23 problem is that when the garbage gets into stable 2015-11-30 15:56:33 then i have to clean up the same garbage 4 times 2015-11-30 15:56:54 so i am thinking that we need be much more strict in what we accept 2015-11-30 16:04:46 yeah, I totally agree on that 2015-11-30 16:04:53 we should have some kind of package guidelines 2015-11-30 16:05:14 and than we could extend abuild sanitycheck to make sure that packages always follow those guidelines 2015-11-30 16:06:13 +2 2015-11-30 16:08:46 we might as well do what voidlinux does and setup a CI to run tests automatically on suggested patches and reject them automatically if they don't pass 2015-11-30 16:08:54 but the first step really is to write guidelines 2015-11-30 16:09:07 s/to write/writing/ 2015-11-30 16:09:56 yes 2015-11-30 16:13:19 perhaps expand on this? http://wiki.alpinelinux.org/wiki/Package_policies 2015-11-30 16:21:09 i have a newbie git/patch question 2015-11-30 16:21:59 my wbar patch for a new aport got rejected (checksum error) but i deleted the orignal files i created to make the patch 2015-11-30 16:22:36 how can i create the APKUILD file and patches again from the git patch? 2015-11-30 16:23:13 it's here -> http://patchwork.alpinelinux.org/patch/913/ 2015-11-30 17:05:52 ScrumpyJack: http://patchwork.alpinelinux.org/patch/913/raw/ 2015-11-30 17:06:01 also: stacked git 2015-11-30 17:07:02 we ned to document our everything more thoroughly tbh 2015-11-30 17:40:04 chris|: stacked git the python thing? 2015-11-30 17:44:37 ScrumpyJack: download the mbox or patch file and apply it with git-am(1) (if you downloaded the mbox file) or git-apply(1) (if you downloaded the patch file) 2015-11-30 17:45:03 ncopa: I think it's up to you to write such a style guide :p 2015-11-30 17:45:12 nmeum: thanks. i'll give that a shot 2015-11-30 18:17:17 are the APKBUILDs supposed to be valid for cross-compilation too, or can they assume they will always be natively compiled? 2015-11-30 18:18:01 ACTION wants to make sure he follows the style guide even before it's a thing 2015-11-30 18:22:27 skarnet: that's a good question…currently most APKBUILDs assume that the latter is the case 2015-11-30 18:23:17 the problem is, different options may be used for the configure stage if the package is cross-compiled 2015-11-30 18:23:22 yeah, sure 2015-11-30 18:24:33 so we should probably say that build() is native-only, and replace it via cross_build() when it's cross-compile, with a default cross_build() equal to build() 2015-11-30 18:25:50 and in the cross_build() function, make a variable available for the crosscompiler prefix 2015-11-30 18:26:12 and maybe another for the target 2015-11-30 18:29:57 ScrumpyJack: it is 2015-11-30 18:31:53 it is basically a patch queue ontop of git, and it has stg undo for all operations, including rm 2015-11-30 18:42:16 what's the appropriate arch= for an apkbuild which contains no binaries but does actually care about the arch 2015-11-30 18:43:13 ammunta: I think that would be "noarch" 2015-11-30 19:45:05 tried compiling curl with, --with-nghttp2 --enable-unix-sockets , it increases its size by 5kb 2015-11-30 19:45:26 would be nice to have it, I can send patch if needed 2015-11-30 19:50:46 but this requires testing/nghttp2 2015-11-30 19:53:04 https://fiche.greata.uk/xxob anyone mind taking a look at this before I submit a patch? 2015-11-30 19:53:36 I don't think there's precedent for luajit-only libs 2015-11-30 19:55:57 lua-turbo I guess 2015-11-30 20:03:15 ncopa: http://tpaste.us/GPEb <- for curl 2015-11-30 20:05:35 good evening 2015-11-30 20:05:49 ok that is weird with nghttp2 1.3.4 diff was about 5kb but not its 200kb 2015-11-30 20:06:19 but now with nghttp2 v1.5.0 its 200k 2015-11-30 20:08:16 hmm.. ~135kb 2015-11-30 20:09:03 ok nghttp2 v1.5.0 has new API 2015-11-30 20:25:32 skarnet: I believe we should try to have build() crosscompiler compatible 2015-11-30 20:36:53 vkris: curl is also a main package while nghttp2 is testing, so that would have to change first before one can depend on the other 2015-11-30 20:41:31 chris|: yes 2015-11-30 20:42:01 main/ cannot have a dep in testing or community 2015-11-30 20:43:10 ok this is very weired, I upgraded my lxc and not curl is ~398kb, earlier it was ~277kb 2015-11-30 20:43:23 and now* curl 2015-11-30 20:44:17 so its not newer nghttp2 causing the size increase, coz I rebuild without it 2015-11-30 20:44:41 something it upgrade, gcc 5.2.0 ? 2015-11-30 21:14:54 barthalion: it would be great if it was, but I'm afraid it will be very hard for a lot of packages