Fix bug in tracking w/ project.refs

[AdrianVovk]

Fixes #1851

[gtristan]

Hi,

Thanks for this fix !

Can we please put this new test somewhere else instead of a new file named regression.py ? (a large portion of our tests stem from regressions, yet we do not name them as such ;-) ).

In this case, the issue itself is that local sources do not store refs, and as such they leave gaps in the project.refs, causing IndexErrors to occur when tracking.

I think it's safe to say that this test is related to storing refs in project.refs when there is a lacking ref, as such I think it would be good to have this test in tests/project/format.py where we do preliminary testing on project.refs, and perhaps we could name the test test_project_refs_gap() and explain in a comment that this tests a gap in the project.refs file due to the local source not producing a ref.

Noting the issue in the comments around the added test case is also appreciated :)

[gtristan]

@AdrianVovk if you don't want to update this PR, I will grab it and munge your commit (just to rename the test case) and apply this over the weekend.

Hope this is ok with you :)

[AdrianVovk]

Sorry bout the delay :) porting to bst2 is taking a bit more work/time than anticipated...

If I don't get to it by then go right ahead!

[gtristan]

If I don't get to it by then go right ahead!

Thanks !

[AdrianVovk]

Alright I just moved the test. I couldn't find a tests/project/format.py, but I did find a tests/format/project.py. Let me know if that's not the place you had in mind.

[gtristan]

Alright I just moved the test. I couldn't find a tests/project/format.py, but I did find a tests/format/project.py. Let me know if that's not the place you had in mind.

Right, my bad typo :)

Looks perfect thanks ! I'll get CI to run first but I did run the test locally and was satisfied...

[gtristan]

Closing in favor of #1858

This simply failed the formatting checks, reformatted with tox -e format and passes CI now.