2022-09-01 05:09:29 Hello! Is "sign in with github / gitlab" busted on https://gitlab-test.alpinelinux.org/alpine ? Both are failing for me with variations of "The redirect URI included is not valid." 2022-09-01 05:36:38 that is the test gitlab and not the real one 2022-09-01 05:36:48 remove the -test part 2022-09-01 06:08:03 oh ha! The search engine kicked the -test one back to me. Real one works. Thank you! 2022-09-01 07:40:19 ncopa: nothing uses poppler in main/ , so we could move it to community and merge it with the -qt5 one 2022-09-01 07:40:53 ah 2022-09-01 07:40:56 no, cups uses the util 2022-09-01 07:41:01 +s 2022-09-01 13:41:18 @ncopa so there's an issue with setup-disk in some situations, specifically on UEFI when using cryptsys - this results in a ESP partition of 34MB being created that as well as being used for Grub EFI files is also used (mounted as /boot rather than /boot/efi) for storing the kernel & initramfs - an install works fine but whenever the kernel is later upgraded then this partition can run out of space (as for a brief period space is needed 2022-09-01 13:41:18 for 2 kernels and 2 initramfs files) 2022-09-01 14:00:44 hmm when building docker-registry with abuild it creates a user and group on the build host it self, is that really OK? 2022-09-01 14:01:39 if you are not using rootbld, yes 2022-09-01 14:26:24 do we package libGLX ? 2022-09-01 14:27:46 in xorg-server apparently 2022-09-01 19:02:03 HRio: it deletes the user/group afterward. in melange we solve this differently 2022-09-01 22:33:47 k 2022-09-01 23:21:34 why do I have an empty /etc/ssl1.1/ and a /etc/ssl1.1/certs.apk-new symlink that points to /etc/ssl/certs/ ? 2022-09-01 23:22:04 is this as it should be and I should just be diligent, remove the empty dir and rename the symlink? 2022-09-01 23:22:04 do you have libssl3 installed? 2022-09-01 23:22:15 and libssl1.1 not installed 2022-09-01 23:22:28 both are installed 2022-09-01 23:22:37 or whichever package owned that dir 2022-09-01 23:23:12 /etc/ssl1.1/certs is owned by ca-certificates-bundle-20220614-r2 2022-09-01 23:23:25 apk info -W says 2022-09-01 23:23:30 but it's empty 2022-09-01 23:23:59 ERROR: /etc/ssl/certs: Could not find owner package 2022-09-01 23:24:08 sorry 2022-09-01 23:24:14 libcrypto1.1 and libcrypto3 2022-09-01 23:24:27 both 2022-09-01 23:25:07 then it's weird that /etc/ssl1.1 is empty 2022-09-01 23:25:39 /etc/ssl1.1/certs/ is empty, /etc/ssl1.1/ is not 2022-09-01 23:25:55 I would expect that libcrypto1.1 is gone and the leftover symlink prevented dir delted 2022-09-01 23:28:12 libcrypto1.1-1.1.1q-r3 is required by: 2022-09-01 23:28:12 libssl1.1-1.1.1q-r3 2022-09-01 23:28:13 openvpn-2.5.7-r0 2022-09-01 23:28:37 /etc/ssl1.1/certs is intentionally there for whatever reason, that's fine 2022-09-01 23:31:38 but should I 'rmdir certs; mv certs-apk-new certs'? 2022-09-01 23:32:39 s,certs-apk-new,certs.apk-new, 2022-09-01 23:33:54 yes unless you have manually modified your cert list 2022-09-01 23:34:08 I probably should, not sure why this had me stumped 2022-09-01 23:34:46 .apk-new is where apk upgrade installs its /etc files (and other files marked as 'config') when they have been modified 2022-09-01 23:35:06 when you get a .apk-new it's an indication that something has changed 2022-09-01 23:35:20 I know, perhaps that's why, since I haven't changed the contents of /etc/ssl1.1/certs/ 2022-09-01 23:35:22 either you have made manual modification to your files, in which case you should review the changes and merge it, ideally 2022-09-01 23:35:34 or you haven't and it's just apk being careful with config files 2022-09-01 23:35:46 that's what I usually do,occationally at least 2022-09-01 23:35:57 and what I did when I saw this 2022-09-01 23:36:47 ncopa has the details on the ssl1.1 thing, he's the one who did the upgrade to OpenSSL 3, I suspect /etc/ssl is now the thing that's used by openssl3 and in order to avoid breaking things, openssl1 files were temporarily moved to /etc/ssl1.1 2022-09-01 23:37:24 now that everything's under openssl3, /etc/ssl1.1 shouldn't be used anymore (except, as always with computers, when it is) 2022-09-01 23:38:20 ah, and there's a comment on openssl3 in main/openvpn/APKBUILD 2022-09-01 23:39:45 the 1.1 certs symlink is intentionally there so that things that do still use 1.1 (openvpn) also have certs, as the certs are in ca-certificates instead as you see 2022-09-01 23:39:58 (since 1.1 is configured with ssl1.1 as the ssldir) 2022-09-01 23:40:06 (and expects to find the certs in it) 2022-09-01 23:40:52 yeah.. it was some time since I used openvpn.. 2022-09-01 23:49:53 can I have Mr Stale removed from !35339 and !35272 please? 2022-09-02 06:41:27 oh joy, another instance of broken-cmake-but-pkgconf-1.8-works 2022-09-02 06:47:30 think this time it's not a pkgconf bug 2022-09-02 06:49:01 Ariadne: https://img.ayaya.dev/39q4coFmAhvI top is 1.9.3, bottom is 1.8.0; cmake for some reason only grabs the first from the list for find_package(Gio); target GIO::GIO. that is the way it invokes it from the log, with two arguments 2022-09-02 06:49:03 what do you think of it 2022-09-02 06:51:37 the pkgconf 1.9.3 behavior is more correct, i think 2022-09-02 06:51:57 in terms of the library order being ordered by dependencies 2022-09-02 06:52:00 aye 2022-09-02 06:52:05 the issue is here https://github.com/obsproject/obs-studio/blob/master/cmake/Modules/FindGio.cmake 2022-09-02 06:52:12 it uses _GIO_LIBRARIES from that list 2022-09-02 06:52:15 then finds the first 2022-09-02 06:52:16 so yeah 2022-09-02 06:52:35 afaik the libraries have to be ordered left to right by dependency 2022-09-02 06:52:51 e.g. -lglib -lgtk is correct 2022-09-02 06:52:58 but maybe i have it backwards :) 2022-09-02 06:54:48 :) it's easy to patch but i wonder what the fix is really meant to be here 2022-09-02 06:55:01 find_library will only grab the first match 2022-09-02 06:55:06 so if the first name is not the expected thing.. 2022-09-02 06:55:13 forcing the name works of course: https://img.ayaya.dev/9HEHgWKdEhXT 2022-09-02 06:55:13 If object or library AA needs a symbol from library BB, then AA should come before library BB in the command-line invocation of the linker. 2022-09-02 06:55:21 yes it’s backwards in 1.9.3 2022-09-02 06:55:45 yeah, i guess pkgconf is correct 2022-09-02 06:55:52 now just wondering how to teach cmake.. that 2022-09-02 06:56:00 where they intentionally want only the name of the lib itself 2022-09-02 06:56:10 no i mean 1.9.3 is wrong :) 2022-09-02 06:56:13 ah 2022-09-02 06:56:19 okay, okay 2022-09-02 06:57:44 it is easy enough to fix there i think :) 2022-09-02 07:09:53 bl4ckb0ne: do you know how to make new cmake pick up EGL without glvnd? the cmake file (FindOpenGL) claims EGL only exists for glvnd and doesn't find it otherwise (for something doing find_package(OpenGL COMPONENTS EGL) then linking OpenGL::EGL) 2022-09-02 07:10:00 but i mean.. libEGL.so obviously exists 2022-09-02 07:10:17 did anyone know i hate cmake 2022-09-02 07:47:40 !37801 2022-09-02 07:48:42 as I got it, a warning was put in the documentation "to make expectations clearer" but there is no "fix" in this release 2022-09-02 07:49:09 did I get it right? 2022-09-02 07:51:20 these cves are meaningless 2022-09-02 08:00:06 I read the discussion the other day, but here I don't think anything has been fixed so putting the CVE-2022-32983 under secfixes would be wrong 2022-09-02 08:07:18 the cve itself says 'version before' and it's after it : ) 2022-09-02 08:14:30 yes, but it hasn't been updated since the 5.5.2 release, perhaps it should 2022-09-02 08:27:20 psykose: just hit this kind of issue 2022-09-02 08:27:25 : ) 2022-09-02 08:27:29 went with pkgconfkg 2022-09-02 08:27:34 yupp 2022-09-02 08:27:41 !38515 2022-09-02 08:27:44 i love to cmake 2022-09-02 08:28:36 but we should probably report it to cmake? i dunno 2022-09-02 08:29:10 i understand the random 'portable egl' weird stuff they have commented in there (lots of prose in the file) but surely they're not stupid enough to not find a /usr/include/EGL /usr/lib/libEGL.so 2022-09-02 08:35:33 bl4ckb0ne: also do you know anything about cmake generator expressions breaking pkgconfig files? ends up causing this https://img.ayaya.dev/OZs4fmVOaUtI 2022-09-02 08:36:01 from something like https://img.ayaya.dev/OprKoU8WS8Xc 2022-09-02 08:36:17 yeah i know thar generator expressions sucks 2022-09-02 08:37:02 wonder why it breaks the output 2022-09-02 08:37:52 pdobably a missed < or > 2022-09-02 08:40:22 they do align, ree 2022-09-02 09:54:04 psykose: yeah not much choice for EGL https://gitlab.kitware.com/cmake/cmake/-/blob/master/Modules/FindOpenGL.cmake#L243 2022-09-02 09:57:11 also that generator expression looks good but its probably badly used and fails to process 2022-09-02 10:27:32 bl4ckb0ne: that cmake works if ${_OPENGL_LIB_PATH} has /usr/lib or whatever 2022-09-02 10:27:55 they either populate it stupidly or return before that, idk 2022-09-02 10:28:01 cmake is dumb 2022-09-02 10:28:31 ah 2022-09-02 10:28:36 they simply don't set it anywhere in this file 2022-09-02 10:28:37 why cant everybody use pkgconfig properly 2022-09-02 10:29:08 : ( 2022-09-02 10:29:21 i'm tempted to hotpatch this into the cmake package 2022-09-02 10:29:29 and just add /usr/lib into that path 2022-09-02 10:29:52 is it a good idea? no, but 2022-09-02 10:44:12 no 2022-09-02 10:44:26 just let cmake simmer in its own juice 2022-09-02 10:44:50 also why dont we have glvnd stuff 2022-09-02 10:52:23 no idea 2022-09-02 10:52:59 also would a symlink work 2022-09-02 10:53:05 afaik it's literally just.. adding it and then building it with mesa? 2022-09-02 10:53:10 yeah, a symlink for the name in the path 2022-09-02 10:53:27 is there anything special for glvnd aside from changing a toggle 2022-09-02 10:53:36 (do keep in mind all the platforms and etc) 2022-09-02 10:54:22 but i also don't see what it even accomplishes, as there aren't multiple gl vendors to begin with 2022-09-02 10:54:34 the others are nvidia (lol) and amdgpu-pro? 2022-09-02 10:56:00 dunno if amd ships a gl driver 2022-09-02 11:07:28 what's the real point of it then if we don't have nvidia drivers anyway 2022-09-02 11:08:23 conformance, i think a few places ask for glvnd 2022-09-02 11:08:59 conformance with who? it just looks like pointless layering to me 2022-09-02 11:09:06 and the exact same libEGL exists either way 2022-09-02 11:09:26 exact same -> it exists and works* 2022-09-02 11:10:05 but cmake cant find it 2022-09-02 11:10:41 for literally zero reason that i can see, aside from 'we care about what provides libGL just because' 2022-09-02 11:11:14 i'll admit i don't know how anything on this layer works, so correct me if i'm wrong 2022-09-02 11:11:40 (it just really looks like weird hardcoded expectations without any actual benefit, since the final loaded library is the same, the platform works, the headers work, etc) 2022-09-02 11:12:03 the same .pc works, the same -lEGL/-GL/whatever works... they just want something different anyway 2022-09-02 11:12:12 so for who is it for? what does it let applications do? 2022-09-02 11:12:27 glvnd itself allows the same setup to provide nvidia libGL or whatever, but that should be transparent 2022-09-02 11:12:44 final builds shouldn't give a shit 2022-09-02 11:12:52 unless.. i just don't know :) 2022-09-02 11:17:00 (i'm also not really arguing against it, so if you get it to work and toggle all the switches and it doesn't break anything i'll probably merge it anyway. but i expect it to take a lot of work to get there) 2022-09-02 11:17:16 just.. really curious 2022-09-02 11:19:34 iirc they (mesa) want to make glvnd the official gl stuff 2022-09-02 11:21:41 sure, works for me 2022-09-02 11:22:05 well, some help would be appreciated then :) 2022-09-02 11:22:06 nothing official so we can still dodge glvnd for now 2022-09-02 11:22:12 on what? 2022-09-02 11:22:25 swapping to it 2022-09-02 11:22:40 in time sure 2022-09-02 11:22:47 but for the moment we can live without 2022-09-02 11:23:45 there's still some issues with it https://gitlab.freedesktop.org/glvnd/libglvnd/-/issues/222 2022-09-02 11:24:28 haha 2022-09-02 11:24:30 that's not great 2022-09-02 11:24:52 also it seems to also do some weird initial-exec tls stuff, but i think that was fixed 2022-09-02 12:28:12 i tried to fix it but admittedly i did not test it extensively 2022-09-02 12:29:02 if you have the time you could look into #14140 2022-09-02 12:29:15 i had removed the elf tls disable for 22 and it works everywhere except there apparently 2022-09-02 12:29:25 the next mesa version doesn't allow that anymore.. so i assume it will just be broken 2022-09-02 12:29:31 it does work on aarch64/x86_64 though 2022-09-02 12:31:00 (22.2 removes all the elf_tls disables as you've seen, and it's always on, per https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17213) 2022-09-02 12:32:05 new mesa release should come out soon 2022-09-02 12:32:16 and will be blocked on that :p 2022-09-02 12:33:44 when's the next wlroots 2022-09-02 12:36:23 someday 2022-09-02 12:36:52 im pushing for a release but theres still stuff to merge 2022-09-02 12:37:17 id like to ship it before the end of the year 2022-09-02 12:40:21 slow n steady 2022-09-02 12:40:22 :3 2022-09-02 12:50:16 lots of stuff in this release 2022-09-02 12:50:18 good ol usual 2022-09-02 12:54:13 psykose: i saw that issue but not the followup 2022-09-02 12:54:35 i'm inclined to suspect that it is not initial-exec tls but some other tls issue 2022-09-02 12:55:02 afaik even 22.1.7 has your patch to ifdef initial-exec to glibc 2022-09-02 12:55:25 is this gcc 11 or 12? 2022-09-02 12:55:31 12 2022-09-02 12:55:43 but we did also test 11, same issue 2022-09-02 12:56:18 (by building on 3.16-stable with same apkbuild, so it was also llvm13, but that doesn't matter) 2022-09-02 12:57:20 and the shimmy disable fa13f5fe4cc573e44eb34bbaabbc7cae393acc9f that fixes it again. i don't think it's initial-exec anymore, but yeah, something else 2022-09-02 12:57:38 hm. if it's not too much work can you try CFLAGS+=-mtls-dialect=gnu CXXFLAGS+=-mtls-dialect=gnu 2022-09-02 12:57:56 (without the "disable elf tls" hack of course) 2022-09-02 12:58:21 glibc has some weird bugs with tlsdesc. i was hoping that musl wouldn't have such bugs but maybe? 2022-09-02 12:58:27 s/maybe/& it does 2022-09-02 13:02:49 hm, or, possibly more likely than musl having a bug, gcc could have a bug with tlsdesc on arm 2022-09-02 13:03:02 or binutils, in fact 2022-09-02 13:04:03 overriding the default gnu2? 2022-09-02 13:07:31 i can possibly help look at this 2022-09-02 13:07:34 what's going on? 2022-09-02 13:08:02 linked issue above of #14140 2022-09-02 13:08:12 ok looking 2022-09-02 13:08:14 it's a bit scattered but tl;dr the mesa elf tls thing causes that (if you disable it it's fine) 2022-09-02 13:08:27 and it does work without the disable on aarch64/x86_64 etc 2022-09-02 13:08:37 i can't test myself as i have no armv7 hardware 2022-09-02 13:09:03 when Newbyte is around he can give the tls dialect a spin if gnu instead of gnu2 works for some reason, could be a bug there 2022-09-02 13:09:39 i rather doubt that 2022-09-02 13:10:11 are we sure they don't have any hard-coded initial-exec-like asm or something? 2022-09-02 13:11:39 what would that look like? 2022-09-02 13:12:01 arm asm files related to tls or dispatch stuff? 2022-09-02 13:14:49 it looks like the mapi entry stuff is still using initial-exec... 2022-09-02 13:15:00 see for example this file: https://gitlab.freedesktop.org/mesa/mesa/-/blob/main/src/mapi/entry_x86_tls.h 2022-09-02 13:16:02 as someone that doesn't know what they're looking at, which part says that (just curious) 2022-09-02 13:18:59 psykose: I can test things in about three horus 2022-09-02 13:19:01 s/horus/hours/ 2022-09-02 13:19:50 sure 2022-09-02 13:20:40 first check the 22 rc 2022-09-02 13:20:44 if it crashes again that's a start 2022-09-02 13:20:48 then we can try random shit 2022-09-02 13:45:25 psykose: you said something about that wxgtk/wxwidgets 3.2 would enable opencpn on wayland, without xwayland I assume? 2022-09-02 13:45:41 for everything with wxwidgets 2022-09-02 13:46:07 did you atempted to package wxwidgets? 2022-09-02 13:46:15 i did it ages ago yeah 2022-09-02 13:46:18 built everything 2022-09-02 13:46:19 everything works 2022-09-02 13:46:21 3.2 2022-09-02 13:46:35 i still have ptsd from using it in a previous dayjob 2022-09-02 13:46:47 except for some random maitreya shit that doesn't build with it and i couldn't be fucked 2022-09-02 13:46:54 and.. codeblocks needs like 15 patches backported 2022-09-02 13:46:58 and then there's some stuff here and there but light 2022-09-02 13:47:17 oh and wxwidgets itself needs to patch the include headers for some *_l functions that don't exist in musl 2022-09-02 13:47:20 that's about it 2022-09-02 13:47:25 nice 2022-09-02 13:47:37 :I 2022-09-02 13:47:47 something for a rainy week 2022-09-02 13:47:49 it was a lot of work and then i rm -rfd my git repo 2022-09-02 13:47:55 really proud of that one 2022-09-02 13:47:57 good job alice 2022-09-02 13:48:05 good job 2022-09-02 13:48:16 so when are you packaging it back :D 2022-09-02 13:48:27 idk, i feel a bit sketch about the 500 codeblocks backports 2022-09-02 13:48:34 though.. those people never release anything 2022-09-02 13:48:47 i think dev stopped 2022-09-02 13:49:01 idk how requested codeblocks is but wxwidgets is used in a few places 2022-09-02 13:49:08 maybe one day we'll get cemu 2022-09-02 13:49:47 wouldnt be surprised 2022-09-02 13:50:02 and yeah wayland wxwidgets is nice 2022-09-02 13:50:03 I'm only interested in OpenCPN 2022-09-02 13:50:07 also there was other stuff in that series i had 2022-09-02 13:50:18 like what 2022-09-02 13:50:19 like whatever latest slightly updated version of audacity 2022-09-02 13:50:22 cough sorry tenacity 2022-09-02 13:50:25 (already unmaintained) 2022-09-02 13:50:43 +- some other stuff updated 2022-09-02 13:50:48 kek 2022-09-02 13:51:05 i'll package cemu for $1 who's buying 2022-09-02 13:51:45 i dont mind doing it 2022-09-02 13:51:53 can prolly do wxwidgets too if you're nice 2022-09-02 13:52:45 nice in what sense :D 2022-09-02 13:53:07 just looking for an excuse to find time 2022-09-02 13:54:56 i'll get to it at some point myself 2022-09-02 13:55:02 guys, is there libcc++ from llvm in alpine? 2022-09-02 13:55:06 *ilbc++ 2022-09-02 13:55:30 no 2022-09-02 13:56:05 Would you accept a community package? 2022-09-02 13:56:17 probably not, i could do it myself in ~10 minutes 2022-09-02 13:56:20 but i don't see a point 2022-09-02 13:56:24 it's quite useless to package 2022-09-02 13:56:28 Why? 2022-09-02 13:56:33 what would the use be? 2022-09-02 13:56:40 you can't mix it with libstdc++ 2022-09-02 13:56:52 and if you care enough about the c++ standard library, you probably know enough how to build it yourself 2022-09-02 13:56:59 having it distro packaged is not very useful 2022-09-02 13:57:06 k 2022-09-02 13:57:25 along with compiler-rt to even use --rtlib=compiler-rt you need to build the toolchain yourself for it to work anyway (otherwise you still have to -lgcc_s, for instance) 2022-09-02 13:57:35 it's just quite pointless, dunno 2022-09-02 13:58:09 if you want an llvm-only toolchain you can build one, it's quite easy with the monorepo build 2022-09-02 13:58:35 Yeah, I know 2022-09-02 13:58:41 libc++ can be used only by software built by clang++, and everything c++ in alpine repo is built by g++ 2022-09-02 13:59:06 My guys from C++ department want that since they are trying to port our software to Alpine 2022-09-02 13:59:16 But I guess it's easier to create s private rep 2022-09-02 14:00:17 you can't really load libc++ and libstdc++ into the same process, can you 2022-09-02 14:00:32 Nope 2022-09-02 14:00:37 so how would that work? if they have any dependencies they won't be able to use any of the packaged ones and then need to built them all against libc++ as well 2022-09-02 14:00:52 It's mostly static from what I've heard 2022-09-02 14:00:58 For some reason libc++ is not 2022-09-02 14:01:05 But yeah, I can see the point 2022-09-02 14:01:36 What is static? 2022-09-02 14:02:06 Afaik libc++ is used by discord, but it's non-free so it's not a valid use case here 2022-09-02 14:05:20 Ermine: its more than non-free, its a malware data mining platform 2022-09-02 14:35:05 Ariadne: friendly ping on !37834. I've made the changes you wanted, psykose has reviewed the MR (and found some nice things to optimize, thank you!), nothing more should be blocking this merge. It's been one month since tsc#52 was opened, that was the deadline you set; I would like this to be done before I leave for vacation in 5 days. 2022-09-02 14:35:24 no algitbot, not clamstmp. 2022-09-02 14:35:29 as a minor thing to note you don't have to constantly rebase it 2022-09-02 14:35:48 but I do. 2022-09-02 14:35:55 what for 2022-09-02 14:36:21 (i'm assuming there's no changes each time) 2022-09-02 14:36:44 I want to make it as easy as possible for anyone to click the merge button. One click, done. That's the level of comfort I'm willing to provide core devs so things have a shadow of a chance of getting done. 2022-09-02 14:36:57 also, if there's ever a conflict, I want to see it asap and fix it right away. 2022-09-02 14:37:10 I don't want to give the slightest pretext for additional delays. 2022-09-02 14:37:22 it's the same effort, haha 2022-09-02 14:37:29 it could be 2k behind and it's the same thing 2022-09-02 14:37:47 but for conflicts, yeah, they only get noticed if someone opens the page 2022-09-02 14:37:50 there's still a rebase button to click 2022-09-02 14:37:59 sure, but that starts at 1 2022-09-02 14:38:03 I wouldn't want people to have to do one additional click 2022-09-02 14:38:11 i don't even click the buttons 2022-09-02 14:38:22 huh? 2022-09-02 14:38:41 i have it macrod into a cli tool instead so i don't have to wait for the endless rebasing 2022-09-02 14:38:49 I see. 2022-09-02 14:39:03 Well, that's smart, and that's you. I'm not sure everyone does the same. 2022-09-02 14:39:08 copy mr number -> `gmm mr1 mr2 mr3 mr4 mr5 mr6 mr7..` 2022-09-02 14:39:28 but the conflict thing is real. 2022-09-02 14:39:29 otherwise i have to sit there and refresh the page constantly until it says merge 2022-09-02 14:39:31 boring! 2022-09-02 14:39:35 yeah, conflicts are manual 2022-09-02 14:39:47 and given how many potentially conflicting things are in the queue 2022-09-02 14:39:57 and I have no idea in what order they will get merged, if ever 2022-09-02 14:40:09 I want to make sure to catch them reaaaally early. 2022-09-02 14:40:26 i am just waiting for one person to look at it and then i'll just merge it myself 2022-09-02 14:40:31 it looks fine to me but perhaps i overlooked something 2022-09-02 14:40:52 agreed 2022-09-02 14:41:02 did you test an upgrade in-place? 2022-09-02 14:41:15 yes, I have turned it off and on again 2022-09-02 14:41:50 try ssh into a machine without anything, then run `apk upgrade -X https://gitlab.alpinelinux.org/alpine/aports/-/jobs/825572/artifacts/raw/packages/main --allow-untrusted` 2022-09-02 14:41:54 but again, that's one anecdote, better have someone else do the same and verify it 2022-09-02 14:42:00 (that's the x86_64 artifact from the latest job) 2022-09-02 14:42:14 or well, i can also do that 2022-09-02 14:42:17 brb nuking my server 2022-09-02 14:42:36 if there's one thing Docker is good for 2022-09-02 14:43:07 also i am bad at links 2022-09-02 14:44:03 skarnet/aports 2022-09-02 14:44:29 first issue: 2022-09-02 14:44:29 ERROR: busybox-openrc-1.35.0-r24: trying to overwrite etc/init.d/acpid owned by acpid-openrc-2.0.33-r2. 2022-09-02 14:44:30 how did you get the alpine link? :D 2022-09-02 14:44:41 i copied some other url and changed the number.. 2022-09-02 14:44:45 ha 2022-09-02 14:45:03 see, that's why this needs other reviewers, I never had that issue 2022-09-02 14:45:10 i merely have acpid installed :p 2022-09-02 14:45:26 sigh 2022-09-02 14:45:52 okay, fuckton of replaces= incoming 2022-09-02 14:45:52 can't find any other issues (but i also don't use busybox extras or w/e) 2022-09-02 14:46:08 i use udev on desktop and mdevd on server and that acpid is the only conflict 2022-09-02 14:46:10 now to reboot 2022-09-02 14:46:56 post-deinstall from acpid: 2022-09-02 14:46:57 * Run the following command to reinstall /etc/init.d/acpid from busybox: 2022-09-02 14:46:57 * > apk fix busybox-initscripts 2022-09-02 14:47:03 ah 2022-09-02 14:47:20 lol. will fix that. 2022-09-02 14:48:02 Btw, how do I pull rebased branch? 2022-09-02 14:49:03 git reset 2022-09-02 14:49:54 see this is exactly what I mean when I'm saying we need to treat all implementations the same way 2022-09-02 14:50:07 it is :) 2022-09-02 14:50:22 having a virtual package for acpid-openrc would have made me notice there was another implementation 2022-09-02 14:50:31 and fix it along the way 2022-09-02 14:51:04 i think that's where they end.. i hope 2022-09-02 14:51:06 Ermine: `git fetch --all && git rebase` while on the branch 2022-09-02 14:51:09 usually does the job 2022-09-02 14:51:16 what was the question 2022-09-02 14:51:28 "Btw, how do I pull rebased branch?" 2022-09-02 14:51:55 git fetch --all && git pull -f 2022-09-02 14:51:55 xd 2022-09-02 14:53:01 git fetch --all takes forever if you have as many remotes as me 2022-09-02 14:54:05 ptrc, panekj, psykose: thank you 2022-09-02 14:54:25 wow all our names start with a p 2022-09-02 14:54:38 not sorted though! 2022-09-02 14:54:38 -1 2022-09-02 14:55:03 panekj, psykose, ptrc 2022-09-02 14:55:26 and Peter 2022-09-02 14:55:47 you have the Power 2022-09-02 14:56:01 we have two Peters 2022-09-02 14:56:13 right 2022-09-02 14:58:36 fix pushed, waiting for psykose to get me the magical URL to test 2022-09-02 14:59:03 pushed 2022-09-02 15:00:06 apk upgrade -X https://gitlab.alpinelinux.org/skarnet/aports/-/jobs/825598/artifacts/raw/packages/main --allow-untrusted 2022-09-02 15:00:17 you need to add 2022-09-02 15:00:22 uhh two -X this time 2022-09-02 15:00:24 cause of acpid 2022-09-02 15:00:29 what 2022-09-02 15:00:31 so also -X of that with /community on the end instead 2022-09-02 15:00:37 what's -X 2022-09-02 15:00:46 extra repo as an apk argument 2022-09-02 15:00:49 so it's two urls 2022-09-02 15:00:50 :p 2022-09-02 15:00:54 since it's also community/ 2022-09-02 15:06:13 looks like Docker can't test that properly because 'apk update' doesn't get the latest world so busybox-openrc isn't found 2022-09-02 15:06:36 the -X trick seems to work with already existing packages 2022-09-02 15:07:06 but when there are new packages, maybe there's another unholy option to spoof world 2022-09-02 15:07:51 if not, it's personal machine breaking time 2022-09-02 15:08:09 Latest world? 