- 28 Jan, 2023 4 commits
-
-
Georges Racinet authored
It is good enough to have them in the prebuild cache, and we don't feel the need to have separate ones for go packages, object files etc hence the renaming.
-
Georges Racinet authored
Because of lack of prefix, it was actually rebundling
-
Georges Racinet authored
It is useful in the gitlab-prebuild job only, and takes about 20% of the extracted size (1.4G over 6.4G). This brings in the splitting of the main cache and the more flexible way to use it (YaML repetitions instead of `extends`, pretty much like upstream is doing)
-
Georges Racinet authored
We've precisely used the rubocop job as a first multi-cache attempt
-
- 25 Jan, 2023 1 commit
-
-
Georges Racinet authored
This will help lifting some doubts
-
- 01 Jan, 2023 2 commits
-
-
Georges Racinet authored
Now that we're on Heptapod Runner 0.6.0rc (GitLab 13.10), it should be doable.
-
Georges Racinet authored
It used to be the default, but now the CC Runner for the Heptapod group uses "S" as default.
-
- 29 Dec, 2022 1 commit
-
-
Georges Racinet authored
With the weighted quota system of heptapod-runner#35, this is now possible. Beside giving us hopefully faster pipelines, we expect this to reduce the odds of side effects between tests, which are a pain to debug, even in the favorable scenario when upstream stumbled on them and fixed it in a later version. The chose value, 50, is the currently highest accepted by the coordinator. It does not seem to be customizable.
-
- 14 Dec, 2022 1 commit
-
-
Georges Racinet authored
-
- 13 Dec, 2022 1 commit
-
-
Georges Racinet authored
Closes #704
-
- 06 Dec, 2022 2 commits
-
-
Georges Racinet authored
It seemes that the only possible solution is to change directory to where the file resides. Closes #715
-
Georges Racinet authored
Due to the sudden death and subsequent rebirth of stator.octobus.net
-
- 26 Nov, 2022 1 commit
-
-
Georges Racinet authored
This is a manual graft of the upstream commit `5832f0d487099c19fdfaa543c4f9f6e720eb6534`, which has not yet been converted to Mercurial. The related tests failures are rather obscure, as seen in [factories_spec.rb](https://foss.heptapod.net/heptapod/heptapod/-/jobs/865790) and [requests/users_controller_spec.rb](https://foss.heptapod.net/heptapod/heptapod/-/jobs/865806) ``` #<ActiveRecord::RecordInvalid: Validation failed: Primary keyid can't be blank> ``` It is likely that some computations for the id fails, leaving it blank instead of raising a straight error.
-
- 22 Nov, 2022 2 commits
-
-
Georges Racinet authored
-
Georges Racinet authored
-
- 20 Nov, 2022 1 commit
-
-
Georges Racinet authored
-
- 21 Nov, 2022 1 commit
-
-
Georges Racinet authored
-
- 20 Nov, 2022 4 commits
-
-
Georges Racinet authored
-
Georges Racinet authored
-
Georges Racinet authored
The disclosure template is not that new, but it is full of GitLab specific instructions.
-
Georges Racinet authored
This is obviously an area untouched by Heptapod. What happens is that upstream's testing certificate expired yesterday: $ openssl x509 -enddate -in spec/fixtures/clusters/leaf_certificate.pem -noout notAfter=Nov 19 18:50:03 2022 GMT This is not the first time we encounter this. Our disabling will likely conflict with a later quarantine statement by upstream, that they will resolve.
-
- 17 Nov, 2022 3 commits
-
-
Georges Racinet authored
In beffba6470df, the integration test in reconcile_existing_runner_versions_service_spec.rb changed to mock the URL call instead of just methods of `Gitlab::Ci::RunnerRelease`. Unfortunately this crashes if the service test is run after `spec/lib/gitlab/ci/runner_releases_spec.rb` (which mocks the URL defined in application settings), because the `RunnerRelease` class memoizes the settings URL. Solution was to add an `after` block to clear memoization from `runner_releases_spec`
-
Georges Racinet authored
-
Georges Racinet authored
The conflict was in `routes.rb`, simply due to a reindentation because of new enclosing block.
-
- 20 Nov, 2022 1 commit
-
-
Georges Racinet authored
It was nice of this test to protect itself against memoizing stubbed values (`before` block). It would have been better to protect its own iterations and any subsequent test (`after` block) This is identical in current trunk (15.5.0.pre) UPSTREAM
-
- 17 Nov, 2022 10 commits
-
-
Georges Racinet authored
-
Georges Racinet authored
The conflict is upstram hitting the same problem we fixed in c57f61293263, with a simpler solution: it was possible to force the task title.
-
Georges Racinet authored
Automatic merge with subsequent fix amendment. The problem lies with spec/lib/gitlab/github_import/importer/events/cross_referenced_spec.rb, again depending on the value of the FactoryBot sequence (see c57f61293263). This time, the offending spec was not fixed as of v15.2.4, but it got quarantined in upstream trunk with 2605918ec4f4 and fixed in the subsequent da9bac3661ce. Our fixing amendment is to apply the quarantine manually, so that the later fix will be merged smoothly in the future.
-
Georges Racinet authored
-
Georges Racinet authored
-
Georges Racinet authored
Trivial resolution, the conflicts are in deleted issue templates
-
Georges Racinet authored
-
Georges Racinet authored
The conflict is in `_tag.html.haml` and resolution was to port the exclusion of Mercurial projects to the change.
-
Georges Racinet authored
-
Georges Racinet authored
For unknown reasons, the failure is not reproducible in HDK, even with the same knapsack parameters. The title of the task depends on the `:title` FactoryBot sequence, (see the `generate` call in spec/factories/work_items.rb). It is therefore not surprising for it to seldom be 'My title 1'. Given that it is passed within the argument of a shared example group, we cannot use `task.title` directly, and instead have to resort to substitution. Could be reported UPSTREAM
-
- 14 Nov, 2022 5 commits
-
-
Georges Racinet authored
Mostly automatic, except that b2e478c70acc activated the `Performance/ConstantRegexp` cop, and that prompted us to make the constant in the native migrator be the compiled regexp instead of its string representation.
-
Georges Racinet authored
This changes GitLab Shell version and we're following along with Heptapod Shell version.
-
Georges Racinet authored
The conflict happened in lib/api/internal/base.rb because they removed the endpoint defined right after our own specific `/hg_access`.
-
Georges Racinet authored
-
Georges Racinet authored
-