This will trigger builds with both mobile and desktop UIs

Part-of: !126

Set the same chassis types for master as for all other branches.

I've finally fixed the running out of space problems. Let's enable
plasma-desktop again!

Fixes: issue 136

Instead of the hack to put the apk cache into tmpfs set the new
PMB_APK_NO_CACHE=1 env var, which will cause pmbootstrap to just not use
a cache during builds of images and packages.

Related: issue 136
Related: https://docs.postmarketos.org/pmbootstrap/environment_variables.html#pmb-apk-no-cache

Part-of: postmarketOS/build.postmarketos.org!124

Part-of: !125

Part-of: !125

Pablo Correa Gomez authored 1 month ago and Oliver Smith committed 3 weeks ago

This is now a dependency of postmarketos-base-ui:
postmarketOS/pmaports!6146

Set it to the past, so new images are triggered immediately. Use the
same date as lots of other entries.

Signed-off-by: Jens Reidel <adrian@travitia.xyz>

Signed-off-by: Jens Reidel <adrian@travitia.xyz>

We don't need to test with pmbv2 anymore as the v24.06 release is EOL.

Our package recipes have recently been changed to include a forked
version of Alpine Linux' abuild package. Unfortunately this causes a lot
of BPO tests to fail. Disable them for now to get the testsuite working
again, then go from there.

Related: postmarketOS/build.postmarketos.org#151

Fix that the mirrors.pmaports and mirrors.systemd arguments are set to
none when building images.

Having the same package in the regular and split repo is supported now.

"updated" makes more sense, when this is printed it is already done.

Attempt to fix / get more information for:

Jan 10 14:35:14 bpo python3[21034]: [SQL: INSERT INTO package_dependency (package_id, dependency_id) VALUES (?, ?)]
Jan 10 14:35:14 bpo python3[21034]: [parameters: ((17445, 17434), (17445, 17434), (17445, 17440), (17445, 17455), (17445, 17458), (17445, 17456), (17445, 17461), (17445, 1746

We filter packages with the splitrepo already to find the existing one,
no need to set the splitrepo afterwards.

Run bpo.helpers.pmb.set_repos_task properly per arch, so it can figure
out whether the mirror exists per arch or not.

With this it should only be needed to set PMB_APK_FORCE_MISSING_REPOSITORIES=1
for pmbv2, adjust the code accordingly.

Configure the repositories as "none" if they don't exist yet.

Change from -s to -S to fix:
  argparse.ArgumentError: argument -s: conflicting option string: -s

This test has been adjusted to have a duplicate package in the regular
and split (systemd) repository. This won't work when using pmbv2, and we
don't intend to use it there so skip the test.

GitLab CI currently fails with timeouts even though the tests pass
locally for me. This might be because I added new migrations, or
because the workers are a bit overloaded right now.

As I'm bringing up the systemd splitrepo, there is no real APKINDEX yet
that could be downloaded from the server:

  + wget https://build.postmarketos.org/wip/extra-repos/systemd/master/x86_64/APKINDEX-symlink-repo.tar.gz -O APKINDEX.tar.gz
  Connecting to build.postmarketos.org ([2a01:4f9:2a:231::2]:443)
  wget: server returned error: HTTP/1.1 404 Not Found

Related: issue 144

For v24.06 we briefly need to be backwards compatible to pmbv2 here.

Use log level "error" instead of "critical", as pytest is configured in
this repo to show critical errors during the log output. Triggering
exceptions in testcases is expected (to test such exceptions), and if we
should get an unexpected exception then the test already aborts. With
log level "error" we don't get cluttered output in regular test runs.