C++20

[Swiftb0y]

Commits contain various fixes for warnings and compile-errors. The last commit actually does the switch.
Tested on GCC12 locally, using CI for everything else.

[Swiftb0y]

Arch build fails because of a GCC bug?
google/benchmark#1398
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105329

[uklotzde]

Arch build fails because of a GCC bug? google/benchmark#1398 https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105329

These are the bugs I have experienced on RPM Fusion during the Fedora 36 Beta phase.

[Swiftb0y]

This doesn't happen on my F36 machine which has gcc 12.1.1? Where does the last .1 come from? Does fedora patch gcc themselves or do they do a rolling release or what is happening here?

[daschuer]

CMake is smart enough to pass "-std=gnu++2a" to Ubuntu Focal GCC Version 9.4.0
This works for me. Thank you.

[Swiftb0y]

Done. Let me autosquash before you merge.

[daschuer]

The Arch Linux issue is:

Error: /usr/include/c++/12.1.0/bits/char_traits.h:431:56: error: ‘void* __builtin_memcpy(void*, const void*, long unsigned int)’ accessing 9223372036854775810 or more bytes at offsets -4611686018427387902 and [-4611686018427387903, 4611686018427387904] may overlap up to 9223372036854775813 bytes at offset -3 [-Werror=restrict]
  431 |         return static_cast<char_type*>(__builtin_memcpy(__s1, __s2, __n));
      |                                        ~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~

Do we have an idea to fix it?

[Swiftb0y]

Not our fault: #4803 (comment)

Its a regression thats only present on GCC 12.1 (edit: maybe 12.0 as well) so we'll just have to wait a bit and upgrade the arch docker container.

[uklotzde]

Just noticed that I haven't backported the patch from RPM Fusion that is still needed: https://github.com/rpmfusion/mixxx/blob/master/disable_werror_in_tests.patch

[Swiftb0y]

I guess that makes sense why the fedora builds succeed while arch doesn't even though they use the same compiler...

[uklotzde]

I guess that makes sense why the fedora builds succeed while arch doesn't even though they use the same compiler...

Only fails with rpmbuild. Manual CMake builds are not affected.

[Swiftb0y]

So should be disable the buggy warnings in our vendored google benchmark?

[Swiftb0y]

@uklotzde can you give me a hint of how to do that? Should I simply just commit the changes to the CMakeLists of google benchmark? Do I supply the .patch from git? Do you instead overwrite some compiler defines in our CMakeLists.txt?

[uklotzde]

I think it's ok to simply inline the patch here. It's only a workaround that should hopefully become obsolete in the future.

[Swiftb0y]

I don't quite understand. So its fine as is or should I also commit a (imo redundant) patch file?

[Swiftb0y]

Not sure how soon this will be fixed though. Neither google seems to make any effort working around this issue, nor the gcc team sees that as very urgent...

[uklotzde]

All fine. Hopefully, we don't need to update the lib anytime soon.

[Swiftb0y]

Great, @daschuer merge?

[daschuer]

This still works for me on Ubuntu Focal. Thank you.