t/t5310-pack-bitmaps.sh

   1 #!/bin/sh
   2
   3 test_description='exercise basic bitmap functionality'
   4 GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME=master
   5 export GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME
   6
   7 . ./test-lib.sh
   8
   9 objpath () {
  10         echo ".git/objects/$(echo "$1" | sed -e 's|\(..\)|\1/|')"
  11 }
  12
  13 # show objects present in pack ($1 should be associated *.idx)
  14 list_packed_objects () {
  15         git show-index <"$1" >object-list &&
  16         cut -d' ' -f2 object-list
  17 }
  18
  19 # has_any pattern-file content-file
  20 # tests whether content-file has any entry from pattern-file with entries being
  21 # whole lines.
  22 has_any () {
  23         grep -Ff "$1" "$2"
  24 }
  25
  26 # To ensure the logic for "maximal commits" is exercised, make
  27 # the repository a bit more complicated.
  28 #
  29 #    other                         second
  30 #      *                             *
  31 # (99 commits)                  (99 commits)
  32 #      *                             *
  33 #      |\                           /|
  34 #      | * octo-other  octo-second * |
  35 #      |/|\_________  ____________/|\|
  36 #      | \          \/  __________/  |
  37 #      |  | ________/\ /             |
  38 #      *  |/          * merge-right  *
  39 #      | _|__________/ \____________ |
  40 #      |/ |                         \|
  41 # (l1) *  * merge-left               * (r1)
  42 #      | / \________________________ |
  43 #      |/                           \|
  44 # (l2) *                             * (r2)
  45 #       \___________________________ |
  46 #                                   \|
  47 #                                    * (base)
  48 #
  49 # We only push bits down the first-parent history, which
  50 # makes some of these commits unimportant!
  51 #
  52 # The important part for the maximal commit algorithm is how
  53 # the bitmasks are extended. Assuming starting bit positions
  54 # for second (bit 0) and other (bit 1), the bitmasks at the
  55 # end should be:
  56 #
  57 #      second: 1       (maximal, selected)
  58 #       other: 01      (maximal, selected)
  59 #      (base): 11 (maximal)
  60 #
  61 # This complicated history was important for a previous
  62 # version of the walk that guarantees never walking a
  63 # commit multiple times. That goal might be important
  64 # again, so preserve this complicated case. For now, this
  65 # test will guarantee that the bitmaps are computed
  66 # correctly, even with the repeat calculations.
  67
  68 test_expect_success 'setup repo with moderate-sized history' '
  69         test_commit_bulk --id=file 10 &&
  70         git branch -M second &&
  71         git checkout -b other HEAD~5 &&
  72         test_commit_bulk --id=side 10 &&
  73
  74         # add complicated history setup, including merges and
  75         # ambiguous merge-bases
  76
  77         git checkout -b merge-left other~2 &&
  78         git merge second~2 -m "merge-left" &&
  79
  80         git checkout -b merge-right second~1 &&
  81         git merge other~1 -m "merge-right" &&
  82
  83         git checkout -b octo-second second &&
  84         git merge merge-left merge-right -m "octopus-second" &&
  85
  86         git checkout -b octo-other other &&
  87         git merge merge-left merge-right -m "octopus-other" &&
  88
  89         git checkout other &&
  90         git merge octo-other -m "pull octopus" &&
  91
  92         git checkout second &&
  93         git merge octo-second -m "pull octopus" &&
  94
  95         # Remove these branches so they are not selected
  96         # as bitmap tips
  97         git branch -D merge-left &&
  98         git branch -D merge-right &&
  99         git branch -D octo-other &&
 100         git branch -D octo-second &&
 101
 102         # add padding to make these merges less interesting
