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

[PATCH 1/3] fast-export: allow dumping the refname mapping

From
Jeff King <peff@peff.net>
Date
Jun 19, 2020, 13:25 UTC
Message-ID
<20200619132546.GA2540774@coredump.intra.peff.net>
In-Reply-To
<20200619132304.GA2540657@coredump.intra.peff.net>

After you anonymize a repository, it can be hard to find which commits correspond between the original and the result, and thus hard to reproduce commands that triggered bugs in the original.

Let's make it possible to dump the mapping separate from the output stream. This can be used by a bug reporter to modify their reproduction recipe without revealing the original names (see the example in the documentation).

The implementation is slightly non-obvious. There's no point in the program where we know the complete set of refs we're going to anonymize. Nor do we have a complete set of anonymized refs after finishing (we have a set of anonymized ref path components, but no knowledge of how those are assembled into complete refs). So we lazily write to the dump file as we anonymize each name, and keep a list of ones that we've output in order to avoid duplicates.

Some possible alternatives:
  - we could just output the mapping of anonymized components (e.g.,
    that "foo" became "ref123"). That works OK when you have short
    refnames (e.g., "refs/heads/foo" becomes "refs/heads/ref123"), but
    longer names would require the user to look up each component to
    assemble the result. For example, "refs/remotes/origin/jk/foo" might
    become "refs/remotes/refs37/refs56/refs102".
  - instead of dumping the mapping, the same problem could be solved by
    allowing the user to leave some refs alone. So if you want to
    reproduce "git rev-list branch~17..HEAD" in the anonymized repo, we
    could allow something like:
      git tag anon-one branch
      git tag anon-two HEAD
      git fast-export --anonymize --all \
                      --no-anonymize-ref=anon-one \
		      --no-anonymize-ref=anon-two \
		      >stream
    and then presumably "git rev-list anon-one~17..anon-two" would
    behave the same in the re-imported repository. This is more
    convenient in some ways, but it does require modifying the
    original repository. And the concept doesn't easily extend to
    other fields (e.g., pathnames, which will be addressed in a
    subsequent patch).
  - we could dump before/after commit hashes; combined with rev-parse,
    that could convert these cases (as well as ones using raw hashes).
    But we don't actually know the anonymized commit hashes; we're just
    generating a stream that will produce them in the anonymized repo.
  - likewise, we probably could insert object names or other markers
    into commit messages, blob contents, etc, in order to let a user
    with the original repo figure out which parts correspond. But using
    this gets complicated (I have to find my commits in the result with
    "git log --all --grep" or similar). It also makes it less clear that
    the anonymized repo didn't leak any information (because we are
    relying on object ids being unguessable).
Signed-off-by: Jeff King <peff@peff.net>
---
 Documentation/git-fast-export.txt | 22 +++++++++++++++++
 builtin/fast-export.c             | 39 +++++++++++++++++++++++++++++++
 t/t9351-fast-export-anonymize.sh  | 12 ++++++++++
 3 files changed, 73 insertions(+)
diff --git a/Documentation/git-fast-export.txt b/Documentation/git-fast-export.txt
index e8950de3ba..e809bb3f18 100644
--- a/Documentation/git-fast-export.txt
+++ b/Documentation/git-fast-export.txt
@@ -119,6 +119,12 @@ by keeping the marks the same across runs.
 	the shape of the history and stored tree.  See the section on
 	`ANONYMIZING` below.
 
+--dump-anonymized-refnames=<file>::
+	Output the mapping of real refnames to anonymized refnames to
+	<file>. The output will contain one line per ref that appears in
+	the output stream, with the original refname, a space, and its
+	anonymized counterpart. See the section on `ANONYMIZING` below.
+
 --reference-excluded-parents::
 	By default, running a command such as `git fast-export
 	master~5..master` will not include the commit master{tilde}5
@@ -238,6 +244,22 @@ collapse "User 0", "User 1", etc into "User X"). This produces a much
 smaller output, and it is usually easy to quickly confirm that there is
 no private data in the stream.
 
+Reproducing some bugs may require referencing particular commits, which
+becomes challenging after the refnames have all been anonymized. You can
+use `--dump-anonymized-refnames` to output the mapping, and then alter
+your reproduction recipe to use the anonymized names. E.g., if you find
+a bug with `git rev-list v1.0..v2.0` in the private repository, you can
+run:
+
+---------------------------------------------------
+$ git fast-export --anonymize --all --dump-anonymized-refnames=refs.out >stream
+$ grep '^refs/tags/v[12].0' refs.out
+refs/tags/v1.0 refs/tags/ref31
+refs/tags/v2.0 refs/tags/ref50
+---------------------------------------------------
+
+which tells you that `git rev-list ref31..ref50` may produce the same
+bug in the re-imported anonymous repository.
 
