1. 01 Apr, 2021 3 commits
  2. 19 Mar, 2021 2 commits
  3. 12 Mar, 2021 3 commits
    • Georges Racinet's avatar
      Setting version for release · 79d8f8a8df49
      Georges Racinet authored
    • Georges Racinet's avatar
      RepositoryService.WriteRef: special ref target can be any revision · 28d8fc467b7a
      Georges Racinet authored
      Found out about this while testing with the Rails app. Of course,
      it is natural for symbolic refs, and doesn't sound very logical
      with normal refs, a result of under-specification ultimately due
      to the fact that Gitaly simply uses `git update-ref`.
      We keep refusing ref target revisions for keep-arounds because
      we are much surer that the Rails app will only use full SHAs as
      target revisions. Also, it swallows exceptions, avoiding failures
      if `WriteRef` refuses to create the keep-around
      (see `Gitlab::Git::KeepAround`).
      We should have a Gitaly comparison test for this, but we already
      have RSpec tests in the Rails app, so this can be in a follow-up.
    • Georges Racinet's avatar
      Bumping future version · 5cc50c9c21ca
      Georges Racinet authored
      The work to support special refs is well worth a minor version
  4. 11 Mar, 2021 4 commits
  5. 09 Mar, 2021 1 commit
  6. 10 Mar, 2021 1 commit
  7. 03 Mar, 2021 3 commits
  8. 10 Mar, 2021 1 commit
  9. 03 Mar, 2021 1 commit
  10. 11 Mar, 2021 1 commit
  11. 09 Mar, 2021 4 commits
  12. 06 Mar, 2021 3 commits
  13. 04 Mar, 2021 2 commits
    • Georges Racinet's avatar
      DiffService: accepting EMPTY_TREE_OID to mean no commit · b6984e1331b3
      Georges Racinet authored
      This is the convention used by the Rails application, for
      instance in
            def diff_from_parent_request_params(commit, options = {})
              parent_id = commit.parent_ids.first || Gitlab::Git::EMPTY_TREE_ID
              diff_between_commits_request_params(parent_id, commit.id, options)
      Without such a special case, HGitaly was returning an error.
      This change gets at least one of the Heptapod functional tests to pass.
      Testing shows that Gitaly also accepts the right commit id to be
      EMPTY_TREE_OID to express the absence of commit. This is probably
      less useful, but we're also implementing it for the sake of
    • Georges Racinet's avatar
      DiffService: better logging of errors from common resolving code · 71d814317a3c
      Georges Racinet authored
      Since `parse_diff_request` is used by several RPCs, it is useful
      for the error logging to include indication of which one was
      called. Using the request class name does the trick.
  14. 05 Mar, 2021 1 commit
  15. 03 Mar, 2021 2 commits
  16. 24 Feb, 2021 1 commit
    • Georges Racinet's avatar
      CommitService: implementation for ListLastCommitsForTree · 9a83d1db67d2
      Georges Racinet authored
      It can probably be improved depending on ordering promises
      of the manifest, and the possibility to use treemanifest at
      some point.
      This is still partial, because `path` can really be any
      Git pathspec, and we don't honour the "literal pathspecs"
      option (the current one is in `GlobalOptions`).
      Closes #14
  17. 01 Mar, 2021 2 commits
    • Georges Racinet's avatar
      message.commit_for_tree · 442a6329c053
      Georges Racinet authored
      This is a shortcut, simpler than the other ones in `message`
      but the fully qualified name,
      makes hard to respect the 80 columns mark.
    • Georges Racinet's avatar
      manifest: new module and classes for manifest mining · f774384f4849
      Georges Racinet authored
      Some repeated patterns of data extraction have been spotted
      in the current ongoing experiments. That's not surprising:
      Gitaly is after all modelled on Git responses.
  18. 24 Feb, 2021 4 commits
  19. 27 Feb, 2021 1 commit