2022-09-02 15:09:08 You can do in upgrade -Ua in Docker before trying to install the packages from the MR 2022-09-02 15:14:33 first upgrade, then do the mr meme 2022-09-02 15:14:48 you're cute 2022-09-02 15:14:52 both of you 2022-09-02 15:17:15 I can't get the -X thing to work with Docker 2022-09-02 15:17:24 but it works on my machine so I'll call it good enough 2022-09-02 15:17:29 one day I'll understand the -X voodoo 2022-09-02 15:19:55 you paste the full line batman 2022-09-02 15:21:06 no, I'm done, and need to go afk anyway 2022-09-02 15:21:14 the newest version works 2022-09-02 15:21:17 test it 2022-09-02 15:21:17 sik 2022-09-02 15:21:24 2late already bricked my machines 2022-09-02 15:21:45 as far as I'm concerned the thing works 2022-09-02 15:21:52 unless there's more than acpid-openrc 2022-09-02 15:22:05 bbl 2022-09-02 15:22:21 yeah seems to work 2022-09-02 15:22:21 gl 2022-09-02 16:56:11 ncopa: what do you want to do with rust? it can't build itself (and hangs ~50% of the time instead of segfaulting), so we either need to bootstrap it by hand into 1.63 again (and hope that works) or disable it 2022-09-02 16:57:21 it always segfaults on stage1 start for reference https://img.ayaya.dev/TgdsL7gnbIpd , or hangs on one of multiple steps. i think i got further on 1.62.1 instead but it was still a segfault/hang in random spots 2022-09-02 17:00:49 psykose, which mesa package/library actually has the TLS stuff in it ? 2022-09-02 17:01:28 ahhh, looks like the current apks in edge still have pthread_getspecific instead ..? 2022-09-02 17:01:33 mesa is just one package, but if you want a reproduction it's in that same thread via mesa-gl libGL i think 2022-09-02 17:01:35 i was trying to get them and read wtf is going on 2022-09-02 17:01:36 probably, yeah 2022-09-02 17:01:44 22.1 -> 22.2 has quite a lot of changes 2022-09-02 17:01:49 do you have a source for a crashing binary i can look at? 2022-09-02 17:02:17 running Xwayland crashes on armv7, or https://gitlab.alpinelinux.org/alpine/aports/-/issues/14140#note_258210 2022-09-02 17:02:25 i want to just analyze the .so files 2022-09-02 17:03:05 that links libGL (mesa-gl) (and uses libepoxy to drive it i guess), not sure of the other deps 2022-09-02 17:03:09 not the test case C program 2022-09-02 17:03:13 the build of mesa it crashes with 2022-09-02 17:03:57 you can pull either the 22.2 artifacts from !38516 (i didn't test if these actually still crash) or rebuild main/mesa removing the elf-tls patch 2022-09-02 17:04:34 i can run the latter through ci if you want 2022-09-02 17:05:43 sure that works i guess 2022-09-02 17:05:52 ok, will be up in 10-15 mins 2022-09-02 17:06:00 i'm hoping just reading the disassembly or reloc tables will show something bogus 2022-09-02 17:06:33 can you run the crashing testcase? 2022-09-02 17:06:40 if so, ltrace of it would be helpful 2022-09-02 17:06:57 i don't have the hardwaare 2022-09-02 17:07:01 if Newbyte was here he could do it 2022-09-02 17:07:13 yes hello 2022-09-02 17:07:15 https://gitlab.alpinelinux.org/alpine/aports/-/pipelines/134940 when they build 2022-09-02 17:07:27 currently trying to put pmOS on a new device 2022-09-02 17:07:33 which also is ARMv7 2022-09-02 17:07:38 Newbyte: first check if 22.2 still crashes (not this link here), then ltrace of crashing 2022-09-02 17:07:39 can test it on there 2022-09-02 17:07:39 sure, sure 2022-09-02 17:08:20 basically i just want the ltrace to narrow down what functions are being called internally and where the bugs might be 2022-09-02 17:08:42 and the crashing .so binaries to do some analysis on them (readelf & objdump) 2022-09-02 17:10:08 pipeline above is the crashing binaries, click armv7 and download, or https://gitlab.alpinelinux.org/alpine/aports/-/jobs/825669/artifacts/download 2022-09-02 17:10:52 from there, it's probably just libGL.so and the things it links to, but yeah the ltrace will be here eventually 2022-09-02 17:13:21 how do I produce an ltrace? 2022-09-02 17:14:24 just ltrace ./a.out i think 2022-09-02 17:24:08 the objdump i'm doing has all sorts of undefined instruction problems... 2022-09-02 17:26:23 in what sense exactly 2022-09-02 17:27:38 hmm maybe the objdump i have is just acting up on the file... 2022-09-02 17:28:13 is it an armv7 objdump 2022-09-02 17:28:56 Anyway I got my new and shiny ARMv7 device set up now 2022-09-02 17:29:02 What do you want me to do? 2022-09-02 17:29:28 ltrace ./a.out of the crashing binary on the above 21.1 artifacts with the tls added again 2022-09-02 17:29:37 then after check if 21.2 also crashes in the other mr 2022-09-02 17:30:30 ok getting the objdump from alpine cross packages 2022-09-02 17:31:30 same error 2022-09-02 17:31:36 00007920 : 2022-09-02 17:31:37 7920: e7fd4778 ; instruction: 0xe7fd4778 2022-09-02 17:31:40 7924: e28fc600 add ip, pc, #0, 12 2022-09-02 17:31:42 etc. 2022-09-02 17:31:56 it's like these binaries have random junk inserted into the instruction stream here and there 2022-09-02 17:32:24 could builder have bad ram or something? 2022-09-02 17:32:42 _glapi_get_dispatch is calling into this corrupt part of the PLT it seems 2022-09-02 17:33:12 strange 2022-09-02 17:33:20 can you check the x86_64 ones too 2022-09-02 17:36:43 wtf 2022-09-02 17:36:51 00015c8a <_glapi_destroy_multithread>: 2022-09-02 17:36:51 15c8a: f000 ba0d b.w 160a8 <_glapi_set_warning_func+0x26e> 2022-09-02 17:36:54 00015c8e <_glapi_check_multithread>: 2022-09-02 17:36:57 15c8e: f000 ba0c b.w 160aa <_glapi_set_warning_func+0x270> 2022-09-02 17:38:27 it looks like something was badly wrong with ld... 2022-09-02 17:40:46 could be a binutils 2.39 bug there 2022-09-02 17:40:46 hm 2022-09-02 17:40:59 there's also the binary bloating for all arm being fucked on 2.39 2022-09-02 17:41:37 tempted to throw lld at it and see if things still fail 2022-09-02 17:42:01 hmm ok some of this messed up looking stuff seems to be something it's doing for TLSDESC... 2022-09-02 17:42:06 let's see if i can figure that out... 2022-09-02 17:44:59 shroomush:~$ ltrace ./a.out 2022-09-02 17:44:59 __libc_start_main(0x4504e1, 1, 0xbecc2c74, 0x450359PTRACE_SINGLESTEP: I/O error 2022-09-02 17:44:59 3833 couldn't continue when handling __libc_start_main (0x450380) at 0x450380 2022-09-02 17:45:05 Not sure what to make of this 2022-09-02 17:45:12 Am I missing some debug info? 2022-09-02 17:47:06 not sure 2022-09-02 17:47:08 given the same thing here https://github.com/postmarketOS/pmbootstrap/issues/1513#issue-326737191 this is probably some weird chroot stuff if it's not metal 2022-09-02 17:47:26 I'm not running in a chroot 2022-09-02 17:47:32 weird then 2022-09-02 17:47:50 also, htc-ace is using some wacky Android kernel 2022-09-02 17:48:16 mhm 2022-09-02 17:51:30 ok the tlsdesc access actually looks ok. just missing symbols makes it look bogus 2022-09-02 17:51:33 https://bugs.launchpad.net/ubuntu/+source/ltrace/+bug/771805 2022-09-02 17:51:40 I found this similar bug where they suggest using latrace instead 2022-09-02 17:51:54 But it doesn't seem we have that in Alpine 2022-09-02 17:52:17 also, this bug is supposedly fixed 2022-09-02 17:53:46 uhg this is stupid. the undefined thing is just objdump not being aware the PLT slot has bogus legacy interworking crap on it 2022-09-02 17:54:08 haha 2022-09-02 17:54:15 it's made to be callable from interworking-incompatible thumb code 2022-09-02 17:54:32 so it starts with bx pc 2022-09-02 17:54:40 in thumb 2022-09-02 17:55:11 Newbyte: it was fixed in 0.7.3-largeubuntupkgrel 2022-09-02 17:55:21 no release since then and we also have 0.7.3 2022-09-02 17:55:33 oh … 2022-09-02 17:55:42 maybe we need to cherrypick some patch then 2022-09-02 17:56:18 if you find it, sure 2022-09-02 17:58:50 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=425fc47adb5bb69f76285be77a09a3341a30799e 2022-09-02 17:58:55 I suppose the reason it doesn't work is this 2022-09-02 17:59:18 PTRACE_SINGLESTEP isn't emulated on 32-bit ARM 2022-09-02 17:59:29 so when ltrace tries to use it, things go bork 2022-09-02 18:00:18 guess u just can't use ltrace 2022-09-02 18:00:38 that's ok i guess 2022-09-02 18:01:21 Newbyte: anyway does it crash, and does 22.2 crash 2022-09-02 18:02:49 let's find out 2022-09-02 18:05:36 ok, i think i found bug and it looks like a compiler bug 2022-09-02 18:05:48 in glGetString.. 2022-09-02 18:06:01 417d2: ee1d 3f70 mrc 15, 0, r3, cr13, cr0, {3} 2022-09-02 18:06:01 417d6: f7d1 ede6 blx 133a4 2022-09-02 18:06:02 417da: 581b ldr r3, [r3, r0] 2022-09-02 18:06:22 afaict that's loading the thread pointer to *r3* then calling the out-of-line tlsdesc call thunk at the end of the PLT 2022-09-02 18:06:35 but that thunk expects the thread pointer in r0 2022-09-02 18:07:17 we did try gcc11 too.. weh 2022-09-02 18:07:21 so, as in the bug report, it's jumping to a random code address from whatever r0 originally pointed to 2022-09-02 18:07:26 aha 2022-09-02 18:07:27 yea 2022-09-02 18:07:50 i think gcc has a wrong/missing rtx constraint that this has to be r0 2022-09-02 18:07:58 so the register allocator allocates whatever it wants... 2022-09-02 18:08:17 or it's wrongly using the r3 from the output as the input... 2022-09-02 18:08:38 psykose: unfortunately 22.2-rc3 segfaults 2022-09-02 18:08:43 sure 2022-09-02 18:08:50 ok, so as a sanity check lets try clang 2022-09-02 18:11:58 on 22.2 in a bit 2022-09-02 18:12:17 staring down that CI right now 2022-09-02 18:12:30 14: error: linker command failed with exit code 1 (use -v to see invocation) 2022-09-02 18:12:32 oh come on 2022-09-02 18:12:57 * clang-14: error: 2022-09-02 18:13:06 haha 2022-09-02 18:13:37 next one 2022-09-02 18:13:40 hmm maybe i'm misinterpreting this.. 2022-09-02 18:14:00 yeah the thread pointer is not the arg... sorry 2022-09-02 18:14:20 is ok, more testing never hurts 2022-09-02 18:14:30 and the arg is in r0 2022-09-02 18:16:00 Newbyte: there 2022-09-02 18:17:30 hmm the argument value seems to be wrong tho 2022-09-02 18:18:08 it seems to be off by 0x3c and thereby pointing at some other junk in the got 2022-09-02 18:18:42 probably address of xcb_present_id 2022-09-02 18:18:42 wow 2022-09-02 18:18:46 this is crazy 2022-09-02 18:18:52 it doesn't crash 2022-09-02 18:18:53 no segfault 2022-09-02 18:19:04 i guess it's a gcc bug then, or.. 2022-09-02 18:19:10 does clang even support tlsdesc/gnu2? 2022-09-02 18:19:11 clang doesn't add -mtls-dialect=gnu2 2022-09-02 18:19:12 no 2022-09-02 18:19:19 i can try gcc and removing that 2022-09-02 18:19:19 i don't think it's a gcc bug 2022-09-02 18:19:28 i think it's ld computing the displacement for a constant wrong 2022-09-02 18:19:41 when it resolves the tlscall reloc 2022-09-02 18:19:44 that would make sense 2022-09-02 18:20:00 someone was trying to wade through assembly in gdb and to him it seemed like a constant was being called as a function 2022-09-02 18:20:03 or something like that 2022-09-02 18:20:03 -mtls-dialect=desc is the default right? just to override it 2022-09-02 18:20:03 we can probably make a minified testcase for this 2022-09-02 18:20:14 -mtls-dialect=gnu 2022-09-02 18:20:24 is the non-tlsdesc choice 2022-09-02 18:20:51 i'm a bit late here but i think compiling mesa with -mtls-dialect=gnu will probably fix it on the basis that afaik alarm and "raspberry pi os" haven't complained about it and i think they're using -mtls-dialect=gnu 2022-09-02 18:21:30 yeah i think alpine's job is just to turn off tlsdesc and report the bug 2022-09-02 18:21:38 ok, pushed just gnu 2022-09-02 18:21:41 not to fix binutils :) 2022-09-02 18:22:01 hopefully we can make a minified testcase that has nothing to do with mesa to report to binutils and get it fixed 2022-09-02 18:22:09 Newbyte: new artifacties for you 2022-09-02 18:22:23 kind of unfortunate that tlsdesc seems broken everywhere though. afaict it's broken in glibc and apparently now also ld 2022-09-02 18:22:31 is it strange it's just v7? aarch64 isn't affected 2022-09-02 18:22:37 yay, more artefacts for me 2022-09-02 18:22:44 psykose, no, not strange at all 2022-09-02 18:22:54 they're completely different and unrelated archs 2022-09-02 18:23:19 psykose: actually, which are you referring to? 2022-09-02 18:23:22 and also tlsdesc is and i think always has been the default on aarch64, so it's much better tested 2022-09-02 18:23:25 new push, same place 2022-09-02 18:23:28 22.2 2022-09-02 18:23:33 sets =gnu 2022-09-02 18:23:52 it seems like build is failing? 2022-09-02 18:24:05 on some arches at least 2022-09-02 18:24:17 not for armv7 2022-09-02 18:24:20 whereas i suspect the only people using -mtls-dialect=gnu2 are gentoo ricers and now also mesa, and the former probably don't use it on arm32 2022-09-02 18:24:24 the only one that matters 2022-09-02 18:25:19 aren't you a gentoo ricer? you tell us 2022-09-02 18:25:20 ;) 2022-09-02 18:25:55 i use -mtls-dialect=gnu2 but on x86 which probably works better than arm 2022-09-02 18:26:18 yeah if just =gnu fixes it i will set it for armhf/v7 2022-09-02 18:26:20 there's probably also some inverse correlation between gentoo ricers and intel gpus 2022-09-02 18:26:31 Intel GPUs? 2022-09-02 18:26:42 what does Intel have to do with anything 2022-09-02 18:26:53 psykose: no segfault 🥳 2022-09-02 18:26:57 nice 2022-09-02 18:27:08 ok, guess it's a weird binutils gnu2 bug 2022-09-02 18:27:10 the combination of glibc + mesa iris + -mtls-dialect=gnu2 is/was broken 2022-09-02 18:28:25 it is "fixed" in mesa 22.2 by deleting the thread-local variables from mesa iris 2022-09-02 18:29:09 is testing this on x86 still relevant? 2022-09-02 18:29:57 it probably works 2022-09-02 18:30:45 anyway, pushed, now just gnu for arm* 2022-09-02 18:30:52 feel free to keep testing 22.2 for other issues ;) 2022-09-02 18:30:58 and someone can make the binutils report 2022-09-02 18:31:17 contemplating whether to use this as an excuse to finally replace Void with Alpine on my x86 netbook 2022-09-02 18:31:36 alpine is just void but things aren't half a year out of date /s 2022-09-02 18:31:51 but void has glibc 2022-09-02 18:32:14 i need my glibc fix 2022-09-02 18:32:23 I installed Void because it was the first Linux distribution I found that still supported x86 2022-09-02 18:32:25 the only point for glibc i can think of is nvidia drivers (i sadly have a much better gpu collecting dust) 2022-09-02 18:32:29 alpine is just void, but it also sucks for desktop distro 2022-09-02 18:32:31 the rest is.. workable in 50 ways 2022-09-02 18:32:36 flatpaks, containers, chroots.. 2022-09-02 18:32:47 ofc, more messy, but eh 2022-09-02 18:32:48 but muh systemd 2022-09-02 18:32:55 psykose: what GPU is it? 2022-09-02 18:32:58 gtx 1070 2022-09-02 18:33:07 aw, that's too bad 2022-09-02 18:33:12 well 2022-09-02 18:33:13 you see 2022-09-02 18:33:18 if it was newer it could've been supported by Nouveau soon 2022-09-02 18:33:18 computers are magic 2022-09-02 18:33:27 like actually well-supported 2022-09-02 18:33:33 and with my measly rx460, i used to get 2022-09-02 18:33:41 maybe 30-40fps on random shitty steam games 2022-09-02 18:33:53 1440p/170hz display, windowed random size 2022-09-02 18:33:55 and then 2022-09-02 18:33:57 half a year passed 2022-09-02 18:34:01 and now i get like 100+fps 2022-09-02 18:34:10 mesa developers be working eh :p 2022-09-02 18:34:36 it's ok these days, but yeah, wish i could use the 1070 instead 2022-09-02 18:34:41 i don't think nouveau would ever support it 2022-09-02 18:34:48 i've never heard of a single person having stable nouveau >60hz 2022-09-02 18:34:54 it's just constant breakage 2022-09-02 18:35:14 the firmware situation is looking grim for the 1000-series 2022-09-02 18:35:16 and 900-series 2022-09-02 18:35:33 perhaps, i don't follow the news there 2022-09-02 18:35:35 2000-series and newer have a good chance of getting good 2022-09-02 18:35:43 like perfectly usable 2022-09-02 18:36:03 unfortunately I too drew the short straw since I have a 970 2022-09-02 18:36:20 there's a 970 in my girlfriends moms house doing nothing too, hah 2022-09-02 18:36:31 i think that weird 3.5g model (was that the scam one?) 2022-09-02 18:36:35 i had gtx 660 which hangs sometimes if using nouveau 2022-09-02 18:36:54 all of them are the scam one 2022-09-02 18:36:56 :) 2022-09-02 18:37:01 iirc one had really broken ram 2022-09-02 18:37:26 last 0,5 GB is super slow 2022-09-02 18:37:27 by design 2022-09-02 18:42:11 yea 2022-09-02 18:48:15 skarnet: ok will check shortly 2022-09-02 19:23:29 would it be possible to get the .o files from the broken tlsdesc build? 2022-09-02 19:31:59 you'd have to build it yourself 2022-09-02 19:32:33 or i can run stuff on the ci and upload garbage intentionally, but then i have to figure out how to just copy all the .o files 2022-09-02 20:11:58 i'll just see if i can come up with something 2022-09-02 20:16:50 ah, almost forgot about it 2022-09-02 20:16:53 i'll try some stuff 2022-09-02 20:21:34 just .o's or any other extensions 2022-09-02 20:23:48 small issue is it's huge and above the 300m limit 2022-09-02 20:23:49 hmm 2022-09-02 20:30:47 shrank it a bit 2022-09-02 20:35:59 dalias: https://gitlab.alpinelinux.org/alpine/aports/-/jobs/826083 .o's in one folder 2022-09-02 20:36:03 no .so's, think that's too big 2022-09-02 20:36:34 no debug info either 2022-09-02 20:43:29 thanks 2022-09-02 21:32:57 psykose, hmm, nothing looks weird in the .o file. so pretty sure it's ld's fault not as's 2022-09-02 21:33:10 mhm 2022-09-02 21:33:51 well, now to just make that smol testcase i suppose 2022-09-02 21:34:00 yeah.. 2022-09-02 21:34:13 btw what ld version? 2022-09-02 21:34:25 so i can test the same trying to reproduce it 2022-09-02 21:34:46 i would assume the bug has been around a while and is the same, but i don't want to get stuck unable to reproduce it because different version lacks the bug :-p 2022-09-02 21:34:48 2.39 binutils 2022-09-02 21:35:06 ok that's what i installed on alpine cross package 2022-09-02 21:35:22 you can look at https://git.alpinelinux.org/aports/tree/main/binutils?id=0a431ba64a46b632ea9fd418ffed3ea0f8fea79d for the ref 2022-09-02 21:36:12 the cross comes from elsewhere: https://git.alpinelinux.org/aports/tree/testing/binutils-cross?id=639b47215ab548b59253798dbcf917c87156688f but it's the same config more or less 2022-09-02 21:36:17 and isn't used for that actual build 2022-09-02 21:36:29 you're just using it for objdump; actual build is the above native one 2022-09-02 21:44:55 ok, just linking this .o file by itself as a .so 2022-09-02 21:44:59 let's see if that gets mislinked 2022-09-02 21:45:59 ok this linked right... 2022-09-02 21:58:05 wait... it might actually be gcc.... 2022-09-02 22:08:38 the offsets, when small, *look like* someone printed a value that's supposed to be hex in decimal 2022-09-02 22:08:56 but they might just be off by 6 from what i expect, and i'm not sure if that's right or wrong 2022-09-02 22:20:23 i think it may just be the wackiness desscribed in "Addend Adjustments" here: https://www.fsfla.org/~lxoliva/writeups/TLS/RFC-TLSDESC-ARM.txt 2022-09-02 22:21:30 certainly reads like magic 2022-09-02 22:25:54 ok with just entry.o linked it comes out fine. the value the linker puts in the .so is what results in computing the right address 2022-09-02 22:26:02 so it's not a compiler or assembler bug 2022-09-02 22:41:47 ok in the mislinked file, all of the offsets are off by the same amount, 0x3c... 2022-09-03 06:48:19 cramfs fails to install 2022-09-03 06:48:21 ERROR: cramfs-1.1-r4: failed to rename .apk.4e0a43cf207ae89d30bad1edda70408d4239a6f180200a4b to sbin. 2022-09-03 07:35:33 uh, so when doing an alpine upgrade yesterday from 3.15 -> 3.16 it overwrote all my config files in /etc 2022-09-03 07:35:43 had to reconfigure the entire server :v 2022-09-03 07:40:01 wasn't me 2022-09-03 07:41:46 (otoh, I used it as an excuse to redo my smtpd config, so not all bad) 2022-09-03 08:09:50 apk seems to reliably overwrite my configs now :/ 2022-09-03 08:10:46 as in, it places an .apk-new, but also overwrites the non .apk-new file 2022-09-03 08:19:54 Huh 2022-09-03 08:20:03 That's unexpected 2022-09-03 08:23:47 yes, it's very annoying :p 2022-09-03 10:55:47 Dodged the bullet with ansible 2022-09-03 13:12:30 did you place a -/etc in /etc/apk/protected_paths.d xD 2022-09-03 15:16:20 back from holiday 2022-09-03 15:24:56 Yay 2022-09-03 18:36:48 the apkindex for v3.16/main is somehow broken on nl.a.o 2022-09-03 18:37:10 nl.a.o is not an official mirror anymore 2022-09-03 18:37:34 It will be removed soon 2022-09-03 20:17:28 Can I get the build timeout of https://gitlab.alpinelinux.org/dhruvin/aports/-/jobs/827571 increased to 2 hours? 2022-09-03 20:17:57 There are three packages in this MR and aarch64 build is taking significant time. 2022-09-03 20:18:18 I can also split the MR into three if increasing timeout is not possible. 2022-09-03 20:19:09 https://gitlab.alpinelinux.org/dhruvin/aports/-/settings/ci_cd 2022-09-03 20:19:28 Note that adjusting it will only affect new jobs 2022-09-03 20:20:34 thanks ikke! 2022-09-03 21:58:36 a16bitsysop: i /almost/ have it working with only 1 last step needed 2022-09-03 21:58:41 this firmware is cursed 2022-09-03 22:01:22 ah, nvm, it builds if you mix toolchains 2022-09-03 23:18:32 psykose: great, I thought with CC=clang it uses all the lvm stuff? 2022-09-03 23:18:59 for that specific thing only, yes 2022-09-04 07:14:46 at last, grep now prints warnings for egrep/fgrep 2022-09-04 08:12:34 Ariadne: thanks for the merge! Now we just need alpine-conf!100 to be merged, to keep consistent with the package naming scheme, and my work will be (temporarily) done. :) 2022-09-04 08:12:48 :D 2022-09-04 08:12:54 no, algitbot. You should really learn the gitlab naming scheme. 2022-09-04 08:12:56 We need to teach algitbot about gl prefixes 2022-09-04 08:12:58 :) 2022-09-04 08:13:13 yeah, for ! *and* for # 2022-09-04 08:13:17 yes 2022-09-04 08:13:28 defaulting to aports is good 2022-09-04 08:13:37 but only addessing aports is meh 2022-09-04 08:13:42 addessing? 2022-09-04 08:13:45 https://gitlab.alpinelinux.org/alpine/infra/compose/algitbot/-/blob/master/sircbot/scripts/sircbot.lua if someone feels inclined 2022-09-04 08:15:43 i added a custom thing into there already hotpatched, hehe 2022-09-04 08:15:49 but yeah, could just throw in a regex match 2022-09-04 08:15:57 feel free skarnet 2022-09-04 08:16:35 i would go for [reposcheme]!/#[number] 2022-09-04 08:16:36 will do when I have functioning brains, which might not happen before two weeks 2022-09-04 08:17:01 so aports ! 300 is the same as ! 300 , but you can also do abuild ! 300 , or infra/xyz ! 300 2022-09-04 08:17:13 that was the plan yes 2022-09-04 08:17:17 aye 2022-09-04 08:17:25 abuild!300 2022-09-04 08:17:31 not yet :) 2022-09-04 08:17:38 indeed 2022-09-04 08:25:19 bl4ckb0ne: do you know anything about glfw being broken as fuck on wayland all the time 2022-09-04 08:25:45 what isnt broken on wayland all the time 2022-09-04 08:25:47 sometimes it works by magic and when it doesn't the glfwterminate segfault trace is hilarious, it can show up anywhere from the application to llvm to mesa 2022-09-04 08:26:22 i dunno i've just never had it work good and i have no idea why 2022-09-04 08:26:47 like it was just failing for 40 minutes.. and now it works 2022-09-04 08:26:48 no changes 2022-09-04 08:26:54 ive used it a bit recently without any issues beside the fact that the api sucks 2022-09-04 08:27:05 i'm not really using it just.. applications and shit 2022-09-04 08:27:20 do you have the stack trace 2022-09-04 08:27:26 it's random every time and useless 2022-09-04 08:27:39 at best you see a glfwterminate in it amongst random shit 2022-09-04 08:29:15 doesn't the randomness point to a multithreading issue? 2022-09-04 08:29:27 mb ur just bad 2022-09-04 08:29:55 bruh i ain't even write the code 2022-09-04 08:29:57 the segfault happens in one thread, the trace serializes in an unpredictable way 2022-09-04 08:30:09 can be 2022-09-04 08:30:43 it's magically fixed itself now so i can't get you anything 2022-09-04 08:41:01 which software stacktraces 2022-09-04 08:57:01 anything with glfw, when it happens 2022-09-04 09:06:42 so since I was waiting for breakfast I took a look at how gitlab provides its information to algitbot and I'm already regretting it 2022-09-04 09:06:58 what is that api even 2022-09-04 09:07:39 /projects doesn't give a list of projects, it gives a list of... something, that's very, very incomplete 2022-09-04 09:08:04 is there even a way to get the list of projects? (I need to match the name against the id, to construct the correct url) 2022-09-04 09:08:28 (I mean make a project name -> project id map) 2022-09-04 09:08:41 I could hardcode such a map but it would be much better to get it from gitlab 2022-09-04 09:09:00 the "docs" don't give the singlest hint on where to find that 2022-09-04 09:09:23 what would you need any of that for 2022-09-04 09:09:41 To get the title of the MR / issue 2022-09-04 09:09:53 skarnet: you can get projects by name 2022-09-04 09:09:56 the url is just https://gitlab.alpinelinux.org/alpine/$leftpart/merge_requestsorissues/$rightpart 2022-09-04 09:10:01 ah 2022-09-04 09:10:03 right 2022-09-04 09:10:04 you need to encode the / in the namespace as %2f 2022-09-04 09:10:06 the title 2022-09-04 09:10:16 oh, I wonder, maybe because in the gitlab_merge_request and gitlab_issue functions there is a url where "1" is helpfully hardcoded, and "1" is aports? 2022-09-04 09:10:27 skarnet: yes, 1 is aports 2022-09-04 09:10:31 but you can do it by name 2022-09-04 09:10:50 https://docs.gitlab.com/ee/api/projects.html#get-single-project 2022-09-04 09:10:51 that's not how algitbot does it currently 2022-09-04 09:10:56 The ID or URL-encoded path of the project. 2022-09-04 09:11:13 oh. 2022-09-04 09:11:17 indeed it was hardcoded 2022-09-04 09:11:32 so projects/1 could be replaced with project/aports? 2022-09-04 09:11:32 can you also port the whole bot to something that doesn't exec a lua script on every request 2022-09-04 09:11:35 skarnet: yes 2022-09-04 09:11:51 oh and can you rewrite it in rust 2022-09-04 09:12:02 oh and i think you want to use graphql for these 2022-09-04 09:12:04 psykose: and a blow job and a Mars bar 2022-09-04 09:12:07 graphql is mehg 2022-09-04 09:12:09 meh 2022-09-04 09:12:14 oh and http/3 only 2022-09-04 09:12:31 make something simple into something really complex 2022-09-04 09:12:44 graphql is nice for multiple sources that you can't get in one request and that's about it 2022-09-04 09:12:48 you know that's my shtick and my whole plan for Alpine 2022-09-04 09:12:54 introduce complexity and tech debt everywhere 2022-09-04 09:13:00 you got it 2022-09-04 09:13:00 skarnet: oh, then go ahead 2022-09-04 09:13:09 i cannot wait to press the merge button 2022-09-04 09:13:25 (i don't think i actually can) 2022-09-04 09:13:36 then again that latter part is already there https://gitlab.alpinelinux.org/alpine/infra/compose/algitbot/-/merge_requests/1 2022-09-04 09:13:48 and already uses the correct path 2022-09-04 09:13:55 you just have to parametrise that 2022-09-04 09:13:56 nothing else 2022-09-04 09:15:01 the things that the gitlab api exposes is still whack 2022-09-04 09:15:02 What is the benefit of that? 2022-09-04 09:15:12 look at https://gitlab.alpinelinux.org/api/v4/users 2022-09-04 09:15:12 I guess a single request instead of multiple 2022-09-04 09:15:21 more faster less slower 2022-09-04 09:15:22 does this look like a list of current gitlab users to you 2022-09-04 09:15:42 it does 2022-09-04 09:15:48 complete? no 2022-09-04 09:15:54 but they certainly are current gitlab users 2022-09-04 09:15:55 ;) 2022-09-04 09:16:04 skarnet: It does to me as well 2022-09-04 09:16:19 >In de Hekken 2022-09-04 09:17:18 why would that endpoint report such a sample 2022-09-04 09:17:32 skarnet: it's paginated 2022-09-04 09:17:51 come again? 