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

[PATCH 18/30] packed-refs: write file format version 2

From
Derrick Stolee via GitGitGadget <gitgitgadget@gmail.com>
Date
Nov 7, 2022, 18:35 UTC
Message-ID
<740c2f6e6d1e628a84dc4e1927fef70b5d8d624c.1667846165.git.gitgitgadget@gmail.com>
In-Reply-To
<pull.1408.git.1667846164.gitgitgadget@gmail.com>
From: Derrick Stolee <derrickstolee@github.com>
TODO: add writing tests.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
---
 refs/packed-backend.c   |   3 +-
 refs/packed-format-v2.c | 108 ++++++++++++++++++++++++++++++++++++++++
 t/t3212-ref-formats.sh  |   6 ++-
 3 files changed, 115 insertions(+), 2 deletions(-)
diff --git a/refs/packed-backend.c b/refs/packed-backend.c
index 09f7b74584f..3429e63620a 100644
--- a/refs/packed-backend.c
+++ b/refs/packed-backend.c
@@ -790,7 +790,8 @@ static int write_with_updates(struct packed_ref_store *refs,
 		break;
 
 	case 2:
-		ok = write_with_updates_v2(refs, updates, err);
+		/* Convert the normal error codes to ITER_DONE. */
+		ok = write_with_updates_v2(refs, updates, err) ? -2 : ITER_DONE;
 		break;
 
 	default:
diff --git a/refs/packed-format-v2.c b/refs/packed-format-v2.c
index ecf3cc93694..044cc9f629a 100644
--- a/refs/packed-format-v2.c
+++ b/refs/packed-format-v2.c
@@ -6,11 +6,30 @@
 #include "../iterator.h"
 #include "../lockfile.h"
 #include "../chdir-notify.h"
+#include "../chunk-format.h"
+#include "../csum-file.h"
+
+#define OFFSET_IS_PEELED (((uint64_t)1) << 63)
+
+#define PACKED_REFS_SIGNATURE          0x50524546 /* "PREF" */
+#define CHREFS_CHUNKID_OFFSETS         0x524F4646 /* "ROFF" */
+#define CHREFS_CHUNKID_REFS            0x52454653 /* "REFS" */
 
 struct write_packed_refs_v2_context {
 	struct packed_ref_store *refs;
 	struct string_list *updates;
 	struct strbuf *err;
+
+	struct hashfile *f;
+	struct chunkfile *cf;
+
+	/*
+	 * As we stream the ref names to the refs chunk, store these
+	 * values in-memory. These arrays are populated one for every ref.
+	 */
+	uint64_t *offsets;
+	size_t nr;
+	size_t offsets_alloc;
 };
 
 struct write_packed_refs_v2_context *create_v2_context(struct packed_ref_store *refs,
@@ -24,15 +43,104 @@ struct write_packed_refs_v2_context *create_v2_context(struct packed_ref_store *
 	ctx->updates = updates;
 	ctx->err = err;
 
+	if (!fdopen_tempfile(refs->tempfile, "w")) {
+		strbuf_addf(err, "unable to fdopen packed-refs tempfile: %s",
+			    strerror(errno));
+		return ctx;
+	}
+
+	ctx->f = hashfd(refs->tempfile->fd, refs->tempfile->filename.buf);
+	ctx->cf = init_chunkfile(ctx->f);
+
 	return ctx;
 }
 
