Description of problem: Advisory: ========== This update of perl-URPM fixes 2 issues: - prevent rpmdrake from crashing if /etc/urpmi/skip.list contains an invalid regular expression (mga#15353) - fix reporting size on big packages on x86_64 (mga#19571) There's a more complete (but more invasive fix) in upcoming Mageia 6 for the later. List of generated packages: ============================= perl-URPM-5.06.2-2.mga5.i586.rpm perl-URPM-5.06.2-2.mga5.x86_64.rpm How to reproduce: ================= For the first bug: - put eg "/*nvidia-current/" in /etc/urpmi/skip.list - run rpmdrake, it'll crash - update perl-URPM - re-run rpmdrake, it won't crash For the second bug: - enable the debug media for Core Updates - Run: urpmq -i kernel-desktop-4.4.16-1.mga5-debuginfo|grep Size It should display an invalid (negative size) Now that perl-URPM has been updated on the servers, the synthesis files contains the right size and locally the new perl-URPM, once updated, should enable to report the right size.
MGA5-32 on Acer D620 Xfce Tested existing config with /etc/urpmi/skip.list as described above. rpmdrake crashes. Commented out the line "/*nvidia-current/" in /etc/urpmi/skip.list to be able to update. Did update of perl-URPM-5.06.2-2.mga5.i586.rpm, closed MCC, opened again, checked correct version is installed. Removed commenting the line "/*nvidia-current/" in /etc/urpmi/skip.list , saved file, ran rpmdrake, results in I think the same crash Crash info below: Can't use an undefined value as a subroutine reference at /usr/lib/perl5/vendor_perl/5.20.1/i386-linux-thread-multi/URPM/Resolve.pm line 1958. Perl's trace: drakbug::bug_handler() called from /usr/lib/perl5/vendor_perl/5.20.1/i386-linux-thread-multi/URPM/Resolve.pm:1958 URPM::compute_flags() called from /usr/lib/perl5/vendor_perl/5.20.1/urpm/media.pm:846 urpm::media::_compute_flags_for_skiplist() called from /usr/lib/perl5/vendor_perl/5.20.1/urpm/media.pm:740 urpm::media::configure() called from /usr/lib/perl5/vendor_perl/5.20.1/Rpmdrake/open_db.pm:157 Rpmdrake::open_db::open_urpmi_db() called from /usr/lib/perl5/vendor_perl/5.20.1/Rpmdrake/pkg.pm:444 Rpmdrake::pkg::get_pkgs() called from /usr/lib/perl5/vendor_perl/5.20.1/Rpmdrake/gui.pm:642 Rpmdrake::gui::pkgs_provider() called from /usr/libexec/drakrpm:804 main::__ANON__() called from /usr/lib/perl5/vendor_perl/5.20.1/rpmdrake.pm:415 rpmdrake::slow_func() called from /usr/libexec/drakrpm:804 main::run_treeview_dialog() called from /usr/libexec/drakrpm:847 Gebruikte thema: oxygen-gtk
CC: (none) => herman.viaene
Not a test, exactly, but activating the updates testing repositories requires me to update this package before revealing the other packages ready for testing. While I did not attempt to test it, so far, on x86-64, no problems noted.
CC: (none) => andrewsfarm
commit 49d5faf692708da20ce26dd4e0b7d6ab6da2b71f Author: Thierry Vignaud <thierry.vignaud@...> Date: Thu Oct 20 16:20:01 2016 +0200 fix crash introduced by mga#15353 fix (mga#19618) rationale: - {debug_URPM} only exists if we're in debug mode - as this is an actual error, let's just always report it bug introduced in commit 070f1be3d5c3aa813f61d061c2a0fd38c2477c08 --- Commit Link: http://gitweb.mageia.org/software/rpm/perl-URPM/commit/?id=49d5faf692708da20ce26dd4e0b7d6ab6da2b71f Bug links: Mageia https://bugs.mageia.org/15353 https://bugs.mageia.org/19618
commit 9921c980dc294b6ce31450c2160a5247d4f04e6d Author: Thierry Vignaud <thierry.vignaud@...> Date: Thu Oct 20 16:20:01 2016 +0200 fix crash introduced by mga#15353 fix (mga#19618) rationale: - {debug_URPM} only exists if we're in debug mode - as this is an actual error, let's just always report it bug introduced in commit 070f1be3d5c3aa813f61d061c2a0fd38c2477c08 (cherry picked from commit 49d5faf692708da20ce26dd4e0b7d6ab6da2b71f) --- Commit Link: http://gitweb.mageia.org/software/rpm/perl-URPM/commit/?id=9921c980dc294b6ce31450c2160a5247d4f04e6d Bug links: Mageia https://bugs.mageia.org/15353 https://bugs.mageia.org/19618
I fixed Angelo's fix and bumped the package version Sorry for the mess, I should have checked more the fix that was commited in git.
