Read about our upcoming Code of Conduct on this issue

  1. 12 Jan, 2022 4 commits
  2. 11 Jan, 2022 2 commits
  3. 10 Jan, 2022 5 commits
  4. 04 Jan, 2022 1 commit
  5. 26 Dec, 2021 4 commits
    • Georges Racinet's avatar
      Heptapod CI: UPDATE_CACHES variable to force · d256885afb1b
      Georges Racinet authored
      Being able to force with a variable will be useful when
      triggering from the web (in case there is no obvious MR to
      bear the `pipeline:update-caches` label).
      
      Before the switch to regular caches, the `gitlab-prebuilds`
      job could be forced with a variable.
      
      --HG--
      branch : heptapod-stable
      d256885afb1b
    • Georges Racinet's avatar
      Heptapod CI: run workhorse jobs only when needed · 02ad47df331a
      Georges Racinet authored
      This is another case of using `changes:`, close to what
      upstream is doing. The new label will allow to force in cases
      it should have run but didn't. Conversely, if it runs when it
      should have not, there are no other consequences than wasted
      resources.
      
      --HG--
      branch : heptapod-stable
      02ad47df331a
    • Georges Racinet's avatar
      Heptapod CI: skipping lints according to a new label · af96e67c745f
      Georges Racinet authored
      In some cases, it can be useful to disable the lints.
      
      In the present case, we are working on the CI definition itself,
      and especially on the interplay between the cache produced by
      `gitlab-prebuilds` and the subsequent RSpec jobs: the `lint`
      stage is unwanted delay and wasted resources for that.
      
      Of course people must not abuse this.
      
      --HG--
      branch : heptapod-stable
      af96e67c745f
    • Georges Racinet's avatar
      Heptapod CI: running gitlab-prebuilds job only when needed · 14879080f956
      Georges Racinet authored
      For that to work, we need to put in cache the rest of what
      the `gitlab-prebuilds` job uploads as artifacts, hoping that it won't
      prevent actual rebuild in next run of `gitlab-prebuilds`.
      
      When we have multiple caches we will be able to separate neatly what
      is cache for `gitlab-prebuilds` and what is cache for the RSpec jobs.
      
      --HG--
      branch : heptapod-stable
      14879080f956
  6. 25 Dec, 2021 2 commits
  7. 23 Dec, 2021 1 commit
    • Georges Racinet's avatar
      Kubernetes platform tests: fixing failed cert store test · 15f2bd1f962f
      Georges Racinet authored
      The root certificate provided was expired since Dec 8,
      this is the simple reason why the test was failing. The
      "external factors" mentioned in 797142d77389 boil down to
      the current date only.
      
      Fixed by taking more recent certs from current head of GitLab,
      including the more proper naming of `leaf_certificate.pem`.
      Not removing `ca_certificate.pem`: its name was inappropriate
      for the failing test at line 221, but it's used in other places.
      
      Closes #590
      
      --HG--
      branch : heptapod-stable
      15f2bd1f962f
  8. 25 Dec, 2021 1 commit
  9. 23 Dec, 2021 2 commits
    • Georges Racinet's avatar
      Heptapod CI: no more pipelines on pushes · 6cc63c52dd98
      Georges Racinet authored
      They were made only of the `gitlab-prebuilds` job and only upon
      changes of `VERSION`. Nowadays, the `gitlab-prebuilds` is part of
      MR and web pipelines, so that is a pure waste of resources.
      
      Besides, because it was so long in the previous system,
      that meant two versions of `gitlab-prebuilds` running in parallel
      each time there was a MR changing the upstream version
      (with major conflict problems).
      
      --HG--
      branch : heptapod-stable
      6cc63c52dd98
    • Georges Racinet's avatar
      Heptapod CI: trigger extended RSpec scope with MR label · ad8725a45fba
      Georges Racinet authored
      Should have done this a long time ago.
      
      --HG--
      branch : heptapod-stable
      ad8725a45fba
  10. 22 Dec, 2021 2 commits
  11. 21 Dec, 2021 1 commit
    • Georges Racinet's avatar
      Heptapod CI: bundle pristine pg only in the prebuilds job · 78c1fb3b6b3e
      Georges Racinet authored
      In Heptapod, we're always testing a single PostgreSQL version.
      The `gitlab-prebuilds` job needs to run the pristine in case
      the version just changed, but the ones downstream are then
      guaranteed to have a correctly built client library.
      
