#archlinux-ports | Logs for 2017-12-17

Back
[00:17:49] -!- guys has joined #archlinux-ports
[00:39:24] -!- guys has quit [Ping timeout: 272 seconds]
[00:44:52] -!- guys has joined #archlinux-ports
[01:16:27] -!- sunshavi has quit [Ping timeout: 256 seconds]
[02:06:15] -!- sunshavi has joined #archlinux-ports
[02:40:33] -!- guys has quit [Ping timeout: 264 seconds]
[02:52:26] -!- guys has joined #archlinux-ports
[04:41:09] -!- guys has quit [Ping timeout: 264 seconds]
[04:41:12] -!- ztrawhcse has joined #archlinux-ports
[04:41:32] ztrawhcse is now known as guys
[06:40:33] -!- guys has quit [Ping timeout: 264 seconds]
[06:54:57] -!- guys has joined #archlinux-ports
[08:24:47] -!- because has quit [Ping timeout: 248 seconds]
[08:26:42] -!- because has joined #archlinux-ports
[08:40:38] -!- ztrawhcse has joined #archlinux-ports
[08:40:54] -!- guys has quit [Killed (adams.freenode.net (Nickname regained by services))]
[08:40:54] ztrawhcse is now known as guys
[08:52:26] -!- deep42thought has joined #archlinux-ports
[09:10:12] -!- deep42thought has quit [Quit: Leaving.]
[09:47:05] -!- abaumann has joined #archlinux-ports
[10:01:41] -!- mrkiko has quit [Quit: Leaving ...]
[10:05:16] -!- hringriin has joined #archlinux-ports
[10:06:50] <abaumann> CalimeroTeknik: agreed, but I also want to cover the generic way of bootstrapping non-backwards-compatible architectures with my approach
[10:07:43] <abaumann> backwards compatibility can be a problem, as things work in chroots, but fail on the real machine because some wrong opcodes have crept in
[10:08:01] <abaumann> linux32, setarch won't work, they fake a i686
[10:08:20] <abaumann> either true cross-compilation or a uname_hack in a chroot (credits to CLFS) do the trick
[10:17:00] <abaumann> in theory compiling everything with -march=i486 in a chroot should work, but I have seen funny probing for SSE2 for instance circumventig all methods above.
[10:41:35] -!- ztrawhcse has joined #archlinux-ports
[10:42:37] guys is now known as Guest14336
[10:42:38] -!- Guest14336 has quit [Killed (hitchcock.freenode.net (Nickname regained by services))]
[10:42:38] ztrawhcse is now known as guys
[10:45:20] -!- deep42thought has joined #archlinux-ports
[11:00:26] <abaumann> deep42thought: my build slave is ready again. :-)
[11:01:02] <deep42thought> :-)
[11:01:38] <deep42thought> if you send me your patch of makepkg to disable the root-check, I can also put this into devtools32, if you like
[11:02:10] <abaumann> nah. I think it's not a good idea to build as root. we should not enforce that.
[11:02:17] <deep42thought> hmm, ok
[11:02:20] <deep42thought> as you like
[11:02:21] <abaumann> I hade an issue with build-package -l and work-tree..
[11:02:26] <abaumann> ..this is more important.
[11:02:34] <deep42thought> yeah, what is it?
[11:02:40] <abaumann> the initial checkout of the repos in conf/default.conf is done with --bare
[11:02:48] <deep42thought> oh
[11:02:49] <deep42thought> yes
[11:02:50] <abaumann> and then work-tree is not working.
[11:02:50] <deep42thought> right
[11:03:09] <deep42thought> well, then work-tree also has no meaning
[11:03:16] <deep42thought> so just use HEAD instead
[11:03:32] <abaumann> ah. that's also an option.
[11:03:48] <deep42thought> or are you suggesting to go away from bare repositories?
[11:04:23] <deep42thought> My reason for changing to bare was, that we could avoid all the house keeping of the work dir and index
[11:04:26] <abaumann> mmh. there is a benefit of non-bare repos.. I don't have to checkout before I can have a look at a PKGBUILD file for instance.
[11:04:44] <abaumann> efficiency vs. usability.
[11:04:45] <abaumann> mmh.
[11:04:46] <abaumann> :-)
[11:04:55] <deep42thought> you can simply use a non-bare on your system
[11:04:59] <deep42thought> (I'm doing this)
[11:05:01] <abaumann> yes.
[11:05:06] <abaumann> I'm doing that too.
[11:05:14] <deep42thought> but I think usual build slaves should use bare
[11:05:18] <abaumann> yep.