 103         # and avoid having them selected for bitmaps
 104         test_commit_bulk --id=file 100 &&
 105         git checkout other &&
 106         test_commit_bulk --id=side 100 &&
 107         git checkout second &&
 108
 109         bitmaptip=$(git rev-parse second) &&
 110         blob=$(echo tagged-blob | git hash-object -w --stdin) &&
 111         git tag tagged-blob $blob &&
 112         git config repack.writebitmaps true
 113 '
 114
 115 test_expect_success 'full repack creates bitmaps' '
 116         GIT_TRACE2_EVENT_NESTING=4 GIT_TRACE2_EVENT="$(pwd)/trace" \
 117                 git repack -ad &&
 118         ls .git/objects/pack/ | grep bitmap >output &&
 119         test_line_count = 1 output &&
 120         grep "\"key\":\"num_selected_commits\",\"value\":\"106\"" trace &&
 121         grep "\"key\":\"num_maximal_commits\",\"value\":\"107\"" trace
 122 '
 123
 124 test_expect_success 'rev-list --test-bitmap verifies bitmaps' '
 125         git rev-list --test-bitmap HEAD
 126 '
 127
 128 rev_list_tests_head () {
 129         test_expect_success "counting commits via bitmap ($state, $branch)" '
 130                 git rev-list --count $branch >expect &&
 131                 git rev-list --use-bitmap-index --count $branch >actual &&
 132                 test_cmp expect actual
 133         '
 134
 135         test_expect_success "counting partial commits via bitmap ($state, $branch)" '
 136                 git rev-list --count $branch~5..$branch >expect &&
 137                 git rev-list --use-bitmap-index --count $branch~5..$branch >actual &&
 138                 test_cmp expect actual
 139         '
 140
 141         test_expect_success "counting commits with limit ($state, $branch)" '
 142                 git rev-list --count -n 1 $branch >expect &&
 143                 git rev-list --use-bitmap-index --count -n 1 $branch >actual &&
 144                 test_cmp expect actual
 145         '
 146
 147         test_expect_success "counting non-linear history ($state, $branch)" '
 148                 git rev-list --count other...second >expect &&
 149                 git rev-list --use-bitmap-index --count other...second >actual &&
 150                 test_cmp expect actual
 151         '
 152
 153         test_expect_success "counting commits with limiting ($state, $branch)" '
 154                 git rev-list --count $branch -- 1.t >expect &&
 155                 git rev-list --use-bitmap-index --count $branch -- 1.t >actual &&
 156                 test_cmp expect actual
 157         '
 158
 159         test_expect_success "counting objects via bitmap ($state, $branch)" '
 160                 git rev-list --count --objects $branch >expect &&
 161                 git rev-list --use-bitmap-index --count --objects $branch >actual &&
 162                 test_cmp expect actual
 163         '
 164
 165         test_expect_success "enumerate commits ($state, $branch)" '
 166                 git rev-list --use-bitmap-index $branch >actual &&
 167                 git rev-list $branch >expect &&
 168                 test_bitmap_traversal --no-confirm-bitmaps expect actual
 169         '
 170
 171         test_expect_success "enumerate --objects ($state, $branch)" '
 172                 git rev-list --objects --use-bitmap-index $branch >actual &&
 173                 git rev-list --objects $branch >expect &&
 174                 test_bitmap_traversal expect actual
 175         '
 176
 177         test_expect_success "bitmap --objects handles non-commit objects ($state, $branch)" '
 178                 git rev-list --objects --use-bitmap-index $branch tagged-blob >actual &&
 179                 grep $blob actual
 180         '
 181 }
 182
 183 rev_list_tests () {
 184         state=$1
 185
 186         for branch in "second" "other"
 187         do
 188                 rev_list_tests_head
 189         done
 190 }
 191
 192 rev_list_tests 'full bitmap'
 193
