Resolve "TestEnv#set_repo_refs fails to update refs sometimes"

What does this MR do?

This merge request solves an error that would happen in development when a new branch/commit/ref was added to gitlab-test but the bare repos under tmp/tests/gitlab-test_bare/tmp/tests/gitlab-test-fork_bare wouldn't get updated when git fetch was run.

The bare repos are removed the first time TestEnv#set_repo_refs fails, so that newly cloned bare repos have the correct refs.

Are there points in the code the reviewer needs to double check?

The branch will be profiled in Redash, in any case.

Why was this MR needed?

Because it was annoying to ask people to rm -rf tmp/tests a few times per day! :)

What are the relevant issue numbers?

Closes #31562 (closed)

Failures look legit...

added 98 commits

• 5989d91a...e3fa5275 - 95 commits from branch master
• f8674ac3 - Factorize TestEvent#copy_repo
• 85777341 - If TestEnv#set_repo_refs fails, clean bare repos and retry once
• b716ef22 - Shorten the last two commit SHA, because CONSISTENCY

Compare with previous version

added 259 commits

• b716ef22...17806d76 - 256 commits from branch master
• 2a767003 - Factorize TestEvent#copy_repo
• 5adb1d98 - If TestEnv#set_repo_refs fails, clean test repos and retry
• 322c4481 - Shorten the last two commit SHA, because CONSISTENCY

Compare with previous version

added 23 commits

• 322c4481...bcc5135b - 20 commits from branch master
• 5b228615 - Factorize TestEvent#copy_repo
• ac6c4af0 - If TestEnv#set_repo_refs fails, clean test repos and retry
• 590f5a69 - Shorten the last two commit SHA, because CONSISTENCY

Compare with previous version

@rspeicher I've implemented a simpler approach as the first MR for this issue: the test repos are removed and recreated if refs cannot be updated (this is done only once of course).

assigned to @rspeicher

@rymai Woops, I did the review backwards. Anyway, pending changes from https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/1905.

assigned to @rymai

added 66 commits

• 590f5a69...a5f06563 - 63 commits from branch master
• eef0ff32 - Factorize TestEvent#copy_repo
• 21df200b - If TestEnv#set_repo_refs fails, clean test repos and retry
• 4f4167f0 - Shorten the last two commit SHA, because CONSISTENCY

Compare with previous version

@rspeicher Should be better now, thanks!!

assigned to @rspeicher

No pipeline for the last commit for some reason, created one manually.

approved this merge request

enabled an automatic merge when the pipeline for 4f4167f0 succeeds

added 93 commits

• 4f4167f0...a67bb1f0 - 90 commits from branch master
• 628cc848 - Factorize TestEvent#copy_repo
• 7e56ac8f - If TestEnv#set_repo_refs fails, clean test repos and retry
• ba441bb6 - Shorten the last two commit SHA, because CONSISTENCY

Compare with previous version

@rspeicher There was a Rubocop offense, I've also rebased!

@rymai Weird, why didn't the EE counterpart have the same offense?

approved this merge request

merged

mentioned in commit c9c4d537

Weird, why didn't the EE counterpart have the same offense?

Because I fixed it there... :)

mentioned in merge request !11939 (merged)