2022-09-04 09:17:56 Do you consider sourcehut? (At least when web-based patches are ready) 2022-09-04 09:18:14 https://docs.gitlab.com/ee/api/#pagination 2022-09-04 09:18:16 the headers tell you x-next-page 2 2022-09-04 09:18:19 x-page 1 2022-09-04 09:18:22 x-per-page 20 2022-09-04 09:18:32 this has been standard for.. 15 years? 2022-09-04 09:18:53 maybe not in implementation but in general 2022-09-04 09:19:34 and of course there's no way to get the whole list in one go? 2022-09-04 09:19:42 because fuck you developers, that's why 2022-09-04 09:19:53 we want you to make 400 https requests 2022-09-04 09:20:07 and complexify your client code 2022-09-04 09:20:25 what? 2022-09-04 09:21:13 Isn't it anti-ddos-related? 2022-09-04 09:21:18 Ermine: also 2022-09-04 09:21:27 love 2 get response of 1,523,698 entries in one go 2022-09-04 09:21:41 skarnet: the alternative is to just return the list of usernames / user ids and having to request details of each and every user as a separate request 2022-09-04 09:22:05 if only 20 people lived in europe maybe it would be reasonable 2022-09-04 09:23:51 well it's either that or force devs to do 15236 requests anyway 2022-09-04 09:24:05 what's the heaviest load on the server, I wonder 2022-09-04 09:24:23 ikke: yes, that's exactly what I'm looking for, that's what would be reasonable 2022-09-04 09:24:36 i don't see why you need to fetch users to get a title of an issue page 2022-09-04 09:24:46 psykose: it was just an example 2022-09-04 09:25:08 more seriously i actually can't think of any case ever where you want 'all the users' 2022-09-04 09:25:31 unless you wanted to index them into something, which is not that common for random clients 2022-09-04 09:25:38 the entire endpoint shouldn't exist if anything 2022-09-04 09:26:13 my point is 2022-09-04 09:26:27 if you're going to provide something, don't force it paginated 2022-09-04 09:26:34 have an option to get the full list 2022-09-04 09:26:51 because arbitrary restrictions like that help nobody 2022-09-04 09:28:15 They help administrators not having to deal with queries that grind your gitlab instance to a halt 2022-09-04 09:28:54 per request ratelimiting is a lot easier than.. network limiting 2022-09-04 09:28:55 graphql does make it easier to select what data you want 2022-09-04 09:28:57 which is next to worthless 2022-09-04 09:29:58 people who want to grind your instance to a halt will grind your instance to a halt 2022-09-04 09:30:11 skarnet: it's not for people who _want_ to 2022-09-04 09:30:33 but about people who fire queries without thinking 2022-09-04 09:30:56 it works on my small dev instance 2022-09-04 09:31:10 and then explodes on production 2022-09-04 09:31:11 you can already do that for cgit because of how slow it is at doing anything, haha 2022-09-04 09:33:47 is there really no urlencode function in the lua stdlib 2022-09-04 09:33:56 lua stdlib is really poor 2022-09-04 09:33:59 nope 2022-09-04 09:34:08 it's meant as an embedded language 2022-09-04 09:34:24 so they have only very basic stdlib, and for the rest you have to defer to other libs 2022-09-04 09:34:38 very basic in the sense of 'worse than c' 2022-09-04 09:34:45 unpossible 2022-09-04 09:34:46 let alone uri-level stuff 2022-09-04 09:34:55 the C stdlib is nonexistent 2022-09-04 09:34:59 with very few exceptions 2022-09-04 09:35:01 and it's better than lua 2022-09-04 09:36:02 well I'm neither yolo downloading a urlencode from the internets nor writing one today, so on to video games 2022-09-04 09:36:16 just sub / with %2f :P 2022-09-04 09:36:27 sure 2022-09-04 09:36:45 why do things well when mediocrity is so much faster and easier 2022-09-04 09:36:54 you get it 2022-09-04 09:36:58 nobody will notice and it will work! 2022-09-04 09:37:23 until it doesn't and then you get people like me to clean shit up and nobody likes it 2022-09-04 09:37:48 if you want to do it well, you'd include something like penlight 2022-09-04 09:37:56 lua doesn't even have regex 2022-09-04 09:37:58 haha 2022-09-04 09:38:17 nope, indeed 2022-09-04 09:38:31 everything to reduce the size 2022-09-04 09:38:32 regex is complex, tbh that was my biggest surprise with C, that regex would be in the stdlib 2022-09-04 09:40:17 https://lunarmodules.github.io/Penlight/libraries/pl.url.html 2022-09-04 09:40:55 that's cool 2022-09-04 09:41:31 not going to import the kitchen sink and associated nuclear plant into algitbot though 2022-09-04 09:42:32 nukes r cool 2022-09-04 10:03:39 psykose: well damn, there is actually an /etc in there thanks to lbu :( 2022-09-04 10:04:14 it has to specifically be `-etc/..` 2022-09-04 10:04:22 it was `-/etc` 2022-09-04 10:04:23 lol 2022-09-04 10:04:24 the default includes it anyway (protection), the - removes it 2022-09-04 10:04:24 ah 2022-09-04 10:04:26 haha, yeha 2022-09-04 10:04:28 unlucky 2022-09-04 10:04:38 good call, thanks 2022-09-04 10:05:05 time to apk fix $(apk info) 2022-09-04 10:06:05 apk fix '*' 2022-09-04 10:07:52 nice :p 2022-09-04 10:12:59 what does apk info without arguments? 2022-09-04 10:14:46 try it and see 2022-09-04 10:15:55 already tried and seen... just want to avoid discovering whats it :P 2022-09-04 10:16:05 I also read 'man apk-info' 2022-09-04 10:16:12 current installed 2022-09-04 10:16:47 ok 2022-09-04 10:17:31 1054 packages, definitevely too many :D 2022-09-04 10:18:57 psykose: I asked on #remmina about the RDP/win7 issue, it's just due openssl3 disabling weak algorithms 2022-09-04 10:19:16 well yes but surely microsoft rdp supports some baseline tls1.2 2022-09-04 10:19:42 or is it like that exact issue where since 1.0 is disabled it aborts everything without then trying to upgrade into 1.2 2022-09-04 10:20:17 uhM, this VM is running offline since... 2022-09-04 10:20:36 maybe 2014 2022-09-04 10:21:41 could be it just too outdated? 2022-09-04 10:24:26 well, I only use it on localhost, disabling it seems fine 2022-09-04 10:24:47 the update button is there for a reason 2022-09-04 10:26:00 it's better running offline... :) 2022-09-04 14:00:03 PureTryOut: !38006 2022-09-04 14:02:43 you might want USE_NFTABLES to make it make lxc use nft by default 2022-09-04 14:02:46 also USE_SYSTEMD=0 2022-09-04 14:34:50 has anyone tried to use libinput with old PS/2 keyboards? I am trying to get swaywm running on an old EeePC laptop. Both keyboard and touchpad are connected internally via PS/2, if I understand correctly. 2022-09-04 14:36:16 i haven't, though i'm curious what the issue is 2022-09-04 14:36:23 The keyboard works fine in the TTY, but libinput won't detect it. Connecting a USB keyboard "solves" the issue, but defeats the purpose of a tiny laptop 2022-09-04 14:36:23 also, did you have an rk3399 for the arm firmware thing 2022-09-04 14:36:30 ah, hmm 2022-09-04 14:38:09 you can try the libinput debugging utilities (/usr/libexec/libinput/libinput-debug-events etc), not sure i know anything specific sadly 2022-09-04 14:38:30 I remember back then I had to add manual X.org configs for getting the internal input running 2022-09-04 14:40:01 https://bbs.archlinux.org/viewtopic.php?id=274130 this looks related 2022-09-04 14:41:31 `libinput debug-events` shows three `DEVICE_ADDED ` events for the USB keyboard, if it is plugged in. If it is not plugged it, I get a warning that I normally `DEVICE_ADDED` events should show up, but there are none. (And indeed, no such events will appear until I plug in the USB keyboard.) 2022-09-04 14:58:26 i think you'd have to do hardware magic to replug ps/2 2022-09-04 14:58:32 but yeah sorry i can't guess anything :( 2022-09-04 15:07:41 interestingly, event the USB keyboard is not detected anymore if already connected during boot. I guess I just somehow messed up udev. For some reason I see busybox mdev being started early at boot, and udev later on. To my understanding it should be mutually exclusive, shouldn't it? 2022-09-04 15:11:51 it's possible to manually enable one 2022-09-04 15:11:56 and end up with both 2022-09-04 15:12:02 afaik they won't super conflict, shrug, but yeah 2022-09-04 15:13:48 hmm, only udev running doesn't change a bit. (Also the keyboard is only detect if plugged out and back in after boot :() 2022-09-04 15:14:08 At least it seems not to be a strange PS/2 issue 2022-09-04 15:16:25 ran into similar issues before, never knew why either 2022-09-04 15:16:28 hardware just sucks sometimes 2022-09-04 15:27:24 PureTryOut: did you get yuzu to work? seems to be fully complete for me, minus actual runtime testing given lack of some illegally dumped keys 2022-09-04 15:28:32 oh also !38596 in case you wanted to have a look, still needs testing 2022-09-04 15:40:15 I built a u-boot blob in .itb format, but according to https://opensource.rock-chips.com/wiki_Boot_option it also needs idbloader.img which is built from a file in the tpl directory which does not exist after compilation. Some websites say you can just flash a ‘u-boot.bin’ file to the start of the sd like armula https://wiki.amarulasolutions.com/bsp/rockchip/rk3399/roc-rk3399-pc.html. But then I think you need a serial cable to 2022-09-04 15:40:15 test it. 2022-09-04 15:41:36 whadya mean doens't exist 2022-09-04 15:43:10 that wiki just says it comes from u-boot 2022-09-04 15:44:05 with those build commands 2022-09-04 15:44:13 the tpl directory has `include u-boot.cfg` it is missing tpl/u-boot-tpl.bin 2022-09-04 15:46:54 i'm sure it's possible to find somewhere 2022-09-04 15:50:48 will try and find why it isn’t built 2022-09-04 15:52:31 well it's in https://github.com/rockchip-linux/rkbin 2022-09-04 15:52:39 i would guess 2022-09-04 15:54:30 this hardware is a disaster 2022-09-04 15:54:54 anyway if you find whatever and need help debugging some build error let me know 2022-09-04 15:56:43 okay thanks, yes hopefully riscv64 is less of a nightmare! 2022-09-04 15:57:44 rv64 is blocked on rust 2022-09-04 15:58:19 nah, gonna be the exact same shit 2022-09-04 15:58:26 rv is just arm but worse and more expensive 2022-09-04 16:00:07 upgrading gitlab 2022-09-04 16:01:17 smol patch 2022-09-04 16:01:33 or did you go for 15.3 2022-09-04 16:01:35 no 2022-09-04 16:01:39 15.1.6 still 2022-09-04 16:01:50 fer shame 2022-09-04 16:01:56 i wanted more random css changes!! 2022-09-04 16:02:04 😥 2022-09-04 16:02:06 :) 2022-09-04 16:02:12 hope nothing breaks 2022-09-04 16:05:01 ok, running again 2022-09-04 16:07:13 ey guys, I would like to create bubblejail-gui subpackage and avoid that main bubblejail depends on qt&dbus, could this be achieved with APKBUILD? bubblejail is mostly python code 2022-09-04 16:09:28 how do you invoke the gui 2022-09-04 16:09:49 bubblejail-config 2022-09-04 16:10:01 it could be named this way too 2022-09-04 16:10:01 ah, i see 2022-09-04 16:10:08 it's already there, just needs 'pyqt5' 2022-09-04 16:10:12 make a subpackage 2022-09-04 16:10:17 mkdir -p "$subpkgdir" 2022-09-04 16:10:29 add depends="$pkgname=$pkgver-rpkgrel py3-pyqt5 .." 2022-09-04 16:10:43 r$pkgrel but you get the idea 2022-09-04 16:10:56 i'd name it config, sure 2022-09-04 16:11:06 uhM 2022-09-04 16:11:25 and you can then amove the /usr/bin/bubblejail-config too instead of mkdir i guess 2022-09-04 16:11:31 I'm gonna try to find something similar 2022-09-04 16:12:02 i'm not sure if I understand the depends= line 2022-09-04 16:12:18 you just add depends= with that in it in the subpackage 2022-09-04 16:12:37 in the subpackage() func? 2022-09-04 16:13:22 yes 2022-09-04 16:13:24 oh great, thanks 2022-09-04 16:22:32 if I need to amove a file on /usr/lib/python3.X/site-packages/foo... , can I do it version independent? 2022-09-04 16:22:53 You should be able to use wildcards 2022-09-04 16:23:02 ok 2022-09-04 16:25:34 exactly what you wrote but make the X a * 2022-09-04 16:25:55 (and escaped) 2022-09-04 16:26:06 nope 2022-09-04 16:26:08 oh 2022-09-04 16:26:34 uhM, let's try 2022-09-04 16:26:36 does that rely on nullglob? 2022-09-04 16:27:01 also nope 2022-09-04 16:27:20 why doesn't that expand to what's present on / rather than $pkgdir? 2022-09-04 16:28:48 i have no idea, but i assume there is nothing setting nullglob (there isn't) 2022-09-04 16:28:58 i guess it's some specificae of passing a glob as a function argument 2022-09-04 16:28:59 mv: can't rename '/home/donoban/aports/community/bubblejail/pkg/bubblejail/usr/bin/bubblejai-config': No such file or directory 2022-09-04 16:29:05 typo 2022-09-04 16:29:08 jai -> jail 2022-09-04 16:29:26 uhM, I thought that was failing the site-package part until paste it 2022-09-04 16:29:28 let me check 2022-09-04 16:30:22 or.. maybe you mean the opposite of nullglob, like it not erasing? 2022-09-04 16:30:29 lol, it's just a typo 2022-09-04 16:30:31 if so it sounds correct haha, it has 0 matches and stays the same 2022-09-04 16:31:07 indeed it's safer to quote, albeit specifically i can't think of a failure case given all the nesting 2022-09-04 16:32:54 psykose: right, apparently it's the default, if it does not match, it remains a '*' 2022-09-04 16:33:06 yeah, that's just normal (non-nullglob) behaviour 2022-09-04 16:33:15 nullglob set would make it turn into '' 2022-09-04 16:33:18 yeah 2022-09-04 16:33:27 so yeah, load bearing on shell specific 2022-09-04 16:44:19 lol, I pushed before commit :D 2022-09-04 16:47:40 !38631 2022-09-04 16:47:57 ouch, another typo 2022-09-04 16:48:59 fixed 2022-09-04 16:55:24 psykose: thx! 2022-09-04 16:55:30 :) 2022-09-04 17:12:50 if you don’t specify a target it builds idbloader.img u-boot.itb u-boot.bin and also u-boot.img 2022-09-04 17:13:07 what does 2022-09-04 17:13:15 u-boot 2022-09-04 17:13:31 no other changes at all, just main/u-boot? 2022-09-04 17:14:20 i am building from source, but yes no changes from upstream u-boot 2022-09-04 17:14:45 you can just call make idbloader.img 2022-09-04 17:14:57 or something with CONFIG_SPL=y 2022-09-04 17:16:02 yes that should work as well 2022-09-04 17:17:08 probably as part of rockpro64-rk3399? no idea 2022-09-04 17:35:09 its in each rockchip build, the current u-boot would build with the `all` target I guess https://git.alpinelinux.org/aports/tree/main/u-boot/APKBUILD#n107 2022-09-04 17:35:30 ye 2022-09-04 19:25:55 psykose: I got it to fully built, but it doesn't actually run any games for me. Just a black screen. It works fine from Flatpak however, I've yet to determine why 2022-09-05 07:52:58 Getting the following error with dabuild: https://paste.debian.net/1252796/ 2022-09-05 07:53:12 I've pulled a new version of alpinelinux/docker-abuild. Any suggestions? 2022-09-05 07:54:03 unless someone did something, dabuild images are outdated for a year now 2022-09-05 07:55:07 Oh. 2022-09-05 07:55:24 also edge keys were rotated not so long ago 2022-09-05 07:55:38 I guessed it was something like that panekj, which was why I tried pulling an update. 2022-09-05 07:55:42 https://alpinelinux.org/posts/Alpine-edge-signing-keys-rotated.html 2022-09-05 07:56:03 adhawkins: you first have to update alpine-keys 2022-09-05 07:56:23 first is to not use dabuild images 2022-09-05 07:57:53 panekj: Any suggestions on how I can do this then? 2022-09-05 07:58:29 I don't build under Alpine, so it's handy to have this docker mechanism. 2022-09-05 07:59:14 use just abuild and alpine image 2022-09-05 08:00:19 adhawkins: alpinelinux/build-base 2022-09-05 08:00:44 But it's just a base image for building packages, there is no automation 2022-09-05 08:01:18 build-base doesn't have all the stuff required for dabuild iirc 2022-09-05 08:01:37 yeah 2022-09-05 08:01:55 So is dabuild abandoned then? 2022-09-05 08:01:56 I can push updated dabuild repo# 2022-09-05 08:02:07 panekj: That would be perfect :) 2022-09-05 08:02:09 adhawkins: well, it's been adandoned for 2 years 2022-09-05 08:04:34 panekj: If you push an update, presumably that'll rebuild all the containers and publish them to docker? 2022-09-05 08:04:47 The problem is that CI is broken 2022-09-05 08:04:54 https://gitlab.alpinelinux.org/alpine/docker-abuild/-/pipelines 2022-09-05 08:04:55 Ah. 2022-09-05 08:04:56 it won't build them 2022-09-05 08:05:00 https://gitlab.alpinelinux.org/pj/docker-abuild/ 2022-09-05 08:05:04 turns out I already pushed it 2022-09-05 08:05:20 pushed where? 2022-09-05 08:05:40 the link I just posted....? 2022-09-05 08:06:18 panekj: Unless the containers are built and published, then I don't think that'll help will it? 2022-09-05 08:06:56 I can fix that but not now, later today I'll have some free time to deal with it 2022-09-05 08:07:14 since I rely on those images I might as well automate that 2022-09-05 08:07:14 That's perfect panekj. I'm just trying to update a couple of my packages that have had new releases. 2022-09-05 13:17:40 Ariadne: did you look into GCC 12.2 already? might make sense to have that for 3.17. would hope that it doesn't require much patching from our side 2022-09-05 13:18:57 looks like a quick rebase to me 2022-09-05 13:23:49 hmm I have a Rust package that in install fails fails to find some `.so`'s. I think they were meant to come from the package itself but I'm not sure, anyone any idea what's up? It probably fails to link them or something 2022-09-05 13:24:07 error? 2022-09-05 13:24:16 `ERROR: File 'libgstcdg.so' could not be found` 2022-09-05 13:25:21 are you building gstreamer plugins? 2022-09-05 13:25:30 ha yup, you guessed it right 2022-09-05 13:26:46 I'm not familiar with error, but gst-plugins-rs should build that file 2022-09-05 13:27:01 yeah well this package is gst-plugins-rs 😂 2022-09-05 13:27:06 maybe post apkbuild hint hint 2022-09-05 13:28:24 https://gitlab.alpinelinux.org/PureTryOut/aports/-/blob/testing_mopidy-spotify/testing/gst-plugins-rs/APKBUILD 2022-09-05 13:31:54 everything seems fine, idk why it fails, maybe build with without package step and check what it looks like in builddir? 2022-09-05 13:36:09 wtf did i just read 2022-09-05 13:36:33 uh oh, mum is angry 2022-09-05 13:37:18 without package step? 🤔 2022-09-05 13:37:49 oh just `abuild build`. Anyway to combine that with rootbld? 2022-09-05 13:39:41 hahahah 2022-09-05 13:41:13 hmm need to manually install the deps :( 2022-09-05 13:41:30 can't use `abuild`? 2022-09-05 13:42:39 I can but `abuild build` doesn't install the deps. `abuild -r` does that, but then does more and does the full build, prepare and packaging steps 2022-09-05 13:42:49 `abuild deps` 2022-09-05 13:43:06 ah useful, thanks! 2022-09-05 13:43:10 PureTryOut: the whole sequence is abuild deps fetch unpack prepare build package 2022-09-05 13:43:22 but I don't see option to run just build 2022-09-05 13:43:30 `abuild build` 2022-09-05 13:43:33 (altough, in practice it uses rootpkg, which does index as well) 2022-09-05 13:43:42 `abuild build` installs into `$pkgdir` 2022-09-05 13:43:46 and forgot check 2022-09-05 13:44:26 is there a reason to be using a random git commit here 2022-09-05 13:44:54 yes there is 2022-09-05 13:45:03 I won't get this into aports as is, don't worry 2022-09-05 13:49:47 the file error can't be found etc etc is most likely just the file not being in rpath but being needed in some form 2022-09-05 13:49:56 iirc those weren't even real errors but i could be wrong 2022-09-05 13:50:25 oh fun this is why I use rootbld for everything. it now complains my Rust is too old, because it comes from rustup rather than using the distro package... 2022-09-05 13:50:47 the easy solution to that is not having rustup installed ;) 2022-09-05 13:50:54 yeah... no 😂 2022-09-05 13:51:22 the easiest solution is to update rust via rustup 2022-09-05 13:54:31 well yeah, but I'd rather have it not use that rust at all lol 2022-09-05 15:17:52 panekj: i wonder how vulnerable rustup is to security issues. 2022-09-05 15:19:14 such as? 2022-09-05 15:19:33 could someone clever find a way to have rustup serve compromised binaries? 2022-09-05 15:19:57 its a much larger attack surface than distro packages 2022-09-05 15:20:06 it's not 2022-09-05 15:21:10 larger as in if someone compromised rustup a lot of people are affected, if someone compromises the alpine package far fewer people are affected 2022-09-05 15:22:00 all rustup tags are signed 2022-09-05 15:22:01 ah, sure 2022-09-05 15:22:05 that's not an attack surface though 2022-09-05 15:22:17 yea, probably used the wrong word 2022-09-05 15:22:35 dunno, wouldn't affect me 2022-09-05 15:22:39 they can hack rustup all they want :) 2022-09-05 15:22:39 you would need to get through many hoops to get rustup or what rustup downloads to be compromised to do something 2022-09-05 15:23:20 it's probably harder than getting into the alpine builders 2022-09-05 15:24:46 Just takes one mistake, but I would hope you are right 2022-09-05 15:25:05 Just like alpine 2022-09-05 18:51:41 panekj: Don't mean to pester, but wondered if there was any update on the dabuild container? 2022-09-05 18:53:11 can't you build an updated one locally and tag it 2022-09-05 18:53:59 i have my own alpine container for aports, you can try that 2022-09-05 19:01:51 a16bitsysop: How would I do that? 2022-09-05 19:02:05 with a little bit of magic 2022-09-05 19:02:15 psykose: I guess I could, would have to mess around with the dabuild script to use the appropriate one. 2022-09-05 19:02:31 didn't had much time to work on it, but I can build up to date images manually for now 2022-09-05 19:02:33 https://gitlab.com/a16bitsysop/alpine-dev-local 2022-09-05 19:02:58 it only has nano installed and of course vim is standard 2022-09-05 19:03:44 a16bitsysop: Ah sorry, I misunderstood. Ideally I want to use dabuild as it automates everything. 2022-09-05 19:04:29 oh, it automates package updates? 2022-09-05 19:04:52 it automates whole abuild handling 2022-09-05 19:05:00 and docker specific stuff 2022-09-05 19:12:09 It looks like its just abuild, by package updates I meant create merge requests etc? Or do you mean use it to update checksums and then use host for rest of it 2022-09-05 19:12:29 no it doesn't do that 2022-09-05 19:14:58 okay, 2022-09-05 19:15:48 but dabuild with its images handles all alpine versions and architectures 2022-09-05 19:15:49 psykose: How would I build the container locally and have the dabuild script use it? 2022-09-05 19:16:00 i've no idea as i've never used it 2022-09-05 19:16:07 you need to do some `make` bullshit in dabuild repo 2022-09-05 19:16:13 and then build with docker 2022-09-05 19:16:30 but if you are willing to wait 5 minutes, I'll have images uploaded to dockerhub 2022-09-05 19:18:38 That's perfect panekj, thanks so much. 2022-09-05 19:19:03 If someone can tell me what I need to run to update them in future locally, I'm happy to continue to do that. 2022-09-05 19:19:58 eh, I'll just need to add it to cron or some CI/CD 2022-09-05 19:20:30 not sure if the image name will exactly in dabuild since I started deviating to my own implementation (docker-abuild-rs) 2022-09-05 19:21:20 https://hub.docker.com/r/panekjdev/docker-abuild/tags 2022-09-05 19:22:51 panekj: Presumably I'll need to edit dabuild to use that container over the 'official' one? 2022-09-05 19:22:58 yeah 2022-09-05 19:23:23 It currently uses alpinelinux/docker-abuild, so just replace that with panekjdev/docker-abuild presumably? 2022-09-05 19:23:29 yep 2022-09-05 19:25:08 clandmeter: I don't remember who, but someone here mentioned they would be willing to fix/support docker-abuild, did anything happen in that regards? 2022-09-05 19:25:26 Thanks panekj, getting some errors in the test phase, but assume that's down to me! Thanks a lot. 2022-09-05 19:26:02 feel free to ping me with paste link, as I mentioned I started deviating a little bit since shell version wasn't working well for me on non-alpine 2022-09-05 19:26:49 (it's currently stitched with exec'ing to `docker`, but will rewrite it to proper API) 2022-09-05 19:27:04 ((if I feel like it, maybe I could write it in python)) 2022-09-05 19:34:59 panekj: Seems Ok now, was just a missing dependency in my package definition. Thanks again 2022-09-05 19:36:26 panekj: I think we first need to fix CI for that project 2022-09-05 19:36:44 clandmeter had dedicated CI runners for it that kinda went awol 2022-09-05 19:44:34 I wouldn't mind chucking in my host, since building those images isn't hard or intensive 2022-09-05 19:45:04 or if you could just plug-in alpine CI 2022-09-05 19:45:31 The current runners that can build docker images are all limited to the docker group 2022-09-05 19:56:07 Anyone advice on this odd packaging issue? It's a python package where the download is v0.5, but the filename of the wheel it creates is 0.5.0 2022-09-05 19:56:40 is it different in fetching stage? 2022-09-05 19:56:41 So I've had to do this in 'package': python3 -m installer -d "$pkgdir" dist/quantiphy_eval-$pkgver.0-py3-none-any.whl 2022-09-05 19:57:04 dist/quantiphy_eval-*.whl 2022-09-05 19:57:46 Would dist/quantiphy_eval-$pkgver*-py3-none-any.whl be better? 2022-09-05 19:58:02 it makes zero difference 2022-09-05 19:58:15 you can do dist/*.whl if you want 2022-09-05 19:58:46 I guess. Ok, thanks psykose 2022-09-05 20:49:37 psykose: What's the difference between py3-flit and py3-flit-core? 2022-09-05 20:49:47 the former is a cli tool 2022-09-05 20:50:03 And the latter? 2022-09-05 20:50:14 if you look at the pyproject.toml you'll see flit_core under build-backend 2022-09-05 20:50:35 the latter is the actual pep517 builder 2022-09-05 20:50:55 Got it, thanks. Will update 2022-09-05 20:50:58 flit_core is meant to have no dependencies 2022-09-05 20:51:15 note that sometimes people /do/ actually put just flit under build-backend, and then it would fail if you don't have the former 2022-09-05 20:51:27 and, if you just.. patch it into the latter, it still works, because people are funny :) 2022-09-05 20:51:31 (in every time i've tested at least) 2022-09-05 20:51:48 same for poetry and poetry-core 2022-09-05 20:51:51 I have a couple of other packages the specify flit rather than flit-core. Should I update them? 2022-09-05 20:51:57 if they work, yeah 2022-09-05 20:52:19 py3-parametrize-from-file and py3-tidyexc 2022-09-05 21:01:04 Ok, one of them specifically asks for flit, the other for core. Ok to add the change to the same MR? 2022-09-05 21:10:27 sure 2022-09-05 21:10:31 Ok, will do. 2022-09-05 21:17:12 ikke: any reason things like https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/38662 still have no ci even though it's not even private 2022-09-05 21:18:57 *pipeline denied* 2022-09-05 21:20:25 psykose: I think that's all done now. Fixed an issue that slipped through my run of lint too. 2022-09-05 21:20:58 ah, did you mean the exact same merge request already 2022-09-05 21:21:03 i merged that one so you'll have to make a new one 2022-09-05 21:22:53 Yeah, just spotted that! Will do. Not to worry psykose. 