 194 test_expect_success 'clone from bitmapped repository' '
 195         git clone --no-local --bare . clone.git &&
 196         git rev-parse HEAD >expect &&
 197         git --git-dir=clone.git rev-parse HEAD >actual &&
 198         test_cmp expect actual
 199 '
 200
 201 test_expect_success 'partial clone from bitmapped repository' '
 202         test_config uploadpack.allowfilter true &&
 203         git clone --no-local --bare --filter=blob:none . partial-clone.git &&
 204         (
 205                 cd partial-clone.git &&
 206                 pack=$(echo objects/pack/*.pack) &&
 207                 git verify-pack -v "$pack" >have &&
 208                 awk "/blob/ { print \$1 }" <have >blobs &&
 209                 # we expect this single blob because of the direct ref
 210                 git rev-parse refs/tags/tagged-blob >expect &&
 211                 test_cmp expect blobs
 212         )
 213 '
 214
 215 test_expect_success 'setup further non-bitmapped commits' '
 216         test_commit_bulk --id=further 10
 217 '
 218
 219 rev_list_tests 'partial bitmap'
 220
 221 test_expect_success 'fetch (partial bitmap)' '
 222         git --git-dir=clone.git fetch origin second:second &&
 223         git rev-parse HEAD >expect &&
 224         git --git-dir=clone.git rev-parse HEAD >actual &&
 225         test_cmp expect actual
 226 '
 227
 228 test_expect_success 'incremental repack fails when bitmaps are requested' '
 229         test_commit more-1 &&
 230         test_must_fail git repack -d 2>err &&
 231         test_i18ngrep "Incremental repacks are incompatible with bitmap" err
 232 '
 233
 234 test_expect_success 'incremental repack can disable bitmaps' '
 235         test_commit more-2 &&
 236         git repack -d --no-write-bitmap-index
 237 '
 238
 239 test_expect_success 'pack-objects respects --local (non-local loose)' '
 240         git init --bare alt.git &&
 241         echo $(pwd)/alt.git/objects >.git/objects/info/alternates &&
 242         echo content1 >file1 &&
 243         # non-local loose object which is not present in bitmapped pack
 244         altblob=$(GIT_DIR=alt.git git hash-object -w file1) &&
 245         # non-local loose object which is also present in bitmapped pack
 246         git cat-file blob $blob | GIT_DIR=alt.git git hash-object -w --stdin &&
 247         git add file1 &&
 248         test_tick &&
 249         git commit -m commit_file1 &&
 250         echo HEAD | git pack-objects --local --stdout --revs >1.pack &&
 251         git index-pack 1.pack &&
 252         list_packed_objects 1.idx >1.objects &&
 253         printf "%s\n" "$altblob" "$blob" >nonlocal-loose &&
 254         ! has_any nonlocal-loose 1.objects
 255 '
 256
 257 test_expect_success 'pack-objects respects --honor-pack-keep (local non-bitmapped pack)' '
 258         echo content2 >file2 &&
 259         blob2=$(git hash-object -w file2) &&
 260         git add file2 &&
 261         test_tick &&
 262         git commit -m commit_file2 &&
 263         printf "%s\n" "$blob2" "$bitmaptip" >keepobjects &&
 264         pack2=$(git pack-objects pack2 <keepobjects) &&
 265         mv pack2-$pack2.* .git/objects/pack/ &&
 266         >.git/objects/pack/pack2-$pack2.keep &&
 267         rm $(objpath $blob2) &&
 268         echo HEAD | git pack-objects --honor-pack-keep --stdout --revs >2a.pack &&
 269         git index-pack 2a.pack &&
 270         list_packed_objects 2a.idx >2a.objects &&
 271         ! has_any keepobjects 2a.objects
 272 '
 273
 274 test_expect_success 'pack-objects respects --local (non-local pack)' '
 275         mv .git/objects/pack/pack2-$pack2.* alt.git/objects/pack/ &&
 276         echo HEAD | git pack-objects --local --stdout --revs >2b.pack &&
 277         git index-pack 2b.pack &&
 278         list_packed_objects 2b.idx >2b.objects &&
 279         ! has_any keepobjects 2b.objects
 280 '
 281
 282 test_expect_success 'pack-objects respects --honor-pack-keep (local bitmapped pack)' '