+static int write_packed_entry_v2(const char *refname,
+				 const struct object_id *oid,
+				 const struct object_id *peeled,
+				 void *write_data)
+{
+	struct write_packed_refs_v2_context *ctx = write_data;
+	size_t reflen = strlen(refname) + 1;
+	size_t i = ctx->nr;
+
+	ALLOC_GROW(ctx->offsets, i + 1, ctx->offsets_alloc);
+
+	/* Write entire ref, including null terminator. */
+	hashwrite(ctx->f, refname, reflen);
+	hashwrite(ctx->f, oid->hash, the_hash_algo->rawsz);
+	if (peeled)
+		hashwrite(ctx->f, peeled->hash, the_hash_algo->rawsz);
+
+	if (i)
+		ctx->offsets[i] = (ctx->offsets[i - 1] & (~OFFSET_IS_PEELED));
+	else
+		ctx->offsets[i] = 0;
+	ctx->offsets[i] += reflen + the_hash_algo->rawsz;
+
+	if (peeled) {
+		ctx->offsets[i] += the_hash_algo->rawsz;
+		ctx->offsets[i] |= OFFSET_IS_PEELED;
+	}
+
+	ctx->nr++;
+	return 0;
+}
+
+static int write_refs_chunk_refs(struct hashfile *f,
+				 void *data)
+{
+	struct write_packed_refs_v2_context *ctx = data;
+	int ok;
+
+	trace2_region_enter("refs", "refs-chunk", the_repository);
+	ok = merge_iterator_and_updates(ctx->refs, ctx->updates, ctx->err,
+					write_packed_entry_v2, ctx);
+	trace2_region_leave("refs", "refs-chunk", the_repository);
+
+	return ok != ITER_DONE;
+}
+
+static int write_refs_chunk_offsets(struct hashfile *f,
+				    void *data)
+{
+	struct write_packed_refs_v2_context *ctx = data;
+	size_t i;
+
+	trace2_region_enter("refs", "offsets", the_repository);
+	for (i = 0; i < ctx->nr; i++)
+		hashwrite_be64(f, ctx->offsets[i]);
+
+	trace2_region_leave("refs", "offsets", the_repository);
+	return 0;
+}
+
 int write_packed_refs_v2(struct write_packed_refs_v2_context *ctx)
 {
+	unsigned char file_hash[GIT_MAX_RAWSZ];
+
+	add_chunk(ctx->cf, CHREFS_CHUNKID_REFS, 0, write_refs_chunk_refs);
+	add_chunk(ctx->cf, CHREFS_CHUNKID_OFFSETS, 0, write_refs_chunk_offsets);
+
+	hashwrite_be32(ctx->f, PACKED_REFS_SIGNATURE);
+	hashwrite_be32(ctx->f, 2);
+	hashwrite_be32(ctx->f, the_hash_algo->format_id);
+
+	if (write_chunkfile(ctx->cf, CHUNKFILE_TRAILING_TOC, ctx))
+		goto failure;
+
+	finalize_hashfile(ctx->f, file_hash, FSYNC_COMPONENT_REFERENCE,
+			  CSUM_HASH_IN_STREAM | CSUM_FSYNC);
+
 	return 0;
+
+failure:
+	return -1;
 }
 
 void free_v2_context(struct write_packed_refs_v2_context *ctx)
 {
+	if (ctx->cf)
+		free_chunkfile(ctx->cf);
 	free(ctx);
 }
diff --git a/t/t3212-ref-formats.sh b/t/t3212-ref-formats.sh
index cd1b399bbb8..03c713ac4f6 100755
--- a/t/t3212-ref-formats.sh
+++ b/t/t3212-ref-formats.sh
@@ -71,7 +71,11 @@ test_expect_success 'extensions.refFormat=files,packed-v2' '
 		# Refuse to parse a v1 packed-refs file.
 		cp ../.git/packed-refs .git/packed-refs &&
 		test_must_fail git rev-parse refs/tags/Q &&
-		rm -f .git/packed-refs
+		rm -f .git/packed-refs &&
+
+		# Create a v2 packed-refs file
+		git pack-refs --all &&
+		test_path_exists .git/packed-refs
 	)
 '
 
