encoding: improve UTF-8 decode strategy

[ralian]

Work related to #1451

[ralian]

I did realize I ran ruff with my settings from another project. I'll have to amend this when I get a chance, but let me know if this change looks ok otherwise.

[jdavid]

Thanks @ralian

PEP-383 says:

File names, environment variables, and command line arguments are defined as being character data in POSIX...

In pygit2 we don't have vars or args I think, so we only care about paths.

maybe_string/to_bytes are general helpers used not only for filesystem paths. I would rather add 2 new helpers: for example decode_fs_path/encode_fs_path; then the work will be to go through the code and use the correct helper (if in doubt keep the old helper call and flag it for discussion).

Also, there is a reference leak in Repository_status :

-        err = PyDict_SetItemString(dict, path, status);
+        PyObject *py_path = PyUnicode_DecodeFSDefault(path);
+        if (py_path == NULL)
+            goto error;
+
+        err = PyDict_SetItem(dict, py_path, status);
+        Py_DECREF(py_path);
         Py_CLEAR(status);

If py_path is NULL then status is not cleared.

[ralian]

Hi David,

Thanks for the feedback. The new helper functions are a good idea, this weekend I will take a crack at that and fix the half broken test too (I have it fixed locally.)

For the reference leak good point. I don't like using goto for this exact reason, but I was trying to match the style of the rest of the codebase. I will think about a better way to handle this, but worst case I can just clear the status in that branch before the goto.

Will