Prevent deadlock in pull_service.GetDiverging(pr)

[6543]

fix #10117 by using a temporary repo to compare

[lafriks]

Could you split cache to different PR, so that first part could be backported to 1.11.x?

[6543]

@lafriks #9784 introduced function GetDiverging() so there is no need for backport

[lafriks]

You are right, sorry 🤦‍♂️

[codecov-io]

Codecov Report

Merging #10905 into master will increase coverage by 0.01%.
The diff coverage is 44.76%.

@@            Coverage Diff             @@
##           master   #10905      +/-   ##
==========================================
+ Coverage   43.43%   43.45%   +0.01%     
==========================================
  Files         593      595       +2     
  Lines       83286    83467     +181     
==========================================
+ Hits        36178    36267      +89     
- Misses      42620    42717      +97     
+ Partials     4488     4483       -5     

Impacted Files	Coverage Δ
models/error.go	30.50% <0.00%> (ø)
models/issue.go	51.17% <0.00%> (-0.25%)	⬇️
models/issue_label.go	61.60% <0.00%> (-1.45%)	⬇️
models/migrations/migrations.go	4.16% <ø> (ø)
models/migrations/v128.go	0.00% <0.00%> (ø)
models/migrations/v134.go	0.00% <0.00%> (ø)
models/migrations/v135.go	0.00% <0.00%> (ø)
modules/private/manager.go	0.00% <0.00%> (ø)
routers/repo/compare.go	40.80% <ø> (-0.14%)	⬇️
routers/repo/pull_review.go	0.00% <0.00%> (ø)
... and 21 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 36d9237...1b92b81. Read the comment docs.

[zeripath]

Hmm... We should be using a temporary repo with shared object DBs to generate these divergences if we genuinely need to look between two repos. I don't like this add remote thing - it's bound to catch us out again - not least in that I think it could cause a repack of git references in the base repo to contain head git references which are later broken. (although we get a bit of that with pull references.)

However do we really need to do that? I don't remember when divergence is shown - if it's only down for actual PRs then can't we just use the refs/pull/ head to generate the divergence as they're then in the same repo.

[6543]

@zeripath do you have time to review :D

[6543]

@guillep2k I'll split things up ...

[6543]

@lafriks in the end it got split up 😅 ...

Cache PR Divergence -> #10914

@guillep2k ready to review

[zeripath]

may not have deleted things properly

[6543]

ping lgtm