Also there's a backported CVE patch for qtdeclarative in Debian and a half-working patch for ffmpeg 8 support for qtwebengine in Fedora.

@eryngion well done for all the work! I have a few comments regarding the commit messages. I'd personally apply the following rebase on the commit log:

pick 051cefa0bfa0 qt5: simplify package functions' definitions and calls
break
# TODO: Check the hash doesn't change
pick f4423ca7f70f qt5.qtbase: remove the unused withQttranslation parameter
break
# TODO: Check the hash doesn't change
pick 636a2757a8c4 qt5.qtbase: remove broken QMAKEMODULES code
pick 3bfdd7d08f77 qt5.qtbase: fix mysql support for cross builds
pick f97b09838819 qt5.qtbase: strictDeps works without mysql in nativeBuildInputs
break
# commit: qt5.qtdeclarative: remove unused qtsvg dependency (split the next commit to 2 separate commits)
reword a3868fe309ef qt5.qtdeclarative: fix building with gcc15 and remove unused qtsvg dependency
pick 7c4d33567998 qt5.qtwebkit: fix build and use hyphen and woff2 from pkgs
pick ed823ab51e4c Revert "qt5.qtwebengine: Pin to GCC 14"
squash a258d0b6f167 qt5.qtwebengine: fix building with gcc 15, use more system librares, and propagate less of them
pick e04ba2746ca8 qt5.qmake: make it usable again for packages that pass file names in qmakeFlags

I also have a question: How come qt5.qtdeclarative builds fine now, with GCC 15 on branch master?

I also have a question: How come qt5.qtdeclarative builds fine now, with GCC 15 on branch master?

I was wondering about something along those lines. So, {NIX_CFLAGS_COMPILE = "-H"; qmakeFlags = ["-d"];} to the resque.

In native builds includes look like this:

. ../qml/compiler/qv4compiler_p.h
.. /nix/store/***-qtbase-5.15.18-dev/include/QtCore/qstring.h
... /nix/store/***-qtbase-5.15.18-dev/include/QtCore/qchar.h
.... /nix/store/***-qtbase-5.15.18-dev/include/QtCore/qglobal.h
..... /nix/store/***-qtbase-5.15.18-dev/include/QtCore/qatomic.h
...... /nix/store/***-qtbase-5.15.18-dev/include/QtCore/qbasicatomic.h
....... /nix/store/***-qtbase-5.15.18-dev/include/QtCore/qatomic_cxx11.h
........ /nix/store/***-gcc-15.2.0/include/c++/15.2.0/atomic
......... /nix/store/***-gcc-15.2.0/include/c++/15.2.0/cstdint

But in cross builds I get this:

..... /nix/store/***-qtbase-aarch64-unknown-linux-gnu-5.15.18-dev/include/QtCore/qatomic.h
...... /nix/store/***-qtbase-aarch64-unknown-linux-gnu-5.15.18-dev/include/QtCore/qbasicatomic.h
....... /nix/store/***-qtbase-aarch64-unknown-linux-gnu-5.15.18-dev/include/QtCore/qatomic_bootstrap.h
........ /nix/store/***-qtbase-aarch64-unknown-linux-gnu-5.15.18-dev/include/QtCore/qgenericatomic.h
......... /nix/store/***-qtbase-aarch64-unknown-linux-gnu-5.15.18-dev/include/QtCore/qtypeinfo.h

Looks like qatomic_cxx11.h is disabled because of QT_BOOTSTRAPPED
https://github.com/qt/qtbase/blob/49adb85d34918034e0d6a4c23817407103fb9f73/src/corelib/thread/qbasicatomic.h#L46-L53

And the bootstrap build is forced by cross compilation:
https://github.com/qt/qtbase/blob/49adb85d34918034e0d6a4c23817407103fb9f73/mkspecs/features/qt_build_config.prf#L85-L86

NB: this is not what I'm doing with bootstrapBuild flag for qtbase package definition.

And maybe we should try and disable this cross_compile --> force_bootstrap dependency at least for some platforms?

Re: splits and merges. I'm also removing a couple of unused dependencies in the big qtwebengine commit, so I don't see why the small qtdeclarative commit should be treated differently.

Re: splits and merges. I'm also removing a couple of unused dependencies in the big qtwebengine commit, so I don't see why the small qtdeclarative commit should be treated differently.

The logic behind the commit separation suggestions is as under discussion here:

• CONTRIBUTING: Clarify commit separation convention #400934

The general rule of thumb I'm following is: Changes that are:

1. Sensible and improve (even minor) details
2. Don't cause a build failure

Should be split to different commits. In this case, the Nixpkgs tree is fully using GCC 15 now, and there is no point in reverting a commit that makes the build fail.

However I did not notice the qtwebengine also mixes changes that are not necessarily related to each other, so now that you are saying that, I think they too should be decoupled, to:

qt5.qtwebengine: apply python 3.12 patch like all other patches
qt5.qtwebengine: fix build with Gcc 15 (reverts 6e27bc915b374bfdc7cabe75018f16b6c623fa91)
qt5.qtwebengine: propagate less buildInputs
qt5.qtwebengine: use more system libraries

Looks like qatomic_cxx11.h is disabled because of QT_BOOTSTRAPPED qt/qtbase@49adb85/src/corelib/thread/qbasicatomic.h#L46-L53

And the bootstrap build is forced by cross compilation: qt/qtbase@49adb85/mkspecs/features/qt_build_config.prf#L85-L86

NB: this is not what I'm doing with bootstrapBuild flag for qtbase package definition.

And maybe we should try and disable this cross_compile --> force_bootstrap dependency at least for some platforms?

TBH I haven't managed to follow you there. I trust you have made the sensible changes, but maybe it would have been nice to explain it better in the commit message.

@eryngion would you like me to help you with this rebase?

Err, if you want a quicker merge, split into different PRs is more acceptable

Here are the 2 first commits of this PR, targeting master and should be very easy to agree upon and merge, because they cause no rebuilds:

• qt5: simplify package functions' definitions and calls #492070

Of course @eryngion is the committer there :).

The qt5.qtweb{engine,kit} commits are irrelevant as these attributes have meta.knownVulnerabilities so I'm not cherry picking these commits.

Here is a qt5.qtbase commit sourced in this PR, but turned out to be simpler:

• qt5.qtbase: fix mysql support for cross builds #492858

Afterwards, the only remaining commit is:

• e04ba27

Afterwards I'll try to open separate PRs for changes from #267311 .