----- Original Message -----
From: Jeff King
Date: 7/16/2009 3:42 AM
Makes sense, though I can't help but wonder what would happen with a
filesystem that did more than just case (like the utf8 normalization
that happens on HFS).
Should we actually be converting the filesystem names into a canonical
format as they are read? IIRC, Linus posted some patches a few weeks ago
about "git path" versus "filesystem path", but I didn't actually look
too closely.
I'm game for whatever. Git actually has a lot of places where it
doesn't pay attention to core.ignorecase, and having a standard and
correct method of comparing filenames would make it easier to handle
core.ignorecase=true in a more global fashion.
If your patch is the right route, it might be nice to collapse the
comparison into its own function. You end up cutting and pasting a lot
of the related conditionals and returns (like above, where 2 lines
become 9), so it might make sense to do something like:
int filename_cmp(const char *a, const char *b, int ignore_case)
{
return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
}
and then just s/strcmp/filename_cmp/ at the appropriate callsites.
I started off with this method, but it required two functions, one with
the strcmp() and one for strncmp(). In fact, in other places in the
code, Git uses memcmp() for comparison. Is that, then, three filename
comparison functions, dependent upon intent? At that point, it felt
like my change wasn't as self contained anymore, so I then wrote what I
posted to the list to get feedback.
I'm hoping someone will offer the most correct method to do this, as I
have a number of patches forthcoming to handle core.ignorecase=true in
other areas. The next one covers 'git status' and its reporting of
'missing' directories due to case differences.
Thanks!
Josh
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html