backfill: default to grabbing edge blobs too
Commit 302aff09223f (backfill: accept revision arguments, 2026-03-26) added support for accepting revision arguments to backfill. This allows users to do things like git backfill --remotes ^v2.3.0 and then run many commands without triggering on-demand downloads of blobs. However, if they have topics based on v2.3.0, they will likely still trigger on-demand downloads. Consider, for example, the command git log -p v2.3.0..topic This would still trigger on-demand blob loadings after the backfill command above, because the commit(s) with A as a parent will need to diff against the blobs in A. In fact, multiple commands need blobs from the lower boundary of the revision range: * git log -p A..B # After backfill A..B * git replay --onto TARGET A..B # After backfill TARGET^! A..B * git checkout A && git merge B # After backfill A...B Add an extra --[no-]include-edges flag to allow grabbing blobs from edge commits. Since the point of backfill is to prevent on-demand blob loading and these are common commands, default to --include-edges. Signed-off-by: Elijah Newren <newren@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Elijah Newren committed
Apr 15, 2026 at 23:58 UTC
a1ad4a0fca14cdeb55ab9fb065551b15cafa8a4f
3 files changed
+119
-8
Documentation/git-backfill.adoc
+8
-1
@@ -9,7 +9,7 @@ git-backfill - Download missing objects in a partial clone
9
SYNOPSIS
10
--------
11
[synopsis]
12
-git backfill [--min-batch-size=<n>] [--[no-]sparse] [<revision-range>]
12
+git backfill [--min-batch-size=<n>] [--[no-]sparse] [--[no-]include-edges] [<revision-range>]
13
14
DESCRIPTION
15
-----------
@@ -63,6 +63,13 @@ OPTIONS
63
current sparse-checkout. If the sparse-checkout feature is enabled,
64
then `--sparse` is assumed and can be disabled with `--no-sparse`.
65
66
+`--include-edges`::
67
+`--no-include-edges`::
68
+ Include blobs from boundary commits in the backfill. Useful in
69
+ preparation for commands like `git log -p A..B` or `git replay
70
+ --onto TARGET A..B`, where A..B normally excludes A but you need
71
+ the blobs from A as well. `--include-edges` is the default.
72
+
73
`<revision-range>`::
74
Backfill only blobs reachable from commits in the specified
75
revision range. When no _<revision-range>_ is specified, it
builtin/backfill.c
+7
-1
@@ -26,7 +26,7 @@
26
#include "path-walk.h"
27
28
static const char * const builtin_backfill_usage[] = {
29
- N_("git backfill [--min-batch-size=<n>] [--[no-]sparse] [<revision-range>]"),
29
+ N_("git backfill [--min-batch-size=<n>] [--[no-]sparse] [--[no-]include-edges] [<revision-range>]"),
30
NULL
31
};
32
@@ -35,6 +35,7 @@ struct backfill_context {
35
struct oid_array current_batch;
36
size_t min_batch_size;
37
int sparse;
38
+ int include_edges;
39
struct rev_info revs;
40
};
41
@@ -116,6 +117,8 @@ static int do_backfill(struct backfill_context *ctx)
117
/* Walk from HEAD if otherwise unspecified. */
118
if (!ctx->revs.pending.nr)
119
add_head_to_pending(&ctx->revs);
120
+ if (ctx->include_edges)
121
+ ctx->revs.edge_hint = 1;
122
123
info.blobs = 1;
124
info.tags = info.commits = info.trees = 0;
@@ -143,12 +146,15 @@ int cmd_backfill(int argc, const char **argv, const char *prefix, struct reposit
146
.min_batch_size = 50000,
147
.sparse = -1,
148
.revs = REV_INFO_INIT,
149
+ .include_edges = 1,
150
};
151
struct option options[] = {
152
OPT_UNSIGNED(0, "min-batch-size", &ctx.min_batch_size,
153
N_("Minimum number of objects to request at a time")),
154
OPT_BOOL(0, "sparse", &ctx.sparse,
155
N_("Restrict the missing objects to the current sparse-checkout")),
156
+ OPT_BOOL(0, "include-edges", &ctx.include_edges,
157
+ N_("Include blobs from boundary commits in the backfill")),
158
OPT_END(),
159
};
160
struct repo_config_values *cfg = repo_config_values(the_repository);
t/t5620-backfill.sh
+104
-6
@@ -257,11 +257,12 @@ test_expect_success 'backfill with revision range' '
257
git -C backfill-revs rev-list --quiet --objects --missing=print HEAD >missing &&
258
test_line_count = 48 missing &&
259
260
- git -C backfill-revs backfill HEAD~2..HEAD &&
260
+ GIT_TRACE2_EVENT="$(pwd)/backfill-trace" git -C backfill-revs backfill HEAD~2..HEAD &&
261
262
- # 30 objects downloaded.
262
+ # 36 objects downloaded, 12 still missing
263
+ test_trace2_data promisor fetch_count 36 <backfill-trace &&
264
git -C backfill-revs rev-list --quiet --objects --missing=print HEAD >missing &&
264
- test_line_count = 18 missing
265
+ test_line_count = 12 missing
266
'
267
268
test_expect_success 'backfill with revisions over stdin' '
@@ -279,11 +280,12 @@ test_expect_success 'backfill with revisions over stdin' '
280
^HEAD~2
281
EOF
282
282
- git -C backfill-revs backfill --stdin <in &&
283
+ GIT_TRACE2_EVENT="$(pwd)/backfill-trace" git -C backfill-revs backfill --stdin <in &&
284
284
- # 30 objects downloaded.
