Re: [PATCH v3 5/9] pack-bitmap: introduce bitmap_walk_contains()

From: Jeff King <peff@peff.net>
Date: Dec 9, 2019, 07:06 UTC
Message-ID: <20191209070642.GE40570@coredump.intra.peff.net>
In-Reply-To: <20191115141541.11149-6-chriscool@tuxfamily.org>

On Fri, Nov 15, 2019 at 03:15:37PM +0100, Christian Couder wrote:

> We will use this helper function in a following commit to
> tell us if an object is packed.

Yeah, makes sense. This is eventually used in have_duplicate_entry() in pack-objects, to check whether an object is already mentioned in reuse_packfile_bitmap. And that's the part that would fix the test failures from the previous commit.

But of course we don't yet have reuse_packfile_bitmap; that comes later.

Show 11 quoted lines

> +int bitmap_walk_contains(struct bitmap_index *bitmap_git,
> +			 struct bitmap *bitmap, const struct object_id *oid)
> +{
> +	int idx;
> +
> +	if (!bitmap)
> +		return 0;
> +
> +	idx = bitmap_position(bitmap_git, oid);
> +	return idx >= 0 && bitmap_get(bitmap, idx);
> +}

This is really a factoring out of code in bitmap_has_oid_in_uninteresting(). So I think you could simplify that like:

diff --git a/pack-bitmap.c b/pack-bitmap.c
index cbfc544411..f5749d0ab3 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -1194,16 +1194,6 @@ void free_bitmap_index(struct bitmap_index *b)
 int bitmap_has_oid_in_uninteresting(struct bitmap_index *bitmap_git,
 				    const struct object_id *oid)
 {
-	int pos;
-
-	if (!bitmap_git)
-		return 0; /* no bitmap loaded */
-	if (!bitmap_git->haves)
-		return 0; /* walk had no "haves" */
-
-	pos = bitmap_position_packfile(bitmap_git, oid);
-	if (pos < 0)
-		return 0;
-
-	return bitmap_get(bitmap_git->haves, pos);
+	return bitmap_git &&
+	       bitmap_walk_contains(bitmap_git, bitmap_git->haves, oid);
 }

One curiosity is that bitmap_has_oid_in_uninteresting() only uses
bitmap_position_packfile(), not bitmap_position(). So it wouldn't find
objects which weren't in the bitmapped packfile (i.e., ones where we
extended the bitmap to handle loose objects, or objects in other packs).

That seems like a bug in the current code to me. I suspect nobody
noticed because the only effect would be that sometimes we fail to
notice that we could reuse a delta against such an object (which isn't
incorrect, just suboptimal). I don't think p5311 would show this,
though, because it simulates a server that is fully packed.

I think it's probably still worth doing this as a preparatory patch,
though:

diff --git a/pack-bitmap.c b/pack-bitmap.c
index e07c798879..6df22e7291 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -1125,7 +1125,7 @@ int bitmap_has_oid_in_uninteresting(struct bitmap_index *bitmap_git,
 	if (!bitmap_git->haves)
 		return 0; /* walk had no "haves" */
 
-	pos = bitmap_position_packfile(bitmap_git, oid);
+	pos = bitmap_position(bitmap_git, oid);
 	if (pos < 0)
 		return 0;
 

-Peff