When a snippet is being created, it inserts a record in the
database and then performs a Gitaly call to create Git data.

The Gitaly call performs HTTP request for permission check and
if replication lag takes place, those HTTP calls may try fetching
data from a replica that is not up-to-date yet.

This commit:

- Sticks to primary before performing the Gitaly call
- Make the internal API requests aware of the sticking

Changelog: fixed

Git considers specific characters part of the author and committer
signature as "cruft" that is to be stripped or disallowed. The rules of
what is considered a cruft character has changed in Git v2.42.0 though
to not include the dot (".") anymore. One test for snippet repositories
still relies on that behaviour though.

Adapt the test to a different character that Git still considers as
cruft.

Fix rubocop offenses

Changelog: other

`multi_action` is not very specific. The RPC that this method uses is
called UserCommitFiles so it makes sense to name the method after this.

When retrieving the default branch of an empty repo, we skip calling
root_ref. This saves us 1 Redis and 1 Gitaly call.

Changelog: performance

Francisco Javier López authored 4 years ago and Jan Provaznik committed 4 years ago

In this commit, we add snippets to GitLab backups. Well, not
exactly snippets but the snippet repositories which weren't
added before.

Markus Koller authored 4 years ago and Francisco Javier López committed 4 years ago

Also add a `Snippet.default_branch` helper with a fallback to the
currently hardcoded 'master'.

- comply to style guide.

The plan for RSpec 4.0 is to disable monkey patching, reference:
https://rspec.info/blog/2013/07/the-plan-for-rspec-3/#zero-monkey-patching-mode
https://relishapp.com/rspec/rspec-core/docs/configuration/zero-monkey-patching-mode

This commit stops using RSpec monkey patching in several specs.

When the user has invalid info like invalid chars
in the name or the email, git operation will fail.

These errors will be raised by libgit2 not us. We
need to capture them and use a reliable user
to perform the migration for those snippets, like
the migration bot.

This commit introduces the background migration
that will be run to create the repository on
every snippet and create as well the file
according to the snippet content.

When the snippet has a repository attached,
the export file will create a bundle per each
one.

When we restore the export file, if the snippet has
a repository, we will import it from the bundle.
If not, we will create a repository from the
snippet file_name and content columns.

In the same operation where the snippet and
its repository is created, commit the file
into it.

We need to create the repository model
for snippets. It will be the same for both
project and personal ones.

This commit also refactors some code in
order to reuse it in projects and snippets