[adrpo]

Fixes #90.

A short outage on one git server (a 504 from the gateway in front of it) made "Update Package Index" fail for every library, and the error didn't say which repository it was.

Changes in getgitrepo (ompackagemanager/updateinfo.py):

• Retries: fetch and clone are tried 3 times, waiting 30 s and then 60 s. Before each clone retry, a half-written clone left by the failed attempt is removed.
• Cached clone: if the fetch still fails, the cached clone is used and a warning is printed. That library keeps its index entries from the previous run, and the other libraries are updated normally. When the caller needs a specific commit (the checkout in main) and the cached clone doesn't have it, the call still fails.
• Which repository: the retry lines, the warning and the final error now include the URL and the cache path. The end of the run lists every repository whose cached clone was used:

  1 repositories could not be fetched, their libraries are as of the previous run:
    https://gitlab.pam-retd.fr/...
      Fetching ... failed after 3 attempts: unexpected http status code: 504
  

Tests: tests/test_git_retry.py builds local repositories and mocks the fetch to fail with the 504. It covers:

• a fetch that succeeds after retries;
• a server that stays down, so the cached clone is used and listed in the summary;
• a server that stays down while the needed commit isn't cached, so the call fails with an error naming the URL and the commit;
• a failed clone that leaves a partial directory and is then retried from scratch.

Unrelated, but it affects the same host: #88 moves TAeZoSysPro from gitlab.pam-retd.fr to GitHub.

---

🤖 Generated with Claude Code