[11:05:32] <abaumann> and the build script works with both types of repos.
[11:05:33] <deep42thought> and the build master does not care (I think, it's also still using non-bare
[11:05:34] <deep42thought> )
[11:05:38] <deep42thought> right
[11:05:50] <deep42thought> it just sets up bare repos for new slaves
[11:05:56] <deep42thought> which is a sane default, I think
[11:06:16] <abaumann> agreed
[11:06:32] <abaumann> https://buildmaster.archlinux32.org
[11:06:41] <deep42thought> yeah
[11:06:45] <abaumann> still a big graph, especially around systemd/cryptsetup
[11:06:56] <deep42thought> it's ceph, blocking the unstaging of all that, I think
[11:07:09] <abaumann> oh.
[11:07:33] <deep42thought> https://ptpb.pw
[11:07:34] <abaumann> well, I have a local build workspace now with the failing test.. but I cannot promise I find fixes for all those bugs very fast.
[11:08:08] <deep42thought> ah, ok now there are even more packages, "not yet built" blocking systemd
[11:08:38] <abaumann> I start to think that the base system has to be rebuilt differently
[11:08:49] <abaumann> sort of in waves..
[11:09:25] <deep42thought> I'm not sure, I understand correctly
[11:10:31] <abaumann> parts of the graph have cycles
[11:10:38] <abaumann> and the ABI dependencies
[11:10:48] <abaumann> nothing new :-)
[11:11:20] <abaumann> there is another thing: if the toolchain changes, in theory everything must be rebuilt
[11:11:31] <deep42thought> yes
[11:11:41] <deep42thought> and if one single package fails then, it will block everything
[11:11:46] <abaumann> I'm not a fan of finding out months from now that the toolchain had a subtle bug causing a KDE program to go belly-up.
[11:11:50] <deep42thought> I think, that is, what we're experiencing currently
[11:11:54] <abaumann> right
[11:13:04] <deep42thought> ceph and deepin-daemon are the packages blocking the unstaging, currently
[11:13:08] <deep42thought> https://ptpb.pw
[11:14:07] <deep42thought> so you think, we should occassionally rebuild everything just to be sure to catch bugs in the toolchain?
[11:14:27] <guys> or bugs everywhere else
[11:14:31] <deep42thought> :-)
[11:14:40] <guys> but #archlinux-reproducible is doing much the same :)
[11:14:45] <deep42thought> like with our first attempt of rebuilding everything at once
[11:15:05] <deep42thought> so you check, that the new toolchain produces the same packages?
[11:15:19] <deep42thought> #archlinux-reproducible, I mean
[11:16:26] <guys> we check that packages built multiple times byte-identical, but mostly that means double builds with varying system times, /bin/sh, hostname, etc.
[11:18:18] <deep42thought> a change in the toolchain might change the package, though
[11:18:20] <deep42thought> right?
[11:19:41] -!- ztrawhcse has joined #archlinux-ports
[11:20:00] guys is now known as Guest92779
[11:20:00] -!- Guest92779 has quit [Killed (cherryh.freenode.net (Nickname regained by services))]
[11:20:00] ztrawhcse is now known as guys
[11:25:20] <guys> sure, but reproducible builds are not currently varying the toolchain
[11:25:46] <jelle> gcc and clang probably (100%) produce different binaries
[11:40:46] -!- deep42thought has quit [Quit: Leaving.]
[11:57:47] <abaumann> sorry. phone call. :-)
[11:58:31] <abaumann> jelle: yes, 100% different. But I'm not aware of a major distribution using clang as their toolchain.
[11:58:40] <jelle> abaumann: mandriva
[11:58:48] <abaumann> ah? ok.
[11:58:50] <abaumann> interesting.
[11:58:53] <jelle> iirc
[12:00:08] <abaumann> I'm aware of FreeBSD and lately OpenBSD
[12:02:04] <abaumann> guys: reproducible builds are a very good thing, I saw the project.
[12:03:19] -!- hringriin_ has joined #archlinux-ports
[13:23:34] -!- guys has quit [Ping timeout: 268 seconds]
[13:36:32] -!- guys has joined #archlinux-ports
[14:07:46] -!- abaumann has quit [Quit: leaving]
[14:32:47] -!- because has quit [Ping timeout: 248 seconds]
[14:34:42] -!- because has joined #archlinux-ports
[14:48:19] -!- deep42thought has joined #archlinux-ports
[15:23:45] -!- guys has quit [Ping timeout: 264 seconds]
[15:31:08] -!- abaumann has joined #archlinux-ports
[15:32:56] <deep42thought> abaumann: I don't really see, how we want to detect bugs introduced by a changing toolchain early
[15:35:11] <abaumann> yeah. sorry for the interruption..