2022-09-05 21:23:01 :) 2022-09-05 21:23:26 That merge had the error in, not sure how it got through the lint check. 2022-09-05 21:23:30 Will fix that in a new MR 2022-09-05 21:24:11 psykose: no, haven't seen that before that it does not even create a pipeline 2022-09-05 21:24:30 i'm also getting tired of telling people to unprivate repositories twice a day again 2022-09-05 21:26:27 The pipeline got created once I rebased it 2022-09-05 21:26:49 Not sure why it doesn't trigger for kogeletey 2022-09-05 21:28:45 i think i actually saw a pipeline on it beforehand 2022-09-05 21:31:50 psykose: https://gitlab.alpinelinux.org/alpine/aports/-/merge_requests/38677 2022-09-05 21:31:55 haha, just saw 2022-09-05 21:31:56 Apologies for missing the typo in the first one. 2022-09-05 21:32:03 np 2022-09-05 21:32:11 i had already fixed the typo after the build failure 2022-09-05 21:33:50 I pulled master again and it seemed to still be there. 2022-09-05 21:35:01 There was a 'gt' on a line by itself before the build function 2022-09-05 21:35:58 check agane 2022-09-05 21:37:18 Ok, got it now. Not sure what happened there. Will amend that latest MR. 2022-09-05 21:39:26 Ok, deleted that commit from the MR 2022-09-06 05:32:25 nmeum: gcc 12.2 seems good to me to have 2022-09-06 05:41:18 we should use 13 instead, more numbers more faster 2022-09-06 10:47:23 can I get a quick look at !38689 so that I look less silly when I publish a blog post on cross-compiling Hare that presumes it's already available in alpine edge 2022-09-06 11:07:44 why cpuburn when you can use gitlab? 2022-09-06 11:51:07 (the web ui) 2022-09-06 13:39:50 ACTION looks at clang being a defactor cross-compiler 2022-09-06 13:40:53 -r 2022-09-06 13:41:17 sh: -r: not found 2022-09-06 13:41:29 :-) 2022-09-06 15:14:55 How is apk configured? It has to know stuff like where to install headers/libraries/binaries, right? 2022-09-06 15:15:10 nope 2022-09-06 15:15:21 .apk is just a tarball 2022-09-06 15:16:13 So when writing the recipe you could decide that `apk install ` will just install it in `./`, for instance? 2022-09-06 15:16:33 there isn't `apk install` 2022-09-06 15:16:40 sorry `apk add` 2022-09-06 15:16:41 and if .apk is build like that, then yes 2022-09-06 15:17:18 Maybe describe your problem 2022-09-06 15:17:37 So I could run `apk` on, say, Arch, and use it with recipes that install locally? Not that it would be useful of course, just for my understanding :D 2022-09-06 15:18:00 I don't have a problem, I'd like to understand better how Alpine is built from scratch 😇 2022-09-06 15:18:15 You technically can 2022-09-06 15:18:32 That's interesting 2022-09-06 15:20:00 And is there some resource somewhere (even if it's CI scripts that do it) where I could learn how Alpine is built? I mean there has to be a rootfs at some point, then apk "adds" packages, but how does apk install the kernel the first time, etc? 2022-09-06 15:20:49 the same way apk adds packages 2022-09-06 15:20:58 it unpacks the tarball and places files 2022-09-06 15:21:44 apk-static can create completely chroot given right packages 2022-09-06 15:21:55 s/completely/complete 2022-09-06 19:47:47 oh, `apk-static`, I need to look into that! 2022-09-06 19:48:14 https://gitlab.alpinelinux.org/alpine/apk-tools/-/releases 2022-09-06 19:48:35 (In case you need to download it on a non-alpine system) 2022-09-06 19:49:08 https://dl-cdn.alpinelinux.org/alpine/latest-stable/main/x86_64/apk-tools-static-2.12.9-r3.apk 2022-09-06 19:49:46 That's so cool! I will definitely play with that :) 2022-09-06 19:59:27 ddevault: done 2022-09-06 20:00:37 thanks psykose! 2022-09-06 21:54:17 psykose: I don't want to do anything about it now, it can wait till the next upgrade, but it looks like systeroid-tui is built a second time during check(), is it due to --no-default-features being passed also there or what gives? 2022-09-06 22:01:20 it's not 2022-09-06 22:01:22 that's just how rust is 2022-09-06 22:14:56 grate 2022-09-06 22:16:13 another thing I noticed, I now installed systeroid-tui first on it's own, and didn't get the -doc package (I have the docs meta package installed) 2022-09-06 22:16:48 but when I then installed systeroid, I got both systeroid-doc and systeroid-tui-doc 2022-09-06 22:17:58 default_doc only applies install_if to pkgname 2022-09-06 22:18:16 you have to manually put another install_if below to change it to the subpackage you want 2022-09-06 22:18:24 it's why i generally don't split docs like that at all 2022-09-06 22:21:03 aha 2022-09-07 06:24:52 Hmm, seems like gitlab 15.3 has a button "rebase without pipeline" 2022-09-07 06:44:00 morning 2022-09-07 06:45:22 morning ncopa 2022-09-07 06:48:06 moin 2022-09-07 06:53:57 gm 2022-09-07 06:54:05 o/ 2022-09-07 06:55:02 Misthios: Good morning 2022-09-07 07:14:12 Good morning everyone 2022-09-07 07:20:36 o/ 2022-09-07 07:34:11 Good morning 2022-09-07 07:34:38 \o 2022-09-07 07:35:45 is there even such a thing as a good morning 2022-09-07 07:37:02 "there are two pleasant moments in the day: at night, when you go to bed, and in the morning, when you don't get up" 2022-09-07 07:37:33 skarnet: no, suffering is eternal 2022-09-07 07:37:58 nothing is eternal, fortunately 2022-09-07 07:38:52 (there's a joke in here about the amount of time my MRs remain pending, but I'm too tired to find it) 2022-09-07 07:39:24 ncopa: would you like to cut mdev-conf release so I can push changes to aports? Or would you prefer to accumulate changes and cut release closer to 3.17? 2022-09-07 07:43:02 Ermine: i backported the input fix 2022-09-07 10:52:13 ncopa: ok, thank you! 2022-09-07 13:17:35 ikke, did a PoC for osp-openvas: https://tpaste.us/OqVm 2022-09-07 13:18:25 check if is a file or dir, and accordingly choose to unlink or rmtree 2022-09-07 13:18:31 but only the subdirs 2022-09-07 13:19:32 You can test it by creating /tmp/{UIID}/ in advance and see if it passes 2022-09-07 13:22:43 you mean like this, locally? 2022-09-07 13:22:43 https://tpaste.us/wYB5 2022-09-07 13:23:15 That's one way 2022-09-07 13:24:52 another option would be creating that dir in the APKBUILD file 2022-09-07 13:25:08 (just for testing purposes) 2022-09-07 18:29:15 Is it possible to have an install_if that gets triggered if either package A or package B is installed? 2022-09-07 18:29:21 or both 2022-09-07 18:32:56 Have both provide a virtual package and install_if on that 2022-09-07 18:34:47 That's the only way? 2022-09-07 18:34:48 Newbyte: some packages declare a install_if with multiple packages listed, I *think* that's an implicit "OR" but not sure 2022-09-07 18:36:07 e.g. device-mapper-udev has install_if="device-mapper eudev" 2022-09-07 18:36:31 Afaik, it's an AND 2022-09-07 18:37:42 install_if="openrc $subpkgname-openrc" 2022-09-07 18:38:00 if that's the case the virtual package covers the "OR" scenario and the multiple listed packages covers the "AND" scenario 2022-09-07 18:44:06 Thanks. 2022-09-07 19:00:47 the better solution is to simply not use install_if 2022-09-07 19:10:33 psykose: what should be used instead? 2022-09-07 19:10:43 nothing 2022-09-07 19:11:11 why? 2022-09-07 19:13:01 because psykose, keeper of the merge gates, says so? ;-) 2022-09-07 19:15:49 given that I have been proposed to join the ranks as a merge gatekeeper it would be good if I understood the rationale in case that proposal comes to fruition 2022-09-07 19:33:13 PureTryOut: did you bumped the libsurvive package with the "latest" *sigh* release 2022-09-07 19:34:02 seems like you did 2022-09-07 19:34:06 i hate libsurvive 2022-09-07 19:38:20 Newbyte: I suppose something akin to KISS :) 2022-09-07 19:40:54 bl4ckb0ne: what's wrong with it 2022-09-07 19:41:38 lots of things 2022-09-07 19:41:45 ohno you're op now 2022-09-07 19:41:49 i have to behave 2022-09-07 19:42:10 ACTION doesnt behave 2022-09-07 19:43:15 🥺 2022-09-07 19:43:43 im getting dumb assertion from libsurvive 2022-09-07 19:44:01 but its mostly due to the poor state of the codebase 2022-09-07 19:44:11 just wanted to know if i had to compile the latest from source 2022-09-07 19:44:28 because they dont know how to release software 2022-09-07 19:45:22 tale as old as time 2022-09-07 20:16:06 Newbyte: install_if is fairly expensive in the solver 2022-09-07 20:16:42 it also violates POLS in a lot of cases 2022-09-07 20:17:09 thats why we try to discourage using it 2022-09-07 20:17:25 in many cases, it is better to declare the dependency in reverse, or to use a metapackage 2022-09-07 20:19:39 (or do nothing and tell people to install something) 2022-09-07 20:20:01 metapackages are probably the least bad, as they can just be removed if unwanted 2022-09-07 20:22:36 you're the least bad 2022-09-07 20:24:08 i'm terrible 2022-09-07 20:24:16 Newbyte: i'd have to see what the usecase is to say 2022-09-07 20:25:54 Newbyte: also would be nice to have !37983 and remove the pmos one since i think the fix is in 4.8 now 2022-09-07 20:28:14 perhaps it even behaves differently for the gtk4 rust armv7 apps 2022-09-07 20:28:18 more stuff to test 2022-09-08 05:55:44 psykose: unfortunately I don't have any devices where the crash issue can be reproduced, but I could try to get someone else to test it for me. I'll make an MR dropping the package from pmOS 2022-09-08 05:56:45 wouldn't it not be an armv7 issue then? you have armv7 devices 2022-09-08 05:57:49 (i'm referring to #14166 and not the gdk hack) 2022-09-08 05:58:06 psykose: the reason we "forked" GTK 4 was because on devices like the PinePhone, all GTK 4 apps crash about half the time on startup 2022-09-08 05:58:13 it's unrelated to the ARMv7 issue 2022-09-08 05:58:39 on my devices no such crashes happen with upstream GTK 4 2022-09-08 05:58:47 i.e. without that hacky patch 2022-09-08 05:58:56 yeah i wasn't referring to any of that :) 2022-09-08 05:59:08 oh, I misunderstood then 2022-09-08 05:59:51 what are you asking for then? If I can test if the GTK 4 + Rust issue has been resolved with this update? 2022-09-08 06:00:02 yeah just in case 2022-09-08 06:00:15 psykose: also, with regards to the install_if question, this is the use-case: https://gitlab.com/postmarketOS/pmaports/-/blob/master/main/postmarketos-ui-phosh/APKBUILD#L95-101 2022-09-08 06:00:47 it's an ugly hack. I would like to replace mobile-config-firefox with firefox or firefox-esr, since there's no reason to use mobile-config-firefox on tablets but I'd still like to have this hack applied 2022-09-08 06:01:23 but now I feel like it's probably better to just fix this in the Firefox APKBUILs 2022-09-08 06:01:26 s/APKBUILs/APKBUILDs/ 2022-09-08 06:01:31 what's there to fix 2022-09-08 06:01:59 making the desktop file name match the app ID for Firefox 2022-09-08 06:02:14 and then patching the manifest to refer to the desktop file by that ID 2022-09-08 06:02:19 I can't think of a better solution 2022-09-08 06:02:32 what would the point be 2022-09-08 06:02:34 alternatively, fix Phosh to not rely on app IDs 2022-09-08 06:03:32 psykose: the point is that Phosh doesn't understand how to find the Firefox icon when displaying it in the running app list for some reason. it's just a ? icon 2022-09-08 06:03:49 unless it's org.mozilla.firefox.desktop? 2022-09-08 06:03:51 yeah 2022-09-08 06:03:54 not a huge deal 2022-09-08 06:03:55 sure, i'll fix it 2022-09-08 06:03:59 also for esr 2022-09-08 06:04:12 and i'll change esr to be /usr/lib/firefox-esr with another profile so they can be parallel 2022-09-08 06:04:26 psykose: https://gitlab.alpinelinux.org/alpine/aports/-/commit/2d291a4afc2c71e25c8fd9fff188eff436b80580 2022-09-08 06:04:34 here's why it was changed originally for what it's worth 2022-09-08 06:05:04 oh, it seems I'm confused 2022-09-08 06:05:15 it's already that 2022-09-08 06:05:34 yeah, sorry. I had it the wrong way around 2022-09-08 06:06:00 so the GTK app ID for Firefox is just firefox 2022-09-08 06:06:05 and not org.mozilla.firefox 2022-09-08 06:06:09 maybe that's what really should be fixed here 2022-09-08 06:06:20 i prefer the short one if anything 2022-09-08 06:06:51 I prefer the long one since it's more consistent with other GTK app IDs, but I wouldn't be opposed to changing it back 2022-09-08 06:07:08 if you (right now) just rename the file to firefox.desktop does the icon fix itself 2022-09-08 06:07:23 let me try, one moment 2022-09-08 06:07:25 for the appid i'm going to have to patch the files for esr anyway 2022-09-08 06:07:35 to make it something else (afaik it defaults to firefox still) 2022-09-08 06:08:34 how do i quickly debug the appid of a running process 2022-09-08 06:11:59 psykose: yeah, the icon works fine if I just rename the desktop file to firefox.desktop 2022-09-08 06:12:07 ACTION uploaded an image: (106KiB) < https://matrix.org/_matrix/media/r0/download/matrix.org/THLkIXCiYcdJoPccyeUDFERn/20220908_08h11m07s_grim.png > 2022-09-08 06:12:08 neat 2022-09-08 06:12:26 can you get the appid of esr for me 2022-09-08 06:12:56 psykose: this is Firefox ESR actually 2022-09-08 06:13:00 so it seems to be "firefox" 2022-09-08 06:13:04 neat, ok 2022-09-08 06:13:06 i'll have to fix those 2022-09-08 06:13:27 can you check if the developer edition icon works, haha 2022-09-08 06:13:43 developer edition is packaged in Alpine? 2022-09-08 06:13:57 found it, let's see 2022-09-08 06:14:12 actually, it's not available on armv7 2022-09-08 06:14:27 I don't have any aarch64 devices where Phosh works unfortunately 2022-09-08 06:14:58 I really need to write that stupid touch driver 2022-09-08 06:15:15 anyway, I'll see if I can get someone to test the GTK 4 issue in the meantime 2022-09-08 06:17:54 ah, right 2022-09-08 06:17:59 yeah, i kept it only on the main ones 2022-09-08 06:18:07 sure sure 2022-09-08 06:20:02 also, i assume the firefox-safe.desktop doesn't show up correctly? 2022-09-08 06:20:28 it would ideally be a [Desktop Action safe-window] 2022-09-08 06:20:38 but most things don't really use that i guess 2022-09-08 06:22:37 psykose: someone who isn't me made an executive decision that it doesn't make sense to show the safe-desktop version of Firefox on phones (or rather, all devices where pmOS runs), so it's hidden. I could try to restore it if you want me to test that. 2022-09-08 06:23:28 the way you word that makes it sound like some weird politics :) 2022-09-08 06:23:34 i don't see a reason to not merge it into an action 2022-09-08 06:24:01 I think it's a bad solution to the problem. I agree that merging it into an action would be better 2022-09-08 06:25:24 there's already actions for e.g. private window 2022-09-08 06:25:28 but, I would guess the argument would go that some desktops don't support actions, or that it's less discoverable 2022-09-08 06:25:36 https://img.ayaya.dev/T8aRjNE7q0wY 2022-09-08 06:25:46 yeah, the support isn't exactly common 2022-09-08 06:29:35 psykose: tested GTK 4.8.0. works fine here, but this is not a PinePhone. Rust issue still happens 2022-09-08 06:29:44 okie 2022-09-08 06:29:49 time for some backtraces 2022-09-08 06:29:53 which I wouldn't call surprising since it seems to error in Rust code, and GTK itself has no Rust code 2022-09-08 06:31:31 did you run into this recently or have you not used the apps before on armv7 2022-09-08 06:31:51 I think it didn't happen in the past, but I'm not 100% sure. It's been happening for a few months at least. 2022-09-08 06:32:27 either way, I'll look into it later. I need to go shower and all that 2022-09-08 06:32:36 pffff 2022-09-08 06:33:03 being stinky is preferable 2022-09-08 06:43:52 i tested a armv7 chroot and got the same error 2022-09-08 06:44:04 if i try 3.16 i instead get a different error, still crashes 2022-09-08 06:46:16 hmm 2022-09-08 06:58:12 aha 2022-09-08 06:58:18 i found one that is gtk4+rust and works 2022-09-08 06:58:27 most of them use libadwaita, that one doesn't 2022-09-08 07:33:03 psykose: which one is it? 2022-09-08 07:33:40 pidif 2022-09-08 07:33:58 also checked 1.2.rc libadwaita, didn't fix 2022-09-08 07:34:06 now gonna try the usual backtrace i guess 2022-09-08 07:36:06 psykose: let me know in the issue if you figure anything out, I'm leaving for now 2022-09-08 07:36:11 have fun 2022-09-08 08:48:57 ERROR: gtk+3.0-doc-3.24.34-r5: trying to overwrite usr/share/man/man1/gtk-update-icon-cache.1.gz owned by gtk+2.0-doc-2.24.33-r6. 2022-09-08 08:50:22 huh, 'apk fix' helped 2022-09-08 08:54:02 aha 2022-09-08 09:00:21 sounds like a replace missing 2022-09-08 09:01:21 yeah, already fixed 2022-09-08 09:49:27 does apk have alterantive handling capabilities? like neovim providing vim so i don't have to symlink it in /usr/bin myself 2022-09-08 09:50:17 No, there is no concept of alternatives built-in 2022-09-08 10:10:46 is there a specific policy regarding when to use supervisor="supervise-daemon" in service files or not ? 2022-09-08 10:11:10 aside from jirutka being really mad at you if you put it in the init.d instead of conf.d, no 2022-09-08 10:13:15 why would he be 2022-09-08 10:14:00 i patched bluez /etc/init.d/bluetooth to make it much more reliable by being supervised 2022-09-08 10:14:19 because he doesn't to use it 2022-09-08 10:14:27 (or being forced to use it) 2022-09-08 10:51:56 is helix not able to use /usr/lib/tree-sitter? symlinked to with /usr/share/helix/runtime/grammars 2022-09-08 10:52:40 omni: there is an issue open about it 2022-09-08 10:53:09 #14162 2022-09-08 11:01:15 it is, but the tree sitter grammars are out of date 2022-09-08 11:01:32 we should package newer ones 2022-09-08 11:02:46 "Please report it to the affected tree-sitter grammars/parsers, to finally start making regular releases. This problem is on their side, we do package the latest released versions of the tree-sitter grammars." 2022-09-08 11:13:27 i would love to agree with that but that's not how packaging works 2022-09-08 11:13:48 if you ship something with "system libraries" and the thing immediately crashes on startup because they don't match, it's a packaging issue 2022-09-08 11:14:03 the solution is to upgrade it to random $gitrev or to not use the system ones 2022-09-08 11:14:15 not ship broken software that literally doesn't work then complain someone didn't make some releases 2022-09-08 11:14:24 yes, they should make some releases anyway, but 2022-09-08 11:15:22 life's hard 2022-09-08 11:16:07 who cares about user experience right 2022-09-08 11:16:16 certainly not some people apparently 2022-09-08 11:16:22 oh u wanna use this software: go complain to x 2022-09-08 11:16:59 nevertheless arguing with jirutka is like arguing with a brick wall, so i guess it shall remain broken 2022-09-08 11:17:27 Piraty: you can have it as a toggle in conf.d 2022-09-08 11:18:03 you have to edit the init.d to be compatible with both (the easiest way to do that is to make bluetoothd stay foreground, add command_background=true, and then make sure whatever logging might have been there still works) 2022-09-08 11:18:16 But it's the kind of pressure necessary to make upstreams to behave 2022-09-08 11:18:23 i agree with all that 2022-09-08 11:18:32 at the same time i'm just saying it how it is 2022-09-08 11:19:46 see issue, fix issue, open upstream issue 'hey please actually tag something every few months instead of every 2 years', done 2022-09-08 11:20:25 where is the part where that upstream issue stays there for x years 2022-09-08 11:20:48 then you jump to that prior solution point and pick one 2022-09-08 11:20:57 it's really not rocket science 2022-09-08 11:28:16 It's a clash of 2 models 2022-09-08 11:29:27 distros like to have everything neatly packaged and predictable 2022-09-08 11:29:41 developers don't want overhead and just expect users to yolo download the latest version 2022-09-08 11:29:54 it's the cause of it but i'm not even referring to that part 2022-09-08 11:30:05 distros like to not ship broken software 2022-09-08 11:30:23 It's the root cause 2022-09-08 11:31:50 I'm not saying that's an excuse for shiping broken software 2022-09-08 11:32:15 i mean we both know what the other is saying and we agree 2022-09-08 18:21:59 is any of the packaged tree-sitter grammars compatible with the packaged helix? 2022-09-08 18:22:48 most of them 2022-09-08 18:23:02 and by most i mean the three i tried worked and i already forgot 2022-09-08 18:23:14 anything that got updated more often than 2 years ago probably work 2022-09-08 18:23:54 the handful I've looked at seem up to date wrt their upstream releases, but not all from the tree-sitter org at msgh are packaged (21 out of 40-ish?) 2022-09-08 18:24:43 and hx --grammar fetch/build track 107 from various repos 2022-09-08 18:25:10 I would like to add an optional helix-grammar sub-package 2022-09-08 18:25:56 why 2022-09-08 18:26:32 hello panekj 2022-09-08 18:26:36 goodbye panekj 2022-09-08 18:27:14 ok, I can just build them myself and set HELIX_RUNTIME where I want to use helix 2022-09-08 18:27:29 the issue is that also points to the rest of the runtime files 2022-09-08 18:27:44 personally i've never used packaged editors, they're very personal to begin with 2022-09-08 18:27:48 random git checkouts is where it's at 2022-09-08 18:28:01 it's fine, I've created a folder with symlinks to them 2022-09-08 18:28:36 you roll your own $EDITOR? 2022-09-08 18:32:02 i build it myself sure 2022-09-08 18:43:14 I'll add my suggestion to the issue 2022-09-08 18:43:47 someone may not want to fetch and build them all on their phone or whatnot 2022-09-08 20:17:26 aha https://wiki.archlinux.org/title/Rust_package_guidelines#Check 2022-09-08 20:46:46 jag har inte sett någon dålig sydkoreansk film än 2022-09-08 20:48:02 Which ikea piece of furniture is it 2022-09-08 20:50:39 Something about a south korean movie 2022-09-08 20:57:40 this again, wrong window 2022-09-08 20:58:06 but yes, I've yet to see a bad south korean movie 2022-09-08 22:06:29 maybe if they made south korean movies in sweden they would be bad 2022-09-08 22:38:50 bork bork 2022-09-09 09:05:23 where's my spork 2022-09-09 10:32:27 added some colors and a 'support level' column: https://wwwtest.alpinelinux.org/releases/ 2022-09-09 10:34:36 The colors being all black? 2022-09-09 10:34:56 At least, for some reason it renders like that for me 2022-09-09 10:35:13 there is a green/yellow/red dot? 2022-09-09 10:35:32 I see the style, but the dot is always rendered black 2022-09-09 10:35:38 it's grey in Firefox at least 2022-09-09 10:35:43 oh... its black on firefox. 2022-09-09 10:35:46 interesting 2022-09-09 10:35:59 however if I drag my mouse over the character the highlight is green 2022-09-09 10:36:10 or blue, or whatever 2022-09-09 10:36:24 oh, never mind 2022-09-09 10:36:25 I was mistaken 2022-09-09 10:36:27 https://imgur.com/a/UXk3awv 2022-09-09 10:36:40 2022-09-09 10:36:45 yes 2022-09-09 10:36:51 is what i added, but firefox does not seem to render that 2022-09-09 10:36:52 hum 2022-09-09 10:37:16 Same in chromium for me 2022-09-09 10:37:21 maybe use an SVG instead? 2022-09-09 10:37:50 firefox here seems to render it 2022-09-09 10:37:51 https://www.w3schools.com/tags/tag_span.asp 2022-09-09 10:37:58 maybe it's a font thing? 2022-09-09 10:38:22 Newbyte: that's my suspicion 2022-09-09 10:39:01 indeed. If I change it to * it shows the color 2022-09-09 10:43:20 it's borked in Chromium for me too 2022-09-09 10:43:35 ACTION uploaded an image: (245KiB) < https://matrix.org/_matrix/media/r0/download/matrix.org/rBrftwCKHlkcgnvGGZlgkkpR/clipboard.png > 2022-09-09 10:44:28 try now 2022-09-09 10:44:44 looks fine now 2022-09-09 10:44:56 👍 2022-09-09 10:45:17 the yellow one is a bit hard to see on the white background 2022-09-09 10:45:25 other than that it's good 2022-09-09 10:45:36 ^^ 2022-09-09 10:45:44 maybe change it to orange? 2022-09-09 10:46:20 yup, for me as well 2022-09-09 10:46:43 yeah, orange as color looks better, more contrast 2022-09-09 10:48:11 sort of off-topic but would you agree that Alpine is "made to be a container OS"? 2022-09-09 10:48:31 Newbyte: that's mostly incidental 2022-09-09 10:48:48 its not made, it just happend to fit very well 2022-09-09 10:48:52 Alpine was made before containers was a thing 2022-09-09 10:49:16 its made to be a router os kind of thing 2022-09-09 10:49:21 we did use containers early though, before linux kernel had support for namespaces, via 3rd party patch for vserver 2022-09-09 10:49:43 ok orange looks better indeed 2022-09-09 10:50:06 thanks for your answers 2022-09-09 10:51:07 orange looks great, nice 2022-09-09 10:51:39 ncopa: what is the cut-off for 'on request' versus 'none'? 2022-09-09 10:51:47 4 years? 2022-09-09 11:40:04 ikke: i was thinking 5 years, but depends on how long we keep the builders on the servers 2022-09-09 11:58:47 Which is something we need to start thinking about pretty soon 2022-09-09 12:03:17 I only see black dots in kristall :B 2022-09-09 12:08:01 🧌 2022-09-09 12:14:47 (looks good in qutebrowser) 2022-09-09 14:44:18 the reason it rendered as black is because you picked a black emoji- the times where it worked is where the emoji font wasn't a real one and had a 'default color' emoji that was affected by colors 2022-09-09 14:44:30 can't color emoji like that 2022-09-09 15:30:19 almost all dots are red though? 2022-09-09 15:30:30 oh 2022-09-09 15:32:18 yes, most releases no longer have active support 2022-09-09 15:32:58 no, the top ones were also looking red, until i looked at the css 2022-09-09 15:33:46 ( the colors and size are not really good for anyone with red-green colorblindness ) 2022-09-09 15:34:26 lmao 2022-09-09 15:34:30 right 2022-09-09 15:34:38 was waiting for you to see that xD 2022-09-09 15:35:02 the dots are also too small 2022-09-09 15:35:22 ptrc: I suppose using different shapes would be helpful? 2022-09-09 15:35:29 yeah, that's what i meant by size, it would be probably a bit easier to distinct if they were bigger 2022-09-09 15:35:31 aand that too 2022-09-09 15:35:46 it also doesn't differentiate community/main so i'm not sure if it's accurate at all 2022-09-09 15:35:52 if the red ones were a solid square, i think it would be a big improvement 2022-09-09 15:36:06 3.14 is green, 3.13 is yellow, both are 'sec fixes', and neither support community 2022-09-09 15:36:29 i think this is just worse than an empty page 2022-09-09 15:36:53 "End of Support for Main" 2022-09-09 15:37:05 hmm 2022-09-09 16:32:15 EOL 🪦 2022-09-09 16:44:39 how about a green tick symbol, and a red cross symbol, and something else (?) in yellow? 