t7703: demonstrate object corruption with pack.packSizeLimit

When doing a `--geometric=<d>` repack, `git repack` determines a
splitting point among packs ordered by their object count such that:

  - each pack above the split has at least `<d>` times as many objects
    as the next-largest pack by object count, and
  - the first pack above the split has at least `<d>` times as many
    object as the sum of all packs below the split line combined

`git repack` then creates a pack containing all of the objects contained
in packs below the split line by running `git pack-objects
--stdin-packs` underneath. Once packs are moved into place, then any
packs below the split line are removed, since their objects were just
combined into a new pack.

But `git repack` tries to be careful to avoid removing a pack that it
just wrote, by checking:

    struct packed_git *p = geometry->pack[i];
    if (string_list_has_string(&names, hash_to_hex(p->hash)))
      continue;

in the `delete_redundant` and `geometric` conditional towards the end of
`cmd_repack`.

But it's possible to trick `git repack` into not recognizing a pack that
it just wrote when `names` is out-of-order (which violates
`string_list_has_string()`'s assumption that the list is sorted and thus
binary search-able).

When this happens in just the right circumstances, it is possible to
remove a pack that we just wrote, leading to object corruption.

Luckily, this is quite difficult to provoke in practice (for a couple of
reasons):

  - we ordinarily write just one pack, so `names` usually contains just
    one entry, and is thus sorted
  - when we do write more than one pack (e.g., due to `--max-pack-size`)
    we have to: (a) write a pack identical to one that already
    exists, (b) have that pack be below the split line, and (c) have
    the set of packs written by `pack-objects` occur in an order which
    tricks `string_list_has_string()`.

Demonstrate the above scenario in a failing test, which causes `git
repack --geometric` to write a pack which occurs below the split line,
_and_ fail to recognize that it wrote that pack.

The following patch will fix this bug.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>

0 files changed, 0 insertions, 0 deletions