[15:35:16] <abaumann> ..phone call. :-)
[15:35:43] <deep42thought> np
[15:35:46] <abaumann> this is correct. we cannot detect them.
[15:35:54] <abaumann> upstream can do a lot about this.
[15:36:09] <abaumann> I would never dream of just upgrading gcc to a new release version..
[15:36:36] <abaumann> I would only change the trio infernale binutils/gcc/glibc together.. with known working releases.
[15:36:57] <abaumann> Now for the toolchain I think this is done since long time in this way..
[15:37:10] <abaumann> .. I 'm thinking about new "toolchains", one of them being systemd.
[15:37:17] <abaumann> the package manager is another one.
[15:37:21] <deep42thought> ah, yeah
[15:37:22] -!- guys has joined #archlinux-ports
[15:38:08] <abaumann> pacman has big dependencies (gpg, libarchive, curl) to name the main ones.
[15:38:28] <abaumann> systemd is more complex.
[15:39:17] <abaumann> I also made good experiences in the past to have static versions of those central things around like a static pacman.
[15:39:36] <deep42thought> might be an option for build-support
[15:39:54] <deep42thought> e.g. have pacman-static in build-support providing pacman
[15:39:54] <abaumann> yes, basically as a fallback. pacman must never break.
[15:40:14] <deep42thought> can you set this up?
[15:40:24] <deep42thought> I'm not that familiar with linker flags :-)
[15:40:35] <abaumann> now, here come problems:
[15:41:32] <abaumann> not even sure if that's still possible with all the PIC/PIE whatever stuff floating around lately for improved security. :-)
[15:42:04] <abaumann> For cross compilation I made a temporary repository with build tools, pacman and base.
[15:42:22] <abaumann> still shared linked.
[15:42:43] <abaumann> There was a static pacman once, IIRC?
[15:43:16] <abaumann> https://aur.archlinux.org
[15:43:16] <phrik> Title: AUR (en) - pacman-static (at aur.archlinux.org)
[15:43:18] <abaumann> aha
[15:44:46] <deep42thought> why should pic/pie block static linking?
[15:45:43] <abaumann> pic is required nowadays also for static libraries, otherwise you cannot link the code to for instance a shared library or a binary. yep. you'right.
[16:00:42] <abaumann> ./build-packages -p cantor
[16:00:47] <abaumann> I get ERROR: pkgrel is not allowed to be empty.
[16:00:59] <abaumann> The package was moved from community to extra.
[16:01:17] <abaumann> I seem not to be able to force building it on my slave
[16:16:13] <abaumann> aha: cantor extra on rechenknecht.. so it recovered. :-)
[16:23:47] <deep42thought> let me check
[16:26:28] <deep42thought> it's currently on buildknecht2, but failing with "error Julia can only be built for architectures above Pentium 4. Pass -march=pentium4, or set MARCH=pentium4 and ensure that -march is not passed separately with an older architecture"
[16:27:08] <abaumann> ah. then my patch is not working. I thought adding march helps.
[16:30:04] <deep42thought> did you move your patch from community to extra, too?
[16:30:12] <deep42thought> yeah, seems so
[16:30:31] <deep42thought> "a" appends in the next line
[16:30:38] <deep42thought> I don't think, that's what you want
[16:31:01] <deep42thought> probably some "s/$/ MARCH=pentium4/" or similar is what you're looking for
[16:31:04] <abaumann> Nae. I have to test again do I need a patch for cmake, for make, for both?
[16:31:40] -!- oaken-source has joined #archlinux-ports
[16:32:05] <deep42thought> huh, what?
[16:33:15] <abaumann> I though it reacts to an environment variable, but it looks more like a make flag..
[16:33:18] <abaumann> easy..
[16:33:49] <abaumann> https://bbs.archlinux32.org
[16:33:51] <phrik> Title: How can I see PKGBUILD's of packages in repository ? / Creating/Maintaining Packages / Arch Linux 32 Forums (at bbs.archlinux32.org)
[16:34:00] <abaumann> there is a xerxec-c depends on wrong icu issue it seems.
[16:34:30] <abaumann> could also be manjaro-specific.
[16:36:53] <abaumann> yep. confirmed.xerces-c-3.2.0-2 libicuuc.so.59 => not found
[16:37:41] <deep42thought> these are stable packages?