2022-09-09 16:45:31 then the colour isn't as important as the symbol also illustrates the status 2022-09-09 16:47:46 yep 2022-09-09 16:51:00 or use something like Edvard Munch's "The Scream" for the unsupported versions ;-) 2022-09-09 16:53:24 i was thinking red square (■, \u25a0) and yellow triangle (▲, \u25b2) 2022-09-09 16:54:36 i don't think 2022-09-09 16:54:57 neither do I 2022-09-09 16:57:23 I tried once, it hurt 2022-09-09 16:57:37 3■, 4▲ 2022-09-09 16:57:43 use rock, paper, scissors? ;-) 2022-09-09 17:00:54 minimal: like this? https://ptrc.gay/SGwXFvAx 2022-09-09 17:03:37 cool ;-) 2022-09-09 17:06:24 nice 2022-09-09 17:14:50 11,13🦦 2022-09-10 22:07:24 should obs depend on qt6-qtwayland if it doesn't start on wayland without it? 2022-09-10 22:12:00 no 2022-09-10 22:32:46 definitely not 2022-09-11 00:47:42 Would perhaps creating an empty 'obs-wayland' subpackage with depends="qt6-qtwayland" and install_if="wayland obs" work? A bit hacky, so psykose has final say, but it might cover that particular use-case. 2022-09-11 00:49:14 the 'correct' way to do it is the same as qt5-qtwayland 2022-09-11 00:49:26 correct in the sense of to do that 2022-09-11 00:49:31 the actual correct way is to do nothing 2022-09-11 00:50:41 if someone wants qt6 wayland support, they can install the literal package 2022-09-11 00:51:15 if we want to make it a bit easier to discover.. a mere 'wayland-meta-huge-list-of-things-only-install-this-if-you-really-want-one-million-deps' with the depends on these random things would work 2022-09-11 00:51:42 the install_ifs lead to a lot of unexpected things all the time and i wish nobody used them for random small things like this 2022-09-11 00:52:32 the previous idea in qt5-qtwayland was the same and then it ended up in 1f5a1494d732af559ea46514acdb513cb7de31e7 2022-09-11 00:52:42 add nothing, install random thing, suddenly you pull in +200 packages 2022-09-11 00:53:06 it's good for "docs" or things truly generic but aside from that you have to know all of aports to use it sanely 2022-09-11 00:54:03 or for very tiny in scope things inside the same aport 2022-09-11 20:13:04 ikke: is there a go installed in armhf/x86? 2022-09-11 20:13:07 seems to be 2022-09-11 20:13:28 on the builders 2022-09-11 20:46:16 every time abuild crashes mid build for whatever reason it keeps around the makedepends forever, it seems 2022-09-12 04:35:33 removed 2022-09-12 04:35:59 (makedepends for hut and hledger-timeclock) 2022-09-12 09:31:28 morning! 2022-09-12 09:32:22 We are getting closer october, when we set up the 3.17 builders and have a feature freeze. what are the things we need to have included for alpine 3.17? 2022-09-12 09:34:36 abuild release 2022-09-12 09:37:37 indeed 2022-09-12 09:37:51 are there any of the current open MRs or issues that should be prioritized? 2022-09-12 09:40:12 the ones that were already merged and fixed months ago 2022-09-12 09:58:58 wdym 2022-09-12 10:10:23 that there are no important MRs or issues that needs to be fixed before tagging new abuild release 2022-09-12 10:13:16 only you would know those, because it's an internal tool and nobody runs the main branch 2022-09-12 10:13:27 if you made any regressions nobody would find out until you released it 2022-09-12 10:13:36 so aside from that.. it has only fixed, from perspective 2022-09-12 10:13:59 if you mean what i think you mean 2022-09-12 10:14:14 if you instead mean "fix even more things before making a release on top of the existing changes", then that doesn't make sense 2022-09-12 10:14:19 releases are cheap, you can make 50 of them 2022-09-12 10:15:13 realistically there should be one after any change, because it's just an internal tool that's used for alpine development and the builders 2022-09-12 10:16:49 is that a bit extreme? maybe, but i'm just saying it as it is. the only testing for it is on the releases themselves and the tests you write, keeping things merged but unreleased doesn't do anything towards qa 2022-09-12 10:18:56 now, separately onto potential things to fix for 3.17... i can't think of anything 2022-09-12 10:19:09 we need a release to even apply the rust/go flags globally for the actual release, that's about it 2022-09-12 10:19:25 the potential dt_relr change would also need a release and commits later for the abuild.conf modification 2022-09-12 10:20:13 everything else is just the few issues i have opened or whatever, but those don't affect 3.17 or anything, just personal small bugs i find 2022-09-12 10:30:45 I think this change should be in 3.17 https://gitlab.alpinelinux.org/alpine/mkinitfs/-/commit/76028dc92b86181202ef2c5302bc04458dd94629 not sure if there are other changes that could go in before tagging a mkinitfs release but stable users shouldn't have to see that warning 2022-09-12 10:34:43 Which is ETA before 3.17 release? 2022-09-12 10:35:58 November 2022-09-12 10:38:13 builds should be ok, just gonna be the usual soup of gcc12 fixes 2022-09-12 10:38:20 circle checking tool is clean for once 2022-09-12 10:39:06 can't wait to be awake for 2 days restarting everything, i should pack some snacks this time 2022-09-12 10:40:07 ACTION brings coffee to psykose 2022-09-12 10:58:58 do we need gcc-cross-embedded for all architectures? 2022-09-12 10:59:21 like, who would want to use riscv64 to crosscompile for embedded environments? 2022-09-12 10:59:26 or s390x? 2022-09-12 10:59:37 or even 32 bit x86 2022-09-12 11:00:29 nobody 2022-09-12 11:00:53 should probably just be ppc64le/x86_64/aarch64 2022-09-12 11:00:54 i guess it would make sense to have that for only x86_64 and aarch64? 2022-09-12 11:00:59 yeah 2022-09-12 11:05:44 aside from or1k the other targets are also supported by clang 2022-09-12 11:05:54 and when that works it's very convenient, haha 2022-09-12 11:08:11 you can also disable newlib on the same disabled arches 2022-09-12 11:11:46 should ask the maintainer 2022-09-12 11:32:06 there's -avr too 2022-09-12 11:37:27 yeah, im looking at it now. it uses different libc 2022-09-12 11:37:50 was wondering if it would make sense to add avr to gcc-cross-embedded 2022-09-12 11:43:25 yeah, just replacing the --with-newlib with the avr one 2022-09-12 11:43:27 +- some other args 2022-09-12 12:03:05 https://gitlab.alpinelinux.org/Newbyte/aports/-/jobs/837294#L40 2022-09-12 12:03:12 why is a commit a volatile source? 2022-09-12 12:04:32 hash only 2022-09-12 12:04:52 $pkgname-startup-notify.patch:: 2022-09-12 12:05:02 why is that better? 2022-09-12 12:06:37 actually has info in it instead of random numbers 2022-09-12 12:06:52 but the reason for the flag is it just checks for ^https gitlab/github .patch/.diff 2022-09-12 12:06:53 lmao 2022-09-12 14:28:14 regarding the patchwork mail on -devel, can we simply drop the whole paragraph: https://wiki.alpinelinux.org/wiki/Patch_Workflow#I.27m_an_alpine_developer._How_can_I_start_to_use_patchwork_.3F 2022-09-12 14:28:58 oh, in fact the whole page could be deleted 2022-09-12 14:41:21 delete it 2022-09-12 14:44:03 think only admins can 2022-09-12 15:08:52 yup, I can't delete the whole page 2022-09-12 15:09:33 there is a ref to it from https://wiki.alpinelinux.org/wiki/Creating_patches#See_also 2022-09-12 15:10:25 ...which i removed 2022-09-12 15:11:36 you missed a grep -E in is_vserver 2022-09-12 15:23:10 do we use buildlab? 2022-09-12 15:23:32 looks like it is somethign for chroot builds, but we have abuild rootbld now 2022-09-12 15:26:23 don't think so 2022-09-12 15:29:58 time for champagne 2022-09-12 15:32:09 :) indeed 2022-09-12 15:32:12 on a monday 2022-09-12 15:32:56 abuild 3.10.0_rc1 pushed. lets see what breaks 2022-09-12 15:33:00 hopefully nothing 2022-09-12 15:33:18 i guess i should continue work on abuild tests 2022-09-12 15:33:30 probably nothin 2022-09-12 15:33:36 ikke: you have to update all the builder flags 2022-09-12 15:34:05 especially because goflags were removed from being hardcoded, so everything in go will break on rebuild 2022-09-12 15:35:20 short timeout 2022-09-12 15:38:02 ikke: i can update GOFLAGS on the builders (except riscv64) 2022-09-12 15:41:45 ok 2022-09-12 15:41:51 i think the test is broken. it is an interactive test 2022-09-12 15:41:58 so it probably asks for something it shouldnt 2022-09-12 15:44:44 there's rust flags too 2022-09-12 15:49:48 hum 2022-09-12 15:50:01 it might actaully be the timeout that is a bit optimistic 2022-09-12 17:50:32 ikke: can you check if all the flags were set 2022-09-12 17:50:48 psykose: I can check later 2022-09-12 17:50:54 sure 2022-09-12 17:51:09 should look like the abuild.conf default for go/rust 2022-09-12 18:34:39 psykose: I see goflags, but no cargo related flags other than CARGO_BUILD_JOBS in the .apk-new file 2022-09-12 18:34:56 well, add them 2022-09-12 18:35:22 They're not even in the .apk-new file, which I would expect 2022-09-12 18:35:56 they were definitely added https://gitlab.alpinelinux.org/alpine/abuild/-/blob/master/abuild.conf 2022-09-12 18:36:18 showed up in apk-new for me 2022-09-12 18:36:26 Oh wait, they're there, I just messed something up in vimdiff 2022-09-12 18:37:05 do the other way around accidentally :P 2022-09-12 18:37:08 'do' 2022-09-12 18:38:18 happens 2022-09-12 18:38:19 : ) 2022-09-12 18:48:44 psykose: fyi, if we get strange issues with cargo downloading crates, it's because I removed CARGO_HTTP_MULTIPLEXING=false which I've added before due to issues 2022-09-12 18:49:32 noted, thanks 2022-09-12 18:51:18 when did you add it before? 2022-09-12 18:51:42 Don't recall, but quite some time ago 2022-09-12 18:52:15 hopefully fixed by now 2022-09-12 18:52:42 yes 2022-09-12 18:55:42 I've updated abuild.conf on every arch now except rv64 2022-09-12 19:04:31 neat :) 2022-09-12 19:05:33 should do rv too for go 2022-09-12 19:06:44 yea, but it hadn't even built the last abuild yet 2022-09-12 19:07:31 It's docker based, so I want to make sure it ends up in the image 2022-09-12 19:11:23 will take another 1-2 hours 2022-09-12 19:12:23 Did you ever got anywhere with rv64 in qemu-system? 2022-09-12 19:12:30 afaik mps had something running 2022-09-12 19:13:02 ah 2022-09-12 19:13:07 nope, got completely sidetracked 2022-09-12 19:13:28 heh, relatable 2022-09-12 19:14:26 But did I understood it correctly the current issue with rust on rv64 is that the current compiler cannot compile itself? 2022-09-12 19:17:34 yes 2022-09-12 19:17:44 gets further on 1.62.1 as opposed to 1.63, but 2022-09-12 19:17:53 anyway, it will need to be bootstrapped from scratch again 2022-09-12 19:17:58 rust 1.64 is like next week 2022-09-12 19:18:18 so bootstrap 1.64 from scratch? 2022-09-12 19:18:33 use whatever method (as before, etc) to make it, yeah 2022-09-12 19:18:41 last time it was cross from gentoo 2022-09-12 19:18:47 they'll have a 1.64 or w/e 2022-09-12 19:19:03 what i'm surprised about is why it couldn't compile itself 2022-09-12 19:19:26 first it got built cross, then it got stuck repeatedly (but no segfaults) until pass 2022-09-12 19:19:36 then we bumped openssl3 and it just never passed again (same version) 2022-09-12 19:19:46 then 1.63, etc 2022-09-12 19:28:12 wish they're release qt 6.4 already 2022-09-12 19:28:18 they'd 2022-09-12 19:28:35 there's been a specific bug with sway for like 2-3 years lol 2022-09-12 19:28:38 finally fixed 2022-09-12 19:29:39 or maybe it's actually a different bug.. 2022-09-12 20:22:49 Hello71: thanks for weighing in on https://discourse.llvm.org/t/configure-script-breakage-with-the-new-werror-implicit-function-declaration/65213 2022-09-12 20:23:04 i don't know how they expected us to realise 2022-09-12 20:23:09 i can't watch every single commit of every project 2022-09-12 20:34:27 sam_: thanks. i am hesitant because i am concerned about this thread turning into a crapshow of piling in, but hopefully i have added some value 2022-09-12 20:34:28 fun, silently breaking changes 2022-09-12 20:34:58 and this really does suck badly 2022-09-12 20:35:30 it's 100x worse than the -fno-common struggle because at least that one failed reliably-ish 2022-09-12 20:35:42 absolutely 2022-09-12 20:35:53 i should have mentioned that, forgot to 2022-09-12 20:35:57 That failed builds which could be most of the time trivially fixed 2022-09-12 20:36:12 (although -fno-common is kind of funny because it's an example of one of the breaking changes GCC is happy to do, but other ones are off-limits; I know this case is different, but there's been other incidents where they refuse to tighten things up) 2022-09-12 20:37:38 >I’ve heard that some people even manage to have real life, friends and family, that kind of stuff. 2022-09-12 20:37:39 where?? 2022-09-12 20:37:59 I thought those were myths :) 2022-09-12 20:38:43 i really do believe it's a good change in a pure sense, i just think the communication was terrible and i'm like 90% sure that nobody involved in the decision making process was aware of the pitfalls 2022-09-12 20:39:16 i've started writing some cheesy bashrc hook to run configure tests twice & filter out useless lines, it sort of works 2022-09-12 20:39:21 but i mean, it's still a huuuuge amount of work 2022-09-12 21:37:31 ikke: riscv ready for more flags 2022-09-12 23:19:07 bl4ckb0ne: wxwidgets 3.2 just for you : ) 2022-09-12 23:22:24 thanks bruv 2022-09-12 23:22:35 maybe ill pack cemu soon 2022-09-13 05:37:12 psykose: recreated the container with the new abuild.conf 2022-09-13 09:48:01 has something changed on 3.16 CI builders wrt cleaning up srcdir? 2022-09-13 09:50:25 !38976 2022-09-13 10:15:10 omni: that's because GOMODCACHE is set in the APKBUILD without adding -modcacherw to GOFLAGS 2022-09-13 10:15:36 go by default makes the modcache read-only, preventing abuild from deleting it if it's inside srcdir 2022-09-13 10:16:52 ikke: aha, is this different on edge? 2022-09-13 10:16:57 no, same 2022-09-13 10:17:19 But, since yesterday -modcacherw is part of abuild.conf GOFLAGS 2022-09-13 10:17:44 so that would be different between 3.16 and edge 2022-09-13 10:32:36 thanks 2022-09-13 10:44:46 psykose: re https://gitlab.alpinelinux.org/alpine/aports/-/issues/14189 . FYI g-c-c still segfaults, the rebuild didn't solve it . https://dpaste.com/ATPY7D25S.txt 2022-09-13 11:40:24 abump pdns-4.7.0-beta2 -> >>> ERROR: pdns: 4.7.0-beta2 is not a valid version 2022-09-13 11:40:30 how do i get around this? 2022-09-13 11:40:39 (not going to submit an MR, just want to test the build) 2022-09-13 11:42:42 Habbie: 4.7.0_beta2 2022-09-13 11:43:55 ah, plus ${pkgver//_/-} in source 2022-09-13 11:43:57 wonderful 2022-09-13 11:44:06 the underscore helped me find the right hints, thanks 2022-09-13 11:44:22 there it goes 2022-09-13 11:44:50 can i slip that ${pkgver//_/-} in with the next real MR? 2022-09-13 11:45:57 Sure 2022-09-13 11:46:02 ta 2022-09-13 11:53:58 plus the same hack in builddir= 2022-09-13 12:29:55 ikke: thanks 2022-09-13 12:30:10 Piraty: i know, not sure what the issue is 2022-09-13 12:30:54 upstream issue ? 2022-09-13 12:39:55 no, it's something related to tooling most likely 2022-09-13 12:48:13 ncopa: re https://gitlab.alpinelinux.org/alpine/aports/-/issues/14189#note_262542 . any flags you want in particular for valgrind ? 2022-09-13 12:49:03 yeez, it started 2022-09-13 12:49:14 a.k.a. didn't segfault 2022-09-13 12:49:20 heisenbug 2022-09-13 12:54:26 does it warn about any uninitialized vars? 2022-09-13 12:56:35 maybe there are some use of uninitialized var somewhere. maybe there are warnings in build log? 2022-09-13 13:00:46 any slowdown breaks the race and it works 2022-09-13 13:00:56 even some gdb watch commands fix it 2022-09-13 13:01:43 build log: https://build.alpinelinux.org/buildlogs/build-edge-x86_64/community/gtk4.0/gtk4.0-4.8.0-r1.log 2022-09-13 13:02:47 ah, so it is a race 2022-09-13 13:02:58 but a consistent one 2022-09-13 13:03:11 with the CI one always winning and the builder one always failing 2022-09-13 13:03:11 fun 2022-09-13 13:05:10 http://ix.io/4anu 2022-09-13 13:05:53 librsvg? 2022-09-13 13:06:21 the log is invalid because of the sigill i think 2022-09-13 13:06:30 can you install librsvg-dbg? 2022-09-13 13:06:51 sure 2022-09-13 13:07:14 ==13345== Conditional jump or move depends on uninitialised value(s) 2022-09-13 13:12:55 https://termbin.com/92jc 2022-09-13 13:14:45 eh, hang on 2022-09-13 13:16:31 it truncated the file and processed dbus-launch x| 2022-09-13 13:16:50 as i passed --trace-children=yes 2022-09-13 13:19:18 https://ttm.sh/qz1.txt 2022-09-13 13:24:49 strange how different the place is compared to anything yesterday 2022-09-13 13:31:30 psykose why are you an op now 2022-09-13 13:32:16 who knows 2022-09-13 13:33:17 i assume there's an oftc autoop that i don't know how to disable 2022-09-13 13:33:50 but aside from that i have it to deal with spammers x) 2022-09-13 13:35:03 psykose, /msg chanserv set autoop #alpine-devel off 2022-09-13 13:35:12 psykose, you'll still be able to /msg chanserv op #alpine-devel 2022-09-13 13:35:24 thankies 2022-09-13 13:35:34 (however, clandmeter/ikke/ncopa may have opinions, they are MASTER where you are merely CHANOP) 2022-09-13 13:35:47 (I also don't know if you you, then, can actually change that flag, because it applies to all four of you) 2022-09-13 13:36:02 except only they can toggle that 2022-09-13 13:36:20 and it's for the whole channel lol 2022-09-13 13:36:24 not per user 2022-09-13 13:36:24 yes :) 2022-09-13 13:36:27 how useless 2022-09-13 13:36:37 as expected from oftc, so i don't mind 2022-09-13 13:36:45 i can do that myself 2022-09-13 13:36:49 it's one of the things the freenode^Wlibera software does better, yes 2022-09-13 13:37:06 what does ^W mean? 2022-09-13 13:37:10 ctrl-w 2022-09-13 13:37:14 backspace word 2022-09-13 13:37:16 what is wrong with beeing +o? 2022-09-13 13:37:17 it removes the previous word 2022-09-13 13:37:38 clandmeter, can't speak for anybody in here, but in some channels people prefer that nobody looks privileged - as some trolls will actually react to it 2022-09-13 13:37:38 sign of power 2022-09-13 13:37:42 so it's meant to represent the same as strikethrough? 2022-09-13 13:37:49 i also prefer telling people to behave without my @ visible 2022-09-13 13:37:51 Newbyte, yes 2022-09-13 13:37:56 I see, thanks 2022-09-13 13:38:05 old fashioned strikethrough 2022-09-13 13:38:13 non-richtext version 2022-09-13 13:38:19 in what editor does Ctrl + W remove a word? 2022-09-13 13:38:22 emacs? 2022-09-13 13:38:25 all of them ime 2022-09-13 13:38:27 emacs, bash, joe, etc. 2022-09-13 13:38:32 or, I guess GNU readline stuff 2022-09-13 13:38:36 bash as well 2022-09-13 13:38:42 readline except when in vi-mode, i guess 2022-09-13 13:38:51 abviously 2022-09-13 13:38:55 some ppl have problems with anything that could cause problems. its called too sensitive. 2022-09-13 13:39:02 all shells as well 2022-09-13 13:39:10 how would anybody know who to call for if shit happens? 2022-09-13 13:39:11 I presume that's all shells that use readline 2022-09-13 13:39:21 clandmeter, in most channels, they can ask chanserv 2022-09-13 13:39:37 clandmeter, some channels have bot commands for notifying ops, which is a hack 2022-09-13 13:39:38 so a troll can do that also? 2022-09-13 13:39:43 most of them don't use readline 2022-09-13 13:39:43 yes, but it turns out they don't 2022-09-13 13:39:51 (speaking from years of experience moderating #raspberrypi) 2022-09-13 13:39:53 Habbie: honestly I have no idea how to ask chanserv 2022-09-13 13:40:01 Newbyte, /msg chanserv access #alpine-devel list 2022-09-13 13:40:06 so we should hide instead of beeing here to help. makes no sense to me. 2022-09-13 13:40:20 clandmeter, i'm not arguing any side of this - just sharing my observations 2022-09-13 13:40:27 (and my personal bias) 2022-09-13 13:40:40 im just telling you my opinion :) 2022-09-13 13:40:42 :) 2022-09-13 13:42:15 oh i read psykose msg wrongly, i thought she did not know how to set auto op, but it was the opposite. 2022-09-13 13:43:03 i know how to use the commands ^_^ 2022-09-13 13:44:01 ok, just ignore me 2022-09-13 13:45:16 ikke: do we have a doc how to setup a builder? 2022-09-13 13:45:42 something that is not related to lxc-copy or similar 2022-09-13 13:47:28 Not that I'm aware of 2022-09-13 13:47:48 I think there was some documentation on the wiki 2022-09-13 13:48:24 one last time. ncopa . 2022-09-13 13:48:29 has there been talk about setting up a stable builder for rv64? 2022-09-13 13:48:53 https://termbin.com/1dv9 2022-09-13 13:48:54 or are we still waiting for decent hw 2022-09-13 13:49:03 another 5y? 2022-09-13 13:49:08 which probably another x hars 2022-09-13 13:49:10 years 2022-09-13 13:49:13 Being able to run tests was a base criteria 2022-09-13 13:49:32 i guess we first get edge running on lxc 2022-09-13 13:49:36 see what the bottlenecks are 2022-09-13 13:50:02 Should be similar to docker, right? 2022-09-13 13:50:21 i dunno, rv64 is fragile 2022-09-13 13:50:32 maybe in recent qemu its better 2022-09-13 13:51:14 We already have the latest qemu on usa5 2022-09-13 13:51:17 Afaik 2022-09-13 13:51:22 7.1? 2022-09-13 13:51:26 was recent 2022-09-13 13:51:27 we do? 2022-09-13 13:51:35 7.0 we did for sure i think 2022-09-13 13:53:39 worth an upgrade 2022-09-13 13:53:48 7.0 2022-09-13 13:54:39 so on 7.0 things still hang? 2022-09-13 13:54:42 like golang 2022-09-13 13:57:09 Yes 2022-09-13 14:02:40 is there some upstream issue about it? 2022-09-13 14:09:17 I don't think we have reported it. 2022-09-13 14:17:59 Piraty: the uninitialized var seems to happen in rust 2022-09-13 14:18:36 clandmeter, ikke: did we try update current riscv64 to use qemu 7.1? maybe it fixes rust build? 2022-09-13 14:19:00 No, not yet 2022-09-13 14:19:08 We'd 2022-09-13 14:19:09 we have a new box ready to have rv64 in lxc 2022-09-13 14:19:49 clandmeter: which one will we use for that? 2022-09-13 14:19:54 The dev box? 2022-09-13 14:20:03 as noted in the issue 2022-09-13 19:22:12 https://gitlab.alpinelinux.org/omni/aports/-/jobs/839358#L139 ? 2022-09-13 19:23:17 yes, known 2022-09-13 19:23:22 rust needs to be rebuilt 2022-09-13 19:23:29 but is kinda broken right now on rv64 2022-09-13 19:23:56 segfaults during build 2022-09-13 19:24:21 ah, ok 2022-09-14 04:11:38 alpine git seems to be very slow 2022-09-14 04:11:50 does anyone have a mirror i can use for cloning? 2022-09-14 04:15:02 I bet you can clone from GitHub 2022-09-14 04:26:00 what should i do with schroot? it's there's a fork that is actively maintained called reschroot 2022-09-14 04:26:30 so, should i rename the package to that or just don't rename it? 2022-09-14 04:39:33 the existing one is already reschroot(?) as it's pulling the debian source which follows that 2022-09-14 04:39:41 nothing to rename i don't think :) 2022-09-14 04:44:11 psykose, i'm thinking of switching to the git tree 2022-09-14 04:44:19 go for it 2022-09-14 04:44:28 the reason why i used debian source was because the thing was not really maintained 2022-09-14 04:46:35 psykose, thx, will do 2022-09-14 04:46:50 you can also rename then 2022-09-14 04:46:55 throw in a provides/replaces for the old name 2022-09-14 04:47:17 does `apk add schroot` after the rename works? 2022-09-14 04:49:26 if you have a provides="schroot=$pkgver-r$pkgrel", then yes 2022-09-14 04:55:39 also ppc might be stuck on openjdk on 3.16 again 2022-09-14 04:55:43 not sure when it last moved 2022-09-14 04:59:51 hmm, must be some time then, I already saw it building it again before 2022-09-14 05:00:21 power macintoshes :) 2022-09-14 05:01:19 seems like lxc now passed on rv64. I reduced the max amout of jobs a bit 2022-09-14 05:02:23 does nayone have a link for cmake standard on alpine? 2022-09-14 05:05:30 testing/qt-creator 2022-09-14 05:05:36 dunno, there's a thousand of them 2022-09-14 06:08:00 anyone here worked with po can help me? 2022-09-14 06:08:25 i got: /home/pmos/build/src/reschroot/man/po4a.cfg:1: too many POT files: 'po//schroot.pot' and 'po//schroot-man.pot'. 2022-09-14 06:08:36 https://codeberg.org/shelter/reschroot/src/branch/main/man 2022-09-14 06:08:41 no idea what it means by that 2022-09-14 07:24:03 Daaanct12: are you trying to package reschroot? 2022-09-14 07:24:14 yes 2022-09-14 07:24:24 could you send the APKBUILD? 2022-09-14 07:24:30 OK 2022-09-14 07:25:29 https://termbin.com/6ofn 2022-09-14 07:28:00 Daaanct12: you also want replaces="schroot" to make sure upgrades go smooth 2022-09-14 07:28:26 i forgot about that, thanks for picking that up 2022-09-14 10:02:49 Howdy... I got a failure to build on s390x in MR !39036 but in APKBUILD I have specified only x86 and x86_64 (due to CPU specific instruction set usage). Have I missed something or is it a known issue? 2022-09-14 10:19:44 EvTheFuture: The pipeline still runs, it normally would not build that package, but in this case, the pipeline failed due to some network issue 2022-09-14 10:20:49 EvTheFuture: Restarted the job, and now it succeeded 2022-09-14 12:50:06 yeah I was very pleased to see schroot revived 2022-09-14 17:27:33 ikke: s390x is hung so you have to restart it 2022-09-14 18:42:36 ikke: oh, any chance i could get the usual armv7 container thing to test some qtwebengine stuff 2022-09-14 19:17:31 psykose: stunnel again.. 2022-09-14 19:39:46 it hung in ci too 2022-09-14 19:41:13 seems like jirutka disabled the test 2022-09-14 19:41:27 after merge 2022-09-14 19:42:12 hmm 2022-09-14 20:46:48 hmm 2022-09-14 20:46:50 traps: fuse mainloop[58174] general protection fault ip:7f6d90cbcab7 sp:7f6d90745a28 error:0 in ld-musl-x86_64.so.1 2022-09-14 20:46:55 been getting a lot of these recently 2022-09-14 20:47:02 they don't do anything but it looks strange 2022-09-15 10:52:37 psykose: just fyi: I will add some comments to the stack haskell mr, I am just on a business trip rn and might need a few days to review it 2022-09-15 10:52:58 no rush ^^ i just don't really know much about if it's a good idea or not 2022-09-15 10:53:18 i'm busy playing factorio anyway 2022-09-15 10:53:48 watch out for them biters 2022-09-15 10:54:18 haha hf :) 2022-09-15 10:54:30 factorio guys never made discounts on steam 2022-09-15 11:46:32 *factory gone* 2022-09-15 12:24:12 Updating the py3-inform packages, and one of the tests is failing because it looks like the test expects the source be in an 'inform' directory, when actually it's in an 'inform-1.27' directory. Any way I can fix that in the package build process? 2022-09-15 12:24:13 https://github.com/KenKundert/inform/issues/7 2022-09-15 12:25:04 You'd rename the source dir en the prepare phase 2022-09-15 12:25:19 And adjust builddir accordingly 2022-09-15 12:25:43 Ok, will see if I can work out how to do that. Thanks ikke 2022-09-15 12:26:13 Or symlink it 2022-09-15 12:31:25 Ok, thanks ikke. 2022-09-15 12:41:14 psykose: btw, tried to create a container for you, but running into an issue where lxc complains armv7 is not recognized 2022-09-15 12:41:29 new lxc? or still 4.0 2022-09-15 12:41:35 confile.c: set_config_personality: 1285 Invalid argument - Unsupported personality "armv7" 2022-09-15 12:41:45 4.0 2022-09-15 12:42:23 If I look at other containers, they use armv7l as arch 2022-09-15 12:42:39 psykose: might also be an issue with the alpine lxc template 2022-09-15 12:42:58 armv7l would be the correct one 2022-09-15 12:43:59 I already had to modify the script where arm*), which sets it as armhf, was before other non-wildcard arches 2022-09-15 12:47:32 hmm, how has this ever worked then I wonder 2022-09-15 12:47:41 I did create armv7 containers before 2022-09-15 12:48:13 the template uses the same arch for the alpine repositories as the lxc arch 2022-09-15 12:52:10 strange, now it works 2022-09-15 12:53:23 oh, right, now it defaulted to armhf again :/ 2022-09-15 12:53:30 sigh 2022-09-15 12:57:13 fixed it 2022-09-15 12:59:21 psykose: 172.16.23.124 2022-09-15 12:59:38 (psykose-edge-armv7.usa9.alpin.pw) 2022-09-15 13:00:32 thanks ^_^ 2022-09-15 13:00:49 need to patch the template 2022-09-15 13:01:39 or, I suppose we need a newer version of the templates :p 2022-09-15 13:01:52 seems like just broken in 3.15 2022-09-15 13:02:23 ah ofcourse 2022-09-15 13:02:33 we patch it broken 2022-09-15 13:19:25 ikke: Think I've done it right, does this look Ok? https://gitlab.alpinelinux.org/alpine/aports/-/blob/d2700d4462be6d0389b422f7a95ce14d61141cc0/community/py3-inform/APKBUILD 2022-09-15 13:19:52 yes 2022-09-15 13:20:06 Perfect, thanks for the advice as ever ikke 2022-09-15 13:20:08 adhawkins: did you consider filing an issue upstream? 