git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 2/2] Copy resolve_ref() return value for longer use

From
Junio C Hamano <gitster@pobox.com>
Date
Nov 13, 2011, 20:41 UTC
Message-ID
<7vbosfoiuy.fsf@alter.siamese.dyndns.org>
In-Reply-To
<1321179735-21890-2-git-send-email-pclouds@gmail.com>
Nguyễn Thái Ngọc Duy <pclouds@gmail.com> writes:
Show 71 quoted lines
> resolve_ref() may return a pointer to a static buffer. Callers that
> use this value longer than a couple of statements should copy the
> value to avoid some hidden resolve_ref() call that may change the
> static buffer's value.
>
> The bug found by Tony Wang <wwwjfy@gmail.com> in builtin/merge.c
> demonstrates this. The first call is in cmd_merge()
>
> branch = resolve_ref("HEAD", head_sha1, 0, &flag);
>
> Then deep in lookup_commit_or_die() a few lines after, resolve_ref()
> may be called again and destroy "branch".
>
> lookup_commit_or_die
>  lookup_commit_reference
>   lookup_commit_reference_gently
>    parse_object
>     lookup_replace_object
>      do_lookup_replace_object
>       prepare_replace_object
>        for_each_replace_ref
>         do_for_each_ref
>          get_loose_refs
>           get_ref_dir
>            get_ref_dir
>             resolve_ref
>
> All call sites are checked and made sure that xstrdup() is called if
> the value should be saved.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  Unfortunately there are still 37 resolve_ref() calls. An API change
>  would touch all of them and potentially introduce more bugs along the
>  way, either leak more memory or free() the wrong way.
>
>  So I'd stick with this for v1.7.8.
>
>  builtin/branch.c        |    5 +++-
>  builtin/checkout.c      |    4 ++-
>  builtin/commit.c        |    3 +-
>  builtin/fmt-merge-msg.c |    6 ++++-
>  builtin/merge.c         |   62 ++++++++++++++++++++++++++++++----------------
>  builtin/notes.c         |    6 ++++-
>  builtin/receive-pack.c  |    3 ++
>  reflog-walk.c           |    5 +++-
>  8 files changed, 66 insertions(+), 28 deletions(-)
>
> diff --git a/builtin/branch.c b/builtin/branch.c
> index 0fe9c4d..5b6d839 100644
> --- a/builtin/branch.c
> +++ b/builtin/branch.c
> @@ -115,8 +115,10 @@ static int branch_merged(int kind, const char *name,
>  		    branch->merge[0] &&
>  		    branch->merge[0]->dst &&
>  		    (reference_name =
> -		     resolve_ref(branch->merge[0]->dst, sha1, 1, NULL)) != NULL)
> +		     resolve_ref(branch->merge[0]->dst, sha1, 1, NULL)) != NULL) {
> +			reference_name = xstrdup(reference_name);
>  			reference_rev = lookup_commit_reference(sha1);
> +		}
>  	}
>  	if (!reference_rev)
>  		reference_rev = head_rev;
> @@ -141,6 +143,7 @@ static int branch_merged(int kind, const char *name,
>  				"         '%s', even though it is merged to HEAD."),
>  				name, reference_name);
>  	}
> +	free((char*)reference_name);
>  	return merged;
>  }

Now reference_name stores the result of xstrdup(), it does not have reason to be of type "const char *". It is preferable to lose the cast here, I think. The same comment applies to the remainder of the patch.

Otherwise the patch looks good. Thanks.
Previous: Nguyễn Thái Ngọc DuyNext: Nguyen Thai Ngoc Duy
Message 12 of 31 in “Re: git bug(?) for commit baf18fc261ca475343fe3cb9cd2c0dded4bc1bb7”
  1. Nguyen Thai Ngoc DuyNov 7, 2011
  2. Tony WangNov 7, 2011
  3. Nguyen Thai Ngoc DuyNov 7, 2011
  4. Tony WangNov 7, 2011
  5. Nguyen Thai Ngoc DuyNov 7, 2011
  6. Copy resolve_ref() return value for longer useNguyễn Thái Ngọc Duy, Nov 8, 2011
  7. Junio C HamanoNov 13, 2011
  8. Nguyen Thai Ngoc DuyNov 13, 2011
  9. Junio C HamanoNov 13, 2011
  10. 1/2 Convert many resolve_ref() calls to read_ref*() and ref_exists()Nguyễn Thái Ngọc Duy, Nov 13, 2011
  11. 2/2 Copy resolve_ref() return value for longer useNguyễn Thái Ngọc Duy, Nov 13, 2011
  12. Junio C HamanoNov 13, 2011
  13. Nguyen Thai Ngoc DuyNov 14, 2011
  14. Junio C HamanoNov 14, 2011
  15. Jeff KingNov 14, 2011
  16. Nguyen Thai Ngoc DuyNov 15, 2011
  17. 01/10 Allow resolve_ref() caller to decide whether to receive static bufferNguyễn Thái Ngọc Duy, Nov 15, 2011
  18. 02/10 cmd_merge: convert to single exit pointNguyễn Thái Ngọc Duy, Nov 15, 2011
  19. 03/10 merge: do not point "branch" to a resolve_ref()'s static bufferNguyễn Thái Ngọc Duy, Nov 15, 2011
  20. 04/10 commit: move resolve_ref() closer to where the return value is usedNguyễn Thái Ngọc Duy, Nov 15, 2011
  21. 05/10 checkout: do not try xstrdup() on NULLNguyễn Thái Ngọc Duy, Nov 15, 2011
  22. 06/10 reflog-walk.c: request allocated buffer from resolve_ref()Nguyễn Thái Ngọc Duy, Nov 15, 2011
  23. 07/10 receive-pack: request resolve_ref() to allocate new bufferNguyễn Thái Ngọc Duy, Nov 15, 2011
  24. 08/10 notes: request resolve_ref() to allocate new bufferNguyễn Thái Ngọc Duy, Nov 15, 2011
  25. 09/10 fmt-merge-msg: request resolve_ref() to allocate new bufferNguyễn Thái Ngọc Duy, Nov 15, 2011
  26. 10/10 branch: request resolve_ref() to allocate new bufferNguyễn Thái Ngọc Duy, Nov 15, 2011
  27. Junio C HamanoNov 15, 2011
  28. Junio C HamanoNov 13, 2011
  29. Tony WangDec 10, 2011
  30. Nguyen Thai Ngoc DuyDec 10, 2011
  31. Tony WangDec 11, 2011

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.