Not-cherry-picked-because: On master, I want to remove nixseparatedebuginfod and replace it with nixseparatedebuginfod2 instead. This is done in https://github.com/NixOS/nixpkgs/pull/452053
We were relying on `composerNoScripts = false` to make sure post-install
command `assets:install` is run. `assets:install` copies assets from
`vendor/` directory into `public/` directory, placing it in appropriate
places.
However, with commit 80bb9aec24 ("kimai: switch to buildComposerProject2
and tag"), we switched to `buildComposerProject2` which has moved
`composer install` step to `composerVendor` derivation. By design,
`composerVendor` ignores anything that happens outside `vendor/`, so
the assets was not copied into final derivation.
So stop relying on `composerNoScripts = false` and run `assets:install`
ourselves in `postInstall` step. A side effect of this is that there is
another post-install step being skipped (`cache:clear`). However we
simply handle caches outside of the derivation (it's handled in the
module), so that's not a problem.
Fixes: https://github.com/NixOS/nixpkgs/issues/442208
(cherry picked from commit 1422ed8801)
This should be a test that will change whenever most likely *all* NixOS
tests change. As such it should use the most "default" configuration
possible. This decreases chances of a change that will affect almost
other NixOS tests to go unnoticed by this test.
With the minimal profile used previously, there could be changes made to
the default behavior of NixOS modules which were explicitly disabled in
this profile. These would not change the simple test, but would change
all other tests.
Removing as many moving parts as possible from this test should make it
the most default it can be. It now contains only pieces that every other
NixOS test has, too: A name, a machine and a testScript. But nothing
fancy.
(cherry picked from commit 6d90bae447)
(cherry picked from commit 8ae6f582ef)
Manually adjusted some mouse click coords in the test, due to different mouse handling between stable and unstable.
Qt5 qtwebengine will be marked insecure, so this would block the entire DE from being usable.
Epiphany seems to work as a replacement until Morph can be built with Qt6.
Content-Hub test will still need adjustment, but keeping it out of this to make future reverting easier.
(cherry picked from commit 5063462bf3)
EOL upstream.
We only have one hardened kernel at the moment now because
LTS == latest available. This situation would've also happened before
the cleanup since 6.13/6.14 were removed in June already[1].
[1] 23b573705d
(cherry picked from commit 510532e9ae)
By the end of the month, I'll leave Flying Circus. Thanks a lot for the
journey together <3
The rootless-test for podman is something I decided to keep since I'm
using parts of the features covered in there myself.
(cherry picked from commit 201cb3e519)
The new `\restrict` migitation creates random keys in the dump file by
default, which breaks a before/after test for the backup module. By
making the restrict key reproducible, the test passes again.
(cherry picked from commit 87e1134406)
prometheus-smokeping-prober was updated to 0.10.0 in #396980
which introduced a new label `tos` in its metrics.
add it to the failing tests to make them match
the expected metric (and pass) again
you could argue that the tests are a bit too greedy with the way
they match metrics, but I actually like it that way
(cherry picked from commit 2103ba2688)
Before this change, the THIRDPARTY_EXTENSIONS_PATH would end up with a
double-slash in the path, which was breaking FreshRSS's is_valid_path
detection.
(cherry picked from commit 637fc36529)
Test for linking had to be removed because now the linking is more
complex and it would take too much mocking to test it.
The test in question was moved to the dashboard codebase that is
closed-source, if that helps in any way.
(cherry picked from commit a667834a5f)