2022-09-15 13:21:46 Aldready done ikke :) 2022-09-15 13:21:52 ah, good 2022-09-15 21:22:51 Ikke: thank you 2022-09-16 13:51:51 its been very quiet in here this week, everyone at the LF events in Dublin? 2022-09-16 13:52:45 seems like matrix woes 2022-09-16 13:53:05 maybe people are speaking on matrix and forgot to reauth 2022-09-16 15:10:11 minimal: noticed the same 2022-09-16 15:10:50 its not just quiet here, the rate of MRs etc also seems lower than usual 2022-09-16 15:11:51 didn't realise so many regulars might be heading to Dublin, if I knew I might have considered going as its only down the road 2022-09-16 15:17:23 everyone is busy playing factorio, obviously 2022-09-16 15:18:11 what is there in dublin? 2022-09-16 15:21:08 Linux Foundation events this week - Embedded Conf, CloudOpen, ContainerCom, etc 2022-09-16 15:21:28 https://events.linuxfoundation.org 2022-09-16 15:22:48 ah 2022-09-16 23:13:05 Any Qt wizard may help me with arch-specific error here? https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/843010#L625 2022-09-16 23:59:00 sounds more like a cmake issue in telegram-desktop 2022-09-16 23:59:16 something is probably if'd in some way to accidentally also use 5 at the same time 2022-09-16 23:59:27 maybe they 'disabled qt6 for mobile arm' or something, and ended up with that with both installed 2022-09-16 23:59:31 i'd read through all the files 2022-09-17 00:03:41 basically in cmake they find qt in some way, and say what version they want (i guess), and something includes both, but only for not-x86_64 2022-09-17 00:06:44 psykose: I had to add the qt5-qtsvg-dev package even though 6 is already installed... 2022-09-17 00:07:01 the error you get without it is the first place to look 2022-09-17 00:07:14 should not be needed 2022-09-17 00:08:08 That error also shows up in x86_64 without the package 2022-09-17 00:10:11 https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/842997#L518 2022-09-17 00:12:13 Maybe qt6-qt5compat-dev broke and doesn't have SVG anymore? 2022-09-17 00:18:45 I think for some reason this is importing Qt5 https://github.com/desktop-app/cmake_helpers/blob/b9f1b6036329e44ec61f51f0b363fe2319158b84/external/qt/package.cmake#L28 2022-09-17 00:19:41 It was changed in this commit https://github.com/desktop-app/cmake_helpers/commit/9615515b069a1d3174fc45f9de69a23b616b80b3 2022-09-17 00:21:00 I have no clue what this means and why it breaks 2022-09-17 00:23:16 I think I got it. I had to add kcoreaddons-dev as it seems like a new dependency, but that also adds qt5-qtbase-dev which makes the scripts understand that Qt5 is being used and fail. 2022-09-17 00:24:13 Maybe a Qt6 kcoreaddons must be made? 2022-09-17 00:24:19 you can try define DESKTOP_APP_QT6 2022-09-17 00:24:20 dunno 2022-09-17 00:26:34 Even if that fixed that part, wouldn't it fail as kcoreaddons links to Qt5? 2022-09-17 00:39:16 not necessarily 2022-09-17 00:39:22 but kcoreaddons is only qt5 anyway, isn't it 2022-09-17 00:40:31 psykose: Nah https://github.com/KDE/kcoreaddons/search?q=qt6&type=commits 2022-09-17 00:43:28 add -DQT_MAJOR_VERSION=6 to kcoreaddons locally and check if that works 2022-09-17 00:43:52 psykose: It would require having a separate qt6-kcoreaddons though, right? 2022-09-17 00:44:00 to actually commit sure 2022-09-17 00:44:15 i have a suspicion it's not the actual issue 2022-09-17 00:44:26 because it doesn't need "qt6 kcoreaddons" on x86_64 2022-09-17 00:44:39 if it did it would fail everywhere all the time 2022-09-17 00:46:05 What is "everywhere all the time"? I haven't seen any other distro package tdesktop 4.2.0 yet 2022-09-17 00:46:24 It does fail in x86_64 if I don't add qt5-qtsvg 2022-09-17 00:54:42 that's what i mean 2022-09-17 00:55:18 if it needed kcoreaddons-qt6 it would not pass on x86_64 without it 2022-09-17 00:55:28 What I believe is that it doesn't break with mixed Qt versions in x86_64 but it does on other arches 2022-09-17 00:55:35 And it's not _intended_ to use mixed Qt versions 2022-09-17 00:55:46 So kcoreaddons-qt6 is the "right thing" 2022-09-17 00:57:22 Would qt6-kcoreaddons or kcoreaddons-qt6 be preferred 2022-09-17 00:57:23 ? 2022-09-17 00:58:23 latter 2022-09-17 01:12:40 ncopa: if you get a chance, could you rebase https://github.com/python/cpython/pull/18380? 2022-09-17 01:42:30 If this works, should I send it as a separate MR? I don't really want to maintain kcoreaddons-qt6, how can I look for a maintainer? 2022-09-17 01:43:08 kcoreaddons-qt6 does fix x86_64: https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/843087 2022-09-17 01:43:23 And it looks like aarch64 too: https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/843091 2022-09-17 01:58:33 you should build it in the same kcoreaddons apkbuild 2022-09-17 02:12:19 psykose: I'm not sure how I would go about doing that 2022-09-17 12:55:42 psykose: in aports you added testing/qbs last month even though community/qbs already existed. Any reason for this or an accident? testing version is newer though than community 2022-09-17 13:06:23 sometimes happens that you overlook an existing package 2022-09-17 13:07:03 https://pkgs.alpinelinux.org/packages?name=qbs&branch=edge&repo=&arch=&maintainer= does not return community/qbs 2022-09-17 13:07:31 ah, arch="!s390x" 2022-09-17 13:07:35 so it's effectively disabled 2022-09-17 15:11:33 oh amazing haha 2022-09-17 15:11:35 I was already wondering why I didn't find it in pkgs site 2022-09-17 15:16:16 ikke: what would that even mean? it gets built for nothing? 2022-09-17 15:17:04 Newbyte: correct, no builder will pick it up 2022-09-17 15:17:17 how did that get into community 2022-09-17 15:19:05 ACTION uploaded an image: (107KiB) < https://matrix.org/_matrix/media/r0/download/matrix.org/aHHahErAKGCPiTqBOOavVQgS/clipboard.png > 2022-09-17 15:19:08 also, what happened here? 2022-09-17 15:20:24 someome merged a merge request without looking at the commits 2022-09-17 15:23:11 re qbs: !5360 2022-09-17 16:33:02 Is it ok if paths from builders leak into packages? 2022-09-17 17:08:51 Ermine: it's not a disaster, but we try to avoid it if possible 2022-09-17 20:12:30 How bad or good is this of an idea? https://github.com/telegramdesktop/tdesktop/issues/25073#issuecomment-1250133715 2022-09-17 20:13:01 As I understand it it's effectively creating a nonstandard path for the Qt6 version of this library, which I would assume Alpine wouldn't like 2022-09-17 20:13:52 Nulo: others already commented that it will conflict 2022-09-17 20:14:11 I guess you are one of them 2022-09-17 20:15:20 ? But what I linked is an approach where the Qt6 package would save things in a different path 2022-09-17 20:16:49 And the other one is statically linking it? Or it's the same approach and I'm confusing it 2022-09-17 20:18:03 This is so messy, damn 2022-09-17 20:22:34 yup, sorry, I cannot really help with this 2022-09-17 20:23:15 I guess I'll wait for it to settle down with the (more experienced) maintainers of other distros and see their approach. 2022-09-17 22:21:52 z3ntu: i legitimately didn't find it, haha 2022-09-17 22:22:23 but yeah, apparently was disabled 2022-09-17 23:10:19 I'm trying to port a piece of code to musl for dotnet's port of s390x which tries to use setcontext. Despite libucontext-dev being installed, it still dies with Undefined reference. Any pointers on what I'm missing? The code in question is here: https://github.com/mono/mono/blob/main/mono/mini/exceptions-s390x.c#L782 2022-09-17 23:10:59 I tried to explicitely include the headers via #include to no avail. I'm a neophyte to C, so maybe I'm missing something obvious 2022-09-17 23:20:02 Found the fix: linker flag -lucontext 2022-09-18 00:01:42 psykose: https://gitlab.alpinelinux.org/alpine/aports/-/issues/14198#note_263319 2022-09-18 00:01:47 I don't think this is right, it definitely does use seccomp at runtime 2022-09-18 00:01:59 (whether or not it should is another question) 2022-09-18 00:02:12 i worded it poorly, what i meant was my understanding of it was it tried, but failing didn't do anything aside from print a warning 2022-09-18 00:02:17 ohhh 2022-09-18 00:02:38 then again, the bpf seccomp stuff is quite deep for me 2022-09-18 00:02:49 maybe there's a side effect where it fails and stuff explodes 2022-09-18 00:03:28 at here https://github.com/gentoo/pax-utils/blob/2d981305b117b669c60bede076557c2d765cf198/security.c#L144 2022-09-18 12:38:33 Where am I supposed to pass -DCMAKE_INSTALL_LIBDIR= to? 2022-09-18 12:43:40 I guess in the build command instead of the install one 2022-09-18 12:44:05 yes 2022-09-18 12:44:36 The initial cmake invocation 2022-09-18 13:37:46 >>> kcoreaddons-qt6*: Preparing subpackage kcoreaddons-qt6... 2022-09-18 13:37:58 abuild: cd: line 2486: can't cd to /home/diablo/proy/aports/community/kcoreaddons/pkg/kcoreaddons-qt6: No such file or directory 2022-09-18 13:38:13 Why is that directory not created? 2022-09-18 13:38:20 You need to do that in the split function 2022-09-18 13:38:31 abuild does not create it itself 2022-09-18 13:39:13 Are you installing / moving stuff to $subpkgdir? 2022-09-18 13:43:05 ikke: I think I incorrectly set it to pkgdir, but it seems to me that abuild itself is trying to cd into a non-existant folder _before_ it runs the split function 2022-09-18 13:44:26 No, this runs after the split functions have already run 2022-09-18 13:44:31 https://gitlab.alpinelinux.org/alpine/abuild/-/blob/master/abuild.in#L2117 2022-09-18 13:44:51 prepare_subpackages runs the split functions 2022-09-18 13:45:21 Line 2486 is part of splitbin, which is called by prepare_package 2022-09-18 14:07:42 Okay, thanks. I have now made the very cursed -qt6 subpackage. 2022-09-18 14:52:05 Okay, I got pretty far, but I'm not sure how to make CMake use desktoptojson-qt6 instead of desktoptojson https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/844442#L3297 2022-09-18 14:58:55 Does it do something like add_executable("desktoptojson" imported)? 2022-09-18 14:59:00 Yep. 2022-09-18 15:07:09 I guess I can use CMAKE_INSTALL_BINDIR 2022-09-18 15:43:25 Okay, it builds. Can someone actually experienced check !39162? 2022-09-18 19:59:34 Going to upgrade gitlab 2022-09-18 20:05:57 Upgrade complete 2022-09-18 21:03:15 pog 2022-09-18 21:09:49 ikke: there's a very high incidence of #1: PUT https://gitlab.alpinelinux.org/api/..etcetcetc: 422 {message: Branch cannot be merged} now 2022-09-18 21:09:52 when merging via glab 2022-09-18 21:10:02 it fails 80% of the time 2022-09-18 21:10:23 :-/ 2022-09-19 09:56:31 Could someone take a 2nd look at !39229 ? 2022-09-19 10:18:10 Cogitri: I assume someone (eg. infra team) needs to take over alpine-qa-bot? 2022-09-19 10:28:00 psykose: seems to only affect glab? 2022-09-19 10:28:55 like it's not waiting long enough before performing the merge or something like that 2022-09-19 10:50:55 "Cogitri: I assume someone (eg..." <- Oh right, I can also keep that if no one else wants it 2022-09-19 11:46:32 wtf. 2b934708063f19dbb228ba46d1bf59a430d76b49 2022-09-19 11:49:13 ah, jirutka rightfully complained in the mr already, sorry to bother :p 2022-09-19 12:14:47 oh noes Cogitri :( 2022-09-19 12:25:57 markand: Yeah unfortunately with work and my master studies starting there’s a lot of stuff going on 2022-09-19 12:27:02 Cogitri: thank you for your help so far! 2022-09-19 12:27:43 <3 2022-09-19 12:28:25 Cogitri: has been nice to have you on the team, and has always been nice work with you. best wishes for your studies 2022-09-19 12:37:22 Thanks a lot for all the help along the way, Alpine has certainly helped me a lot to understand the ins and outs of linux, distros and communities in general :) 2022-09-19 12:37:47 Cogitri: we could move the project to the infra namespace 2022-09-19 12:38:03 And thanks from my side as well 2022-09-19 12:41:17 psykose: c384b4ce6eb4955503cbd2f62d0db2c0170dd507 appearently broke phosh for me, a downgrade to glib-2.72 fixed it. https://paste.rs/whn 2022-09-19 12:42:54 That would be nice, yes 👍 2022-09-19 15:16:28 PureTryOut: https://build.alpinelinux.org/buildlogs/build-edge-aarch64/community/pure-maps/pure-maps-3.1.1-r1.log 2022-09-19 19:20:24 Hi yall, I want to pin a package dependancy to ver 0.6 2022-09-19 19:20:31 I dont care about the revision 2022-09-19 19:20:47 do I do $pkgname~=0.6 2022-09-19 19:20:55 or $pkgname~0.6 ? 2022-09-19 19:21:29 anjan: and what if the repo doesn't contain version 0.6 of that package? 2022-09-19 19:21:38 or am I completely wrong? I cant find documentation anywhere 2022-09-19 19:21:50 minimal: I dont want the package to be installed 2022-09-19 19:21:55 I want it to throw an error. 2022-09-19 19:22:12 (this is for postmarketos) 2022-09-19 19:22:18 and it's a ui package 2022-09-19 19:22:35 anjan: and if it is installed successfully and then later an update occurs of the dep package, what then? 2022-09-19 19:23:01 I want apk to trhwo unsatisfied constraints when apk -U upgrade is ran 2022-09-19 19:23:08 and it cant satisfy the constraints 2022-09-19 20:50:30 ikke: yeah, probably more of a glab issue 2022-09-19 20:51:42 anjan: when apk upgrade would run, it wouldn't print anything but refuse to upgrade the pinned thing, unless something explicitly wants a newer than 0.6 version 2022-09-19 20:52:01 Piraty: can you give any more info than 'it broken' (and perhaps open an issue) :) 2022-09-19 20:52:46 psykose: ok, how do I do that? $pkgname~=0.6 or $pkgname~0.6 2022-09-19 20:56:21 =~0.6 2022-09-19 20:57:10 well, that would also allow 0.6.1 2022-09-19 20:57:59 psykose: I just wanna make sure revisions are ignored 2022-09-19 20:58:18 but I can communicate with the package maintainer to make sure they dont tag 0.6.1 heh 2022-09-19 20:58:31 ~0.6.0 should work then? 2022-09-19 20:58:56 then =~0.6.0 2022-09-19 20:59:21 you can see what does what by running apk add with random packages and versions right now 2022-09-19 21:00:12 i.e. apk add firefox=~104.0.2 or 104.0 would work, == would fail without adding -r 2022-09-19 21:00:49 ok sounds good 2022-09-19 21:00:56 can I document this somewhere? 2022-09-19 21:01:01 the manpage for APKBUILD? 2022-09-19 21:01:08 would yall accept a patch for that? 2022-09-19 21:01:15 or is the wiki good 2022-09-19 21:02:49 yes from me to all of it 2022-09-19 21:08:52 psykose: from my experience, ~0.6.0 and =~0.6.0 are the same 2022-09-19 21:09:57 yeah, most likely 2022-09-19 21:10:08 not sure why i added the = 2022-09-19 21:11:55 psykose: re glab: it appears like glab is not waiting for the rebase to be finished, it returns the 422 error immedately 2022-09-19 21:12:00 after then retrying, it works 2022-09-19 21:12:54 yeah, it does 2022-09-19 21:12:56 noticed that just now 2022-09-19 21:13:05 call once -> wait 10 seconds -> call again -> works 2022-09-19 21:13:11 yes, exactly 2022-09-19 21:13:17 strange it's triggered only by the new update - the old gitlab waited for rebase 2022-09-19 21:13:21 i guess they changed some minor api 2022-09-19 21:13:28 probably some changes in behavior 2022-09-19 21:13:32 glab moved repos from github to gitlab proper now, but no new release yet 2022-09-19 21:13:37 yes 2022-09-19 21:13:40 gitlab-org/cli 2022-09-19 21:13:42 mhm 2022-09-19 21:14:09 I already looked at existing issues 2022-09-19 21:14:10 i wonder if rebase-without-ci is supported by our gitlab 2022-09-19 21:14:18 next minor release 2022-09-19 21:14:22 neat 2022-09-19 21:14:31 if glab also supports that it makes it more friendly to ci to merge things 2022-09-19 21:15:03 note that there is some unexpected behavior atm icm with 'pipelines must succeed' 2022-09-19 21:15:13 meaning the merge button disappears 2022-09-19 21:15:17 and you cannot merge it 2022-09-19 21:15:21 at all? strange 2022-09-19 21:15:23 haven't noticed that yet 2022-09-19 21:15:29 via glab at least 2022-09-19 21:15:31 no, it would at the next release 2022-09-19 21:15:33 ah 2022-09-19 21:15:37 if you use that feature 2022-09-19 21:16:28 https://gitlab.com/gitlab-org/gitlab/-/issues/372218 2022-09-19 22:03:55 Piraty: ah, i missed the paste. i found even more issues, so i downgraded it for now 2022-09-19 22:39:01 can someone here check if they can open notepad or whatever in wine? it seems to work for me but not others 2022-09-19 22:55:04 it didn't work for me several months ago 2022-09-19 22:55:58 which is strange, since i always check at least that and use it to run a game mod installer too 2022-09-19 22:56:06 the other reported issue was wine-in-root, i know that is fucked 2022-09-19 22:56:15 but even regular dir fails for some reason, just.. not for me 2022-09-19 22:56:20 i wonder why my pc is special :-) 2022-09-19 22:58:40 ah, works as root too now 2022-09-19 22:59:13 but only in a container.. 2022-09-19 23:04:56 Hello71: is your root on a raw device, or a device mapper (lvm luks etc)? 2022-09-19 23:05:37 looking at the strace it's reading /dev/dm-0 on a non-working setup, maybe there's something funny there 2022-09-19 23:05:55 i just use a raw device myself (UUID= to a /dev/nvme0n1p2) 2022-09-19 23:06:59 the others i know with broken wine are also on lvm 2022-09-19 23:07:52 (and then i guess running a containerd container abstracts that in some way and it works again..) 2022-09-19 23:08:02 then the root thing seems like another issue (works in container root but fails for me) 2022-09-20 01:30:01 psykose: iirc it was broken in podman with overlayfs lowerdir=tmpfs upperdir=tmpfs 2022-09-20 01:30:50 i think rr might be a really good tool for debugging wine strictly because of rr replay -p 2022-09-20 01:31:08 does that even build on musl 2022-09-20 01:31:11 iirc it doesn't 2022-09-20 01:31:26 the biggest problem with debugging early-start wine issues is that it starts like half a dozen processes and each one has like a dozen threads 2022-09-20 01:32:22 huh, i misrememebered, i thought alpine had rr but i was thinking of void 2022-09-20 01:32:33 (glibc only) 2022-09-20 01:34:40 beep boop? 2022-09-20 01:34:42 what's rr? 2022-09-20 01:34:44 yeah, i tried to package rr two or three times, it's incredibly broken on musl 2022-09-20 01:34:55 https://rr-project.org/ 2022-09-20 01:35:17 basically a fancy debugger 2022-09-20 01:35:44 https://github.com/rr-debugger/rr/issues/2236 seems like for part 1 you could just do s/64\>/_64/ 2022-09-20 01:36:34 i still don't know what exactly libthread_db does but iirc musl implements something compatible or at least good enough for gdb? 2022-09-20 01:41:27 PureTryOut: why are some qt5 libraries using a commit hash rather than a tag? i've looked at a few and they all have a tag like v5.15.5-lts-lgpl 2022-09-20 02:00:32 theoretically the tag can be changed but the commit cannot 2022-09-20 02:01:35 it would be nice if there was some way to specify a tag and (tag or commit) hash and it would automatically verify that they are the same 2022-09-20 02:04:21 that applies to everything ever though, and the same strategy is not used there, so i don't think that's the reason 2022-09-20 02:04:32 i mean, that's why we have checksums for files 2022-09-20 02:04:55 so even if someone moves the tag, abuild just fails, because that's clearly not the source that was intended to be used 2022-09-20 02:05:10 s/$/ by the packager/ 2022-09-20 02:07:37 hm, yes. i recall some similar reason but this specific explanation is not correct 2022-09-20 03:57:22 hey yall, thanks for the help earlier today with ~. I sent a MR to APKBUILD man pages and updated the wiki on APKBUILDs. I mightve not chosen the right words but something is better than nothing hopefully. 2022-09-20 04:50:09 ptrc: they are following the KDE patches, which they do not tag 2022-09-20 04:50:50 I update them to a newer commit every once and while 2022-09-20 05:32:25 Hmm, strange. For some reason, when using abump, srcdir is empty in the global scope, resulting go packages to fail 2022-09-20 08:44:30 would anyone opposed to splitting libpurlple-dev from pidgin-dev ? 2022-09-20 09:06:14 psykose: thanks for dealing with glib 2022-09-20 09:35:24 Piraty, that makes sense 2022-09-20 09:36:27 yeah it does :p 2022-09-20 09:37:05 can apk list files of a package that is not installed? 2022-09-20 09:37:15 like xbps-query -R -f foo 2022-09-20 09:53:31 no. we usually use https://pkgs.alpinelinux.org/contents for this purpose 2022-09-20 10:52:11 Piraty: what would the point be? splitting dev packages is usually pointless 2022-09-20 10:52:37 ikke: yep, noticed that too 2022-09-20 10:53:09 hi guys, I'm trying again to build purple-gowhatsapp 2022-09-20 10:53:14 any clue with this: (12:52:37) plugins: /usr/lib/purple-2/libwhatsmeow.so is not loadable: Error relocating /usr/lib/purple-2/libwhatsmeow.so: sqlite3_set_authorizer: symbol not found 2022-09-20 10:53:17 psykose: building libpurple plugins won't pull gtk2-dev et al 2022-09-20 10:53:40 sqlite is installed as dependency 2022-09-20 10:53:56 that sounds fine to me 2022-09-20 10:54:03 where do you even see that 2022-09-20 10:54:14 it's a plugin so it's probably just dlopened, and sqlite is already loaded 2022-09-20 10:54:21 ldd is worthless to see if things resolve there 2022-09-20 10:54:35 I see it at pidgin startup 2022-09-20 10:54:42 now if normally the thing doesn't even use sqlite.. different issues 2022-09-20 10:56:23 I'm trying to build without tag=libsqlite3, I think that it fails similar but differently 2022-09-20 10:57:05 (12:56:57) plugins: /usr/lib/purple-2/libwhatsmeow.so is not loadable: Error relocating /usr/lib/purple-2/libwhatsmeow.so: sqlite3_autovacuum_pages: initial-exec TLS resolves to dynamic definition in /usr/lib/purple-2/libwhatsmeow.so 2022-09-20 10:57:57 so it's a problem with purple plugin loading and musl? 2022-09-20 11:02:10 someone already complained about this issue, uh... last year? 2022-09-20 11:02:44 oh, it was you 2022-09-20 11:03:54 you said "2022-04-14 16:31:04 donoban wow Hello71 , now I get a segfault :\" and then didn't paste the backtrace 2022-09-20 11:05:02 thanks Hello71 , I did but the backtrace was confusing, it's on github issue, let me find it 2022-09-20 11:05:20 https://github.com/hoehermann/purple-gowhatsapp/issues/119 2022-09-20 11:05:56 they released 3 new versions and I hoped that it just works now :D 2022-09-20 11:06:11 you need to install musl-dbg for backtrace 2022-09-20 11:06:29 uhM, I'm gonna try again with buildmode=c-archive 2022-09-20 11:06:32 it's a GOFLAG? 2022-09-20 11:06:48 that doesn't sound like a good idea but you can try if you want 2022-09-20 11:08:32 yeah modifying a build paramater that I don't know what does it's not a good idea :D 2022-09-20 11:10:28 go: parsing $GOFLAGS: non-flag "buildmode=c-archive" 2022-09-20 11:12:20 it has to be passed to 'go build' 2022-09-20 11:13:00 "As an alternative I tried to build it with system sqlite3 lib, but it ended:" 2022-09-20 11:13:07 I don't remeber how I did this 2022-09-20 11:13:20 well, you have to at least put it in flag format. i don't think i recommended that though 2022-09-20 11:19:25 same problem with c-archive 2022-09-20 11:20:32 wops lol, line was #commented 2022-09-20 11:24:54 meh, it doesn't help 2022-09-20 11:26:47 "On AIX, when linking a C program that uses a Go archive built with -buildmode=c-archive, you must pass -Wl,-bnoobjreorder to the C compiler." 2022-09-20 11:27:51 that -Wl flag is specific to AIX 2022-09-20 11:28:09 maybe there is another go based plugin that also uses sqlite and I could compare it 2022-09-20 11:28:36 ah, ok, I'm gonna try PIE mode 2022-09-20 11:29:00 but I'm not really sure if the flag is being used 2022-09-20 11:29:44 export GOFLAGS="$GOFLAGS -buildmode=pie" 2022-09-20 11:30:10 looks ok? 2022-09-20 11:32:40 that's already there 2022-09-20 11:33:21 uhM, and specifying c-archive will override it? 2022-09-20 11:47:18 -Wl,-b NOOB JRE order . lulz 2022-09-20 15:41:58 https://tpaste.us/RxJr 2022-09-20 15:42:06 any idea of how I try to build it static? 2022-09-20 15:42:51 ouch, maybe I'm overwrting linkmode with GOLDFLAGS.. 2022-09-20 15:45:01 [ 13%] Linking C static library libpurple-dummy.a 2022-09-20 15:46:33 but then, [100%] Linking C shared library libwhatsmeow.so 2022-09-20 15:46:56 and finally: (17:45:13) plugins: /usr/lib/purple-2/libwhatsmeow.so is not loadable: Error relocating /usr/lib/purple-2/libwhatsmeow.so: sqlite3_autovacuum_pages: initial-exec TLS resolves to dynamic definition in /usr/lib/purple-2/libwhatsmeow.so 2022-09-20 18:48:39 never had that :o 2022-09-21 10:53:28 Hello71: I returned to same segfault, a patch that I had already added sqlite to CMake 2022-09-21 10:55:34 wow, now backtrace looks more intereting 2022-09-21 10:56:13 http://ix.io/4b2g 2022-09-21 10:59:16 it looks that crashes while trying to print the loading error 2022-09-21 10:59:30 "Error relocating %s: %s: initial-exec TLS resolves to dynamic definition in %s" 2022-09-21 11:41:44 i just added a wrapper around go on the riscv64 builder. it will make qemu strace the output to stderr but the wrapper will redirect it to a separate logfile. 2022-09-21 11:42:23 so any error msg to stderr will not be logged in the regular builder log file. 2022-09-21 11:43:14 i hope we can get some more information why golang pkgs will sometimes hang. so after we get that log we will remove the wrapper again. 2022-09-21 13:41:51 donoban: most likely one of those %s is null 2022-09-21 13:42:19 donoban: would be curious to see the object file in question (libwhatsmeow.so) 2022-09-21 14:06:04 Ariadne: it's these https://github.com/golang/go/issues/54805 https://github.com/golang/go/issues/13492 most likely 2022-09-21 14:09:49 psykose: ah 2022-09-21 14:10:24 well, go should not do that. there is no requirement in ELF that DSO init functions get passed argc/argv/envp 2022-09-21 14:11:20 unfortunately, it appears that they have built the c-archive/c-shared support around the assumption that this holds true 2022-09-21 14:11:25 bummer :) 2022-09-21 14:18:16 sadly golang is just a small indie project 2022-09-21 14:18:27 if only they had a large enterprise behind them to fund resolving such issues :/ 2022-09-21 14:19:20 We should start a fundraiser :( 2022-09-21 14:25:57 Hmm, the patch from upstream doesn't work for some reason. It tries to use a header file from an .xml DBus thingy, that it seemingly does build (https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/847932#L819) but it says it can't find it (https://gitlab.alpinelinux.org/Nulo/aports/-/jobs/847932#L829) 2022-09-21 14:26:40 This is probably part of the "include errors" that they warned about (https://github.com/telegramdesktop/tdesktop/issues/25073#issuecomment-1253263014) 2022-09-21 14:28:41 yes, it's wip 2022-09-21 14:28:47 just wait for the next release (hopefully 4.2.1) 2022-09-21 14:41:38 As far as I understand from the issue, the dev isn't particularily interested in continuing that patch though. 