On Sun, May 23, 2010 at 03:40:52AM -0400, Jeff King wrote:
By the way, there is one minor bug remaining after this patch:
We may write some new filename into opt_file in the follow_parent_rename
call, but setup_blame_parent_line always diffs the original file. Which
means we lose the line position when following a rename.
We need to do the equivalent of:
git diff -U0 \
opt_ref:opt_file \
blame->commit->id:blame->commit->filename
IOW, to blame directly between the two blobs. Sadly, I don't think there
is a plumbing command to do this, so we are stuck using regular "git
diff", which may have surprises in the config.
The patch below works for my simple tests. I think we probably want to
be doing this anyway for the multiple-parent case. I didn't test, but I
don't think that diff-tree invocation is going to produce any output for
a merge commit.
diff --git a/tig.c b/tig.c
index cfa26ce..4388c2f 100644
--- a/tig.c
+++ b/tig.c
@@ -5177,15 +5177,21 @@ check_blame_commit(struct blame *blame, bool check_null_id)
static void
setup_blame_parent_line(struct view *view, struct blame *blame)
{
+ char from[SIZEOF_REF+SIZEOF_STR];
+ char to[SIZEOF_REF+SIZEOF_STR];
const char *diff_tree_argv[] = {
- "git", "diff-tree", "-U0", blame->commit->id,
- "--", blame->commit->filename, NULL
+ "git", "diff", "--no-textconv", "--no-extdiff", "--no-color",
+ "-U0", from, to, "--", NULL
};
struct io io = {};
int parent_lineno = -1;
int blamed_lineno = -1;
char *line;
+ snprintf(from, sizeof(from), "%s:%s", opt_ref, opt_file);
+ snprintf(to, sizeof(to), "%s:%s", blame->commit->id,
+ blame->commit->filename);
+
if (!io_run(&io, diff_tree_argv, NULL, IO_RD))
return;
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html