[PATCH] dir.c: clean the entire struct in clear_exclude_list()

Nguyễn Thái Ngọc Duy· Dec 27, 2015, 01:54 UTC · lore

Make sure "el" can be reuseable again. The problem was el->alloc is not cleared and may cause segfaults next time because add_exclude() thinks el->excludes (being NULL) has enough space. Just clear the entire struct to be safe.

Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
---
 This was needed in the worktree config split series because the same
 exclude_list could be reused. But that series is changing and will not
 need exclude_list any more. So instead of dropping this patch, I
 submit it separately. It's a good fix anyway.

 dir.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

Show changes to dir.c +1 −3

diff --git a/dir.c b/dir.c
index d2a8f06..b8b4576 100644
--- a/dir.c
+++ b/dir.c
@@ -564,9 +564,7 @@ void clear_exclude_list(struct exclude_list *el)
 	free(el->excludes);
 	free(el->filebuf);
 
-	el->nr = 0;
-	el->excludes = NULL;
-	el->filebuf = NULL;
+	memset(el, 0, sizeof(*el));
 }
 
 static void trim_trailing_spaces(char *buf)

-- 
2.3.0.rc1.137.g477eb31

Junio C Hamano· Dec 28, 2015, 20:47 UTC · re: Nguyễn Thái Ngọc Duy · lore

Re: [PATCH] dir.c: clean the entire struct in clear_exclude_list()

Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:

> Make sure "el" can be reuseable again. The problem was el->alloc is
> not cleared and may cause segfaults next time because add_exclude()
> thinks el->excludes (being NULL) has enough space. Just clear the
> entire struct to be safe.

OK.

If somebody ever adds a field to this structure that should not be cleared by filling it with '\0', that person would notice and turn this memset() back to individual field assignment, but until then, I think memset is a reasonable thing to do, as that is how it is initialized by add_exclude_list() in the first place anyway.

Will apply.

Show 25 quoted lines

> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> ---
>  This was needed in the worktree config split series because the same
>  exclude_list could be reused. But that series is changing and will not
>  need exclude_list any more. So instead of dropping this patch, I
>  submit it separately. It's a good fix anyway.
>
>  dir.c | 4 +---
>  1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/dir.c b/dir.c
> index d2a8f06..b8b4576 100644
> --- a/dir.c
> +++ b/dir.c
> @@ -564,9 +564,7 @@ void clear_exclude_list(struct exclude_list *el)
>  	free(el->excludes);
>  	free(el->filebuf);
>  
> -	el->nr = 0;
> -	el->excludes = NULL;
> -	el->filebuf = NULL;
> +	memset(el, 0, sizeof(*el));
>  }
>  
>  static void trim_trailing_spaces(char *buf)

patchdir.c: clean the entire struct in clear_exclude_list()