René Scharfe <rene.scharfe@xxxxxxxxxxxxxx> writes:
> Hmm, let's see if we can help the compiler follow the code without
> making it harder for people to understand. The patch looks a bit
> jumbled, but the resulting code is OK in my biased opinion.
I actually think the result is much better than a mere "OK"; the
duplicated "at this point we know path1 (or path2) is missing from
the other side" has been bothering me and I was about to suggest a
similar rewrite before I read your message ;-)
However, the same compiler still thinks {elem,path,mode}1 can be
used uninitialized (but not {elem,path,mode}2). The craziness I
reported in the previous message is also the same. With this patch
on top to swap the side we inspect first, the compiler thinks
{elem,path,mode}2 can be used uninitialized but not the other three
variables X-<.
So I like your change for readability, but for GCC 4.4.5 we still
need the unnecessary initialization.
match-trees.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/match-trees.c b/match-trees.c
index c0c66bb..9ea2c80 100644
--- a/match-trees.c
+++ b/match-trees.c
@@ -77,16 +77,16 @@ static int score_trees(const unsigned char *hash1, const unsigned char *hash2)
unsigned mode1, mode2;
int cmp = 0;
- if (one.size)
- elem1 = tree_entry_extract(&one, &path1, &mode1);
- else
- /* two has more entries */
- cmp = 1;
if (two.size)
elem2 = tree_entry_extract(&two, &path2, &mode2);
else
/* two lacks this entry */
cmp = -1;
+ if (one.size)
+ elem1 = tree_entry_extract(&one, &path1, &mode1);
+ else
+ /* two has more entries */
+ cmp = 1;
if (!cmp)
cmp = base_name_compare(path1, strlen(path1), mode1,
--
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