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

[PATCH 11/18] pack-objects: support --blob-max-bytes

From
JTJonathan Tan <jonathantanmy@google.com>
Date
Sep 29, 2017, 20:11 UTC
Message-ID
<b9ea93edabc42754dc3643d6307c22a947eabaf3.1506714999.git.jonathantanmy@google.com>
In-Reply-To
<cover.1506714999.git.jonathantanmy@google.com>

As part of an effort to improve Git support for very large repositories in which clients typically have only a subset of all version-controlled blobs, teach pack-objects to support --blob-max-bytes, packing only blobs not exceeding that size unless the blob corresponds to a file whose name starts with ".git". upload-pack will eventually be taught to use this new parameter if needed to exclude certain blobs during a fetch or clone, potentially drastically reducing network consumption when serving these very large repositories.

Since any excluded blob should not act as a delta base, I did this by avoiding such oversized blobs from ever going into the "to_pack" data structure, which contains both preferred bases (which do not go into the final packfile but can act as delta bases) and the objects to be packed (which go into the final packfile and also can act as delta bases). To that end, add_object_entry() has been modified to exclude the appropriate non-preferred-base objects. (Preferred bases, regardless of size, still enter "to_pack" - they are something that the client indicates that it has, not something that the server needs to serve, so no exclusion occurs.)

If bitmaps are to be used, we would not know if a blob corresponded to a file whose name starts with ".git". For this reason, when invoked with --blob-max-bytes, pack-objects will not use bitmaps.

Signed-off-by: Jonathan Tan <jonathantanmy@google.com>
---
 Documentation/git-pack-objects.txt | 12 +++++++++-
 builtin/pack-objects.c             | 33 ++++++++++++++++++++++++++--
 t/t5300-pack-object.sh             | 45 ++++++++++++++++++++++++++++++++++++++
 t/test-lib-functions.sh            | 12 ++++++++++
 4 files changed, 99 insertions(+), 3 deletions(-)
diff --git a/Documentation/git-pack-objects.txt b/Documentation/git-pack-objects.txt
index 473a16135..ddce2d4c1 100644
--- a/Documentation/git-pack-objects.txt
+++ b/Documentation/git-pack-objects.txt
@@ -12,7 +12,8 @@ SYNOPSIS
 'git pack-objects' [-q | --progress | --all-progress] [--all-progress-implied]
 	[--no-reuse-delta] [--delta-base-offset] [--non-empty]
 	[--local] [--incremental] [--window=<n>] [--depth=<n>]
-	[--revs [--unpacked | --all]] [--stdout | base-name]
+	[--revs [--unpacked | --all]]
+	[--stdout [--blob-max-bytes=<n>] | base-name]
 	[--shallow] [--keep-true-parents] < object-list
 
 
@@ -236,6 +237,15 @@ So does `git bundle` (see linkgit:git-bundle[1]) when it creates a bundle.
 	With this option, parents that are hidden by grafts are packed
 	nevertheless.
 
+--blob-max-bytes=<n>::
+	This option can only be used with --stdout. If specified, a blob
+	larger than this will not be packed unless a to-be-packed tree
+	has that blob with a filename beginning with ".git".  The size
+	can be suffixed with "k", "m", or "g", and may be "0".
++
+If specified, after printing the packfile, pack-objects will print, in packet
+format (pkt-line), the names of excluded blobs and their sizes.
+
 SEE ALSO
 --------
 linkgit:git-rev-list[1]
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index ef0b61d5f..8686b4351 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -26,6 +26,8 @@
 #include "argv-array.h"
 #include "mru.h"
 #include "packfile.h"
+#include "pkt-line.h"
+#include "varint.h"
 
 static const char *pack_usage[] = {
 	N_("git pack-objects --stdout [<options>...] [< <ref-list> | < <object-list>]"),
@@ -81,6 +83,8 @@ static unsigned long cache_max_small_delta_size = 1000;
 
 static unsigned long window_memory_limit = 0;
 
+static long blob_max_bytes = -1;
+
 /*
  * stats
  */
@@ -1072,6 +1076,27 @@ static const char no_closure_warning[] = N_(
 "disabling bitmap writing, as some objects are not being packed"
 );
 
