This is a tracking bug for self-compare bugs The general command to reproduce these failures is: $ fedabipkgdiff --self-compare -a --from fc36 $PACKAGE
With trunk as of df28c220976f41620b0bd7b22000815e11f66b75 Test are still running but so far the packages which have self-compare bugs are: 389-ds-base arm-none-eabi-gcc-cs binutils crash darktable eccodes enlightenment freefem++ gnumeric gsoap hercules iaito infinipath-psm insight kcov kdepim-addons kf5-libktorrent kf5-mailcommon kim-api kscope lammps libcouchbase libomp libXt llvm nest pocl stratagus wimlib
29406 - 389-ds-base 29407 - arm-none-eabi-gcc-cs - 29299 - binutils - which may also be the reason behind crash infinipath-psm libXt
The following seem new to me and so I think that they were likely introduced between 2b8b8bc5a8868d3ab9eae487fa64b4d4374e5297 and df28c220976f41620b0bd7b22000815e11f66b75 crash darktable eccodes enlightenment freefem++ gnumeric gsoap hercules iaito insight kcov kdepim-addons kf5-libktorrent kf5-mailcommon kim-api kscope lammps libcouchbase nest pocl stratagus wimlib
some additional ones that have popped out while testing has continued to run: v8-314 - this has both an assert and a self-compare error 29324 These were previously reported as part of the btrfs-progs self-compare problem but evidently they have a different root cause even though they have similar morphology: qt6-qtbase zfs-fuse openmpi These are new: lighttpd redis
I think from a morphological level, I’ve got basically 4 clumps There are the ones which have no information like binutils and eccodes sometimes they are arch specific which makes me wonder if there are two things in there. The original bug which manifested this is the one for binutils 29299 however probably the best example is eccodes. Then there are the ones like darktable 29414 which are confusion between typdefs and structs, template instantiations, or classes. I think that these are regressions from the past couple of patches. Then there are the ones which are C++ related and show up as a rather large number of filtered errors and also vtable changes 29407 Finally there are ones which seem to be some size confusion in the types. One package that exhibits this is hercules. I want to see how the other fixes finally play out before I file that one. With the 2.5 test run nearing completion we currently have 42 packages which have self-compare problems. Dodji is expecting rather significant changes to the C comparison engine when he lands the binutils fix. I'll do another test run at that point see how many of these were resolved. Unsorted list of packages which have self-compare problems: 389-ds-base arm-none-eabi-gcc-cs binutils crash darktable eccodes enlightenment freefem++ gsoap hercules iaito infinipath-psm insight kcov kdepim-addons kf5-libktorrent kf5-mailcommon kim-api kscope lammps libcouchbase libomp libXt lighttpd llvm mono nest octave OpenMolcas openmpi pocl proxysql qgis qt6-qtbase redis smesh stratagus unicorn v8-314 vsomeip3 wimlib zfs-fuse
2.1 has been released many of the bugs have been addressed. I am currently testing all of F37 with trunk as of 4c5f44473e9f7c0a561174fdc6cc89f40dd9892a. When that test is done I will verify that 29299 29324 29406 29407 29414 have been taken care of.
Another one that looks similar which is arch specific is: Comparing the ABI of binaries between crash-8.0.1-3.fc37.s390x.rpm and crash-8.0.1-3.fc37.s390x.rpm: ======== comparing'crash' to itself wrongly yielded result: =========== Functions changes summary: 0 Removed, 1 Changed, 0 Added function Variables changes summary: 0 Removed, 0 Changed, 0 Added variable 1 function with some indirect sub-type change: [C] 'function regcache* get_thread_regcache(thread_info*)' at regcache.c:408:1 has some indirect sub-type changes: ===SELF CHECK FAILED for 'crash'
Posted to the wrong bug.
Also kcov, nest, pocl.
With the current trunk as of: commit 7952a748de3578964b2c6b8d3bb94e0bbcfb653e (HEAD -> master, origin/master, origin/HEAD) Author: Dodji Seketeli <dodji@redhat.com> Date: Mon Mar 27 00:10:21 2023 +0200 Bug 29345 - abipkgdiff is confused by symlinked binaries in RPMs The following packages continue to have self-compare bugs: bpftrace btrfs-progs cppcheck-gui darktable enlightenment gdb-headless gdb-minimal getdns glibc groonga-libs insight kscope mono-complete mono-core mono-data mono-devel mono-extras mono-mvc mono-reactive mono-wcf mono-web mono-winforms mono-winfx nest-mpich nest-openmpi nest openldap-servers OpenMolcas pocl qt6-qtbase ruby-libs xorg-x11-server-Xdmx They seem to be reasonably split amongst the still open self-compare bugs. The one exception is the Bug 29673 which is starting to look like there may be more than one problem in there.
Results from first F38 run: OpenMolcas - Bug 29683 dia - Bug 30463 darktable - Bug 30464 flang - Bug 30465 harfbuzz - Bug 30466 enlightenment - Bug 30467 qt6-qtbase - Bug 29603 kscope - Bug 29407 insight - Bug 30461 mono getdns unicorn btrfs-progs - no real data but I'm guessing it is a duplicate of something above nest - no real data and so maybe a duplicate of something above but only happens on aarch64 and ppc64le