#archlinux32 | Logs for 2022-10-10

Back
[00:30:17] <bill-auger> i must ask if you can estimate whern ruby 3.0.4 will come out of staging
[00:31:31] <bill-auger> i have it packaged now; but there were some major changes which cause conflicts - i probably should not release it until the same day as arch32
[00:33:06] <bill-auger> it seems that you have wrestled with it quite a lot - is it significantly different than the way arch packages it (them)? - maybe i need to merge the arch32 changes, and rebuild?
[03:25:17] <KillerWasp> https://archlinux.org
[03:25:17] <KillerWasp> https://paste.rs
[03:25:18] <phrik> Title: Arch Linux - News: QEMU >= 7.0.0 changes split package setup (at archlinux.org)
[03:25:18] <KillerWasp> I don't know what are the steps to follow to solve it.
[03:29:54] <KillerWasp> i solve with uninstall both qemu packages before to update
[03:51:38] <bill-auger> KillerWasp: the obvious solution is to remove qemu-emulators-full and qemu-system-aarch64
[03:52:27] <KillerWasp> bill-auger: not obvious for me...
[03:52:33] <KillerWasp> :$
[03:52:42] <bill-auger> then maybe open a bug report so the conflict can be fixed, and you will be able to re-install qemu-system-aarch64 again someday
[03:53:18] <bill-auger> np - it is obvious to me - i look at bug reports like this almost every day
[03:55:18] <bill-auger> the cause of it, is that qemu-system-aarch64 depends on a package named 'edk2-armvirt' ; but there is no such package with that name - someone goofed up somewhere
[03:55:50] <bill-auger> so opening a bug report is a great idea
[03:55:52] <KillerWasp> I solved it by deleting those packages just to guess.
[03:56:27] <bill-auger> yea but its not a solution really - it is a work-around - no one can use qemu-system-aarch64 now, until it gets fixed
[07:41:06] -!- abaumann has joined #archlinux32
[07:41:06] <buildmaster> Hi abaumann!
[07:41:07] <buildmaster> !rq abaumann
[07:41:07] <phrik> buildmaster: <abaumann> very soon only a machine learning algorithm will be able to devise a build plan for a Linux distribution..
[07:41:29] <abaumann> Yeah, qemu has a non-rebuilding firmware package..
[07:41:41] <abaumann> bill-auger: ruby still has issues.
[07:42:07] <abaumann> I had to rebuild the whole bunch (I went with a ruby-bootstrap in /opt/ruby-XXX in order to get out of the source tree of Archlinux).
[07:43:03] <abaumann> Then I rebuilt all packages from ruby-stdlib, finally ruby itself. It goes in 2-3 stages, all of them need patches (first using ruby-bootstrap instead of ruby, then RUBYLIB needs to be set).
[07:43:25] <abaumann> Currently the _platform split is still wrong for native code as in ruby-digest. I'll make a patch this evening.
[07:43:52] <abaumann> I commented and vented the ideas on the chat protocols.. all individual steps are recorded in the IRC chatlog.
[07:44:22] <abaumann> Do you really have to rebuild them? My gut feeling says, they should be ok in terms of GPL and proprietary code..
[07:46:09] <abaumann> https://git.archlinux32.org (also ruby-rake-bootstrap and ruby-rake-compiler-bootstrap)
[07:46:11] <phrik> Title: PKGBUILD « ruby-bootstrap « build-support - packages - Archlinux32 package modifications (at git.archlinux32.org)
[07:46:56] <abaumann> https://git.archlinux32.org
[07:46:57] <phrik> Title: packages - Archlinux32 package modifications (at git.archlinux32.org)
[07:47:46] <abaumann> The list of all modules is in ruby's PKGBUILD https://github.com
[07:48:19] <abaumann> additionally you will have fun with rubygems conflicting with ruby-rdoc..
[12:48:50] <bill-auger> i did not have any such trouble )bootstrapping or whatever), but i only needed to package the 'ruby' pkgbase set - those packages built happily against whatever is in the standard repos now
[12:50:58] <bill-auger> "Do you really have to rebuild them?" unfortunately yes, and it is a very stupid reason - there is one C file which does not meet FSF standards
[12:54:00] <bill-auger> its in a vendored library - the license of that project actually changed in its upstream about 8 years ago, and the new license is acceptable; but the library maintainer is completely unresponsive, the ruby devs will not change upstreams
[12:57:10] <bill-auger> and whats more dumb, that library only replaces functionality which is in the standard ruby libraries (its a JSON parser) - so long story short, we need to rebuild each ruby for like 10 years now, only to replace that "ancient" third-party library with the native ruby library
[13:22:11] <abaumann> Uh, yeah, that's not nice..
[13:22:46] <abaumann> ..yes, I did too many rounds, I think. Also, some of the ruby modules where quite old, that's why I maybe had some clashes..
