[Commented On] D8733: overlayworkingctx: remove unused `nofilechanges()` and `_compact()` methods

[martinvonz (Martin von Zweigbergk)]

martinvonz added inline comments.

INLINE COMMENTS

> martinvonz wrote in context.py:2549
> It seems that this was indeed a problem -- one of Google's internal tests failed when we import D8732 <https://phab.mercurial-scm.org/D8732>. The test case passes again if I add back a call to `_compact()` before we call `overlayworkingctx.files()` (though I'm not sure if that's the best place to call it. I'll try to add a test case that reproduces the problem without our internal extensions. Then I'll most likely roll back this patch and add back a call to `_compact()` somewhere.

Actually, running the test suite as I suggested comes up with examples we already had. I ran the test suite before your patch with `--extra-config-opt rebase.experimental.inmemory=1` finds 43 failing tests. I then re-ran those with `-j1` to get stable output and sent stderr to a file. Then I ran the same 43 tests in the same way after your patch. The diff of the outputs shows that `test-rebase-named-branches.t` and `test-rebase-parameters.t` change behavior. If you add the `self._compact()` call in `files()`, then change in `test-rebase-parameters.t` goes away. I haven't analyzed it further, but that seems to suggest that the change in `test-rebase-named-branches.t` might be a correct fix, but the change in `test-rebase-parameters.t` is undesirable.

REPOSITORY
  rHG Mercurial

CHANGES SINCE LAST ACTION
  https://phab.mercurial-scm.org/D8733/new/

REVISION DETAIL
  https://phab.mercurial-scm.org/D8733

To: mjacob, #hg-reviewers
Cc: martinvonz, mercurial-patches
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mercurial-scm.org/pipermail/mercurial-patches/attachments/20200727/3bd9ee06/attachment-0002.html>