t/t5314-pack-cycle-detection.sh

   1 #!/bin/sh
   2
   3 test_description='test handling of inter-pack delta cycles during repack
   4
   5 The goal here is to create a situation where we have two blobs, A and B, with A
   6 as a delta against B in one pack, and vice versa in the other. Then if we can
   7 persuade a full repack to find A from one pack and B from the other, that will
   8 give us a cycle when we attempt to reuse those deltas.
   9
  10 The trick is in the "persuade" step, as it depends on the internals of how
  11 pack-objects picks which pack to reuse the deltas from. But we can assume
  12 that it does so in one of two general strategies:
  13
  14  1. Using a static ordering of packs. In this case, no inter-pack cycles can
  15     happen. Any objects with a delta relationship must be present in the same
  16     pack (i.e., no "--thin" packs on disk), so we will find all related objects
  17     from that pack. So assuming there are no cycles within a single pack (and
  18     we avoid generating them via pack-objects or importing them via
  19     index-pack), then our result will have no cycles.
  20
  21     So this case should pass the tests no matter how we arrange things.
  22
  23  2. Picking the next pack to examine based on locality (i.e., where we found
  24     something else recently).
  25
  26     In this case, we want to make sure that we find the delta versions of A and
  27     B and not their base versions. We can do this by putting two blobs in each
  28     pack. The first is a "dummy" blob that can only be found in the pack in
  29     question.  And then the second is the actual delta we want to find.
  30
  31     The two blobs must be present in the same tree, not present in other trees,
  32     and the dummy pathname must sort before the delta path.
  33
  34 The setup below focuses on case 2. We have two commits HEAD and HEAD^, each
  35 which has two files: "dummy" and "file". Then we can make two packs which
  36 contain:
  37
  38   [pack one]
  39   HEAD:dummy
  40   HEAD:file  (as delta against HEAD^:file)
  41   HEAD^:file (as base)
  42
  43   [pack two]
  44   HEAD^:dummy
  45   HEAD^:file (as delta against HEAD:file)
  46   HEAD:file  (as base)
  47
  48 Then no matter which order we start looking at the packs in, we know that we
  49 will always find a delta for "file", because its lookup will always come
  50 immediately after the lookup for "dummy".
  51 '
  52 . ./test-lib.sh
  53
  54
  55
  56 # Create a pack containing the the tree $1 and blob $1:file, with
  57 # the latter stored as a delta against $2:file.
  58 #
  59 # We convince pack-objects to make the delta in the direction of our choosing
  60 # by marking $2 as a preferred-base edge. That results in $1:file as a thin
  61 # delta, and index-pack completes it by adding $2:file as a base.
  62 #
  63 # Note that the two variants of "file" must be similar enough to convince git
  64 # to create the delta.
  65 make_pack () {
  66         {
  67                 printf '%s\n' "-$(git rev-parse $2)"
  68                 printf '%s dummy\n' "$(git rev-parse $1:dummy)"
  69                 printf '%s file\n' "$(git rev-parse $1:file)"
  70         } |
  71         git pack-objects --stdout |
  72         git index-pack --stdin --fix-thin
  73 }
  74
  75 test_expect_success 'setup' '
  76         test-genrandom base 4096 >base &&
  77         for i in one two
  78         do
  79                 # we want shared content here to encourage deltas...
  80                 cp base file &&
  81                 echo $i >>file &&
  82
  83                 # ...whereas dummy should be short, because we do not want
  84                 # deltas that would create duplicates when we --fix-thin
  85                 echo $i >dummy &&
  86
  87                 git add file dummy &&
  88                 test_tick &&
  89                 git commit -m $i ||
  90                 return 1
  91         done &&
  92
  93         make_pack HEAD^ HEAD &&
  94         make_pack HEAD HEAD^
  95 '
  96
  97 test_expect_success 'repack' '
  98         # We first want to check that we do not have any internal errors,
  99         # and also that we do not hit the last-ditch cycle-breaking code
 100         # in write_object(), which will issue a warning to stderr.
 101         >expect &&
 102         git repack -ad 2>stderr &&
 103         test_cmp expect stderr &&
 104
 105         # And then double-check that the resulting pack is usable (i.e.,
 106         # we did not fail to notice any cycles). We know we are accessing
 107         # the objects via the new pack here, because "repack -d" will have
 108         # removed the others.
 109         git cat-file blob HEAD:file >/dev/null &&
 110         git cat-file blob HEAD^:file >/dev/null
 111 '
 112
 113 test_done