Re: [PATCH/RFC][GSoC] diff-no-index: transform "$directory $file" args to "$directory/$file $file"

[Junio C Hamano <gitster@pobox.com>]

Yurii Shevtsov <ungetch@gmail.com> writes:

> diff --git a/diff-no-index.c b/diff-no-index.c
> index 265709b..9a3439a 100644
> --- a/diff-no-index.c
> +++ b/diff-no-index.c
> @@ -97,8 +97,39 @@ static int queue_diff(struct diff_options *o,
>      if (get_mode(name1, &mode1) || get_mode(name2, &mode2))
>          return -1;
>
> -    if (mode1 && mode2 && S_ISDIR(mode1) != S_ISDIR(mode2))
> -        return error("file/directory conflict: %s, %s", name1, name2);
> +    if (mode1 && mode2 && S_ISDIR(mode1) != S_ISDIR(mode2)) {
> +        struct strbuf path;
> +        const char *dir, *file;
> +        char *filename, *dirname = 0;
> +        int i, ret = 0;

If you have two directories, a and b, under which there are two
files a/sub/file and b/sub (i.e. 'sub' in a/ is a directory and b/
is a file), and if you say "git diff --no-index a b", what happens?

 - the caller of this function gives a and b in name1 and name2;

 - we do not come in this codepath as both are directories;

 - we read from a/ and b/ and fill p1 and p2 with names of paths in
   the directories -- p1 and p2 will both have 'sub';

 - queue_diff() is recursively called to compare a/sub and b/sub;

   - now we have name1 = a/sub and name2 = b/sub;

   - we come in this codepath, and they are turned into comparison
     between a/sub/sub and b/sub.

The last step is simply crazy.

Hmmmm, is vger reinjecting an old message, or you sent an older and
wrong version of a patch by mistake?  We discussed why doing this in
queue_diff() is wrong in the thread that has $gmane/265543 in it,
and I was expecting to see a logic like this in the caller.

Puzzled...