The libstdc++ additions are intentional, libstdc++ is symbol versioned and new symbols have been added to the latest symbol version. For go it feels like abidiff has issues with gccgo.

RHEL doesn't ship gccgo nor gnat I think.

.symtab and .debug_info sections in libgo and gccgo binaries are intentional, go unfortunately isn't able to deal with stripped or stripped to file executables or libraries. Missing .gdb_index must be some GDB issue, perhaps it gives up on Ada debuginfo. The Wbidi-chars tests intentionally test how the compiler behaves on the problematic Unicode chars. If it would be possible to waive some of this and be remembered for all future gcc builds, I'd waive it somewhere, but I'm certainly not willing to waive it for each single build when it reports the same things again and again.

Must be https://gcc.gnu.org/PR109304

LGTM.

I need an errata together with annobin.

Note, the PR104099 change just buys the basic_json using packages some time, as nothing changed for C++23 which requires the behavior that gcc previously had for older standards as an extension. So at latest when g++ switches to -std=gnu++23 by default (right now the default is still -std=gnu++17), it will not work again.

PR104099 is actually fixed in 12-0.1-0.2.fc36, I just forgot to mention it in the %changelog (created snapshot, then noticed the PR was fixed and merged again from release branch to cherry pick that fix, but didn't update %changelog for it).

We want gcc-12.0.1-0.1.1.fc36 or even better gcc-12.0.1-0.2.fc36, rather than this build.

That is up to the annobin maintainer. As gcc-11.1.1-1.fc34 is compatible with the annobin that is in f34, there is no need to include it in the same errata.

I had to update annobin again for the gcc f33 errata https://bodhi.fedoraproject.org/updates/FEDORA-2021-3a908d6827 so I think it would be best to withdraw this update.

Are you sure you had the corresponding annobin installed too? With annobin-9.27-1.fc32 or older the above is certainly expected on armv7hl.

That wasn't really a good idea to push this into F32, as it will make #1830472 happen there too.

No, this is subsumed by https://bodhi.fedoraproject.org/updates/FEDORA-2020-b2de059578

This update has been unpushed.

Sure, see above. The errata is locked and I can't do any changes.

annobin-9.13-2.fc33 should be part of this errata, but bodhi failed to add that to it multiple times, nor allowed me to revoke it.