fix: fallback to HEAD when remote tag lookup fails

[saschabuehrle]

Bug

#2559 — when git ls-remote --tags --refs fails for a credentialed private template URL, Copier raises immediately while resolving the latest tag, instead of continuing with a valid fallback path.

Fix

get_latest_tag() now catches ProcessExecutionError from git ls-remote and gracefully falls back to HEAD. This preserves copy behavior for templates that are still cloneable with embedded credentials even when remote tag listing is denied.

I also added a focused regression test to ensure get_latest_tag() returns HEAD on ls-remote failures.

Testing

• python3 -m py_compile copier/_vcs.py tests/test_vcs.py
• python3 -m pytest ... could not be executed here because pytest is not installed in this environment.

Happy to address any feedback.

Greetings, saschabuehrle

[pawamoy]

In #2559 both @sisp and @valentinoli confirmed the git ls-remote command worked fine, so I'm not sure to understand why this PR tries to fix get_latest_tag.

The PR body makes me think it was AI-generated, esp.

python3 -m pytest ... could not be executed here because pytest is not installed in this environment.

[pawamoy]

Looking at their profile, they seem to open a lot of PRs in many various repositories, all looking the same, some of them closed because no issue was created before sending the PR or because the PR template was overridden. So, AI slop to me. I'll let you decide what to do @sisp (blocking the user, or something else).

[sisp]

Thanks for the quick sanity check review and researching user activity to get more context, @pawamoy. I'd wait for a reply by @saschabuehrle and give the opportunity to update the PR; blocking should be a last resort to protect against spam. Let's see. 🙂

[pawamoy]

Unfortunately the replies by @saschabuehrle also seem to be AI-generated.

• fix: normalize artist timestamp units for New additions pSpitzner/beets-flask#291 (comment)
• fix: sort variable lists by pressure level to match sorted pl ecmwf/WeatherGenerator#2031 (comment)
• fix: validate max_bytes parameter in ByteReceiveStream.receive methods agronholm/anyio#1089 (comment)
• etc.

[saschabuehrle]

Fair points — re-examining the issue, the root cause isn't in get_latest_tag. Closing this. Sorry for the noise.

Greetings, saschabuehrle