Please describe the problem.
I have a repository that contains thousands (30000) of unlocked files. I see the problem that
git keeps refreshing its index despite the modified time did not change compared to the git index.
I think this should not happen. It significantly slows down (delay of ~10s in my case, but
potentially worse for Windows or NFS file systems) any git operation, and this includes
the display of a bash prompt that shows git information (via PROMPT_COMMAND env var.) and bash
completion after things like git add, therefore it is quite annoying.
What steps will reproduce the problem?
I have set up a repository to reproduce the problem. However, it works as expected. I am looking for ideas on how to find the problem with my repository.
git init .
git annex init
git annex config --set annex.addunlocked include="py_env/*"
python -m venv py_env
. py_env/bin/activate
pip install datalad # some package with dependencies that creates many files in py_env
git annex add py_env
git status
git status
git status
...
The only strange thing that I have found is that git annex will forward all __init__.py files to the smudge/clean-filter (pipes to filter-process), but only those! It does not do this for all other files in py_env, as you can check with
strace git status 2>&1 | grep "^write.*pathname="
this shows only the __init__.py files, despite find py_env shows many more files in there and they are skipped. I think that git skips files during git status if the mtime and the ctime are the same. But they are always the same, as can be checked by comparing
git ls-files --debug "(any file that shows up in strace)"
stat -c $'ctime: %.9Z\nmtime: %.9Y' "(same file)"
I do not know how to debug further and are grateful for suggestions!
What version of git-annex are you using? On what operating system?
git-annex version: 10.20250115-ga55e1da1aa0eb07b418e899cc200027e082eb82a git version 2.47.1 Rocky Linux