Re: [PATCH] time: strftime_l: Use malloc rather than an unbounded alloca.

[Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>]

On 10/05/23 16:59, Joe Simmons-Talbott via Libc-alpha wrote:
> Avoid possible stack overflow by replacing alloca() with malloc().
> ---
>  time/strftime_l.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/time/strftime_l.c b/time/strftime_l.c
> index 402c6c4111..59d3e1a3b2 100644
> --- a/time/strftime_l.c
> +++ b/time/strftime_l.c
> @@ -273,8 +273,9 @@ static const CHAR_T zeroes[16] = /* "0000000000000000" */
>      const char *__s = os;						      \
>      memset (&__st, '\0', sizeof (__st));				      \
>      l = __mbsrtowcs_l (NULL, &__s, 0, &__st, loc);			      \
> -    ws = alloca ((l + 1) * sizeof (wchar_t));				      \
> -    (void) __mbsrtowcs_l (ws, &__s, l, &__st, loc);			      \
> +    ws = malloc ((l + 1) * sizeof (wchar_t));				      \
> +    if (ws != NULL)							      \
> +      (void) __mbsrtowcs_l (ws, &__s, l, &__st, loc);			      \
>    }
>  #endif
>  
> @@ -1346,7 +1347,10 @@ __strftime_internal (CHAR_T *s, size_t maxsize, const CHAR_T *format,
>  	    wchar_t *wczone;
>  	    size_t len;
>  	    widen (zone, wczone, len);
> +	    if (wczone == NULL)
> +	      return 0;
>  	    cpy (len, wczone);
> +	    free (wczone);
>  	  }
>  #else
>  	  cpy (strlen (zone), zone);

Do we have a practical maximum size for the abbreviate timezone name?  The
internal tz_rule 'name' field is just a pointer, but I think all timezones
uses a maximum name size.