      --HG--
      branch : heptapod-stable
      78c1fb3b6b3e
  12. 12 Dec, 2021 11 commits
    • Georges Racinet's avatar
      Added tag heptapod-0.27.3 for changeset 797142d77389 · 194c5fd6c8c0
      Georges Racinet authored
      --HG--
      branch : heptapod-0-27
      194c5fd6c8c0
    • Georges Racinet's avatar
      Disabling failing certificate validation test · 797142d77389
      Georges Racinet authored
      Disabling a failing test is always lame, but in this case
      there is strong evidence that it is failing due to a change
      of external factors (it was passing for Heptapod 0.27.0).
      
      The disabling is done to avoid something
      worse: getting used to force releases with failing tests.
      
      Being on a release branch, it won't propagate automatically
      to other Heptapod versions, for which we can hope that a
      subsequent GitLab update will bring in a proper fix. Indeed
      the problem might be due in CI to updates in base CI images
      done for further package installs needed by Heptapod. Later
      GitLab versions may have the same updates in their base CI
      images.
      
      --HG--
      branch : heptapod-0-27
      797142d77389
    • Georges Racinet's avatar
      Setting version for release · 6747f645f9aa
      Georges Racinet authored
      --HG--
      branch : heptapod-0-27
      6747f645f9aa
    • Georges Racinet's avatar
      Merged heptapod-stable branch into its release branch · f1c8c9247111
      Georges Racinet authored
      --HG--
      branch : heptapod-0-27
      f1c8c9247111
    • Georges Racinet's avatar
      Gitlab::Workhorse: dispatch blobs to HGitaly when appropriate · 76c72c819af5
      Georges Racinet authored
      This time, we need a knob to call HGitaly only for fully native
      projects: `GetBlobRequest` works by oid and `send_git_blob` takes
      a Blob object as argument whose id is really a Git oid in the
      HGitaly1 case, and of course a HGitaly oid in the fully native case.
      
      In the HGitaly1 case, it may be frightening that the public side of
      the call can have a Mercurial SHA (navigation trough a commit instead
      of a branch). Yet internally it is correctly resolved to a Git
      blob oid, but that is with hindsight not surprising: it boils down
      to the fact that the HGitaly1 mode works, as this raw download system
      is obviously by far not the only use case for blobs in GitLab.
      
      --HG--
      branch : heptapod-stable
      76c72c819af5
    • Georges Racinet's avatar
      Repository: note for future refactoring · af732fd48e73
      Georges Racinet authored
      --HG--
      branch : heptapod-stable
      af732fd48e73
    • Georges Racinet's avatar
      Heptapod CI/Knapsack: added Gitlab::Workhorse to the regular suite · ae11fddc37a6
      Georges Racinet authored
      We want to be alerted early of API breakage for this pivotal class.
      
      --HG--
      branch : heptapod-stable
      ae11fddc37a6
    • Georges Racinet's avatar
      Gitlab::Workhorse: dispatch raw_patch to HGitaly when appropriate · 31ecaff32011
      Georges Racinet authored
      Also attached to #555, this is the same as the diff problem
      with its sibling RPC: RawPatch.
      
      --HG--
      branch : heptapod-stable
      31ecaff32011
    • Georges Racinet's avatar
      Gitlab::Workhorse: dispatch diffs to HGitaly when appropriate · b3c8155e97b6
      Georges Racinet authored
      This fixes #555 as it was originally reported. The diff
      could not be served on native Mercurial repositories, because
      the given Mercurial hash are unknown to Gitaly.
      
      On the other hand, for non-native repositories, the method is
      called with a Git hash, so must be relayed to Gitaly.
      
      --HG--
      branch : heptapod-stable
      b3c8155e97b6
    • Georges Racinet's avatar
      Gitlab::Workhorse: dispatch to HGitaly only for native Mercurial · a7138c690119
      Georges Racinet authored
      This is not a bug fix. The optional `vcs_dispatch` parameter was
      and still is in use for the `send_git_archive` class method only.
      In the special case of archives, we decided to let HGitaly produce
      them whether the project is native or not. This is now allowed
      by passing `hg_native_only: true`.
      
      The default behavior is now to dispatch according to nativity status,
      which we're about to use for other workhorse proxified responses.
      
      --HG--
      branch : heptapod-stable
      a7138c690119
    • Georges Racinet's avatar
      Gitlab::Workhorse: RSpec tests for archives on Mercurial projects · 6872743721e2
      Georges Racinet authored
      It doesn't test much, just like upstream's test for the Git case,
      but will help with internal refactorings, and we're about to make
      one.
      
      Upstream's test is actually rewrapped in a shared examples group.
      
      --HG--
      branch : heptapod-stable
      6872743721e2
  13. 11 Dec, 2021 1 commit
  14. 08 Dec, 2021 3 commits