Submodules have a name in the configuration, but this name is simply
the path at which the submodule was initially checked in (by default
-- the name is totally arbitrary). If a submodule is moved, it
retains its original name, but its path changes. Since we discover
submodules inside trees, we have their path but not necessarily their
name.
Make the submodules() function return the submodule hash indexed by
path rather than name, so that renamed submodules can be looked up.
Signed-off-by: David Turner <novalis@novalis.org>
This adds an ID-less table containing one row per file, per merge request
diff. It has a column for each attribute on Gitlab::Git::Diff that is serialised
currently, with the advantage that we can easily query the attributes of this
new table.
It does not migrate existing data, so we have fallback code when the legacy
st_diffs column is present instead. For a merge request diff to be valid, it
should have at most one of:
* Rows in this new table, with the correct merge_request_diff_id.
* A non-NULL st_diffs column.
It may have neither, if the diff is empty.
This is controlled with the feature flag gitlab_git_diff_size_limit_increase.
Both of these limits were basically picked arbitrarily in the first place;
disabling the feature flag reverts to the old limits.
This solves a problem where commits populated with Gitaly were not equal
to commits populated with Rugged. This is because Gitaly may not return
all fields of a commit for optimizations purposes, which resulted in
false negatives when comparing the same commit (commits with the same
sha) with different sources.
It seems that bad things happen when two gRPC stubs share one gRPC
channel so let's stop doing that. The downside of this is that we
create more gRPC connections; one per stub.
- Previously, we sorted commits by date, which seemed to work okay.
- The one edge case where this failed was when multiple commits have the same
commit date (for example: when a range of commits are cherry picked with a
single command, they all have the same commit date [and different author
dates]).
- Commits with the same commit date would be sorted arbitrarily, and usually
break the network graph.
- This commit solves the problem by both sorting by date, and by sorting
topographically (parents aren't displayed until all their children are
displayed)
- Include review comments from @adamniedzielski
A more detailed explanation is present here:
https://gitlab.com/gitlab-org/gitlab-ce/issues/30973#note_28706230
Robert Speicher
Ahmad Sherif
Sean McGivern
Jacob Vosmaer
Rémy Coutable
David Turner
Sean McGivern
Douwe Maan
Douwe Maan
Bob Van Landuyt
Alejandro Rodríguez
Timothy Andrew
Tiago Botelho