Description of problem: After gcc-4.4.0-8 change: ------------------------------------------- * Fri Jun 12 2009 Jakub Jelinek <jakub> 4.4.0-8 - update from gcc-4_4-branch - PRs c++/40381, libfortran/40330 - add -mcrc32 support on ix86 - support -gdwarf-3 and default to it, emit DW_OP_call_frame_cfa - fix up ix86 padding for branch mispredicts - improve .debug_loc generation -------------------------------------------- rpm cannot create debuginfo rpm correctly. For example: http://koji.fedoraproject.org/koji/buildinfo?buildID=106022 i586 build log says: -------------------------------------------- + /usr/lib/rpm/find-debuginfo.sh --strict-build-id /builddir/build/BUILD/jd-2.4.1-svn2883_trunk extracting debug info from /builddir/build/BUILDROOT/jd-2.4.1-0.1.svn2883_trunk.fc12.i386/usr/bin/jd /usr/lib/rpm/debugedit: /builddir/build/BUILDROOT/jd-2.4.1-0.1.svn2883_trunk.fc12.i386/usr/bin/jd: Unknown debugging section .debug_pubtypes /usr/lib/rpm/debugedit: /builddir/build/BUILDROOT/jd-2.4.1-0.1.svn2883_trunk.fc12.i386/usr/bin/jd: DWARF version 3 unhandled + /usr/lib/rpm/check-buildroot --------------------------------------------- and the created debuginfo rpm http://koji.fedoraproject.org/koji/rpminfo?rpmID=1275099 does not have any debug information. Version-Release number of selected component (if applicable): rpm-4.7.0-6.fc12.i586 gcc-4.4.0-8.i586 How reproducible: 100%
I'm working on a prelink patch to handle all the recent DWARF3 stuff, and as debugedit is based on prelink's dwarf2 adjustment code, I guess most of it will apply to debugedit as well.
Created attachment 347759 [details] prelink-dwarf3.patch So far only lightly tested prelink patch. debugedit.c needs similar changes.
Created attachment 347764 [details] debugedit-dwarf3.patch Totally untested debugedit.c patch.
Would be nice to have this fixed as soon as possible to avoid mass rebuilds due to broken debuginfos and to make it easier to track ones that are results of packaging bugs...
Jakub's patch from comment #3 now built into rpm-4.7.0-7.fc12 (and applied upstream too) so this should be fixed as soon as the new version hits the builders.