[13:25:23] <bill-auger> alright, i will put it in testing for now
[13:26:12] <bill-auger> WRT qemu, KillerWasp found a package conflict yesterday - qemu-system-aarch64 depends on a package named 'edk2-armvirt' ; but there is no such package with that name
[13:28:50] <abaumann> The edk2 firmware failed to build, I think. But most likely it built in the past. I'll have a look this evening. :-)
[13:29:42] <bill-auger> if it built in the past, wouldnt there be a package in the repos? - it does not depend on a specific version, so probably any version that compiles would do
[13:31:06] -!- abaumann has quit [Quit: Client closed]
[13:32:57] -!- abaumann has joined #archlinux32
[13:33:43] <abaumann> uh. the diff-PKGBUILD says '# no aarch64'
[13:33:48] <abaumann> so I hacked it out.
[13:33:57] <abaumann> so I have to remove it also in qemu..
[13:34:10] <abaumann> or fix the bug in aarch64 support in edk2
[13:38:46] <abaumann> mmh. no cross-compiler IA-32 to AARCH64, that's the root cause..
[13:39:04] <abaumann> ..anyway, many gcc-cross-compilers don't work currently on Arch32 (if not all of them..)
[13:39:33] <abaumann> Ah, now I have the one for AARCH64, so let's try to rebuild edk2 with no patching..
[13:41:46] <bill-auger> our aarch64 support in the armv7h port is stuck for that reason - we can only build the bootstrapping cross-compiler for armv7h->aarch64, and it is not good enough to build the kernel
[13:42:38] <bill-auger> had to cross-compile the aarch64 kernel from x86_64, with "CARCH=armv7h" explicitly in the PKGBUILD - very ugly
[13:44:31] <bill-auger> not good enough to build the kernel ... , or to build itself, the proper cross-compiler which probably could build the kernel
[13:53:04] <abaumann> cc1: error: code model ‘small’ not supported in the 32 bit mode
[13:53:04] <abaumann> cc1: sorry, unimplemented: 64-bit mode not compiled in
[13:53:26] <abaumann> oups, yes, I'm building a 64-bit firmware with a 32-bit cross-compiler..
[13:54:04] <abaumann> I was thinking that maybe we have to switch to use 64-bit packages from upstream (we can still rebuild them with 64-bit chroots, if we like).
[13:54:32] <abaumann> Also some packages might behave better when cross-compiled than when compiled from 32-bit chroots..
[13:55:11] <abaumann> I'll make empty stub libraries for all thinks not building in edk2. Then at least, we can install it..
[14:33:49] -!- lithiumpt has quit [Ping timeout: 252 seconds]
[14:39:00] -!- abaumann has parted #archlinux32
[15:04:49] -!- lithiumpt has joined #archlinux32
[16:34:20] -!- GNUtoo has quit [Remote host closed the connection]
[16:39:43] -!- GNUtoo has joined #archlinux32
[16:50:35] -!- sunshavi has quit [Ping timeout: 252 seconds]
[16:56:32] -!- drathir_tor has quit [Ping timeout: 258 seconds]
[17:15:48] -!- abaumann has joined #archlinux32
[17:15:49] <buildmaster> Hi abaumann!
[17:15:49] <buildmaster> !rq abaumann
[17:15:49] <phrik> buildmaster: <abaumann> Why not copy postfix into systemd?
[17:20:56] <abaumann> uh, any packages get scheduled also to 486?
[17:21:03] <abaumann> poor slaves :-)
[17:33:34] -!- drathir_tor has joined #archlinux32
[17:46:04] <abaumann> bill-auger: actually, if ruby-json is the problematic module, it should be sufficient to patch and build that one only now, ruby is split up in packages..
[18:11:23] <abaumann> warning: cannot resolve "libcelt0.so=2-32", a dependency of "jack2"
[18:11:30] <abaumann> now also for all of qemu, nice..
[18:46:57] -!- sunshavi has joined #archlinux32
[18:55:01] -!- sunshavi has quit [Ping timeout: 260 seconds]
[19:01:03] -!- sunshavi has joined #archlinux32
[21:05:28] <abaumann> so, qemu-full installs.
[21:07:45] <abaumann> The ruby patching works (tested locally), but it takes a while for it to rebuild now..
[21:17:33] <abaumann> ..and then comes another round of rebuilding, this time without the RUBYLIB patches for x86-linux..
[21:17:49] <abaumann> ..finally I want to retrigger a complete rebuild of all of ruby, to make sure, everything builds..
[21:23:07] -!- abaumann has quit [Quit: leaving]
[21:23:29] -!- GNUtoo has quit [Quit: leaving]
[21:25:42] -!- GNUtoo has joined #archlinux32