Fix creating references containing '@'

[tobiashenkel]

References containing the '@' char are legal in git but fail with
[1]. This fails because as soon as '@' is encountered we stop
searching for a ref name, assume it is there and fail while retrieving
it. This can be fixed by catching this exception and continue with the
search for a valid ref.

[1] Trace:
Traceback (most recent call last):
  File "/opt/zuul/lib/python3.7/site-packages/zuul/merger/merger.py", line 108, in __init__
   self._ensure_cloned()
  File "/opt/zuul/lib/python3.7/site-packages/zuul/merger/merger.py", line 178, in _ensure_cloned
    repo.create_head(ref.remote_head, ref, force=True)
  File "/opt/zuul/lib/python3.7/site-packages/git/repo/base.py", line 373, in create_head
    return Head.create(self, path, commit, force, logmsg)
  File "/opt/zuul/lib/python3.7/site-packages/git/refs/symbolic.py", line 546, in create
    return cls._create(repo, path, cls._resolve_ref_on_create, reference, force, logmsg)
  File "/opt/zuul/lib/python3.7/site-packages/git/refs/symbolic.py", line 497, in _create
    target = repo.rev_parse(str(reference))
  File "/opt/zuul/lib/python3.7/site-packages/git/repo/fun.py", line 211, in rev_parse
    ref = name_to_object(repo, rev[:start], return_ref=True)
  File "/opt/zuul/lib/python3.7/site-packages/git/repo/fun.py", line 142, in name_to_object
    raise BadObject("Couldn't find reference named %r" % name)
gitdb.exc.BadObject: <unprintable BadObject object>

[Byron]

Thanks a lot for the contribution! Do you think it's possible to have another look?
Also I am happy to hear your thoughts, as I am (in the review) expressing a preference, not a hard requirement.

[Byron]

A better solution might be to stop treating @ as special outside of the spots that it is allowed in.
I am a hesitant to let it make assumptions unless there absolutely is no other way. Or in other words, I believe the git or C implementation is deterministic here

[tobiashenkel]

Thanks for your feedback, I'll try it when I have time in the next weeks.

[brondsem]

Ping - any updates?

The changes here work for me and the situation I've run into this with. It would be nice to see this (or a better solution) merged.

[Byron]

Having had a look at it again, I still feel uneasy about making an assumption warranting this comment:

                  # If this doesn't result in an object we most likely
                  # encounter a reference containing the '@' character.
                  # In this case we just to continue the search.

I wonder if it's safe that way, or if it leaves us open to causing unexpected breakage.

Maybe @Harmon758 can provide their opinion on this - generally this PR seems like a very desirable fix.

[Byron]

As part of spring cleaning, I am closing this PR as it was opened more than half a year ago while still being open due to unresolved or remarks.

If this is a mistake or any other action should rather be taken, please let me know in the comments or create a new PR.
Thanks for your understanding and the work that undoubtedly went into this.