285
+ # 36 objects downloaded, 12 still missing
286
+ test_trace2_data promisor fetch_count 36 <backfill-trace &&
287
git -C backfill-revs rev-list --quiet --objects --missing=print HEAD >missing &&
286
- test_line_count = 18 missing
288
+ test_line_count = 12 missing
289
'
290
291
test_expect_success 'backfill with prefix pathspec' '
@@ -398,6 +400,102 @@ test_expect_success 'backfill with --since' '
400
test_line_count = 6 missing
401
'
402
403
+test_expect_success 'backfill range with include-edges enables fetch-free git-log' '
404
+ git clone --no-checkout --filter=blob:none \
405
+ --single-branch --branch=main \
406
+ "file://$(pwd)/srv.bare" backfill-log &&
407
+
408
+ # Backfill the range with default include edges.
409
+ git -C backfill-log backfill HEAD~2..HEAD &&
410
+
411
+ # git log -p needs edge blobs for the "before" side of
412
+ # diffs. With edge inclusion, all needed blobs are local.
413
+ GIT_TRACE2_EVENT="$(pwd)/log-trace" git \
414
+ -C backfill-log log -p HEAD~2..HEAD >log-output &&
415
+
416
+ # No promisor fetches should have been needed.
417
+ ! grep "fetch_count" log-trace
418
+'
419
+
420
+test_expect_success 'backfill range without include edges causes on-demand fetches in git-log' '
421
+ git clone --no-checkout --filter=blob:none \
422
+ --single-branch --branch=main \
423
+ "file://$(pwd)/srv.bare" backfill-log-no-bdy &&
424
+
425
+ # Backfill WITHOUT include edges -- file.3 v1 blobs are missing.
426
+ git -C backfill-log-no-bdy backfill --no-include-edges HEAD~2..HEAD &&
427
+
428
+ # git log -p HEAD~2..HEAD computes diff of commit 7 against
429
+ # commit 6. It needs file.3 v1 (the "before" side), which was
430
+ # not backfilled. This triggers on-demand promisor fetches.
431
+ GIT_TRACE2_EVENT="$(pwd)/log-no-bdy-trace" git \
432
+ -C backfill-log-no-bdy log -p HEAD~2..HEAD >log-output &&
433
+
434
+ grep "fetch_count" log-no-bdy-trace
435
+'
436
+
437
+test_expect_success 'backfill range enables fetch-free replay' '
438
+ # Create a repo with a branch to replay.
439
+ git init replay-src &&
440
+ (
441
+ cd replay-src &&
442
+ git config uploadpack.allowfilter 1 &&
443
+ git config uploadpack.allowanysha1inwant 1 &&
444
+ test_commit base &&
445
+ git checkout -b topic &&
446
+ test_commit topic-change &&
447
+ git checkout main &&
448
+ test_commit main-change
449
+ ) &&
450
+ git clone --bare --filter=blob:none \
451
+ "file://$(pwd)/replay-src" replay-dest.git &&
452
+
453
+ # Backfill the replay range: --onto main, replaying topic~1..topic.
454
+ # For replay, we need TARGET^! plus the range.
455
+ main_oid=$(git -C replay-dest.git rev-parse main) &&
456
+ topic_oid=$(git -C replay-dest.git rev-parse topic) &&
457
+ base_oid=$(git -C replay-dest.git rev-parse topic~1) &&
458
+ git -C replay-dest.git backfill \
459
+ "$main_oid^!" "$base_oid..$topic_oid" &&
460
+
461
+ # Now replay should complete without any promisor fetches.
462
+ GIT_TRACE2_EVENT="$(pwd)/replay-trace" git -C replay-dest.git \
463
+ replay --onto main topic~1..topic >replay-out &&
464
+
465
+ ! grep "fetch_count" replay-trace
466
+'
467
+
468
+test_expect_success 'backfill enables fetch-free merge' '
469
+ # Create a repo with two branches to merge.
470
+ git init merge-src &&
471
+ (
472
+ cd merge-src &&
473
+ git config uploadpack.allowfilter 1 &&
474
+ git config uploadpack.allowanysha1inwant 1 &&
475
+ test_commit merge-base &&
476
+ git checkout -b side &&
477
+ test_commit side-change &&
478
+ git checkout main &&
479
+ test_commit main-side-change
480
+ ) &&
481
+ git clone --filter=blob:none \
482
+ "file://$(pwd)/merge-src" merge-dest &&
483
+
484
+ # The clone checked out main, fetching its blobs.
485
+ # Backfill the three endpoint commits needed for merge.
486
+ main_oid=$(git -C merge-dest rev-parse origin/main) &&
487
+ side_oid=$(git -C merge-dest rev-parse origin/side) &&
488
+ mbase=$(git -C merge-dest merge-base origin/main origin/side) &&
489
+ git -C merge-dest backfill --no-include-edges \
490
+ "$main_oid^!" "$side_oid^!" "$mbase^!" &&
491
+
492
+ # Merge should complete without promisor fetches.
493
+ GIT_TRACE2_EVENT="$(pwd)/merge-trace" git -C merge-dest \
494
+ merge origin/side -m "test merge" &&
495
+
496
+ ! grep "fetch_count" merge-trace
497
+'
498
+
499
. "$TEST_DIRECTORY"/lib-httpd.sh
500
start_httpd
501