Stop failing CI upon codecov error 503

[A5rocks]

#3172 took 5 attempts to merge.......

cc @jakkdl

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00000%. Comparing base (3de7996) to head (4d1d54c).

Additional details and impacted files

@@               Coverage Diff               @@
##                 main        #3173   +/-   ##
===============================================
  Coverage   100.00000%   100.00000%           
===============================================
  Files             124          124           
  Lines           18458        18458           
  Branches         1215         1215           
===============================================
  Hits            18458        18458           

[webknjaz]

This would make Codecov less reliable. It'd be better to figure out the version bump.

[CoolCat467]

While I agree that it's unfortunate that this would make Codecov less reliable, I think stopping pull requests from going through just because Codecov is buggy is a worse in the end.

[A5rocks]

This would make Codecov less reliable. It'd be better to figure out the version bump.

Does a new uploader really decrease the amount of 503s codecov returns? I note that the failures I checked were due to that, not due to some uploader issue.

[webknjaz]

Yes, in my experience, v4 eliminated this problem. I think the retries are better, but also the API is new. I don't really have the stats for v5, though.

The only problem is that my attempt to bump to v5 has some weird coverage computation issue: #3161. I haven't really looked into that. But it's weird that I haven't seen this behavior elsewhere, so my hunch is that it might be related to the structure of the automation in this repo.

[A5rocks]

Ok should we merge this as a temporary patch (and a comment saying as much), until someone gets the time to look at v5?

I'm willing to wait a few days just in case we do it quickly but after that I think it's better to be able to get PRs merged on the first attempt. (PR statuses are fine because individual jobs can be rerun, but merge queue needs to rerun every job)

[jakkdl]

I think we should merge this as a temp fix due to how massively painful it is to work with.

It's def suboptimal, but /shrug