+/*
+ * Returns 1 if the given blob does not meet any defined blob size
+ * limits.  Blobs that appear as a tree entry whose basename begins with
+ * ".git" are never considered oversized.
+ */
+static int oversized(const unsigned char *sha1, const char *name) {
+	const char *last_slash, *basename;
+	unsigned long size;
+
+	if (blob_max_bytes < 0)
+		return 0;
+
+	last_slash = strrchr(name, '/');
+	basename = last_slash ? last_slash + 1 : name;
+	if (starts_with(basename, ".git"))
+		return 0;
+
+	sha1_object_info(sha1, &size);
+	return size > blob_max_bytes;
+}
+
 static int add_object_entry(const unsigned char *sha1, enum object_type type,
 			    const char *name, int preferred_base)
 {
@@ -1082,7 +1107,8 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
 	if (have_duplicate_entry(sha1, preferred_base, &index_pos))
 		return 0;
 
-	if (ignore_object(sha1, preferred_base, &found_pack, &found_offset)) {
+	if (ignore_object(sha1, preferred_base, &found_pack, &found_offset) ||
+	    (!preferred_base && type == OBJ_BLOB && oversized(sha1, name))) {
 		/* The pack is missing an object, so it will not have closure */
 		if (write_bitmap_index) {
 			warning(_(no_closure_warning));
@@ -2956,6 +2982,8 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 			 N_("write a bitmap index together with the pack index")),
 		OPT_BOOL(0, "exclude-promisor-objects", &exclude_promisor_objects,
 			 N_("do not pack objects in promisor packfiles")),
+		OPT_MAGNITUDE(0, "blob-max-bytes", &blob_max_bytes,
+			      N_("exclude blobs larger than this")),
 		OPT_END(),
 	};
 
@@ -3054,7 +3082,8 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
 		use_bitmap_index = use_bitmap_index_default;
 
 	/* "hard" reasons not to use bitmaps; these just won't work at all */
-	if (!use_internal_rev_list || (!pack_to_stdout && write_bitmap_index) || is_repository_shallow())
+	if (!use_internal_rev_list || (!pack_to_stdout && write_bitmap_index) ||
+	    is_repository_shallow() || (blob_max_bytes >= 0))
 		use_bitmap_index = 0;
 
 	if (pack_to_stdout || !rev_list_all)
diff --git a/t/t5300-pack-object.sh b/t/t5300-pack-object.sh
index 9c68b9925..697cc20f2 100755
--- a/t/t5300-pack-object.sh
+++ b/t/t5300-pack-object.sh
@@ -457,6 +457,51 @@ test_expect_success !PTHREADS,C_LOCALE_OUTPUT 'pack-objects --threads=N or pack.
 	grep -F "no threads support, ignoring pack.threads" err
 '
 
