Re: Possible --boundary bug

Subsystems: the rest

2 messages, 2 authors, 2016-06-15 · open the first message on its own page

Re: Possible --boundary bug

From: Junio C Hamano <hidden>
Date: 2016-06-15 22:42:22

"Marco Costalba" [off-list ref] writes:
Sorry, the good description is below, please ignore the wrong previous one.
I think this patch should fix it.

-- >8 --
rev-list --boundary: fix re-injecting boundary commits.

Marco reported that

	$ git rev-list --boundary --topo-order --parents 5aa44d5..ab57c8d

misses these two boundary commits.

        c649657501bada28794a30102d9c13cc28ca0e5e
        eb38cc689e84a8fd01c1856e889fe8d3b4f1bfb4

Indeed, we can see that gitk shows these two commits at the
bottom, because the --boundary code failed to output them.

The code did not check to avoid pushing the same uninteresting
commit twice to the result list.  I am not sure why this fixes
the reported problem, but this seems to fix it.

Signed-off-by: Junio C Hamano <redacted>
---
diff --git a/revision.c b/revision.c
index abc8745..c2a95aa 100644
--- a/revision.c
+++ b/revision.c
@@ -420,24 +420,33 @@ static void limit_list(struct rev_info *
 		p = &commit_list_insert(commit, p)->next;
 	}
 	if (revs->boundary) {
-		list = newlist;
-		while (list) {
+		/* mark the ones that are on the result list first */
+		for (list = newlist; list; list = list->next) {
 			struct commit *commit = list->item;
+			commit->object.flags |= TMP_MARK;
+		}
+		for (list = newlist; list; list = list->next) {
+			struct commit *commit = list->item;
 			struct object *obj = &commit->object;
-			struct commit_list *parent = commit->parents;
-			if (obj->flags & (UNINTERESTING|BOUNDARY)) {
-				list = list->next;
-				continue;
-			}
-			while (parent) {
+			struct commit_list *parent;
+			if (obj->flags & UNINTERESTING)
+				continue;
+			for (parent = commit->parents;
+			     parent;
+			     parent = parent->next) {
 				struct commit *pcommit = parent->item;
-				parent = parent->next;
 				if (!(pcommit->object.flags & UNINTERESTING))
 					continue;
 				pcommit->object.flags |= BOUNDARY;
+				if (pcommit->object.flags & TMP_MARK)
+					continue;
+				pcommit->object.flags |= TMP_MARK;
 				p = &commit_list_insert(pcommit, p)->next;
 			}
-			list = list->next;
+		}
+		for (list = newlist; list; list = list->next) {
+			struct commit *commit = list->item;
+			commit->object.flags &= ~TMP_MARK;
 		}
 	}
 	revs->commits = newlist;

Re: Possible --boundary bug

From: Marco Costalba <hidden>
Date: 2016-06-15 22:42:22

On 3/31/06, Junio C Hamano [off-list ref] wrote:
"Marco Costalba" [off-list ref] writes:
quoted
Sorry, the good description is below, please ignore the wrong previous one.
I think this patch should fix it.
Yes. It works for me.

Thanks
Marco
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help