Re: [PATCH v3 0/7] reftable: avoid reading and writing empty keys

[Junio C Hamano <gitster@pobox.com>]

"Han-Wen Nienhuys via GitGitGadget" <gitgitgadget@gmail.com> writes:

> this series makes sure that the object record does not have to consider
> empty keys (and therefore, a NULL memcpy destination)
>
> while we're at it add some more tests, and fix a naming mistake.

Looking good.  Let's mark it for 'next' and below soonish.

Thanks.

>
> Han-Wen Nienhuys (7):
>   Documentation: object_id_len goes up to 31
>   reftable: reject 0 object_id_len
>   reftable: add a test that verifies that writing empty keys fails
>   reftable: avoid writing empty keys at the block layer
>   reftable: ensure that obj_id_len is >= 2 on writing
>   reftable: add test for length of disambiguating prefix
>   reftable: rename writer_stats to reftable_writer_stats
>
>  Documentation/technical/reftable.txt |   2 +-
>  reftable/block.c                     |  27 ++++---
>  reftable/block_test.c                |   5 ++
>  reftable/reader.c                    |   5 ++
>  reftable/readwrite_test.c            | 105 ++++++++++++++++++++++++++-
>  reftable/reftable-writer.h           |   2 +-
>  reftable/writer.c                    |   9 ++-
>  7 files changed, 136 insertions(+), 19 deletions(-)
>
>
> base-commit: 45fe28c951c3e70666ee4ef8379772851a8e4d32
> Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1185%2Fhanwen%2Fobj-id-len-v3
> Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1185/hanwen/obj-id-len-v3
> Pull-Request: https://github.com/git/git/pull/1185
>
> Range-diff vs v2:
>
>  1:  80d29e8f269 = 1:  80d29e8f269 Documentation: object_id_len goes up to 31
>  2:  4c1a19fc4ae ! 2:  68e7bc32ff8 reftable: reject 0 object_id_len
>      @@ Commit message
>           reftable: reject 0 object_id_len
>       
>           The spec says 2 <= object_id_len <= 31. We are lenient and allow 1,
>      -    but we forbid 0, so we can we can be sure that we never read a
>      -    0-length key.
>      +    but we forbid 0, so we can be sure that we never read a 0-length key.
>       
>           Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
>       
>  3:  600b115f8b1 = 3:  8b5aebdb07a reftable: add a test that verifies that writing empty keys fails
>  4:  ba036ee8543 ! 4:  a9372cacd1b reftable: avoid writing empty keys at the block layer
>      @@ reftable/block.c: int block_reader_first_key(struct block_reader *br, struct str
>        	if (n < 0)
>        		return n;
>       +	if (!key->len)
>      -+		return -1;
>      ++		return REFTABLE_FORMAT_ERROR;
>        
>        	return 0;
>        }
>  5:  2bd3d44ba57 = 5:  0b8a42399dd reftable: ensure that obj_id_len is >= 2 on writing
>  6:  82d36ee0e0d = 6:  bdccd969475 reftable: add test for length of disambiguating prefix
>  7:  c6ffdb3471c = 7:  72499a14e38 reftable: rename writer_stats to reftable_writer_stats