 LIMITATIONS
 -----------
diff --git a/builtin/fast-export.c b/builtin/fast-export.c
index 85868162ee..6caea6f290 100644
--- a/builtin/fast-export.c
+++ b/builtin/fast-export.c
@@ -24,6 +24,7 @@
 #include "remote.h"
 #include "blob.h"
 #include "commit-slab.h"
+#include "khash.h"
 
 static const char *fast_export_usage[] = {
 	N_("git fast-export [rev-list-opts]"),
@@ -45,6 +46,7 @@ static struct string_list extra_refs = STRING_LIST_INIT_NODUP;
 static struct string_list tag_refs = STRING_LIST_INIT_NODUP;
 static struct refspec refspecs = REFSPEC_INIT_FETCH;
 static int anonymize;
+static FILE *anonymized_refnames_handle;
 static struct revision_sources revision_sources;
 
 static int parse_opt_signed_tag_mode(const struct option *opt,
@@ -118,6 +120,32 @@ static int has_unshown_parent(struct commit *commit)
 	return 0;
 }
 
+KHASH_INIT(strset, const char *, int, 0, kh_str_hash_func, kh_str_hash_equal);
+
+struct seen_set {
+	kh_strset_t *set;
+};
+
+static int check_and_mark_seen(struct seen_set *seen, const char *str)
+{
+	int hashret;
+	if (!seen->set)
+		seen->set = kh_init_strset();
+	if (kh_get_strset(seen->set, str) < kh_end(seen->set))
+		return 1;
+	kh_put_strset(seen->set, xstrdup(str), &hashret);
+	return 0;
+}
+
+static void maybe_dump_anon(FILE *out, struct seen_set *seen,
+			    const char *orig, const char *anon)
+{
+	if (!out)
+		return;
+	if (!check_and_mark_seen(seen, orig))
+		fprintf(out, "%s %s\n", orig, anon);
+}
+
 struct anonymized_entry {
 	struct hashmap_entry hash;
 	const char *orig;
@@ -515,6 +543,8 @@ static const char *anonymize_refname(const char *refname)
 	};
 	static struct hashmap refs;
 	static struct strbuf anon = STRBUF_INIT;
+	static struct seen_set seen;
+	const char *full_refname = refname;
 	int i;
 
 	/*
@@ -533,6 +563,8 @@ static const char *anonymize_refname(const char *refname)
 	}
 
 	anonymize_path(&anon, refname, &refs, anonymize_ref_component);
+	maybe_dump_anon(anonymized_refnames_handle, &seen,
+			full_refname, anon.buf);
 	return anon.buf;
 }
 
@@ -1144,6 +1176,7 @@ int cmd_fast_export(int argc, const char **argv, const char *prefix)
 	char *export_filename = NULL,
 	     *import_filename = NULL,
 	     *import_filename_if_exists = NULL;
+	const char *anonymized_refnames_file = NULL;
 	uint32_t lastimportid;
 	struct string_list refspecs_list = STRING_LIST_INIT_NODUP;
 	struct string_list paths_of_changed_objects = STRING_LIST_INIT_DUP;
@@ -1177,6 +1210,9 @@ int cmd_fast_export(int argc, const char **argv, const char *prefix)
 		OPT_STRING_LIST(0, "refspec", &refspecs_list, N_("refspec"),
 			     N_("Apply refspec to exported refs")),
 		OPT_BOOL(0, "anonymize", &anonymize, N_("anonymize output")),
+		OPT_STRING(0, "dump-anonymized-refnames",
+			   &anonymized_refnames_file, N_("file"),
+			   N_("output anonymized refname mapping to <file>")),
 		OPT_BOOL(0, "reference-excluded-parents",
 			 &reference_excluded_commits, N_("Reference parents which are not in fast-export stream by object id")),
 		OPT_BOOL(0, "show-original-ids", &show_original_ids,
@@ -1213,6 +1249,9 @@ int cmd_fast_export(int argc, const char **argv, const char *prefix)
 		string_list_clear(&refspecs_list, 1);
 	}
 
+	if (anonymized_refnames_file)
+		anonymized_refnames_handle = xfopen(anonymized_refnames_file, "w");
+
 	if (use_done_feature)
 		printf("feature done\n");
 
diff --git a/t/t9351-fast-export-anonymize.sh b/t/t9351-fast-export-anonymize.sh
index 897dc50907..0c5dd2a4fb 100755
--- a/t/t9351-fast-export-anonymize.sh
+++ b/t/t9351-fast-export-anonymize.sh
@@ -46,6 +46,18 @@ test_expect_success 'stream omits tag message' '
 	! grep "annotated tag" stream
 '
 
+test_expect_success 'refname mapping can be dumped' '
+	git fast-export --anonymize --all \
+		--dump-anonymized-refnames=refs.out >/dev/null &&
+	# we make no guarantees of the exact anonymized names,
+	# so just check that we have the right number and
+	# that a sample line looks sane.
+	# Note that master is not anonymized, and so not included
+	# in the mapping.
+	test_line_count = 6 refs.out &&
+	grep "^refs/heads/other refs/heads/" refs.out
+'
+
 # NOTE: we chdir to the new, anonymized repository
 # after this. All further tests should assume this.
 test_expect_success 'import stream to new repository' '

