@jtojnar This is problematic for me because gala will fail to compile with this version of mutter. Is there some nix'y way I can work around this until they have a fix?

@worldofpeace You could re-add the old mutter expression as is done with VTE for pantheon-terminal, for example.

You mean this?

Mind just including that in nixpkgs here? I don't know when they'll get around to fixing it, and it's already a lot to build.

Taking a note.

Should we just add the meson upgrade here? Then we can debug it with hydra.

Yeah, I did that, just was pushing to a wrong branch.

pushed some fixes to https://github.com/hedning/nixpkgs/commits/gnome-3.30, marked with fixup! or squash! where appropriate (so it's easy to squash things down).

• pipewire to 0.2.2, I had this laying around, 0.2.3 didn't build without any effort
• epiphany needs libdazzle now. It does build, but fails one test libdazzle builds, but fails one test
• gtksourceview should be updated to 3.24.8

Some more fixes at https://github.com/hedning/nixpkgs/commits/gnome-3.30:

• xdg-desktop-portal: fails due to new pipewire, updated this and flatpak and related packages
• evince now depends on gspell
• sysprof is fixed by updating to 3.30
• gnome-system-monitor require meson
• libgda has patch upstream now

Some notes on other failures:

• tracker: libsoup upgrade is the cause here it seems, build is fixed by reverting (might be good to do as there's a lot of packages that depend on tracker).
• libunique fails on a warning
• libdazzle fails on a test as noted above
• gjs: we need 60 now obviously
• gnome-online-accounts fails to produce the man output (release notes also mentions telepathy no longer being supported) not sure what the problem is

I'm stuck on spidermonkey here, getting this error in the configure phase:

Traceback (most recent call last):
  File "./../../configure.py", line 127, in <module>
    sys.exit(main(sys.argv))
  File "./../../configure.py", line 34, in main
    return config_status(config)
  File "./../../configure.py", line 122, in config_status
    return config_status(args=[], **encode(sanitized_config, encoding))
  File "/build/firefox-60.2.0/python/mozbuild/mozbuild/config_status.py", line 144, in config_status
    definitions = list(definitions)
  File "/build/firefox-60.2.0/python/mozbuild/mozbuild/frontend/emitter.py", line 184, in emit
    objs = list(emitfn(out))
  File "/build/firefox-60.2.0/python/mozbuild/mozbuild/frontend/emitter.py", line 1134, in emit_from_context
    full_path), context)
mozbuild.frontend.reader.SandboxValidationError:
==============================
FATAL ERROR PROCESSING MOZBUILD FILE
==============================

The error occurred while processing the following file or one of the files it includes:

    /build/firefox-60.2.0/js/src/shell/moz.build

The error occurred when validating the result of the execution. The reported error is:

    Path specified in LOCAL_INCLUDES is not allowed: .. (resolved to /build/firefox-60.2.0/js/src)

I suspect the tracker failure is caused by the meson rpath issue (cannot reproduce with libsoup built with autotools).

IIRC libdazzle tests are non-deterministic, I should have disabled them long time ago. Done.

Looking at the GOA diff, I see no reason for man pages stopping being built, fixed.

Reading https://git.archlinux.org/svntogit/packages.git/tree/repos/gnome-unstable-x86_64/PKGBUILD?h=packages/js60, I tried dropping the autoconf and building from a different directory but it did not help. Maybe ask at irc://irc.mozilla.org/jsapi.

I suspect the tracker failure is caused by the meson rpath issue (cannot reproduce with libsoup built with autotools).

Right, any idea how to debug this? Doing a readelf -d doesn't seem to produce anything weird at least (freenode is being weird and requiring registered nicks, which doesn't seem to work with matrix...).

I am not really sure. I am still going through the meson commits.

Looked at spidermonkey some more. The test that's failing is new in 60, commenting it out «works». Running configure from another directory also works, which is what Arch do.

Pushed working spidermonkey to https://github.com/hedning/nixpkgs/commits/gnome-3.30

gjs fails due some path weirdness where it tries to link to a file starting with .//nix/store:

config.status: linking /nix/store/gjxbz4mbyd67vfia9gljwwiiyj6lm4na-gobject-introspection-1.58.0-dev/share/gobject-introspection-1.0/tests/gitestmacros.h to installed-tests/gitestmacros.h
config.status: linking /nix/store/gjxbz4mbyd67vfia9gljwwiiyj6lm4na-gobject-introspection-1.58.0-dev/share/gobject-introspection-1.0/tests/regress.c to installed-tests/regress.c
config.status: linking /nix/store/gjxbz4mbyd67vfia9gljwwiiyj6lm4na-gobject-introspection-1.58.0-dev/share/gobject-introspection-1.0/tests/regress.h to installed-tests/regress.h
config.status: linking .//nix/store/gjxbz4mbyd67vfia9gljwwiiyj6lm4na-gobject-introspection-1.58.0-dev/share/gobject-introspection-1.0/tests/warnlib.c to installed-tests/warnlib.c
config.status: error: .//nix/store/gjxbz4mbyd67vfia9gljwwiiyj6lm4na-gobject-introspection-1.58.0-dev/share/gobject-introspection-1.0/tests/warnlib.c: file not found

Merged. Many thanks @hedning, @worldofpeace, @grahamc and any other people who contributed.

I don't use gnome, but I've been vaguely following this and I think you all deserve a huge thank you for your tenacious work!

I want to echo @michaelpj's sentiment here. I do use gnome, and I really appreciate all of the effort that goes into keeping these packages up to date. This branch has been working well for me as a daily driver for a while, and I'm excited that the changes are merged. Thanks @jtojnar (and other contributors) for the great efforts here.