2022-09-21 14:42:39 It seems there's also an optional dependency in kimageformats that's also built with Qt5 in Alpine currently T.T 2022-09-21 14:43:31 Whatever. It seems to me that the only distros that have packaged 4.2.0 use Qt5 (which isn't supported upstream...) so I'll just wait 2022-09-21 14:43:42 Thanks for the help anyway! 2022-09-21 14:48:50 :) i hope there'll be the fix soon 2022-09-21 14:49:21 (and kimageformats needs karchive..) 2022-09-21 15:22:15 Plot twist: the upstream person is fixing the patch, including the broken condition that psykose mentioned 2022-09-21 15:25:42 Ariadne: how could I share it to you? 2022-09-21 15:26:30 uhm, could this be fixed trying to build it statically? 2022-09-21 15:26:36 well, "fixed" ;) 2022-09-21 15:27:06 well, you cannot really dlopen a static file 2022-09-21 15:27:35 uhM, but could it have no .so dependencies? 2022-09-21 15:27:53 I'm not sure if it fails due the plugin itself or sqlite 2022-09-21 15:28:05 it's a go bug and there is no workaround 2022-09-21 15:28:08 there, we solved it 2022-09-21 15:28:31 hehe 2022-09-21 15:28:37 great! 2022-09-21 16:52:02 is this PR https://github.com/golang/go/pull/52541 currently used in Alpine? if no, could I try to add it? 2022-09-21 18:18:16 ariadne, the %s being null wasn't expected by me but is a possibility when the offending reloc is local not symbolic. i'll fix that. but it's not the cause of crash 2022-09-21 18:18:36 dalias: yeah 2022-09-21 18:18:59 re: the wrong init_array thing, i think alpine should either apply that PR or just apply a simpler version that blocks argv access 2022-09-21 18:19:17 (e.g. give the library a fake argv with just program_invocation_name) 2022-09-21 18:19:38 i don't like the 2022-09-21 18:19:48 i don't like the "let's read procfs to emulate the wrong behavior" thing 2022-09-21 18:20:14 and i really doubt any dynamic-loaded library actually has reason to read argv[1..argc-1] 2022-09-21 18:26:46 i think i covered that with "go should not do that" :p 2022-09-21 19:06:20 yeah 2022-09-21 21:37:54 saw on upx issue thread that alpine supposedly used 64k page size on aarch64.. is this true? 2022-09-21 21:38:09 if that's being done by choice, it's a choice that pessimizes musl so probably not a good idea 2022-09-21 23:08:28 ikke: fyi approving mrs is broken for assignees (and reviewers) 2022-09-21 23:09:07 dalias: 64k for upx in some fashion or 'in general'? 2022-09-21 23:09:46 afaik you can't even use <16k on hardware like the m1, so it doesn't make sense to not pick the higher page size 2022-09-21 23:09:59 jemalloc picks 64k too 2022-09-21 23:25:54 psykose, upx isn't related, i just saw it there 2022-09-21 23:26:41 aiui m1 can use smaller, apple just doesn't 2022-09-21 23:27:05 you need smaller for emulating guests with smaller 2022-09-21 23:27:11 and musl strongly prefers smaller 2022-09-21 23:34:17 there is a 4k kernel patch that is 10% slower, sure 2022-09-21 23:34:25 it does not make sense to me to use unless you like things to be slow for no reason 2022-09-22 09:46:15 dalias: no, # CONFIG_ARM64_64K_PAGES is not set 2022-09-22 09:46:25 CONFIG_ARM64_4K_PAGES=y 2022-09-22 09:46:54 but we have 64k pages for ppc64le 2022-09-22 09:47:12 and I believe that custom kernels for apple m1 chip has 64k pages 2022-09-22 09:48:20 haven't checked what the RPI config has 2022-09-22 10:56:36 any tip for local building go? 2022-09-22 10:56:49 my system seems stucked with 7,600 tasks and 24k threads 2022-09-22 11:00:29 well, it worked disabling test 2022-09-22 11:05:07 hmm.. now rootbld uses official go instead my built version (pkg was bumped) 2022-09-22 11:10:44 uhM, only testing repo is passed to rootbld 2022-09-22 11:10:50 is this expected/desired? 2022-09-22 11:11:16 '/etc/apk/repositories' is read only 2022-09-22 11:11:19 ouch 2022-09-22 11:19:26 (73/237) Installing go (1.19.1-r1) 2022-09-22 11:19:29 let's try 2022-09-22 11:19:49 let's go 2022-09-22 11:33:51 (13:32:54) plugins: probing /usr/lib/purple-2/libwhatsmeow.so 2022-09-22 11:33:54 IT WORKS! 2022-09-22 11:34:10 so you applied that MR? 2022-09-22 11:34:18 well, it needs LD_PRELAODING yet 2022-09-22 11:34:32 yes, I fastported it to 1.19 2022-09-22 11:35:28 should I create a separated PR for go? I'm considering rework it for adjust to dalias position 2022-09-22 12:52:31 ncopa, thanks 2022-09-22 12:53:24 musl will tolerate larger pages but it generally assumes/wants pages to be a reasonable granularity for fine-grained control of usage 2022-09-22 13:40:53 Kinda OT but I'm happy to be one of the first to package tdesktop 4.2.0 with Qt6 2022-09-22 13:41:19 Gentoo and Debian still use Qt5 which is unsupported upstream 2022-09-22 13:42:44 (and the problem with 4.2.0 only manifested with Qt6) 2022-09-22 13:44:04 wouldn't say that having a newer package than Debian is an achievement 2022-09-22 13:44:21 I'm talking about Debian sid, which usually is pretty updated 2022-09-22 13:44:45 But particularily, Debian sid already has 4.2.0, but they are still using Qt5 which is explicitly unsupported by upstream 2022-09-22 13:44:47 and it never stopped Debian from "fixing" packages even when unsupported upstream 2022-09-22 17:56:34 Ariadne: would you be interested in giving me push access to your alpine-gcc repo? would love to do the gcc 12.2 upgrade and update some of our gcc-go patches 🥺👉👈 2022-09-22 18:04:36 nmeum: we should move it to alpine org 2022-09-22 18:04:57 yes, I think that would be a good idea 2022-09-22 18:07:27 Do you want me to move it? 2022-09-22 18:10:36 (dumb question) go binary from gcc-go can be used as replacement of go? 2022-09-22 18:11:27 ikke: sounds good to me 2022-09-22 18:12:01 donoban: yes 2022-09-22 18:12:34 oh, so just s/go/gcc-go on makedepends could work? 2022-09-22 18:13:09 maybe. haven't tried that with more complex aports. I am primarly interested in using gcc-go to bootstrap go to have a cleaner bootstrap path, gcc-go upstream doesn't fully support musl targets yet I am currently upstreaming our patches 2022-09-22 18:13:27 Ariadne: nmeum done 2022-09-22 18:13:39 ty 2022-09-22 18:14:11 it's building, it looks identical for now 2022-09-22 18:16:16 yes, it works quite well actually. can even be used to boostrap go itself already (see !27944) 2022-09-22 18:17:02 strangely it failed due x86_64-alpine-linux-musl/bin/ld: cannot find -lsqlite3: No such file or directory 2022-09-22 18:17:53 sqlite-dev is a makedepend 2022-09-22 18:20:12 the cgo stuff may still need some love, haven't tested that yet since it is not really used by go itself 2022-09-22 18:20:23 if you have something that is easy to reproduce then feel free to create a bug report in the alpine gitlab 2022-09-22 18:20:25 ouch, without -tags=libsqlite3 -> http://ix.io/4b9z 2022-09-22 18:20:29 it was close :D 2022-09-22 18:31:14 nmeum: do you think it's related to gcc-go or likely something wrong with cmake? 2022-09-22 18:32:37 I don't know how the cmake build system calls gcc-go. you would have to debug it further and maybe came up with a minimal example to reproduce the linking error 2022-09-22 18:33:21 ok 2022-09-22 18:33:53 if it can't find -lsqlite then there might be an issue with the gcc-go search path 2022-09-22 18:40:54 I will check tomorrow, thanks 2022-09-22 18:49:11 donoban: np, if you have more information just create a bug report in our gitlab :) 2022-09-23 04:37:36 [14:41:19] Gentoo and Debian still use Qt5 which is unsupported upstream 2022-09-23 04:37:40 (we do support Qt 6 for it now) 2022-09-23 04:38:20 not that it takes away any achievement; telegram is a pain to wrangle 2022-09-23 05:37:20 I'm having issues reaching pkgs.a.o over IPv6 2022-09-23 05:43:24 can confirm, need to check 2022-09-23 05:48:14 oh 2022-09-23 05:48:15 hmm 2022-09-23 09:10:08 Ariadne: is there a reason why our gcc aport uses manually created git-archives instead of relying on the weekly snapshots provided by gcc upstream as tarballs on the gcc mirrors? e.g. https://ftp.fu-berlin.de/unix/languages/gcc/snapshots/ 2022-09-23 13:04:21 a lot of python riscv will be quite fucked now with rust missing but py3-cryptography needing it 2022-09-23 13:04:23 unfortunate 2022-09-23 13:23:52 i suppose we need to bootstrap runst again for riscv64? 2022-09-23 13:24:08 im working on samba 4.16 right now 2022-09-23 13:24:35 samba has a horribly difficult dependency chain 2022-09-23 13:24:45 but i think i have it properly done now 2022-09-23 13:25:39 last i checked it was just ldb/tdb/talloc/tevent at very specific versions 2022-09-23 13:25:55 but.. literally all the packaging submodule stuff changed 2022-09-23 13:26:06 i spent an hour on it and gave up cause literally all the files moved or some shit 2022-09-23 13:26:25 ncopa: yeah, we need to rebootstrap 2022-09-23 13:27:05 1.64 came out yesterday (and i bumped to it), so 1.63 is needed. riscv was still at 1.62.1 after failing to build 1.63 for a month (or even 1.62.1 with the openssl3 rebuild, actually) 2022-09-23 13:27:52 not sure how you want to approach it this time, maybe i'd cross 1.63 from gentoo again then try build 1.64 with 1.63 to see if it even works 2022-09-23 13:28:04 instead of straight to 1.64 and self-compile only 2022-09-23 13:28:12 just to make sure it even compiles a newer version itself 2022-09-23 13:28:43 im pushing samba-4.16 2022-09-23 13:29:19 i think i cross-complied it on my x86_64 desktop last time. it was s390x that was built on s390x gentoo env with help from IBM people 2022-09-23 13:29:21 well done 2022-09-23 13:29:28 aha, right 2022-09-23 13:29:42 well, whatever works 2022-09-23 13:29:47 hopefully i didn't break anything in there 2022-09-23 13:29:55 don't think i changed anything in rust since then 2022-09-23 13:30:21 are there anything riscv64 related in the rust changelog? 2022-09-23 13:30:49 not in the short notes 2022-09-23 13:30:51 https://github.com/rust-lang/rust/releases/tag/1.64.0 2022-09-23 13:30:55 also, suspiciously 2022-09-23 13:30:59 the package became ~10% smaller 2022-09-23 13:31:02 with no changes 2022-09-23 13:31:12 which makes me think something broke.. 2022-09-23 13:31:18 even though it works for compiling 2022-09-23 13:31:40 can it build itself? 2022-09-23 13:32:03 you know, i didn't check that specifically, but i'll bet a yes 2022-09-23 13:32:08 (just started a build, check in 30 minutes if you want) 2022-09-23 13:32:25 i suppose it builds istself already 2022-09-23 13:32:40 yeah, the build is staged with itself 2022-09-23 13:32:49 no guarantee on output artifacts i guess 2022-09-23 13:55:48 mv: cannot stat '/home/ncopa/aports/main/gcc/pkg/gcc-riscv64/usr/lib/libstdc++.a': No such file or directory 2022-09-23 14:02:48 i think we need something like: https://tpaste.us/NOe6 2022-09-23 14:03:24 aha, yes 2022-09-23 14:03:31 i forgot to fully test that 2022-09-23 14:03:35 except we probably dont need to bump pkgrel there 2022-09-23 14:03:45 right 2022-09-23 14:03:50 im testing it now 2022-09-23 14:04:02 apologies, i added that change 2022-09-23 14:04:03 native build, and i'll do checkapk 2022-09-23 14:04:07 np. happens 2022-09-23 14:04:12 i do test bootstraps here and there, just not in the past week 2022-09-23 14:04:16 hasn't been very long since that 2022-09-23 14:05:14 that change shouldn't change anything at all for normal builds, but it does allow using clang without gcc installed at all (it needs libstdc++-dev which was in g++ before) 2022-09-23 14:25:02 gcc-12.1.1_git20220630-r5: 2022-09-23 14:25:02 libstdc++-dev-12.1.1_git20220630-r5: 2022-09-23 14:25:02 breaks: g++-riscv64-12.1.1_git20220630-r6[libstdc++-dev=12.1.1_git20220630-r6] 2022-09-23 14:25:02 breaks: g++-riscv64-12.1.1_git20220630-r6[gcc=12.1.1_git20220630-r6] 2022-09-23 14:25:02 satisfies: clang-14.0.6-r4[libstdc++-dev] g++-12.1.1_git20220630-r5[libstdc++-dev=12.1.1_git20220630-r5] 2022-09-23 14:25:04 .makedepends-build-base-riscv64-20220923.141145: 2022-09-23 14:25:04 satisfies: g++-12.1.1_git20220630-r5[gcc=12.1.1_git20220630-r5] rust-1.64.0-r0[gcc] build-base-0.5-r3[gcc] go-1.19.1-r0[gcc] .makedepends-build-base-riscv64-20220923.141145[gcc] 2022-09-23 14:25:06 satisfies: world[.makedepends-build-base-riscv64=20220923.141145] 2022-09-23 14:25:06 masked in: cache 2022-09-23 14:26:15 that's mismatched pkgrel 2022-09-23 14:26:35 right. 2022-09-23 14:26:41 happens for bootstrap roots all the time 2022-09-23 14:26:49 have to delete all the gcc artifacts from cross repo and rebuild it 2022-09-23 14:27:01 not sure what a better solution would be 2022-09-23 14:28:56 gcc-12.1.1_git20220630-r5: 2022-09-23 14:28:56 breaks: g++-riscv64-12.1.1_git20220630-r6[gcc=12.1.1_git20220630-r6] 2022-09-23 14:29:25 i wonder why g++-riscv64 depends on gcc and not gcc-riscv64 2022-09-23 14:30:16 probably just an overlooked issue given it doesn't break anything aside from these very rare cases 2022-09-23 14:31:05 yeah 2022-09-23 14:56:34 now i get 2022-09-23 14:56:39 >>> ERROR: gmp: build failed 2022-09-23 14:56:39 configure: error: C++ compiler not available, see config.log for details 2022-09-23 14:57:02 i think i need to add a libstdc++dev$_target dependency to g++$_target 2022-09-23 15:04:28 oof 2022-09-23 15:04:30 yeah, that too 2022-09-23 16:11:30 fyi, changing some dns records for pkgs.a.o and security.a.o 2022-09-23 16:41:20 this is interesting: 2022-09-23 16:41:23 ERROR: Metadata for package util-linux-dev-2.38.1-r0 is too long. 2022-09-23 16:41:23 ERROR: Index generation failed: No buffer space available 2022-09-23 16:41:23 >>> gcc-riscv64: Updating the main/riscv64 repository index... 2022-09-23 16:44:49 hah 2022-09-23 16:44:52 i've never seen that before! 2022-09-23 16:44:56 same 2022-09-23 16:45:57 i think its the abuild license fix 2022-09-23 16:46:05 that triggers it 2022-09-23 16:46:27 hmm 2022-09-23 16:46:54 sounds like an apk bug to print such a strange error though, for an invalid input of some fashion 2022-09-23 16:47:09 no, its not invalid input 2022-09-23 16:47:27 apk does not think any metadata will be over 512 bytes 2022-09-23 16:47:42 ahh.. and the license fix makes some multiline licenses over 512 2022-09-23 16:47:42 right 2022-09-23 16:47:45 the insanely long licenses makes them longer than that for the first time 2022-09-23 16:48:05 it is reproducible in native build as well 2022-09-23 16:48:06 i've seen the recent apk3 chimera usage and there's plenty of limits like that :/ 2022-09-23 16:48:15 just hardcoded values that are trivially hit 2022-09-23 16:48:36 fabled: do you remember if 512 bytes is some hard limit for apk metadata, or is it ok to just bump the buffer? 2022-09-23 16:48:53 like this: https://tpaste.us/x5Nw 2022-09-23 16:52:08 i think it should be ok 2022-09-23 16:55:36 oh, this error is nasty 2022-09-23 16:56:12 if it hits the builders it will require manual intervention to fix 2022-09-23 16:56:53 omni: pkgs.a.o should work again on ipv6 2022-09-23 16:56:58 at least for me it does 2022-09-23 17:12:01 we need report the error to apk-tools upstream, but its weekend for me now. need to go. 2022-09-23 17:12:11 i pushed a quickfix for aports 2022-09-23 17:12:11 o/ 2022-09-23 17:17:12 ncopa: we could upgrade to .10 again 2022-09-23 17:17:19 i fixed all the unversioned/unprioritised packages 2022-09-23 20:11:34 sam_: I do know that they have a Qt6 overlay, but it's not the default (although it seems they are going to bundle it, probably) 2022-09-23 20:12:17 I'm happy with myself as I'm a noob, and being "the first" to do something (even though most of the work was done by other folks :) is nice 2022-09-23 20:12:37 I did spend way too many hours trying things, so getting it to work is nice 2022-09-23 20:12:48 And all I got is more emoji for reactions /// 2022-09-23 20:13:14 Nulo: good work :) 2022-09-23 20:14:58 I think I need to check why Arch did a different sed than me https://github.com/archlinux/svntogit-community/commit/8a106f2090c946e2fd0eddd99d3d14d50b04a0ef but it seemingly works 2022-09-23 20:17:06 Does using glibc significantly increase (~20%) binary size? 2022-09-23 20:17:39 I'd have to read the source :wink: 2022-09-23 20:18:23 I'm doing a very unscientific comparison between the Arch and the Alpine packaage 2022-09-23 20:18:36 I believe the main difference might be that Alpine's bundles less 2022-09-23 20:18:50 sorry, I meant to write that in #alpine-linux 2022-09-23 20:19:11 :P 2022-09-24 01:16:27 Nulo: no, it's in the main repo 2022-09-24 01:16:42 (but it is masked for now) 2022-09-24 01:16:55 psykose: what's the deal with new glib? lots of stuff broke? 2022-09-24 04:38:26 ncopa: I am currently running gstreamer via Qt6 on a RaspberryPi3. That combination causes a 'green screen'. I got a Pi4+Qt5/Qml+Video to work in the past, but I don't know if the problematic variable is the hardware on the newer Qt6 version. Playback using gst-play-1.0 works. GST_DEBUG seems to indicate a missing imxvideoconvert_g2d. 2022-09-24 05:26:35 ncopa: according to the QtMultimedia sourcecode, the probe is for some special imx platforms. And causes an 'identity' element iff not present. At this moment playout via Qml causes a green screen. 2022-09-24 10:38:07 nmeum: It worked! duplicated definitions were caused by a tricky workaround (dummy lib) for cgo, removing it from cmake it builds fine 2022-09-24 10:38:43 yeah! it starts without LD_PRELOADING :) 2022-09-24 10:43:15 ouch: (12:42:54) plugins: /usr/lib/purple-2/libwhatsmeow.so is not loadable: Error relocating /usr/lib/purple-2/libwhatsmeow.so: runtime.mapaccess2__fast64: symbol not found 2022-09-24 13:12:29 donoban: hm, that symbol should be provided by libgo / go's runtime 2022-09-24 14:41:13 hmmm... manual installing it doesn't help 2022-09-24 14:44:28 maybe I need to explicit add it to the linker 2022-09-24 15:27:00 sam_: i thought that initially but it was just the js compat regex flag 2022-09-25 00:11:28 psykose: got it, thanks! 2022-09-25 00:12:21 maybe there's more stuff, we'll see 2022-09-25 00:12:23 reupgraded it again 2022-09-25 06:59:04 foof ~.~ 2022-09-25 10:17:42 psykose: fyi, void downgraded glib as well . https://github.com/void-linux/void-packages/pull/39394 2022-09-25 10:20:07 and we reupgraded it :) 2022-09-25 10:20:14 yeah i saw 2022-09-25 10:20:48 just so you know. bug reports will fly in directly to you about broken gimp ;) 2022-09-25 10:20:52 and others 2022-09-25 10:21:07 which is all fine, if everyone waits it out, noone will find the bugz 2022-09-25 15:03:43 for us, only gimp was broken (add test and change its color), there was a single user reporting that chromium, firefox and all gtk applications broke, but that person has their own repository with very weird broken software, so I wouldn't trust that 2022-09-25 15:03:55 s/test/text/ 2022-09-25 15:07:09 and hey, fixed already https://gitlab.gnome.org/GNOME/glib/-/merge_requests/2921 2022-09-25 15:14:27 oh nice 2022-09-25 15:58:05 !39438 is lgtm 2022-09-26 05:25:51 something up with the s390x ci builder? 2022-09-26 05:27:25 omni: it's building dotnet 2022-09-26 05:27:32 https://gitlab.alpinelinux.org/ayakael/aports/-/jobs/852702 2022-09-26 05:28:07 ah 2022-09-26 05:28:26 I suppose it's constrained on disk i/o 2022-09-26 09:33:35 huh.. the nsswitch.conf addition actually breaks ssh 2022-09-26 09:33:52 if i remove it ssh something.alpin.pw works 2022-09-26 09:34:02 if it's there (hosts: files dns) then it can't resolve 2022-09-26 09:35:38 nvm, it just magically fixed itself the moment i touched that file lmao 2022-09-26 09:35:39 classic 2022-09-26 09:44:26 was actually a musl dns issue where there are two server in resolv.conf but only one can resolve the address 2022-09-26 09:48:33 ikke: ah, i see now the issues that were reported before with usa9 not having very good numa scheduling.. https://img.ayaya.dev/8bwvCv1PMvmW 2022-09-26 11:26:46 probably I have a problem with cmake on linking but I found this "This library should not be stripped when it is installed. Go code relies on being able to look up file/line information, which comes from the debugging info using the libbacktrace library." 2022-09-26 11:27:16 and I tried to determine if it's being stripped for alpine reading the APKBUILD 2022-09-26 11:28:14 everything is stripped unless options=!strip 2022-09-26 11:28:21 which library 2022-09-26 11:28:26 libgo 2022-09-26 15:05:01 uhMM, finally I linked libgo but now it segfaults 2022-09-26 15:05:28 http://ix.io/4buF 2022-09-26 15:06:26 looks same problem with it trying to access ARGV 2022-09-26 15:31:25 gcc-go should link to libgo by default 2022-09-26 15:32:16 the rest of the discussion is in #musl 2022-09-26 15:36:46 btw: gccgo is also developed by google and basically re-uses most stuff from google's go implementation for libgo 2022-09-26 15:37:13 so if you are trying to workaround a bug in go by using gcc-go you will probably hit the same issue in gcc-go 2022-09-26 15:38:27 yes, it seems to be the exact same bug 2022-09-26 15:39:46 i guess that aforementioned patch for fixing it in regular go (forgot where it is) and a fix for a completely unrelated issue (https://github.com/golang/go/issues/54805) would be the real solution 2022-09-26 16:06:45 nmeum: maybe the problem is probably this CMakelist mess, it has different c and go directories. Paradoxically I had to add '-lgo' on the C CMakelist 2022-09-26 16:08:16 here the author explains the messy config https://github.com/hoehermann/purple-gowhatsapp/issues/119#issuecomment-1255504445 2022-09-26 16:09:57 if other apps are linking proprerly I won't worry due this 2022-09-26 19:29:59 is it just me, or is gitlab slower today? 2022-09-26 19:31:02 Haven't noticed anything 2022-09-26 19:32:08 poland internet moment 2022-09-26 19:32:31 works on my machine 2022-09-26 19:32:42 south poland internet moment 2022-09-26 19:32:51 >:( 2022-09-26 19:33:07 try pulling over https instead 2022-09-26 19:33:19 i'm pulling over https already 2022-09-26 19:33:22 hah 2022-09-26 19:33:39 could also be ipv6 and random broken ipv6 routing 2022-09-26 19:33:47 otherwise, just broken internet 2022-09-26 19:33:49 : ) 2022-09-26 19:33:52 ah 2022-09-26 19:33:54 no, nevermind 2022-09-26 19:33:55 it was over ssh 2022-09-26 19:34:02 and it was 200Kbps at peak 2022-09-26 19:34:10 i changed to https, 30Mbps, pulled instantly 2022-09-26 19:34:47 i wonder, what openssh configuration is there on the gitlab instance 2022-09-26 19:36:18 https://gitlab.alpinelinux.org/alpine/infra/docker/gitlab/-/blob/15.2-stable/gitlab-shell/build/setup-image 2022-09-26 19:36:26 Pretty default config 2022-09-26 19:37:14 ssh is going through traefik though 2022-09-26 19:37:25 not sure if that has impact 2022-09-26 19:37:39 ew 2022-09-26 19:39:11 Only for IPv6 iirc 2022-09-26 19:39:55 ptrc: do you see a difference between IPv4 and IPv6 with ssh? 2022-09-26 19:40:21 i don't have ipv6 sadly 2022-09-26 19:40:49 ok, so it must be IPv6, so no traefik should be involved then 2022-09-26 19:40:56 ipv4* 2022-09-26 19:43:19 i don't think a bare default opensshd config would affect 200k->30Mb compared to https 2022-09-26 19:43:30 but maybe the way it's passed to gitlab and the way gitlab uses it then.. 2022-09-26 19:44:05 Receiving objects: 100% (1301/1301), 1.37 MiB | 10.15 MiB/s, done. 2022-09-26 19:44:14 ssh 2022-09-26 20:11:21 fun, someone wants python2 back 2022-09-26 20:17:21 A "poor deciession 2022-09-26 20:17:26 " to drop python2 2022-09-26 20:17:29 right 2022-09-26 20:22:58 did i word that too harshly 2022-09-26 20:24:29 >I was told that it was, I had no clue 2022-09-26 20:24:34 ok i am logging off this issue and not replying lmao 2022-09-26 20:24:36 wtf 2022-09-26 20:24:43 yeah, better to 2022-09-26 20:24:54 you didn't!! even!! check!! if it!! was!! needed!! 2022-09-26 20:24:55 what!!! 2022-09-26 20:25:17 (to be fair, they probably ran into 'python' not existing because of lack of symlink, which is in edge only) 2022-09-26 20:25:27 since actual python2 is only dropped in edge.. 2022-09-26 20:25:38 but next release has that fixed i guess for those expecting python to be python3 2022-09-26 20:25:44 https://pkgs.alpinelinux.org/packages?name=python2&branch=v3.16&repo=&arch=&maintainer= 2022-09-26 20:26:23 It was in testing, but that ofourse is not included in v3.16 2022-09-26 20:26:26 that too. but also the literal command- when things fail in a build, sometimes even when they're 3-compatible they just invoke 'python' 2022-09-26 20:26:35 npm stuff is common for that 2022-09-26 20:26:38 sure 2022-09-26 20:27:10 and to be fair, i've fixed like two packages that had python2 in makedeps just because of that, when 3 sufficed :) at the time of the removals 2022-09-26 20:27:16 it's easy to get confused there 2022-09-26 20:27:16 I just mentioned in the GH issue that nodejs builds fine with python3, and just ignore the other remark 2022-09-26 20:28:02 no need to be upset? 2022-09-26 20:28:46 The only thing I'm mildly upset about is calling it a poor decission to drop python2 2022-09-26 20:28:56 tell them to use pypy2 2022-09-26 20:29:08 yes, that's what I was thinking as well, if it's needed 2022-09-26 20:29:08 that funnily doesn't even work for a lot of stuff 2022-09-26 20:29:15 but that's only in edge/testing 2022-09-26 20:29:23 psykose: like what? 2022-09-26 20:29:32 my experience with pypy was that it was brutally slow 2022-09-26 20:29:35 ironically 2022-09-26 20:29:38 don't start with that xD 2022-09-26 20:29:45 idk, but i feel like i've suggested it a few times and every time it didn;t work for the usecase 2022-09-26 20:29:49 like the dart bootstrap 2022-09-26 20:29:49 huh, pypy was supposed to be faster 2022-09-26 20:29:49 I was told it was supposed to be fast 2022-09-26 20:29:52 but then it wasn't 2022-09-26 20:29:58 it's much slower for python scripts 2022-09-26 20:30:04 a jit is only good for actual programs 2022-09-26 20:30:14 right, due to slow startup time I suppose 2022-09-26 20:30:22 Node.js does way better 2022-09-26 20:30:44 surely pypy doesn't have to be this slow 2022-09-26 20:30:46 but, off-topic 2022-09-26 20:31:55 sure, there's a thousand other factors 2022-09-26 20:32:37 the constraints of the language, the actual care for optimising each case (or preferring one), the potential of our pypy2 actually not being built correctly (it happens), etc 2022-09-26 21:17:24 the nvm description of Alpine is 'interesting': "Busybox (and thus Alpine Linux) uses a different C/C++ stack to most mainstream/traditional Linux distributions - musl." 2022-09-27 17:32:14 I'm trying to build gcc with jemalloc, should I expect just: 'LD_PRELOAD... abuild' to work? 2022-09-27 17:32:38 or should I edit the APKBUILD or abuild.conf? 2022-09-27 18:56:51 it might work 2022-09-27 19:10:26 I don't notice too much difference, but I should 'time' it hehe 2022-09-28 05:51:24 psykose: abump sources the APKBUILD, causing exports in the APKBUILD ending up in its env, but because srcdir is not set there, the exports with srcdir become invalid 2022-09-28 05:51:59 sounds like an abump issue 2022-09-28 05:52:39 needs to also do what abuild does from 2790: 2022-09-28 05:52:58 or abuild can put it in something like functions.sh and that can be sourced 2022-09-28 05:54:29 psykose: congratulations on your @ :D 2022-09-28 05:55:12 thanks homie 2022-09-28 05:55:32 i have used the power zero times because the only thing to use it on is bot spammers 2022-09-28 05:55:44 which is a waste of time as they get banned by oftc after 1 message regardless of what you do 2022-09-28 05:56:35 it's not a waste of time, bringing down the hammer on a bot is a *pleasure* 2022-09-28 05:57:13 is it? it's quite literally just automated weird bots 2022-09-28 05:57:17 not a person or anything 2022-09-28 05:57:49 which is why it's a pleasure. All the catharsis of violence, without the guilt. :P 2022-09-28 06:07:27 i guess that's what videogames are all about 2022-09-28 06:07:29 : ) 2022-09-28 06:07:58 not all vidya, but Doom, definitely :D 2022-09-28 07:10:16 morning! i'm building rust on riscv64 and I have a deadlock. any ideas how to debug it? 2022-09-28 07:10:40 nope, sounds like the usual deadlock fail 2022-09-28 07:10:52 let me guess, it's in 'building rustbuild' before stage0 2022-09-28 07:10:56 they all happen on lzma-sys 2022-09-28 07:11:12 there is another 2 spots for lzma-sys in the other stages where it can also lock 2022-09-28 07:12:02 actually if you're in cross the first step should still be 'native' and not riscv64, unless you already finished the cross part and this is the build-self-on-riscv64 part 2022-09-28 07:12:11 Hi skarnet ! 