git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH 1/5] index-pack: group the delta-base array entries also by type

From
Junio C Hamano <gitster@pobox.com>
Date
Feb 28, 2011, 09:49 UTC
Message-ID
<1298886583-30965-2-git-send-email-gitster@pobox.com>
In-Reply-To
<1298886583-30965-1-git-send-email-gitster@pobox.com>

Entries in the delta_base array are only grouped by the bytepattern in the delta_base union, some of which have 20-byte object name of the base object (i.e. base for REF_DELTA objects), while others have sizeof(off_t) bytes followed by enough NULs to fill 20-byte. The loops to iterate through a range inside this array still needs to inspect the type of the delta, and skip over false hits.

Group the entries also by type to eliminate the potential of false hits.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
---
 * This is a resend of an earlier patch we already discussed; it is not a
   wrong thing to do per-se, but is somewhat difficult to trigger.  You
   would need to come up with an object name that consists of almost all
   NULs and other bytes that happen to be the same as an offset of some
   other object.
 builtin/index-pack.c |   61 ++++++++++++++++++++++++++++++++-----------------
 1 files changed, 40 insertions(+), 21 deletions(-)
diff --git a/builtin/index-pack.c b/builtin/index-pack.c
index 8dc5c0b..1b5d83a 100644
--- a/builtin/index-pack.c
+++ b/builtin/index-pack.c
@@ -391,7 +391,18 @@ static void *get_data_from_pack(struct object_entry *obj)
 	return data;
 }
 
-static int find_delta(const union delta_base *base)
+static int compare_delta_bases(const union delta_base *base1,
+			       const union delta_base *base2,
+			       enum object_type type1,
+			       enum object_type type2)
+{
+	int cmp = type1 - type2;
+	if (cmp)
+		return cmp;
+	return memcmp(base1, base2, UNION_BASE_SZ);
+}
+
+static int find_delta(const union delta_base *base, enum object_type type)
 {
 	int first = 0, last = nr_deltas;
 
@@ -400,7 +411,8 @@ static int find_delta(const union delta_base *base)
                 struct delta_entry *delta = &deltas[next];
                 int cmp;
 
-                cmp = memcmp(base, &delta->base, UNION_BASE_SZ);
+		cmp = compare_delta_bases(base, &delta->base,
+					  type, objects[delta->obj_no].type);
                 if (!cmp)
                         return next;
                 if (cmp < 0) {
@@ -413,9 +425,10 @@ static int find_delta(const union delta_base *base)
 }
 
 static void find_delta_children(const union delta_base *base,
-				int *first_index, int *last_index)
+				int *first_index, int *last_index,
+				enum object_type type)
 {
-	int first = find_delta(base);
+	int first = find_delta(base, type);
 	int last = first;
 	int end = nr_deltas - 1;
 
@@ -543,11 +556,13 @@ static void find_unresolved_deltas(struct base_data *base,
 		union delta_base base_spec;
 
 		hashcpy(base_spec.sha1, base->obj->idx.sha1);
-		find_delta_children(&base_spec, &ref_first, &ref_last);
+		find_delta_children(&base_spec,
+				    &ref_first, &ref_last, OBJ_REF_DELTA);
 
 		memset(&base_spec, 0, sizeof(base_spec));
 		base_spec.offset = base->obj->idx.offset;
-		find_delta_children(&base_spec, &ofs_first, &ofs_last);
+		find_delta_children(&base_spec,
+				    &ofs_first, &ofs_last, OBJ_OFS_DELTA);
 	}
 
 	if (ref_last == -1 && ofs_last == -1) {
@@ -559,24 +574,24 @@ static void find_unresolved_deltas(struct base_data *base,
 
 	for (i = ref_first; i <= ref_last; i++) {
 		struct object_entry *child = objects + deltas[i].obj_no;
-		if (child->real_type == OBJ_REF_DELTA) {
-			struct base_data result;
-			resolve_delta(child, base, &result);
-			if (i == ref_last && ofs_last == -1)
-				free_base_data(base);
-			find_unresolved_deltas(&result, base);
-		}
+		struct base_data result;
+
+		assert(child->real_type == OBJ_REF_DELTA);
+		resolve_delta(child, base, &result);
+		if (i == ref_last && ofs_last == -1)
+			free_base_data(base);
+		find_unresolved_deltas(&result, base);
 	}
 
 	for (i = ofs_first; i <= ofs_last; i++) {
 		struct object_entry *child = objects + deltas[i].obj_no;
-		if (child->real_type == OBJ_OFS_DELTA) {
-			struct base_data result;
-			resolve_delta(child, base, &result);
-			if (i == ofs_last)
-				free_base_data(base);
-			find_unresolved_deltas(&result, base);
-		}
+		struct base_data result;
+
+		assert(child->real_type == OBJ_OFS_DELTA);
+		resolve_delta(child, base, &result);
+		if (i == ofs_last)
+			free_base_data(base);
+		find_unresolved_deltas(&result, base);
 	}
 
 	unlink_base_data(base);
@@ -586,7 +601,11 @@ static int compare_delta_entry(const void *a, const void *b)
 {
 	const struct delta_entry *delta_a = a;
 	const struct delta_entry *delta_b = b;
-	return memcmp(&delta_a->base, &delta_b->base, UNION_BASE_SZ);
+
+	/* group by type (ref vs ofs) and then by value (sha-1 or offset) */
+	return compare_delta_bases(&delta_a->base, &delta_b->base,
+				   objects[delta_a->obj_no].type,
+				   objects[delta_b->obj_no].type);
 }
 
 /* Parse all objects and return the pack content SHA1 hash */
-- 
1.7.4.1.249.g4aa72
Previous: Junio C HamanoNext: Junio C Hamano
Message 2 of 15 in “Using index-pack in place of verify-pack”
  1. 0/5 Using index-pack in place of verify-packJunio C Hamano, Feb 28, 2011
  2. 1/5 index-pack: group the delta-base array entries also by typeJunio C Hamano, Feb 28, 2011
  3. 2/5 write_idx_file: introduce a struct to hold idx customization optionsJunio C Hamano, Feb 28, 2011
  4. 3/5 index-pack: --verifyJunio C Hamano, Feb 28, 2011
  5. Shawn O. PearceFeb 28, 2011
  6. Junio C HamanoFeb 28, 2011
  7. 4/5 write_idx_file: need_large_offset() helper functionJunio C Hamano, Feb 28, 2011
  8. 5/5 index-pack --verify: read anomalous offsets from v2 idx fileJunio C Hamano, Feb 28, 2011
  9. Sverre RabbelierFeb 28, 2011
  10. Junio C HamanoFeb 28, 2011
  11. Sverre RabbelierFeb 28, 2011
  12. Jeff KingFeb 28, 2011
  13. Sverre RabbelierFeb 28, 2011
  14. Shawn O. PearceFeb 28, 2011
  15. Junio C HamanoFeb 28, 2011

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.