Re: [PATCH 6/6] pack-objects: reuse on-disk deltas for thin "have" objects

On Tue, Aug 21, 2018 at 04:07:47PM -0400, Jeff King wrote:

On Tue, Aug 21, 2018 at 12:50:09PM -0700, Junio C Hamano wrote:

quoted

quoted

Sorry for commenting on something completely off-topic, but when
applied with "git am -s", I get a resulting commit with 3 S-o-b (the
above two, plus the one added by "-s"), with a blank line in between
them.  I can understand the first blank line (the one between your
two S-o-b), as the first S-o-b does not even appear to be part of
the trailer block, but cannot explain why I get an extra one before
the one added by "-s".  Puzzled...

I think your original "two s-o-b with a blank line in between" was
caused by the same problem, and "git commit --amend -s" perhaps
added an extra one at the end, and added a blank line before the
last "paragraph" while at it?

My suspicion is the long horizontal line at the beginning of the
table, triggers it.  I haven't followed the code closely yet,
though.

Ah, yeah, I think you're right. We call find_patch_start(), which thinks
the "---" line is the end of the commit message. That makes sense when
parsing trailers out of "format-patch" output, but not when we know we
have just the commit message.

So one obvious fix is a new option for the trailer code to tell it we
have _just_ a commit message. That would still leave this obvious false
positive for the format-patch case, but I'm not sure it can be helped.

Another is to tighten the check. Something like this seems more
sensible:

diff --git a/trailer.c b/trailer.c
index 4e309460d1..92ec5cae82 100644
--- a/trailer.c
+++ b/trailer.c

@@ -793,7 +793,8 @@ static int find_patch_start(const char *str)
 	const char *s;
 
 	for (s = str; *s; s = next_line(s)) {
-		if (starts_with(s, "---"))
+		const char *v;
+		if (skip_prefix(s, "---", &v) && isspace(*v))
 			return s - str;
 	}
 

as it would catch "--- /some/file", "--- ", "---\n", and "---\r\n", but
not longer dashed lines. I wondered what "git am" does, though, and I
think it is mailinfo.c:patchbreak(), which has a few other cases to
handle things that git itself would not have generated. I don't know if
that's worth supporting or not.

I think there really are two bugs here, though. The find_patch_start()
check is overly lax, but we also should not have to use it at all when
we know there is no patch.

-Peff