fscache: 3 tests in t7063-status-untracked-cache.sh fail with fscache enabled

[benpeart]

With the recent addition of the GIT_TEST_FSCACHE=true environment variable, running the test suite with the fscache turned on causes 3 tests in t7063-status-untracked-cache.sh to fail.

Test Summary Report

./t7063-status-untracked-cache.sh (Wstat: 256 Tests: 50 Failed: 3)
Failed tests: 29-31
Non-zero exit status: 1
Files=839, Tests=19846, 1937 wallclock secs ( 9.70 usr 17.19 sys + 1362.22 cusr 4527.29 csys = 5916.40 CPU)
Result: FAIL

• [x ] I was not able to find an open or closed issue matching what I'm seeing

Setup

• Which version of Git for Windows are you using? Is it 32-bit or 64-bit?

$ git --version --build-options

git version 2.18.0.windows.1
cpu: x86_64
built from commit: cd1a74fc9dc8a07626c216940db9a51f25206e03
sizeof-long: 4

• Which version of Windows are you running? Vista, 7, 8, 10? Is it 32-bit or 64-bit?

$ cmd.exe /c ver

Microsoft Windows [Version 10.0.17763.55]

• What options did you set as part of the installation? Or did you choose the
  defaults?

NA

• Any other interesting things about your environment that might be related
  to the issue you're seeing?

NA

Details

• Which terminal/shell are you running Git from? e.g Bash/CMD/PowerShell/other

Bash

• What commands did you run to trigger this issue? If you can provide a
  Minimal, Complete, and Verifiable example
  this will help us understand the issue.

GIT_TRACE_FSCACHE=true sh t7063-status-untracked-cache.sh

• What did you expect to occur after running these commands?

The tests to pass

• What actually happened instead?

failed 3 among 50 test(s)

• If the problem was occurring with a specific repository, can you provide the
  URL to that repository to help us with testing?

NA

[benpeart]

I've looked at this a little but still don't have a root cause. When I run the steps manually, they succeed which makes me wonder if it is timing related. Another area to investigate is the fact that the untracked cache is explicitly invalidating a directory but the fscache is not impacted by that. Given the error is that the new/modified directory doesn't show up in the fscache, it makes me wonder if those two are somehow related.

[dscho]

That's really curious. It seems that the untracked cache has some bugs. What I see in wt_status_collect_untracked() is that the first time I call git status --porcelain, the following entries come in from the untracked cache:

.gitignore
done/five
dtwo/
dtwo/

Note not only the absence of done/sub/ but the double entry dtwo/.

Now, if I run the command immediately after that, there is an additional entry:

.gitignore
done/five
done/sub/
dtwo/
dtwo/

Note that now there is an entry for done/sub/.

Even more curious: if I run this without FSCache, the first time it looks like this (i.e. like the second time with FSCache):

.gitignore
done/five
done/sub/
dtwo/
dtwo/

and the second time it looks like this:

.gitignore
done/five
done/sub/
done/sub/
dtwo/
dtwo/

(Note the double done/sub/ entry).

There seems to be something completely wrong with the untracked cache, and I think the next step should be to see whether the same symptoms happen on Linux (i.e. whether this is anything Windows-specific). I will try to find the time to investigate further next week, unless you beat me to it.

[dscho]

So there is now also t7300.51 that fails with FSCache, strangely with an "Illegal instruction"...

[dscho]

It would be so grand if a volunteer showed up and magically took care of this...