+lcut () {
+	perl -e '$/ = undef; $_ = <>; s/^.{'$1'}//s; print $_'
+}
+
+test_expect_success '--blob-size-limit works with multiple excluded' '
+	rm -rf server &&
+	git init server &&
+	printf a > server/a &&
+	printf b > server/b &&
+	printf c-very-long-file > server/c &&
+	printf d-very-long-file > server/d &&
+	git -C server add a b c d &&
+	git -C server commit -m x &&
+
+	git -C server rev-parse HEAD >objects &&
+	git -C server pack-objects --revs --stdout --blob-max-bytes=10 <objects >my.pack &&
+
+	# Ensure that only the small blobs are in the packfile
+	git index-pack my.pack &&
+	git verify-pack -v my.idx >objectlist &&
+	grep $(git hash-object server/a) objectlist &&
+	grep $(git hash-object server/b) objectlist &&
+	! grep $(git hash-object server/c) objectlist &&
+	! grep $(git hash-object server/d) objectlist
+'
+
+test_expect_success '--blob-size-limit never excludes special files' '
+	rm -rf server &&
+	git init server &&
+	printf a-very-long-file > server/a &&
+	printf a-very-long-file > server/.git-a &&
+	printf b-very-long-file > server/b &&
+	git -C server add a .git-a b &&
+	git -C server commit -m x &&
+
+	git -C server rev-parse HEAD >objects &&
+	git -C server pack-objects --revs --stdout --blob-max-bytes=10 <objects >my.pack &&
+
+	# Ensure that the .git-a blob is in the packfile, despite also
+	# appearing as a non-.git file
+	git index-pack my.pack &&
+	git verify-pack -v my.idx >objectlist &&
+	grep $(git hash-object server/a) objectlist
+'
+
 #
 # WARNING!
 #
diff --git a/t/test-lib-functions.sh b/t/test-lib-functions.sh
index 1701fe2a0..07b79c7ad 100644
--- a/t/test-lib-functions.sh
+++ b/t/test-lib-functions.sh
@@ -1020,3 +1020,15 @@ nongit () {
 		"$@"
 	)
 }
+
+# Converts big-endian pairs of hexadecimal digits into bytes. For example,
+# "printf 61620d0a | hex_pack" results in "ab\r\n".
+hex_pack () {
+	perl -e '$/ = undef; $input = <>; print pack("H*", $input)'
+}
+
+# Converts bytes into big-endian pairs of hexadecimal digits. For example,
+# "printf 'ab\r\n' | hex_unpack" results in "61620d0a".
+hex_unpack () {
+	perl -e '$/ = undef; $input = <>; print unpack("H2" x length($input), $input)'
+}

-- 
2.14.2.822.g60be5d43e6-goog
Previous: Jonathan TanNext: Johannes Schindelin
Message 21 of 28 in “Partial clone (from clone to lazy fetch in 18 patches)”
  1. 00/18 Partial clone (from clone to lazy fetch in 18 patches)Jonathan Tan, Sep 29, 2017
  2. 01/18 fsck: introduce partialclone extensionJonathan Tan, Sep 29, 2017
  3. 04/18 fsck: support promisor objects as CLI argumentJonathan Tan, Sep 29, 2017
  4. 06/18 introduce fetch-object: fetch one promisor objectJonathan Tan, Sep 29, 2017
  5. 07/18 sha1_file: support lazily fetching missing objectsJonathan Tan, Sep 29, 2017
  6. Christian CouderOct 12, 2017
  7. Christian CouderOct 12, 2017
  8. 08/18 rev-list: support termination at promisor objectsJonathan Tan, Sep 29, 2017
  9. 02/18 fsck: support refs pointing to promisor objectsJonathan Tan, Sep 29, 2017
  10. 13/18 fetch: refactor calculation of remote listJonathan Tan, Sep 29, 2017
  11. 17/18 unpack-trees: batch fetching of missing blobsJonathan Tan, Sep 29, 2017
  12. 15/18 clone: support excluding large blobsJonathan Tan, Sep 29, 2017
  13. 14/18 fetch: support excluding large blobsJonathan Tan, Sep 29, 2017
  14. 16/18 clone: configure blobmaxbytes in created reposJonathan Tan, Sep 29, 2017
  15. 18/18 fetch-pack: restore save_commit_buffer after useJonathan Tan, Sep 29, 2017
  16. 03/18 fsck: support referenced promisor objectsJonathan Tan, Sep 29, 2017
  17. 12/18 fetch-pack: support excluding large blobsJonathan Tan, Sep 29, 2017
  18. 05/18 index-pack: refactor writing of .keep filesJonathan Tan, Sep 29, 2017
  19. 09/18 gc: do not repack promisor packfilesJonathan Tan, Sep 29, 2017
  20. 10/18 pack-objects: rename want_.* to ignore_.*Jonathan Tan, Sep 29, 2017
  21. 11/18 pack-objects: support --blob-max-bytesJonathan Tan, Sep 29, 2017
  22. Johannes SchindelinSep 29, 2017
  23. Junio C HamanoOct 2, 2017
  24. Christian CouderOct 3, 2017
  25. Junio C HamanoOct 3, 2017
  26. Jeff HostetlerOct 3, 2017
  27. Jonathan TanOct 3, 2017
  28. Jeff HostetlerOct 4, 2017

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.