Thread (2 messages) flat view 2 messages, 2 authors, 2016-06-15
DORMANTno replies

[PATCH] Optimize sha1_object_info for loose objects, not concurrent repacks

From: Steven Grimm <hidden>
Date: 2016-06-15 22:45:06
Subsystem: the rest · Maintainer: Linus Torvalds

When dealing with a repository with lots of loose objects, sha1_object_info
would rescan the packs directory every time an unpacked object was referenced
before finally giving up and looking for the loose object. This caused a lot
of extra unnecessary system calls during git pack-objects; the code was
rereading the entire pack directory once for each loose object file.

This patch looks for a loose object before falling back to rescanning the
pack directory, rather than the other way around.

Signed-off-by: Steven Grimm <redacted>
---

	I discovered this by running strace on a pack-objects that was
	taking especially long to run; it was making more system calls
	to scan the pack directory than to do stuff with the loose
	objects, which didn't seem right.

 sha1_file.c |    9 ++++++++-
 1 files changed, 8 insertions(+), 1 deletions(-)
diff --git a/sha1_file.c b/sha1_file.c
index e281c14..32e4664 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -1929,11 +1929,18 @@ static int sha1_loose_object_info(const unsigned char *sha1, unsigned long *size
 int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
 {
 	struct pack_entry e;
+	int status;
 
 	if (!find_pack_entry(sha1, &e, NULL)) {
+		/* Most likely it's a loose object. */
+		status = sha1_loose_object_info(sha1, sizep);
+		if (status >= 0)
+			return status;
+
+		/* Not a loose object; someone else may have just packed it. */
 		reprepare_packed_git();
 		if (!find_pack_entry(sha1, &e, NULL))
-			return sha1_loose_object_info(sha1, sizep);
+			return status;
 	}
 	return packed_object_info(e.p, e.offset, sizep);
 }
-- 
1.6.0.rc1.66.gc78d7
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help