[PATCH 0/6] harden tree-walking against integer overflow

[Jeff King <peff@peff.net>]

I noticed that it's possible to get funny integer over/underflows with
tree-walk's traverse_info (with specially-crafted absurdly-sized tree
paths). I wasn't able to turn this into an actual buffer overflow
because the funky sizes cause allocation failures way before we ever get
into make_traverse_path(). But it makes sense to protect ourselves
anyway.

The first two patches are an unrelated memory problem I found (and
they're here in the same series because I build on the cleanups). I
don't think it's security-relevant, though; it involves reading from
uninitialized heap memory, but we don't actually _do_ anything with the
result. We just copy uninitialized bytes from one heap buffer to the
other, and then never look at them again.

  [1/6]: setup_traverse_info(): stop copying oid
  [2/6]: tree-walk: drop oid from traverse_info
  [3/6]: tree-walk: use size_t consistently
  [4/6]: tree-walk: accept a raw length for traverse_path_len()
  [5/6]: tree-walk: add a strbuf wrapper for make_traverse_path()
  [6/6]: tree-walk: harden make_traverse_path() length computations

 Documentation/technical/api-tree-walking.txt |  8 ++-
 builtin/merge-tree.c                         |  5 +-
 cache-tree.c                                 |  2 +-
 tree-walk.c                                  | 64 +++++++++++------
 tree-walk.h                                  | 18 +++--
 unpack-trees.c                               | 74 +++++++++++---------
 6 files changed, 103 insertions(+), 68 deletions(-)

-Peff