Re: [PATCH] Potential memory leak in argz_replace.c

[Corinna Vinschen <vinschen at redhat dot com>]

On May 31 21:27, David Stacey wrote:
> On 27/05/2015 11:16, Corinna Vinschen wrote:
> >However, there appear to be more issues with this function.  E.g., when
> >allocating new_argz, it's never tested if the allocation worked.  Also,
> >the expression `*argz = (char *)realloc(*argz, new_argz_len)', when
> >failing, will overwrite *argz with NULL.  When that happens the caller
> >potentially loses its (un-free'd) argz pointer with no way to recover.
> >
> >Care to fix those as well?
> 
> 'Twould be a pleasure.
> 
> Dave.
> 
> newlib/ChangeLog
> 2015-05-31  David Stacey <...>
> 
>     * libc/argz/argz_replace.c: Corrected behaviour when memory is
> exhausted.
> 

Thanks but, uhm, that didn't quite work out:

  libc/argz/argz_replace.c: In function âargz_replaceâ:
  libc/argz/argz_replace.c:75:26: error: assignment makes integer from pointer without a cast [-Werror]
	 if(!(*argz_realloc = (char *)realloc(*argz, new_argz_len)))
			    ^
  libc/argz/argz_replace.c:80:12: error: assignment from incompatible pointer type [-Werror]
	 argz = argz_realloc;
	      ^
  cc1: all warnings being treated as errors

I fixed that and applied the patch.


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer
Red Hat

Attachment: pgpzgkeGhrutr.pgp
Description: PGP signature