`get_object` return value is not checked

[kmsquire]

We're occasionally getting an error,

  File "/usr/local/lib/python3.9/site-packages/gitinfo/gitinfo.py", line 108, in get_git_info
    return get_git_info_dir(gitdir)
  File "/usr/local/lib/python3.9/site-packages/gitinfo/gitinfo.py", line 86, in get_git_info_dir
    r = get_pack_info(idx_file, gi)
  File "/usr/local/lib/python3.9/site-packages/gitinfo/pack_reader.py", line 176, in get_pack_info
    data = get_object(fin, pack_idx)
  File "/usr/local/lib/python3.9/site-packages/gitinfo/pack_reader.py", line 149, in get_object
    data, r = decode_delta(fin, data, delta_res, pack_idx, offset)
  File "/usr/local/lib/python3.9/site-packages/gitinfo/pack_reader.py", line 110, in decode_delta
    r += oo[offset : offset + size]
  TypeError: 'NoneType' object is not subscriptable

This seems to be caused by get_object returning None to decode_delta (in these lines). Unfortunately, this seems to only happen in CI and is not locally reproducible, so I don't have any specific information about the contents of the git repo.

That said, because get_info can return None, calls to it should probably be checked for None and handled appropriately.

[spapas]

Hello this is not easy to debug and also notice that handling None there will result in that error appearing somwhere else ...

The only way get_object can return None is from here: https://github.com/spapas/python-git-info/blob/master/gitinfo/pack_reader.py#LL129C5-L130

So it seems that it tries to get an object that is not a commit (1) or an ofs_delta (6). I thought that the missing object typesd weren't needed but probably I am wrong ... Unfortunately I can't implement them unless I find a repo that has that behavior :(

Is the offending repo open source so I can also test it to see in case it throws the error for me also?

[kmsquire]

Unfortunately, no, it's a private repo. It's also hard to reproduce on our end. Anyway, when I get the chance, I'll try to dig in and see if I can figure anything out.

[spapas]

I'll try to check out some big repos to see if I also encounter the problem.

Also, another thing you may want to check is to create a clean clone of the problematic repo (I guess that's what the CI does) to see if you can reproduce the problematic behavior?