[16:38:00] <abaumann> yes.
[16:38:27] <abaumann> Nov 18 11:23
[16:38:33] <abaumann> mirror/i686/extra/xerces-c-3.2.0-2-i686.pkg.tar.xz
[16:38:40] <deep42thought> yeah, I see
[16:39:04] <deep42thought> I'll reschedule xerces
[16:39:08] <abaumann> thanks.
[16:39:44] <deep42thought> ok, done
[16:41:00] <abaumann> Sholdn't the PKGBUILD of xerces-c contain icu? or is icu in base?
[16:41:11] <deep42thought> icu is in base iirc
[16:41:54] <abaumann> https://www.archlinux.org
[16:41:55] <phrik> Title: Arch Linux - base (x86_64) - Group Details (at www.archlinux.org)
[16:42:09] <deep42thought> hmm, it isn't
[16:42:16] <abaumann> I suspect, everything which is drawn in by a package in the list, is also considered base?
[16:42:26] <deep42thought> then it is
[16:42:37] <abaumann> I noticed that before: packages are not in base, but it's drawn in by a package in base.
[16:42:42] <deep42thought> but it's a clear case of "links against a not-direct dependency"
[16:42:47] -!- oaken-source has quit [Quit: leaving]
[16:42:53] <abaumann> yes.
[16:43:01] <deep42thought> which breaks our (current) dependency detection
[16:43:37] <abaumann> the xerces PKGBUILD has ' ./configure --prefix=/usr --sysconfdir=/etc', so it's built with icu if icu is there and not if icu is not there.
[16:43:51] <abaumann> yeah. only detecteable with a readelf -d.
[16:54:31] <deep42thought> link-dependencies are still on the TODO
[16:57:27] <abaumann> ah: cantor uses julia and the pentium4 patch must be applied there
[17:00:49] <abaumann> mmh. there is a MARCH=pentium4 patch in julia.. ok. all fine
[17:13:44] <deep42thought> hmm, repo-add fails on the buildmaster with a lot of "Write failed" :-(
[17:14:10] <deep42thought> during the bsdtar extraction of community.{db,files}.tar.gz
[17:16:13] <deep42thought> hmm, some temp. cache seems to have overflown :-/
[17:38:21] -!- guys has quit [Quit: A random quit message]
[17:38:27] -!- ztrawhcse has joined #archlinux-ports
[17:38:43] ztrawhcse is now known as guys
[18:34:24] -!- isacdaavid has quit [Ping timeout: 248 seconds]
[18:55:41] -!- isacdaavid has joined #archlinux-ports
[19:09:17] <deep42thought> abaumann: you broke the buildmaster via extra/kdepim-runtime
[19:09:24] <deep42thought> "/home/slave/builder/work/tmp.216bEU/PKGBUILD: line 42: sha256sums[0]: cannot assign list to array member"
[19:10:31] <deep42thought> Note, that there is currently no safeguard for failing PKGBUILDs: if it's invalid bash, then the buildmaster can't create the metadata and won't continue with determining the package updates
[19:31:29] -!- jemadux has joined #archlinux-ports
[19:32:11] -!- jemadux has quit [Client Quit]
[19:38:31] -!- guys has quit [Quit: A random quit message]
[19:38:52] -!- guys has joined #archlinux-ports
[19:50:50] <abaumann> deep42thought: sorry about that. :-(
[19:50:56] <deep42thought> yeah, np
[19:51:15] <deep42thought> it's just, that the buildmaster sent me an email every 10 minutes :-)
[19:51:25] <abaumann> there was a weird problem with that package anyway: it had a completly different sum in x86_64-testing, there was no trunk.
[19:51:36] <deep42thought> O.o
[19:51:38] <abaumann> building on 64-worked, checksuming with build-package not..
[19:52:12] <deep42thought> huh? I see a trunk, there
[19:52:33] <deep42thought> and also no testing
[19:53:47] <deep42thought> but anyway: problem is solved now :-)
[19:54:47] <deep42thought> issues like FS32#22 might become a real problem in the future
[19:56:04] <guys> firefox-developer-edition?
[19:56:07] <deep42thought> yes
[19:56:13] <deep42thought> it fails due to out-of-memory
[19:56:18] <guys> I guess you could build on a better server
[19:56:27] <guys> Does regular firefox not have this issue though?
[19:56:38] <deep42thought> but we're limited to 4G ram on 32 bit aren't we?