 283         ls .git/objects/pack/ | grep bitmap >output &&
 284         test_line_count = 1 output &&
 285         packbitmap=$(basename $(cat output) .bitmap) &&
 286         list_packed_objects .git/objects/pack/$packbitmap.idx >packbitmap.objects &&
 287         test_when_finished "rm -f .git/objects/pack/$packbitmap.keep" &&
 288         >.git/objects/pack/$packbitmap.keep &&
 289         echo HEAD | git pack-objects --honor-pack-keep --stdout --revs >3a.pack &&
 290         git index-pack 3a.pack &&
 291         list_packed_objects 3a.idx >3a.objects &&
 292         ! has_any packbitmap.objects 3a.objects
 293 '
 294
 295 test_expect_success 'pack-objects respects --local (non-local bitmapped pack)' '
 296         mv .git/objects/pack/$packbitmap.* alt.git/objects/pack/ &&
 297         rm -f .git/objects/pack/multi-pack-index &&
 298         test_when_finished "mv alt.git/objects/pack/$packbitmap.* .git/objects/pack/" &&
 299         echo HEAD | git pack-objects --local --stdout --revs >3b.pack &&
 300         git index-pack 3b.pack &&
 301         list_packed_objects 3b.idx >3b.objects &&
 302         ! has_any packbitmap.objects 3b.objects
 303 '
 304
 305 test_expect_success 'pack-objects to file can use bitmap' '
 306         # make sure we still have 1 bitmap index from previous tests
 307         ls .git/objects/pack/ | grep bitmap >output &&
 308         test_line_count = 1 output &&
 309         # verify equivalent packs are generated with/without using bitmap index
 310         packasha1=$(git pack-objects --no-use-bitmap-index --all packa </dev/null) &&
 311         packbsha1=$(git pack-objects --use-bitmap-index --all packb </dev/null) &&
 312         list_packed_objects packa-$packasha1.idx >packa.objects &&
 313         list_packed_objects packb-$packbsha1.idx >packb.objects &&
 314         test_cmp packa.objects packb.objects
 315 '
 316
 317 test_expect_success 'full repack, reusing previous bitmaps' '
 318         git repack -ad &&
 319         ls .git/objects/pack/ | grep bitmap >output &&
 320         test_line_count = 1 output
 321 '
 322
 323 test_expect_success 'fetch (full bitmap)' '
 324         git --git-dir=clone.git fetch origin second:second &&
 325         git rev-parse HEAD >expect &&
 326         git --git-dir=clone.git rev-parse HEAD >actual &&
 327         test_cmp expect actual
 328 '
 329
 330 test_expect_success 'create objects for missing-HAVE tests' '
 331         blob=$(echo "missing have" | git hash-object -w --stdin) &&
 332         tree=$(printf "100644 blob $blob\tfile\n" | git mktree) &&
 333         parent=$(echo parent | git commit-tree $tree) &&
 334         commit=$(echo commit | git commit-tree $tree -p $parent) &&
 335         cat >revs <<-EOF
 336         HEAD
 337         ^HEAD^
 338         ^$commit
 339         EOF
 340 '
 341
 342 test_expect_success 'pack-objects respects --incremental' '
 343         cat >revs2 <<-EOF &&
 344         HEAD
 345         $commit
 346         EOF
 347         git pack-objects --incremental --stdout --revs <revs2 >4.pack &&
 348         git index-pack 4.pack &&
 349         list_packed_objects 4.idx >4.objects &&
 350         test_line_count = 4 4.objects &&
 351         git rev-list --objects $commit >revlist &&
 352         cut -d" " -f1 revlist |sort >objects &&
 353         test_cmp 4.objects objects
 354 '
 355
 356 test_expect_success 'pack with missing blob' '
 357         rm $(objpath $blob) &&
 358         git pack-objects --stdout --revs <revs >/dev/null
 359 '
 360
 361 test_expect_success 'pack with missing tree' '
 362         rm $(objpath $tree) &&
 363         git pack-objects --stdout --revs <revs >/dev/null
 364 '
 365
 366 test_expect_success 'pack with missing parent' '
 367         rm $(objpath $parent) &&
 368         git pack-objects --stdout --revs <revs >/dev/null
 369 '
 370
 371 test_expect_success JGIT,SHA1 'we can read jgit bitmaps' '
