odd behavior with conflicting outputs with native_test on windows

[novas0x2a]

So, to start with, I should say I'm not sure what was supposed to happen here. The circumstance that led to this was my accidentally setting the out parameter of native_test to the same value in two different targets (in the same package); I acknowledge that this is not the correct thing to do :). I was surprised that bazel let me do that without erroring immediately?

With that said, this seems to be triggering a race condition that leads to an odd failure on windows, so I figured I'd report it in case this pathological behavior is triggering a different bug.

The reproduction is here: https://github.com/novas0x2a/native-test-repro. The repo script is run.sh. The HEAD as I write this is novas0x2a/native-test-repro@761b23c

• On linux, you can run run.sh and it will loop forever just fine.
• On windows, it appears to race. After a few successful loops, the loop will terminate with:

INFO: Analyzed 4 targets (0 packages loaded, 0 targets configured).
ERROR: C:/<redacted>/native-test-repro/BUILD.bazel:13:12: Testing //:windows-one (run 3 of 5) failed: java.io.IOException: F:/<redacted>/kcgnk7y7/execroot/__main__/_tmp/a2722bcc0b9902c7324c6c5148102354 (Permission denied)
Use --verbose_failures to see the command lines of failed build steps.
INFO: Elapsed time: 0.265s, Critical Path: 0.04s
INFO: 11 processes: 11 internal.
ERROR: Build did NOT complete successfully
//:unix-one                                                             SKIPPED
//:unix-two                                                             SKIPPED
//:windows-one                                                        NO STATUS
//:windows-two                                                        NO STATUS

Executed 0 out of 4 tests: 4 were skipped.

It appears to be racing on the deletion of the TEST_TMPDIR, though I'm not sure about that. Windows doesn't like it when you delete directories that have files in them that another process has open; I assume that's what's happening here.