-- 
gitgitgadget
Previous: Derrick Stolee via GitGitGadgetNext: Derrick Stolee via GitGitGadget
Message 24 of 56 in “[RFC] extensions.refFormat and packed-refs v2 file format”
  1. 00/30 [RFC] extensions.refFormat and packed-refs v2 file formatDerrick Stolee via GitGitGadget, Nov 7, 2022
  2. 01/30 hashfile: allow skipping the hash functionDerrick Stolee via GitGitGadget, Nov 7, 2022
  3. 02/30 read-cache: add index.computeHash config optionDerrick Stolee via GitGitGadget, Nov 7, 2022
  4. Elijah NewrenNov 11, 2022
  5. Derrick StoleeNov 14, 2022
  6. Ævar Arnfjörð BjarmasonNov 17, 2022
  7. 03/30 extensions: add refFormat extensionDerrick Stolee via GitGitGadget, Nov 7, 2022
  8. Elijah NewrenNov 11, 2022
  9. Derrick StoleeNov 16, 2022
  10. 06/30 refs: allow loose files without packed-refsDerrick Stolee via GitGitGadget, Nov 7, 2022
  11. 07/30 chunk-format: number of chunks is optionalDerrick Stolee via GitGitGadget, Nov 7, 2022
  12. 04/30 config: fix multi-level bulleted listDerrick Stolee via GitGitGadget, Nov 7, 2022
  13. 05/30 repository: wire ref extensions to ref backendsDerrick Stolee via GitGitGadget, Nov 7, 2022
  14. 08/30 chunk-format: document trailing table of contentsDerrick Stolee via GitGitGadget, Nov 7, 2022
  15. 09/30 chunk-format: store chunk offset during writeDerrick Stolee via GitGitGadget, Nov 7, 2022
  16. 11/30 chunk-format: parse trailing table of contentsDerrick Stolee via GitGitGadget, Nov 7, 2022
  17. 10/30 chunk-format: allow trailing table of contentsDerrick Stolee via GitGitGadget, Nov 7, 2022
  18. 13/30 packed-backend: extract add_write_error()Derrick Stolee via GitGitGadget, Nov 7, 2022
  19. 12/30 refs: extract packfile format to new fileDerrick Stolee via GitGitGadget, Nov 7, 2022
  20. 14/30 packed-backend: extract iterator/updates mergeDerrick Stolee via GitGitGadget, Nov 7, 2022
  21. 16/30 config: add config values for packed-refs v2Derrick Stolee via GitGitGadget, Nov 7, 2022
  22. 15/30 packed-backend: create abstraction for writing refsDerrick Stolee via GitGitGadget, Nov 7, 2022
  23. 17/30 packed-backend: create shell of v2 writesDerrick Stolee via GitGitGadget, Nov 7, 2022
  24. 18/30 packed-refs: write file format version 2Derrick Stolee via GitGitGadget, Nov 7, 2022
  25. 19/30 packed-refs: read file format v2Derrick Stolee via GitGitGadget, Nov 7, 2022
  26. 20/30 packed-refs: read optional prefix chunksDerrick Stolee via GitGitGadget, Nov 7, 2022
  27. 21/30 packed-refs: write prefix chunksDerrick Stolee via GitGitGadget, Nov 7, 2022
  28. 22/30 packed-backend: create GIT_TEST_PACKED_REFS_VERSIONDerrick Stolee via GitGitGadget, Nov 7, 2022
  29. 24/30 t5312: allow packed-refs v2 formatDerrick Stolee via GitGitGadget, Nov 7, 2022
  30. 23/30 t1409: test with packed-refs v2Derrick Stolee via GitGitGadget, Nov 7, 2022
  31. 26/30 t3210: require packed-refs v1 for some testsDerrick Stolee via GitGitGadget, Nov 7, 2022
  32. 25/30 t5502: add PACKED_REFS_V1 prerequisiteDerrick Stolee via GitGitGadget, Nov 7, 2022
  33. 27/30 t*: skip packed-refs v2 over http testsDerrick Stolee via GitGitGadget, Nov 7, 2022
  34. 28/30 ci: run GIT_TEST_PACKED_REFS_VERSION=2 in some buildsDerrick Stolee via GitGitGadget, Nov 7, 2022
  35. 29/30 p1401: create performance test for ref operationsDerrick Stolee via GitGitGadget, Nov 7, 2022
  36. 30/30 refs: skip hashing when writing packed-refs v2Derrick Stolee via GitGitGadget, Nov 7, 2022
  37. Derrick StoleeNov 9, 2022
  38. Elijah NewrenNov 11, 2022
  39. Derrick StoleeNov 14, 2022
  40. Elijah NewrenNov 15, 2022
  41. Derrick StoleeNov 16, 2022
  42. Elijah NewrenNov 17, 2022
  43. Junio C HamanoNov 18, 2022
  44. Elijah NewrenNov 19, 2022
  45. Taylor BlauNov 19, 2022
  46. Derrick StoleeNov 30, 2022
  47. Han-Wen NienhuysNov 28, 2022
  48. Derrick StoleeNov 30, 2022
  49. Phillip WoodNov 30, 2022
  50. Taylor BlauNov 30, 2022
  51. Han-Wen NienhuysNov 30, 2022
  52. Sean AllredNov 30, 2022
  53. Derrick StoleeDec 1, 2022
  54. Han-Wen NienhuysDec 2, 2022
  55. Ævar Arnfjörð BjarmasonDec 2, 2022
  56. Junio C HamanoNov 30, 2022

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.