Just an FYI. Today I worked on tracking down the problems that we've
had with GitHub Actions working on a tag (YETUS-966). I think I've got it
licked, but one of the outcomes was that prior to the PR (which is now waiting
for review) tag checkout was slightly non-deterministic. In the _vast_
majority of cases, the CI system + test-patch did the correct thing and
operations were done on the tag. But in some cases, FETCH_HEAD was pointing to
something else entirely and the branch compilation was actually being done on
some other SHA. This bug would have only impacted qbt or test-patch
--empty-patch on tags in real-world scenarios.
Anyway, I'm fairly confident https://github.com/apache/yetus/pull/138
will fix the tag situation while mostly maintaining expected behavior in other
uses. At least based upon my quick tests on Cirrus CI, Circle CI, and GitHub
Actions. Note that this fix will now _overwrite_ any local tags that conflict
with remote tags when --resetrepo is configured.