-- 
2.27.0.480.g4f98dbcb10
Previous: Jeff KingNext: Eric Sunshine
Message 2 of 64 in “fast-export: allow dumping anonymization mappings”
  1. 0/3 fast-export: allow dumping anonymization mappingsJeff King, Jun 19, 2020
  2. 1/3 fast-export: allow dumping the refname mappingJeff King, Jun 19, 2020
  3. Eric SunshineJun 19, 2020
  4. Jeff KingJun 19, 2020
  5. Eric SunshineJun 19, 2020
  6. Jeff KingJun 19, 2020
  7. Eric SunshineJun 19, 2020
  8. Jeff KingJun 22, 2020
  9. Junio C HamanoJun 19, 2020
  10. Jeff KingJun 22, 2020
  11. 2/3 fast-export: anonymize "master" refnameJeff King, Jun 19, 2020
  12. 3/3 fast-export: allow dumping the path mappingJeff King, Jun 19, 2020
  13. Eric SunshineJun 19, 2020
  14. Junio C HamanoJun 19, 2020
  15. Jeff KingJun 22, 2020
  16. Johannes SchindelinJun 19, 2020
  17. Junio C HamanoJun 22, 2020
  18. 0/4 fast-export: allow dumping anonymization mappingsJeff King, Jun 22, 2020
  19. 1/4 fast-export: allow dumping the refname mappingJeff King, Jun 22, 2020
  20. 2/4 fast-export: anonymize "master" refnameJeff King, Jun 22, 2020
  21. 3/4 fast-export: refactor path printing to not rely on stdoutJeff King, Jun 22, 2020
  22. 4/4 fast-export: allow dumping the path mappingJeff King, Jun 22, 2020
  23. 0/10 fast-export: allow seeding the anonymized mappingJeff King, Jun 23, 2020
  24. 01/10 t9351: derive anonymized tree checks from original repoJeff King, Jun 23, 2020
  25. 02/10 fast-export: use xmemdupz() for anonymizing oidsJeff King, Jun 23, 2020
  26. 03/10 fast-export: store anonymized oids as hex stringsJeff King, Jun 23, 2020
  27. SZEDER GáborJun 24, 2020
  28. Jeff KingJun 24, 2020
  29. Jeff KingJun 25, 2020
  30. SZEDER GáborJun 25, 2020
  31. Jeff KingJun 25, 2020
  32. Johannes SchindelinJun 29, 2020
  33. Jeff KingJun 30, 2020
  34. 04/10 fast-export: tighten anonymize_mem() interface to handle only stringsJeff King, Jun 23, 2020
  35. 05/10 fast-export: stop storing lengths in anonymized hashmapsJeff King, Jun 23, 2020
  36. 06/10 fast-export: use a flex array to store anonymized entriesJeff King, Jun 23, 2020
  37. 07/10 fast-export: move global "idents" anonymize hashmap into functionJeff King, Jun 23, 2020
  38. 08/10 fast-export: add a "data" callback parameter to anonymize_str()Jeff King, Jun 23, 2020
  39. Junio C HamanoJun 24, 2020
  40. 09/10 fast-export: allow seeding the anonymized mappingJeff King, Jun 23, 2020
  41. Eric SunshineJun 23, 2020
  42. Jeff KingJun 23, 2020
  43. Eric SunshineJun 23, 2020
  44. Jeff KingJun 24, 2020
  45. Eric SunshineJun 23, 2020
  46. Jeff KingJun 23, 2020
  47. Eric SunshineJun 23, 2020
  48. Jeff KingJun 24, 2020
  49. 10/10 fast-export: anonymize "master" refnameJeff King, Jun 23, 2020
  50. Junio C HamanoJun 23, 2020
  51. Jeff KingJun 23, 2020
  52. 0/11 fast-export: allow seeding the anonymized mappingJeff King, Jun 25, 2020
  53. 01/11 t9351: derive anonymized tree checks from original repoJeff King, Jun 25, 2020
  54. 02/11 fast-export: use xmemdupz() for anonymizing oidsJeff King, Jun 25, 2020
  55. 03/11 fast-export: store anonymized oids as hex stringsJeff King, Jun 25, 2020
  56. 05/11 fast-export: stop storing lengths in anonymized hashmapsJeff King, Jun 25, 2020
  57. 06/11 fast-export: use a flex array to store anonymized entriesJeff King, Jun 25, 2020
  58. 04/11 fast-export: tighten anonymize_mem() interface to handle only stringsJeff King, Jun 25, 2020
  59. 07/11 fast-export: move global "idents" anonymize hashmap into functionJeff King, Jun 25, 2020
  60. 08/11 fast-export: add a "data" callback parameter to anonymize_str()Jeff King, Jun 25, 2020
  61. 09/11 fast-export: allow seeding the anonymized mappingJeff King, Jun 25, 2020
  62. 11/11 fast-export: use local array to store anonymized oidJeff King, Jun 25, 2020
  63. 10/11 fast-export: anonymize "master" refnameJeff King, Jun 25, 2020
  64. Junio C HamanoJun 25, 2020

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.