Re: [PATCH v2 01/10] argp-parse.c (__argp_input): Don't crash if pstate is NULL

[Bruno Haible]

Robbie Harwood wrote:
> From: Colin Watson <cjwatson@ubuntu.com>
> 
> [rharwood@redhat.com: tweaked commit message]
> Signed-off-by: Robbie Harwood <rharwood@redhat.com>
> ---
>  lib/argp-parse.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/lib/argp-parse.c b/lib/argp-parse.c
> index 053495ec0..4f1c65d73 100644
> --- a/lib/argp-parse.c
> +++ b/lib/argp-parse.c
> @@ -940,7 +940,7 @@ weak_alias (__argp_parse, argp_parse)
>  void *
>  __argp_input (const struct argp *argp, const struct argp_state *state)
>  {
> -  if (state)
> +  if (state && state->pstate)
>      {
>        struct group *group;
>        struct parser *parser = state->pstate;

I don't think this patch is needed, because:

1) The application cannot construct a 'struct argp_state' by itself, since [1]
   says that the 'struct argp_state' contains a member 'pstate' that is
   "Private, for use by the argp implementation.".

2) The only place in the gnulib / glibc code where a 'struct argp_state' is
   being constructed, is in function parser_init, invoked from 'argp_parse',
   and there a non-NULL value is assigned.

In other words, there is no way, compliant with the documented API, that a
NULL pointer can arise as state->pstate.

Bruno

[1] https://www.gnu.org/software/libc/manual/html_node/Argp-Parsing-State.html