 372         git clone --bare . compat-jgit.git &&
 373         (
 374                 cd compat-jgit.git &&
 375                 rm -f objects/pack/*.bitmap &&
 376                 jgit gc &&
 377                 git rev-list --test-bitmap HEAD
 378         )
 379 '
 380
 381 test_expect_success JGIT,SHA1 'jgit can read our bitmaps' '
 382         git clone --bare . compat-us.git &&
 383         (
 384                 cd compat-us.git &&
 385                 git repack -adb &&
 386                 # jgit gc will barf if it does not like our bitmaps
 387                 jgit gc
 388         )
 389 '
 390
 391 test_expect_success 'splitting packs does not generate bogus bitmaps' '
 392         test-tool genrandom foo $((1024 * 1024)) >rand &&
 393         git add rand &&
 394         git commit -m "commit with big file" &&
 395         git -c pack.packSizeLimit=500k repack -adb &&
 396         git init --bare no-bitmaps.git &&
 397         git -C no-bitmaps.git fetch .. HEAD
 398 '
 399
 400 test_expect_success 'set up reusable pack' '
 401         rm -f .git/objects/pack/*.keep &&
 402         git repack -adb &&
 403         reusable_pack () {
 404                 git for-each-ref --format="%(objectname)" |
 405                 git pack-objects --delta-base-offset --revs --stdout "$@"
 406         }
 407 '
 408
 409 test_expect_success 'pack reuse respects --honor-pack-keep' '
 410         test_when_finished "rm -f .git/objects/pack/*.keep" &&
 411         for i in .git/objects/pack/*.pack
 412         do
 413                 >${i%.pack}.keep
 414         done &&
 415         reusable_pack --honor-pack-keep >empty.pack &&
 416         git index-pack empty.pack &&
 417         git show-index <empty.idx >actual &&
 418         test_must_be_empty actual
 419 '
 420
 421 test_expect_success 'pack reuse respects --local' '
 422         mv .git/objects/pack/* alt.git/objects/pack/ &&
 423         test_when_finished "mv alt.git/objects/pack/* .git/objects/pack/" &&
 424         reusable_pack --local >empty.pack &&
 425         git index-pack empty.pack &&
 426         git show-index <empty.idx >actual &&
 427         test_must_be_empty actual
 428 '
 429
 430 test_expect_success 'pack reuse respects --incremental' '
 431         reusable_pack --incremental >empty.pack &&
 432         git index-pack empty.pack &&
 433         git show-index <empty.idx >actual &&
 434         test_must_be_empty actual
 435 '
 436
 437 test_expect_success 'truncated bitmap fails gracefully (ewah)' '
 438         test_config pack.writebitmaphashcache false &&
 439         git repack -ad &&
 440         git rev-list --use-bitmap-index --count --all >expect &&
 441         bitmap=$(ls .git/objects/pack/*.bitmap) &&
 442         test_when_finished "rm -f $bitmap" &&
 443         test_copy_bytes 256 <$bitmap >$bitmap.tmp &&
 444         mv -f $bitmap.tmp $bitmap &&
 445         git rev-list --use-bitmap-index --count --all >actual 2>stderr &&
 446         test_cmp expect actual &&
 447         test_i18ngrep corrupt.ewah.bitmap stderr
 448 '
 449
 450 test_expect_success 'truncated bitmap fails gracefully (cache)' '
 451         git repack -ad &&
 452         git rev-list --use-bitmap-index --count --all >expect &&
 453         bitmap=$(ls .git/objects/pack/*.bitmap) &&
 454         test_when_finished "rm -f $bitmap" &&
 455         test_copy_bytes 512 <$bitmap >$bitmap.tmp &&
 456         mv -f $bitmap.tmp $bitmap &&
 457         git rev-list --use-bitmap-index --count --all >actual 2>stderr &&
 458         test_cmp expect actual &&
 459         test_i18ngrep corrupted.bitmap.index stderr
 460 '
 461
 462 # have_delta <obj> <expected_base>
 463 #
 464 # Note that because this relies on cat-file, it might find _any_ copy of an
 465 # object in the repository. The caller is responsible for making sure
 466 # there's only one (e.g., via "repack -ad", or having just fetched a copy).
