Re: [arch-dev-public] gcc 4.8 breaking libdrm for me
On 13/07/13 17:40, Andreas Radke wrote: > Uploaded tarballs with gcc47 and 48 with -O0 and -save-temps: > > https://docs.google.com/file/d/0B8HSjV2qdYV1VXhmdnlxc29RZDQ/edit?usp=sharing > https://docs.google.com/file/d/0B8HSjV2qdYV1aE80NWh4S1Yzdlk/edit?usp=sharing > > Feel free to run meld over .s files. > The first thing to be done is to limit the issue to a single file. I.e. which file do you need to compile with gcc-4.8 to generate the issue. You could also bisect gcc to find the broken commit. The gcc-git PKGBUILD in the AUR will bisect quite easily (add #commit=xx at the end of the source) if you uncomment the lines after the "XXX" markers. Building with --nocheck will not take too long to build. Allan
Re: [arch-dev-public] gcc 4.8 breaking libdrm for me
On Sat, Jul 13, 2013 at 1:21 AM, Andreas Radke wrote: > I've built libdrm now with CLANG compiler and so far also no problems. > Clean dmesg, no hangs and no glitches over Firefox tabs. > > I suggest to push a "fixed" package built with clang to testing and > report this one to gcc people. > > Opinions? > > -Andy Sounds fine to me, we have two production C/C++ compilers in the repositories so there's no need to be blocked by bugs in one. Bisecting gcc and identifying the commit causing the regression would be easier, but also a *huge* time waster :). Comparing the assembly doesn't sound worth doing, if you can't get a nice stack trace to narrow down the problem code.
[arch-dev-public] Signoff report for [testing]
=== Signoff report for [testing] === https://www.archlinux.org/packages/signoffs/ There are currently: * 7 new packages in last 24 hours * 0 known bad packages * 0 packages not accepting signoffs * 4 fully signed off packages * 78 packages missing signoffs * 2 packages older than 14 days (Note: the word 'package' as used here refers to packages as grouped by pkgbase, architecture, and repository; e.g., one PKGBUILD produces one package per architecture, even if it is a split package.) == New packages in [testing] in last 24 hours (7 total) == * openresolv-3.5.6-1 (any) * cups-1.6.3-1 (i686) * liferea-1.10.0-1 (i686) * mesa-9.1.4-5 (i686) * cups-1.6.3-1 (x86_64) * liferea-1.10.0-1 (x86_64) * mesa-9.1.4-5 (x86_64) == Incomplete signoffs for [core] (4 total) == * openresolv-3.5.6-1 (any) 0/2 signoffs * tzdata-2013d-1 (any) 1/2 signoffs * syslinux-6.01-3 (i686) 0/1 signoffs * syslinux-6.01-3 (x86_64) 1/2 signoffs == Incomplete signoffs for [extra] (74 total) == * archboot-2013.07-1 (any) 0/2 signoffs * texlive-bibtexextra-2013.30944-1 (any) 1/2 signoffs * texlive-core-2013.30962-1 (any) 1/2 signoffs * texlive-fontsextra-2013.30955-1 (any) 1/2 signoffs * texlive-formatsextra-2013.30214-1 (any) 1/2 signoffs * texlive-games-2013.30959-1 (any) 0/2 signoffs * texlive-genericextra-2013.29803-1 (any) 0/2 signoffs * texlive-htmlxml-2013.29725-1 (any) 0/2 signoffs * texlive-humanities-2013.30815-1 (any) 0/2 signoffs * texlive-langcjk-2013.30856-2 (any) 0/2 signoffs * texlive-langcyrillic-2013.30653-1 (any) 0/2 signoffs * texlive-langextra-2013.30816-1 (any) 0/2 signoffs * texlive-langgreek-2013.30912-1 (any) 0/2 signoffs * texlive-latexextra-2013.30964-1 (any) 0/2 signoffs * texlive-music-2013.30005-1 (any) 0/2 signoffs * texlive-pictures-2013.30637-1 (any) 0/2 signoffs * texlive-plainextra-2013.30377-1 (any) 0/2 signoffs * texlive-pstricks-2013.30956-1 (any) 0/2 signoffs * texlive-publishers-2013.30958-1 (any) 0/2 signoffs * texlive-science-2013.30910-1 (any) 0/2 signoffs * akonadi-1.10.0-2 (i686) 0/1 signoffs * asio-1.4.8-3 (i686) 0/1 signoffs * avogadro-1.1.0-5 (i686) 0/1 signoffs * boost-1.54.0-2 (i686) 0/1 signoffs * clucene-2.3.3.4-7 (i686) 0/1 signoffs * cups-1.6.3-1 (i686) 0/1 signoffs * ekiga-4.0.1-5 (i686) 0/1 signoffs * enblend-enfuse-4.1.1-2 (i686) 0/1 signoffs * hugin-2012.0.0-7 (i686) 0/1 signoffs * libindi-0.9.6-3 (i686) 0/1 signoffs * libkolabxml-0.8.4-2 (i686) 0/1 signoffs * liborigin2-20110829-3 (i686) 0/1 signoffs * libtorrent-rasterbar-1:0.16.10-2 (i686) 0/1 signoffs * licq-1.7.1-5 (i686) 0/1 signoffs * liferea-1.10.0-1 (i686) 0/1 signoffs * lirc-1:0.9.0-51 (i686) 0/1 signoffs * lyx-2.0.6-3 (i686) 0/1 signoffs * mesa-9.1.4-5 (i686) 0/1 signoffs * mkvtoolnix-6.3.0-2 (i686) 0/1 signoffs * nvidia-319.32-3 (i686) 0/1 signoffs * nvidia-304xx-304.88-11 (i686) 0/1 signoffs * qtiplot-0.9.8.9-5 (i686) 0/1 signoffs * scribus-1.4.2-3 (i686) 0/1 signoffs * source-highlight-3.1.7-6 (i686) 0/1 signoffs * strigi-0.7.8-4 (i686) 0/1 signoffs * texlive-bin-2013.30973-2 (i686) 0/1 signoffs * xsd-3.3.0-9 (i686) 0/1 signoffs * akonadi-1.10.0-2 (x86_64) 0/2 signoffs * asio-1.4.8-3 (x86_64) 0/2 signoffs * avogadro-1.1.0-5 (x86_64) 0/2 signoffs * boost-1.54.0-2 (x86_64) 0/2 signoffs * clucene-2.3.3.4-7 (x86_64) 0/2 signoffs * cups-1.6.3-1 (x86_64) 0/2 signoffs * ekiga-4.0.1-5 (x86_64) 0/2 signoffs * enblend-enfuse-4.1.1-2 (x86_64) 0/2 signoffs * hugin-2012.0.0-7 (x86_64) 0/2 signoffs * libindi-0.9.6-3 (x86_64) 0/2 signoffs * libkolabxml-0.8.4-2 (x86_64) 0/2 signoffs * liborigin2-20110829-3 (x86_64) 0/2 signoffs * libtorrent-rasterbar-1:0.16.10-2 (x86_64) 0/2 signoffs * licq-1.7.1-5 (x86_64) 0/2 signoffs * liferea-1.10.0-1 (x86_64) 0/2 signoffs * lirc-1:0.9.0-51 (x86_64) 0/2 signoffs * lyx-2.0.6-3 (x86_64) 0/2 signoffs * mesa-9.1.4-5 (x86_64) 0/2 signoffs * mkvtoolnix-6.3.0-2 (x86_64) 0/2 signoffs * nvidia-319.32-3 (x86_64) 0/2 signoffs * nvidia-304xx-304.88-11 (x86_64) 0/2 signoffs * qtiplot-0.9.8.9-5 (x86_64) 0/2 signoffs * scribus-1.4.2-3 (x86_64) 0/2 signoffs * source-highlight-3.1.7-6 (x86_64) 0/2 signoffs * strigi-0.7.8-4 (x86_64) 0/2 signoffs * texlive-bin-2013.30973-2 (x86_64) 0/2 signoffs * xsd-3.3.0-9 (x86_64) 0/2 signoffs == Completed signoffs (4 total) == * dhcpcd-6.0.2-1 (i686) * linux-3.10-1 (i686) * dhcpcd-6.0.2-1 (x86_64) * linux-3.10-1 (x86_64) == All packages in [testing] for more than 14 days (2 total) == * dhcpcd-6.0.2-1 (i686), since 2013-06-25 * dhcpcd-6.0.2-1 (x86_64), since 2013-06-25 == Top five in signoffs in last 24 hours ==
Re: [arch-dev-public] gcc 4.8 breaking libdrm for me
Uploaded tarballs with gcc47 and 48 with -O0 and -save-temps: https://docs.google.com/file/d/0B8HSjV2qdYV1VXhmdnlxc29RZDQ/edit?usp=sharing https://docs.google.com/file/d/0B8HSjV2qdYV1aE80NWh4S1Yzdlk/edit?usp=sharing Feel free to run meld over .s files. -Andy signature.asc Description: PGP signature
Re: [arch-dev-public] gcc 4.8 breaking libdrm for me
Am Sat, 13 Jul 2013 08:45:38 +0200 schrieb Alexander Rødseth : > Well done narrowing the issue down. I haven't had the chance to > compare the assembly output yet, but does it work if you compile with > -fno-aggressive-loop-optimizations? Ref: > http://postgresql.1045698.n5.nabble.com/Back-branches-vs-gcc-4-8-0-td5750997.html > > - Alexander / xyproto > No. I've tried this one already and it doesn't fix it. When built with -O0 all optimization should be turned off and it's still failing. I guess I should build libdrm with gcc 4.7 and 4.8 both with -O0 and compare the full trees with meld. -Andy signature.asc Description: PGP signature