Source RPM: perl-URPM-5.06.2-2.mga5 => perl-URPM-5.06.3-1.mga5
On mga5-64 added /*nvidia-current/ to skip.list rpmdrake crashed on opening $ urpmq -i kernel-desktop-4.4.16-1.mga5-debuginfo|grep Size ftp://www.mirrorservice.org/sites/mageia.org/pub/mageia/distrib/5/x86_64/media/debug/core/updates/media_info/20161021-080950-info.xml.lzma Size : -1102491331 Architecture: x86_64 Updated perl-URPM: # urpmi --media "Core Updates Testing" perl-URPM ftp://www.mirrorservice.org/sites/mageia.org/pub/mageia/distrib/5/x86_64/media/core/updates_testing/perl-URPM-5.06.3-1.mga5.x86_64.rpm installing perl-URPM-5.06.3-1.mga5.x86_64.rpm from /var/cache/urpmi/rpms added /*nvidia-current/ to skip.list rpmdrake opens normally $ urpmq -i kernel-desktop-4.4.16-1.mga5-debuginfo|grep Size Size : -1102491331 Architecture: x86_64 Only the first of the two bugs described seems to have been fixed.
CC: (none) => jim
And this is the one installed on BS: [root@duvel ~]# rpm -qa |grep perl-URPM perl-URPM-5.06.3-1.mga5
CC: (none) => tmb
(In reply to James Kerr from comment #6) > Only the first of the two bugs described seems to have been fixed. For the pkg size, synthesis need to be rebuild with the fixed perl-URPM in order to report the right sizes...
Well, the first fixed perl-URPM was installed on Oct 15th. The second on: Oct 18 22:11:07 duvel.mageia.org [RPM][19827]: install perl-URPM-5.06.2-2.mga5.x86_64: success and the third on: # rpm -qa --last |grep perl-URPM perl-URPM-5.06.3-1.mga5.x86_64 Thu 20 Oct 2016 06:07:31 PM CEST And for example kernel-4.4.22-1.mga5 got pushed on mirrors on Oct 18th: http://advisories.mageia.org/MGASA-2016-0345.html And kernel-4.4.26-1.mga5 got built and pushed on mirrors on Oct 20th: http://advisories.mageia.org/MGASA-2016-0347.html but still: ]$ urpmq -i kernel-desktop-4.4.26-1.mga5-debuginfo|grep Size Size : -1102108237 Architecture: x86_64 Now I know genhdlist2 does not update metadata on already indexed rpms (as found out when we removed kernel module versions way back to fix upgrade bug), but that should not be the case atleast with the 4.4.26 one...
CC: andrewsfarm => (none)
(In reply to Thierry Vignaud from comment #8) > For the pkg size, synthesis need to be rebuild with the fixed perl-URPM in > order to report the right sizes... Thanks for this clarification. How/when is this likely to happen? Just asking, not pressing.
CC: (none) => lewyssmith
(In reply to Lewis Smith from comment #10) > (In reply to Thierry Vignaud from comment #8) > > For the pkg size, synthesis need to be rebuild with the fixed perl-URPM in > > order to report the right sizes... > > Thanks for this clarification. How/when is this likely to happen? Just > asking, not pressing. Every time something is uploaded to testing or moved to updates, the hdlists and synthesis are rebuilt... so something is still missing from this update...
MGA5-32 on Acer D620 Xfce No installation issues Tested version perl-URPM-5.06.3-1. again no crash anymore with line "/*nvidia-current/" in /etc/urpmi/skip.list , OK at CLI as root # urpmq -i kernel-desktop-4.4.16-1.mga5-debuginfo|grep Size reg ex problem: Quantifier follows nothing in regex; marked by <-- HERE in m/(?:* <-- HERE nvidia-current)/ at /usr/lib/perl5/vendor_perl/5.20.1/i386-linux-thread-multi/URPM/Resolve.pm line 1950. $MIRRORLIST: media/debug/core/updates/media_info/20161101-002945-info.xml.lzma Size : 2114678058 Architecture: i586 seems OK
Whiteboard: (none) => MGA5-32-OK
Yeah, the real fix for very big packages will be mga6. So I suggests to remove this from the advisory: Advisory: ========== This update of perl-URPM prevents rpmdrake from crashing if /etc/urpmi/skip.list contains an invalid regular expression (mga#15353) List of generated packages: ============================= perl-URPM-5.06.2-2.mga5.i586.rpm perl-URPM-5.06.2-2.mga5.x86_64.rpm
Puzzled by comment 13, which cites version 5.06.2-2; I think this should be 5.06.3-1. Herman Comment 1 showed that version 5.06.2-2 did *not* fix the regex problem. James Comment 6 and Herman Comment 12 showed that 5.06.3-1 did fix it, but not the wrong length one. The pruned Advisory in Comment 13 now omits the 'wrong length' issue, but the version given 5.06.2-2 is that which did *not* fix the error cited (regex). So the Advisory text should be as per comment 13, but the SRPM version 5.06.3-1 . If you agree, I will do the Advisory accordingly & OK & validate this update.
(In reply to Lewis Smith from comment #14) > So the Advisory text should be as per comment 13, but the SRPM version > 5.06.3-1 . > > If you agree, I will do the Advisory accordingly & OK & validate this update. I agree.
Thanks James. OK'd for x64; validated; advisory uploaded.
Keywords: (none) => validated_updateWhiteboard: MGA5-32-OK => MGA5-32-OK MGA5-64-OK advisoryCC: (none) => sysadmin-bugs
An update for this issue has been pushed to the Mageia Updates repository. http://advisories.mageia.org/MGAA-2016-0132.html
Status: NEW => RESOLVEDResolution: (none) => FIXED