 467 have_delta () {
 468         echo $2 >expect &&
 469         echo $1 | git cat-file --batch-check="%(deltabase)" >actual &&
 470         test_cmp expect actual
 471 }
 472
 473 # Create a state of history with these properties:
 474 #
 475 #  - refs that allow a client to fetch some new history, while sharing some old
 476 #    history with the server; we use branches delta-reuse-old and
 477 #    delta-reuse-new here
 478 #
 479 #  - the new history contains an object that is stored on the server as a delta
 480 #    against a base that is in the old history
 481 #
 482 #  - the base object is not immediately reachable from the tip of the old
 483 #    history; finding it would involve digging down through history we know the
 484 #    other side has
 485 #
 486 # This should result in a state where fetching from old->new would not
 487 # traditionally reuse the on-disk delta (because we'd have to dig to realize
 488 # that the client has it), but we will do so if bitmaps can tell us cheaply
 489 # that the other side has it.
 490 test_expect_success 'set up thin delta-reuse parent' '
 491         # This first commit contains the buried base object.
 492         test-tool genrandom delta 16384 >file &&
 493         git add file &&
 494         git commit -m "delta base" &&
 495         base=$(git rev-parse --verify HEAD:file) &&
 496
 497         # These intermediate commits bury the base back in history.
 498         # This becomes the "old" state.
 499         for i in 1 2 3 4 5
 500         do
 501                 echo $i >file &&
 502                 git commit -am "intermediate $i" || return 1
 503         done &&
 504         git branch delta-reuse-old &&
 505
 506         # And now our new history has a delta against the buried base. Note
 507         # that this must be smaller than the original file, since pack-objects
 508         # prefers to create deltas from smaller objects to larger.
 509         test-tool genrandom delta 16300 >file &&
 510         git commit -am "delta result" &&
 511         delta=$(git rev-parse --verify HEAD:file) &&
 512         git branch delta-reuse-new &&
 513
 514         # Repack with bitmaps and double check that we have the expected delta
 515         # relationship.
 516         git repack -adb &&
 517         have_delta $delta $base
 518 '
 519
 520 # Now we can sanity-check the non-bitmap behavior (that the server is not able
 521 # to reuse the delta). This isn't strictly something we care about, so this
 522 # test could be scrapped in the future. But it makes sure that the next test is
 523 # actually triggering the feature we want.
 524 #
 525 # Note that our tools for working with on-the-wire "thin" packs are limited. So
 526 # we actually perform the fetch, retain the resulting pack, and inspect the
 527 # result.
 528 test_expect_success 'fetch without bitmaps ignores delta against old base' '
 529         test_config pack.usebitmaps false &&
 530         test_when_finished "rm -rf client.git" &&
 531         git init --bare client.git &&
 532         (
 533                 cd client.git &&
 534                 git config transfer.unpackLimit 1 &&
 535                 git fetch .. delta-reuse-old:delta-reuse-old &&
 536                 git fetch .. delta-reuse-new:delta-reuse-new &&
 537                 have_delta $delta $ZERO_OID
 538         )
 539 '
 540
 541 # And do the same for the bitmap case, where we do expect to find the delta.
 542 test_expect_success 'fetch with bitmaps can reuse old base' '
 543         test_config pack.usebitmaps true &&
 544         test_when_finished "rm -rf client.git" &&
 545         git init --bare client.git &&
 546         (
 547                 cd client.git &&
 548                 git config transfer.unpackLimit 1 &&
 549                 git fetch .. delta-reuse-old:delta-reuse-old &&
 550                 git fetch .. delta-reuse-new:delta-reuse-new &&
 551                 have_delta $delta $base
 552         )
 553 '
 554
 555 test_done