[19:57:24] <abaumann> yes. per process limit is 4G. There is PAE, but this is per machine, not per process
[19:57:50] <abaumann> firefox-developer-edition, yeah. It can only be built on a machine with a lot of memory in a chroot, I'm afraid
[19:58:01] <guys> wait, so Firefox takes more than 4GB RAM to build now, even on i686?
[19:58:16] <abaumann> firefox-developer-edition is not firefox, as I understood.
[19:58:20] <guys> That's a shame, I'd built it for a long time on a 2GB system
[19:58:26] <abaumann> there is a lot of rust stuff in there.
[19:58:49] <guys> abaumann: developer includes some stuff that stable will eventually include, I think
[19:59:03] <abaumann> I had a similar issue on a 64MB virtual machine with i486: C compiler, no problem, C++ for a hello world runs out of memory.
[19:59:08] <deep42thought> it's not like "plus *.h files"?
[19:59:51] <abaumann> yes: normal development: modern software development relies more on the paging capabilities of the OS and on a 64-bit address room.
[20:00:44] <abaumann> there is still the option to build on chroots with enough memory on a 64-bit machine. I don't think the limitation should kick in there..
[20:01:00] <deep42thought> ok
[20:01:03] <abaumann> last resort is -m32 it or cross-compile.
[20:01:11] <deep42thought> yes
[20:01:14] <deep42thought> but it's tedious
[20:01:21] <abaumann> yes.
[20:08:40] <guys> Surely extra-i686-build on a 64-bit machine would do it?
[20:08:55] <deep42thought> that's what we do
[20:10:05] <deep42thought> hmm, it looks like the bugreport, I refered to, is outdated
[20:10:25] <deep42thought> firefox-developer-edition is currently "testing" :-)
[20:10:33] <abaumann> :-)
[20:10:50] -!- yokel has quit [Remote host closed the connection]
[20:10:55] -!- because has quit [Ping timeout: 248 seconds]
[20:12:40] -!- because has joined #archlinux-ports
[20:18:21] -!- because has quit [Ping timeout: 264 seconds]
[20:20:08] <deep42thought> abaumann: I added the stuff from the original i686 PKGBUILD for extra/cantor - let's see if that builds :-)
[20:22:18] -!- because has joined #archlinux-ports
[20:28:28] -!- hringriin_ has quit [Ping timeout: 240 seconds]
[20:33:58] <deep42thought> yup, looks good
[20:44:14] -!- yokel has joined #archlinux-ports
[20:49:24] -!- yokel has quit [Remote host closed the connection]
[20:49:48] -!- yokel has joined #archlinux-ports
[20:53:00] -!- deep42thought has quit [Quit: Leaving.]
[21:03:05] -!- Aero_ has joined #archlinux-ports
[21:03:08] <Aero_> Hi
[21:03:31] -!- Aero_ has quit [Client Quit]
[21:07:57] -!- oaken-source has joined #archlinux-ports
[21:09:58] -!- yokel has quit [Ping timeout: 272 seconds]
[21:17:45] <abaumann> deep42thought: cantor, ah. ok. seems to have worked. :-)
[21:23:21] -!- abaumann has quit [Quit: leaving]
[21:24:46] -!- oaken-source has quit [Ping timeout: 260 seconds]
[21:34:12] -!- yokel has joined #archlinux-ports
[21:42:16] -!- guys has quit [Ping timeout: 272 seconds]
[21:49:22] -!- yokel has quit [Ping timeout: 265 seconds]
[21:55:05] -!- yokel has joined #archlinux-ports
[21:56:27] -!- guys has joined #archlinux-ports
[22:08:08] -!- yokel has quit [Ping timeout: 240 seconds]
[22:23:38] -!- yokel has joined #archlinux-ports
[22:31:02] -!- yokel has quit [Ping timeout: 272 seconds]
[22:46:44] -!- yokel has joined #archlinux-ports
[22:53:36] -!- yokel has quit [Ping timeout: 248 seconds]
[22:55:09] -!- yokel has joined #archlinux-ports
[22:58:33] -!- p71 has quit [Read error: Connection reset by peer]
[23:11:43] -!- yokel has quit [Ping timeout: 248 seconds]
[23:19:35] -!- yokel has joined #archlinux-ports
[23:20:06] -!- p71 has joined #archlinux-ports
[23:40:43] -!- ztrawhcse has joined #archlinux-ports
[23:42:28] -!- guys has quit [Ping timeout: 240 seconds]
[23:44:53] -!- yokel has quit [Ping timeout: 265 seconds]
[23:46:40] ztrawhcse is now known as guys
[23:57:33] -!- yokel has joined #archlinux-ports