2022-09-28 07:12:51 but yeah sadly i have no idea what causes them 2022-09-28 07:13:03 you can attach a debugger and backtrace 2022-09-28 07:13:21 i doubt you'll find anything that meaningful.. probably spinning a spinlock forever 2022-09-28 07:19:41 Fidget spinlock 2022-09-28 07:22:39 i have succesfully built rust using crosscompiling. im trying to build it natively now in qemu user 2022-09-28 07:22:52 Compiling rustc_ast_pretty v0.0.0 (/home/ncopa/aports/community/rust/src/rustc-1.64.0-src/compiler/rustc_ast_pretty) 2022-09-28 07:22:52 Building [===================> ] 213/258: ryu(build) 2022-09-28 07:23:21 I cannot connect debugger 2022-09-28 07:23:47 /usr/bin/strace: test_ptrace_get_syscall_info: PTRACE_TRACEME: Function not implemented 2022-09-28 07:23:47 /usr/bin/strace: attach: ptrace(PTRACE_ATTACH, 33612): Function not implemented 2022-09-28 07:24:04 Attaching to process 33612 2022-09-28 07:24:04 ptrace: Function not implemented. 2022-09-28 07:28:50 ptrace is not implemented 2022-09-28 07:28:58 you can use gdb to connect to qemu 2022-09-28 07:29:02 but 2022-09-28 07:29:14 not sure how you do that with binfmt 2022-09-28 07:29:36 you can also export QEMU_STRACE 2022-09-28 07:36:46 right, yeah 2022-09-28 07:36:53 usual qemu rust hang 2022-09-28 07:40:19 we should probably ask upstream qemu-riscv64 for help 2022-09-28 07:40:36 i suppose i cannot do anything with QEMU_STRACE after the deadlock happened, right? 2022-09-28 07:41:28 most likely not 2022-09-28 07:42:21 correct 2022-09-28 07:42:47 i tried it with golang, but the output was not really helpful 2022-09-28 07:43:18 it's strange because i'm not sure who's issue it is 2022-09-28 07:43:29 it's only really present for rust building itself and golang... 2022-09-28 07:43:37 yet at the same time is almost definitely an issue in qemu-user 2022-09-28 07:43:45 but i would think far more things would have it? 2022-09-28 07:43:54 i guess there's a few more, like ocaml that can't build themselves 2022-09-28 07:43:54 do I set QEMU_STRACE in the lxc container? 2022-09-28 07:44:05 and ocaml just uses a c compiler 2022-09-28 07:44:27 QEMU_STRACE=1 ls 2022-09-28 07:44:30 i suppose there is not many users who runs qemu user with -j48 2022-09-28 07:44:30 :) 2022-09-28 07:44:44 ah, interesting 2022-09-28 07:45:12 i suppose building rust will generate a bit of output though 2022-09-28 07:45:21 like QEMU_DEBU=9000 would listen on port 9000 2022-09-28 07:45:26 +G 2022-09-28 07:45:38 and then I can attach to that? 2022-09-28 07:46:06 you should but im unsure with binfmt how it would work 2022-09-28 07:46:25 i mean you normally would exec qemu-riscv64 yourself 2022-09-28 07:46:36 you probably need gdb cross for riscv64 to attach with 2022-09-28 07:46:55 normally you could attach with the hosts own qemu 2022-09-28 07:47:00 err gdb 2022-09-28 07:47:20 im not sure where the port will be opened 2022-09-28 07:47:21 can i do that from within lxc container? 2022-09-28 07:47:33 could be its from the host not container 2022-09-28 07:47:45 qemu-user is not present in the container 2022-09-28 07:48:26 ncopa: it will also need to be started afterwards 2022-09-28 07:48:36 the process is paused 2022-09-28 07:49:47 netstat does not show anything listening to port 9000 2022-09-28 07:52:52 i never tried it so i dont know exactly 2022-09-28 07:53:01 i think i saw the env option in the src 2022-09-28 07:53:20 docs only talk about QEMU_STRACE 2022-09-28 07:56:06 ncopa: i think its QEMU_GDB 2022-09-28 07:56:21 ok 2022-09-28 07:57:12 but the port will be open in the container as it shows now 2022-09-28 07:57:51 what happens if there are more than one process? won't port 9000 be occupied? 2022-09-28 08:00:23 exactly, this is why i said binfmt would be problematic 2022-09-28 08:00:44 im not sure how it works 2022-09-28 08:02:07 i was able to connect to htop using QEMU_GDB 2022-09-28 08:02:16 but not sure it will be useful 2022-09-28 08:04:46 maybe we should try llvm15 2022-09-28 08:05:52 Changes to the RISC-V Backend (in llvm15): Support for the unratified Zbe, Zbf, Zbm, Zbp, Zbr, and Zbt extensions have been removed. 2022-09-28 08:06:42 don't think it will make any difference for building rust in qemu 2022-09-28 08:10:39 lack of extension implementation doesn't cause infinite loops in qemu-user 2022-09-28 08:13:39 well, if the currently use those extensions, and they are not working well, they may cause deadlock? 2022-09-28 08:14:21 i dont know 2022-09-28 08:14:35 there are a couple of other things we can try 2022-09-28 08:14:36 maybe it would solve rust, but what about golang? 2022-09-28 08:14:45 i mean its kind of useless now 2022-09-28 08:14:59 most of the golang packages will hang/deadlock 2022-09-28 08:15:20 have we tried to upgrade qemu-riscv64? 2022-09-28 08:15:25 qemu-riscv64 --version 2022-09-28 08:15:25 qemu-riscv64 version 6.1.0 2022-09-28 08:15:42 from what i understand we are one of the only ones who are using qemu user mode with binfmt 2022-09-28 08:15:49 yes i am running 7 2022-09-28 08:16:01 3.16 is 7.0.0 2022-09-28 08:16:15 seems like nld5-dev1 needs a few updates.... 2022-09-28 08:16:27 no it needs to go away :) 2022-09-28 08:16:44 there is a new kid in town 2022-09-28 08:16:49 ok 2022-09-28 08:16:59 rustc is a compiler, if it "used those extensions" the cpu it was running on (or qemu-user) would have to support them 2022-09-28 08:17:04 qemu-user most definitely does not support any of that 2022-09-28 08:17:10 so it would just crash instantly 2022-09-28 08:17:25 ncopa: we need to move infra 2022-09-28 08:17:36 so those user containers need to move to the new box 2022-09-28 08:17:40 ok 2022-09-28 08:17:43 same box as current rv64 builder 2022-09-28 08:17:46 can i move mine now? 2022-09-28 08:17:53 regardless it will be llvm15 soon in any case 2022-09-28 08:17:58 not like there's a reason for it to not be 2022-09-28 08:18:09 just waiting on one thing to support it 2022-09-28 08:18:15 psykose: i also wonder if it woudl be an idea to build llvm with -O2 2022-09-28 08:18:18 ncopa: 147.75.87.25 2022-09-28 08:18:27 you can rsync your container to it if you like 2022-09-28 08:18:30 else we will do it later 2022-09-28 08:18:31 thanks! 2022-09-28 08:18:46 what for 2022-09-28 08:19:09 (it's not built with -Os) 2022-09-28 08:19:11 but please rm the origin container if you do it 2022-09-28 08:19:17 its not? 2022-09-28 08:19:22 ok then 2022-09-28 08:19:23 else i could rysnc it again and... 2022-09-28 08:19:33 clandmeter: ok 2022-09-28 08:21:41 i don't exactly remember but i think when we tried building rust on real hardware it didn't hang 2022-09-28 08:21:58 it did still segfault, same as the other failure mode, of course (at a random step) 2022-09-28 08:22:20 looks like there is qemu 7.1.0 in edge, and it has a lot of rv additions 2022-09-28 08:22:23 and fixes 2022-09-28 08:22:27 i don't think there's anything we can do about the qemu hangs aside from trying 7.1 2022-09-28 08:22:27 yeah 2022-09-28 08:22:45 worth a shot 2022-09-28 08:23:00 but how to get it? 2022-09-28 08:23:17 not sure im happy to switch to edge 2022-09-28 08:23:31 mixing repos... 2022-09-28 08:23:33 upgrade host that hosts the qemu-user containers to edge, or build it manually (clone aports, build community/qemu, upgrade on 3.16) 2022-09-28 08:23:44 latter should work 2022-09-28 08:23:59 i think there's no other requirements 2022-09-28 08:24:04 i guess we could build it on 3.16 2022-09-28 08:24:18 ikke: what do you think? 2022-09-28 08:24:48 https://wiki.qemu.org/ChangeLog/7.1#RISC-V 2022-09-28 08:25:24 ncopa: when you were trying this, was it in a qemu-user session on 7.0? 2022-09-28 08:25:39 you could also try locally on edge with 7.1 first if you don't want to wait for above ^ 2022-09-28 08:25:49 qemu-riscv64 version 6.1.0 2022-09-28 08:25:52 ah 2022-09-28 08:25:55 ok that is ancient 2022-09-28 08:25:59 yup 2022-09-28 08:26:11 im moving the lxc container to the new host now 2022-09-28 08:26:15 relatively speaking to riscv improvements at least 2022-09-28 08:26:16 yeah 2022-09-28 08:26:18 with newer qemu 2022-09-28 08:27:10 ncopa: this box will give you the same issues 2022-09-28 08:27:10 I also saw that during rust build libunwind was built with gcc and I saw -Os -fomit-frame-pointer 2022-09-28 08:27:22 at least it does with golang 2022-09-28 08:27:41 i will try build rust with -O2 2022-09-28 08:28:08 clandmeter: new box will give me same qemu-riscv64 as the current build-edge-riscv64, right? 2022-09-28 08:28:12 which is a good thing 2022-09-28 08:28:18 ncopa: maybe you can give qemu 7.1.0 a try 2022-09-28 08:28:23 will do that too 2022-09-28 08:29:06 first I need to figure out how to copy the container 2022-09-28 08:29:21 rsync? 2022-09-28 08:30:27 btw, this host does not have dmvpn installed yet 2022-09-28 08:30:41 ok 2022-09-28 08:37:48 qemu 7.1 builds fine on 3.16 world makedeps 2022-09-28 08:37:55 if you want to go for that too 2022-09-28 08:37:59 should just be an abuild -r :) 2022-09-28 08:53:37 right, well we have to think about when all things fail how we want to proceed. the current status is kind of... crap. 2022-09-28 08:53:47 other distros just use qemu-system 2022-09-28 08:54:20 at least thats what i think what fedora does 2022-09-28 08:55:23 there was some benchmarks made by q66 yesterday 2022-09-28 08:55:38 https://gist.github.com/q66/ec52e90c53e54eb293ebd6de8d58c435 2022-09-28 08:55:47 on chimera and not alpine, but for the exact same thing 2022-09-28 08:56:05 interestingly, going from -j16 to -j32 barely changes the time 2022-09-28 08:56:09 even though the user time almost doubles 2022-09-28 08:56:16 meaning it's very innefficient for not much gain 2022-09-28 08:56:39 so, technically.. the best setup is being able to do parallel builds (more than one builder) on something lik j16, but that's impossible and unfeasible for a lot of reasons for us 2022-09-28 08:56:56 and rvvm turns out to still be slower anyway 2022-09-28 08:57:23 if that qemu-system doesn't have any improvements it's over 5x slower 2022-09-28 08:57:39 if it was just '2x' it would be fine, as long as there was no qemu-system bugs instead 2022-09-28 08:57:42 but 5x is too much 2022-09-28 08:58:02 so we kinda have the best setup already, toolchain/qemu bugs notwithstanding 2022-09-28 08:58:23 so i hope 7.1 improves it at least a little bit 2022-09-28 09:09:59 psykose: nice list 2022-09-28 09:10:08 and its exactly what i experienced before 2022-09-28 09:10:17 qemu user vs system 2022-09-28 09:10:32 yep, i thought it would be less, but didn't figure out how to do the same benchmark a bit ago 2022-09-28 09:10:36 nice to see i was wrong 2022-09-28 09:10:40 the performance diff is so much, its hard to swtich back 2022-09-28 09:10:42 and it indeed is slow as fuck 2022-09-28 09:10:44 yep :/ 2022-09-28 09:11:17 would be nice to talk to the qemu riscv64 ppl and see if they have some ideas 2022-09-28 09:14:10 yeah, it's a good idea 2022-09-28 09:14:13 have to find them though 2022-09-28 09:26:12 i think there are 2 main contacts 2022-09-28 09:26:25 one from wd, and one from i dont remember :) 2022-09-28 09:27:13 do you have the time to build 7.1 now on the host? 2022-09-28 09:27:18 i can also do it with access if you want me to 2022-09-28 09:28:08 i dont have the time 2022-09-28 09:28:18 im ok if you want to do it 2022-09-28 09:28:26 not sure what ncopa is up to 2022-09-28 09:30:33 im setting up dmvpn on the new machine and will move my container to there 2022-09-28 09:30:47 and then try upgrade qemu to 7.1 2022-09-28 09:30:51 sure, np 2022-09-28 09:31:10 ping me if you want help- i checked just 3.16 /etc/apk/repositories only and abuild -r worked on master aports qemu 2022-09-28 09:31:31 then adding the usual local /home repo and upgrading should be completely ok 2022-09-28 09:31:52 i'd recommend remembering it- when we upgrade edge qemu we should match it on that host 2022-09-28 09:31:58 and also we should do the same for the builder i think 2022-09-28 09:32:08 makes more sense to run latest qemu as long as it's not completely broken, given all the improvements 2022-09-28 09:52:01 <@psykose> on chimera and not alpine, but for the exact same thing 2022-09-28 09:52:03 on ubuntu 2022-09-28 09:52:12 indeed 2022-09-28 09:52:13 smol typo 2022-09-28 09:52:29 tiny 2022-09-28 09:52:38 nano sized 2022-09-28 11:42:07 ncopa: ah, what error do you get for brotli cross? 2022-09-28 11:42:26 (before the change) 2022-09-28 12:01:02 i dont remember 2022-09-28 12:02:18 sure, i'll check 2022-09-28 12:16:17 Building [=======================> ] 81/89: lzma-sys 2022-09-28 12:16:22 is that where it usually hangs? 2022-09-28 12:28:34 yep 2022-09-28 12:52:25 ok it happens with qemu 7.1 too 2022-09-28 12:56:08 fun 2022-09-28 12:56:13 it passes that spot sometimes 2022-09-28 12:56:18 you should keep retrying it i guess 2022-09-28 12:56:27 as long as it doesn't segfault in some other spot, it's.. ok 2022-09-28 15:09:32 I would need to use jemalloc on oldest alpine version to have oldest ruby versions. I builded it on 3.13 and I'm running it without issues. It looks possible to backport jamelloc on oldest aports tags ? Or maybe it would be better to maintains multiple ruby version (as with php) ? 2022-09-28 16:00:30 <@psykose> meaning it's very innefficient for not much gain 2022-09-28 16:00:30 <@psykose> even though the user time almost doubles 2022-09-28 16:00:37 btw, the reason for that is that the 5950x is a 16-core processor 2022-09-28 16:00:49 hah, yeah, you're right 2022-09-28 16:00:58 it has 32 threads, but HT is a relatively small gain 2022-09-28 16:01:11 yep 2022-09-28 16:01:14 i assume on a natively 32-core CPU the performance difference would be far greater 2022-09-28 16:01:25 yeah, good point 2022-09-28 16:01:36 the -j32/smt=32 tests are there mostly just to see what it does 2022-09-28 16:01:45 also because i could (qemu 7.0 enabled smt=32 for riscv) 2022-09-28 16:01:48 er, smp=32 2022-09-28 16:01:56 overall i bet that's better performing than our riscv builder 2022-09-28 16:02:05 idk 2022-09-28 16:02:07 what's your riscv builder 2022-09-28 16:02:08 it's -j48 from 24 cores but they're much slower ones compared to a 5950x 2022-09-28 16:02:28 i did these tests mostly to figure out the options 2022-09-28 16:03:02 so i guess next i'll add cbuild support for bootstrapping foreign containers and try building the distro with qemu-user 2022-09-28 16:03:11 yeah, i guess it makes it pretty clear 2022-09-28 16:03:22 good luck on the builds 2022-09-28 16:03:23 but first i'll have to cross-compile the packages needed to assemble the build container 2022-09-28 16:05:32 but yeah i guess the 3 main outcomes really are 1) qemu-user is fast for builds 2) qemu-system scales linearly until 8 cores and then does the opposite 3) qemu-system with smp=8 is actually faster than best native hardware 2022-09-28 16:05:46 though the individual cores in qemu-system are slower than native hardware 2022-09-28 16:05:56 but you have twice as many of them 2022-09-28 16:07:55 the qemu-user figures for -j16 are roughly similar to my aarch64 builder (16-core too) btw 2022-09-28 16:08:05 i managed to build the whole repo on that, so 2022-09-28 16:48:14 💨Don’t miss this chance to be Rich... (full message at ) 2022-09-28 17:16:39 Anybody interested in https://gitlab.alpinelinux.org/alpine/tsc/-/issues/46 could maybe have a look at https://gitlab.alpinelinux.org/alpine/infra/aports-qa-bot/-/merge_requests/38? Not sure if the approach is the correct one, but at least it's a start 2022-09-28 17:16:50 Also avoids having to setup email redirects 2022-09-28 19:36:13 pabloyoyoista_: I think that's the most straight-forward way to handle it 2022-09-28 20:49:57 psykose: do you know where qbittorrent-nox keeps the BT_backup folder? 2022-09-28 20:50:14 im trying to migrate from the qt client and reimporting my torrent looses the trackers 2022-09-28 20:51:33 nnnowhere it seems 2022-09-28 20:51:55 ah 2022-09-28 20:51:56 no 2022-09-28 20:51:58 tools bad 2022-09-28 20:52:07 /var/lib/qbittorrent/.local/share/qBittorrent 2022-09-28 20:52:11 yes, you read it correctly 2022-09-28 20:52:26 the dir is $cwd/.local/share/qBittorrent 2022-09-28 20:52:32 settings is $cwd/.config .. 2022-09-28 20:52:37 and cwd is invocation point 2022-09-28 20:52:48 it's really stupid but i guess they started with that and it's impossible to change it 2022-09-28 20:52:59 or at least no one really cares 2022-09-28 20:53:04 it's not very disk configurable 2022-09-28 20:53:13 though there is at least one option you have to do on disk, funnily 2022-09-28 20:53:55 (that cwd comes from directory= in init.d which cd's into it) 2022-09-28 20:58:02 cursed 2022-09-28 20:58:03 ty 2022-09-28 21:01:40 yay migration works 2022-09-28 21:03:43 : ) 2022-09-28 21:03:56 there are a fuckton of people going around complaining about libtorrent 2.x memory use 2022-09-28 21:04:03 to the point i got personally emailed to change it 2022-09-28 21:05:40 yeah it sucks 2022-09-28 21:05:56 im regularly at +1.5TB virt mem 2022-09-28 21:06:07 yeah, and it makes no difference 2022-09-28 21:06:27 i run hundreds of torrents on a shitty bulldozer on top of 10 other services on a 7200rpm 1tb drive 2022-09-28 21:06:49 i think qute is bothering me more with memleak 2022-09-28 21:06:58 idk what people are complaining about when the memory isn't really used 2022-09-28 21:07:00 hm, what leak specifically 2022-09-28 21:07:17 idk, it just clogs mem 2022-09-28 21:07:24 it was worst before, not its just meh 2022-09-28 21:07:32 but i dont think its leaking anymore 2022-09-28 21:07:44 in what sense 2022-09-28 21:07:50 post numbers 2022-09-28 21:08:46 unless you have something like vm.overcommit_memory=2 the virt doesn't ""really"" matter, and the RSS maps are all a lie because of mmap 2022-09-28 21:09:09 e.g. if i do a full recheck htop will say qbittorrent is using 93% of the memory 2022-09-28 21:09:17 but the memory bar will be at 400mb/16gb 2022-09-28 21:09:21 and that is indeed the real use 2022-09-28 21:09:35 i don't know how it actually calculates it, but it is the real one 2022-09-28 21:10:14 i actually did downgrade to 1.2 and checked again 2022-09-28 21:10:30 it actually uses more memory on 1.2 because instead of mmaping into disk cache it reserves real memory for it 2022-09-28 21:10:43 it just looks less absurd in the rss 2022-09-28 21:10:54 but the real usage is maybe 50% worse, and the caching will be worse too 2022-09-28 21:11:40 no number, its all classified 2022-09-28 21:11:44 :) 2022-09-28 21:11:49 im just using my french heritage to complain 2022-09-28 21:12:00 fix my software psykose 2022-09-28 21:12:01 i mean, i've used everything i've looked for 2022-09-28 21:12:17 in lots of cases (not just qbittorrent, but my girlfriends laptop), i just have no idea how to debug memory 2022-09-28 21:12:30 if you google anything you get shitty ass stackoverflow links on how to run x coreutil 2022-09-28 21:12:38 none of the numbers add up to anything 2022-09-28 21:13:00 pmap doesn't have any information, etc 2022-09-28 21:13:35 or well, for pmap -XX i must really just not know how to read it, there's like 30 columns 2022-09-28 21:18:36 but yeah idk, if someone that knows how all that stuff interacts with a real analysis/solution (no armchair shit), they can go fix it so it doesn't look so strange or something 2022-09-28 21:18:47 in practice i don't see any issue, but some people report infinite hangs/freezing/whatever 2022-09-28 21:19:02 idk how cause i always look at their specs and they have 10x the computer that i do 2022-09-28 21:19:15 maybe they tweaked the advanced settings too much 2022-09-28 21:19:24 i just open htop and complain on irc 2022-09-28 21:19:29 20 hashing threads on hdd i bet 2022-09-28 21:19:31 that must be it 2022-09-28 21:19:34 i also have a dumb script to clear the mem 2022-09-28 21:19:41 drop caches, clear swap, etc 2022-09-28 21:19:50 it doesn't really do anything unless you've actually noticed something else is negatively affected 2022-09-28 21:19:56 if anything that always harms performance for 0 gain 2022-09-28 21:22:08 qute regularly gets on the swap 2022-09-28 21:22:18 with visible perf degradation 2022-09-28 21:22:23 but i bought more sticks 2022-09-28 21:26:22 visible by what metric 2022-09-28 21:26:44 being on swap is a good thing, unless it's on swap because there is no memory 2022-09-28 21:26:59 i'm currently at 5/16. but also my swap is at 1.2G 2022-09-28 21:27:02 magic in a sense 2022-09-28 21:35:08 was sluggish af 2022-09-28 21:35:29 im at 4/32 and about a G of swap 2022-09-28 21:36:06 i assume you mean the web interface is sluggish? 2022-09-28 21:39:02 on qutebrowser yeah 2022-09-28 21:39:15 qbittorrent no issue 2022-09-28 21:39:31 lmao this whole time i thought you were just referring to qbit as qute 2022-09-28 21:39:44 ah no 2022-09-28 21:39:48 you're just bad 2022-09-28 21:39:49 yeah sorry webkit be bad, apologies :) 2022-09-28 21:39:56 yeah i'm terrible 2022-09-29 16:26:26 any clue with this https://gitlab.alpinelinux.org/donoban/aports/-/jobs/856826 ? other archs built fine 2022-09-29 16:29:32 slice bounds out of range [:10] with length 10 2022-09-29 16:30:17 I've ready seen it but don't sure what means 2022-09-29 16:30:27 comes from a 'go build' ? 2022-09-30 06:36:47 morning! 2022-09-30 06:36:55 what is blocking us to upgrade llvm to 15? 2022-09-30 06:38:44 waiting for a reverse dependency to support it 2022-09-30 06:39:13 it's going to make zero difference to qemu-user so i'm not sure what the rush is 2022-09-30 06:39:25 what results did you get with the other qemu-user static 2022-09-30 06:42:09 didnt make any diff 2022-09-30 06:42:29 im planning to start set up 3.17 builders soon. its october tomorrow 2022-09-30 06:42:37 fun 2022-09-30 06:42:55 would be nice to have llvm15 for alpine 3.17, but if thats not a goal, then no rush 2022-09-30 06:43:28 it will be there for sure 2022-09-30 06:43:41 what would really be nice is dt_relr 2022-09-30 06:48:54 i merged llvm anyway since it's so hotly demanded :) 2022-09-30 06:49:05 just one thing that will temporarily not work (include-what-you-use) 2022-09-30 08:53:37 thanks! 2022-09-30 09:33:37 question regarding bootstrap.sh: if I change community/go to bootstrap using gcc-go: do I need to enable LANG_GO for our bootstrapped/cross-compiled gcc as well? because as far as I can tell there is nothing on the bootstrap path that needs go, I think community/go is only in bootstrap.sh because it needs to be cross-compiled presently, right? 2022-09-30 09:34:11 sounds correct to me 2022-09-30 09:34:30 so yeah, you'd need LANG_GO 2022-09-30 09:34:34 and make sure it actually works 2022-09-30 09:34:52 and can then later, natively on target, build the real go 2022-09-30 09:35:05 or something like that 2022-09-30 09:35:43 do I really need LANG_GO in the cross-compiled gcc though? I don't think go is really needed for anything on the bootstrap path 2022-09-30 09:36:00 so it should be fine to compile community/go later when main/gcc has been rebuild using the cross-compiled gcc 2022-09-30 09:36:07 or am I missing something? 2022-09-30 09:36:42 ah right 2022-09-30 09:36:56 yeah i guess you don't, when the real gcc rebuilds itself natively it will have it 2022-09-30 09:44:15 I guess I will move forward with the gccgo-based community/go bootstrap for 3.17 then :) 2022-09-30 09:44:31 sounds good to me :) 2022-09-30 09:48:50 Much appreciated, saves me or ncopa work later :-) 2022-09-30 11:25:04 llvm15 does not cross build for some reason 2022-09-30 11:25:06 error: Record `IntrNoCallback' does not have a field named `ArgNo'! 2022-09-30 11:25:23 that's what it says when you try to build it 2022-09-30 11:25:28 Arg! No! 2022-09-30 11:54:58 :) 2022-09-30 12:06:29 ncopa: it build for me 2022-09-30 12:06:39 but fails on something else 2022-09-30 12:07:38 i think it builds now. after apk upgrade here 2022-09-30 12:08:46 for some reason it tried linking build zstd from /usr/lib haha 2022-09-30 12:38:23 indeed 2022-09-30 12:39:32 trying a fix for it, if it works i'll push it 2022-09-30 13:18:12 fixed by disabling, good enough for bootstrap 2022-09-30 13:18:29 you'll probably want to manually edit out 14/ghc to save yourself some time for the rust stuff 2022-09-30 13:20:26 thanks 2022-09-30 17:55:14 what is the purpose of abuild-apk? 2022-09-30 18:00:34 suid wrapper 2022-09-30 18:00:52 abuild needs to install deps, installing deps needs root 2022-09-30 18:01:05 it does a check that the user is in abuild group, then allows adding deps, and that's about it 2022-09-30 18:01:14 the code is quite short, you could read it 2022-09-30 18:01:29 (and maybe even find a vulnerability.. hehe..) 2022-09-30 18:11:40 Ah well, this is how abuild calls apk 2022-09-30 18:13:13 hi. where i can check which environment variables abuild sets before calling build(), check() and so on? 2022-09-30 18:15:30 There are multiple places they can come from 2022-09-30 18:15:48 /usr/lib/abuid/functions.sh /etc/abuild.conf 2022-09-30 18:17:38 tyvm! /usr/lib/abuid/functions.sh is missing on my system btw 2022-09-30 18:18:11 sorry, /usr/share/abuild/functions.sh 2022-09-30 18:20:00 that one exists, thanks 2022-09-30 18:32:07 Not common, but they can also be defined in ~/.abuild/abuild.conf