Re: master eb37e4814e 3/3: Fix unlikely null pointer dereference

Re: master eb37e4814e 3/3: Fix unlikely null pointer dereference

[Po Lu]

Paul Eggert <eggert@cs.ucla.edu> writes:

> branch: master
> commit eb37e4814e354befaa12f80dc5e75368ad489a1e
> Author: Paul Eggert <eggert@cs.ucla.edu>
> Commit: Paul Eggert <eggert@cs.ucla.edu>
>
>     Fix unlikely null pointer dereference
>     
>     * src/xselect.c (Fx_get_atom_name): Fix unlikely core dump when
>     build_string is called on a null pointer.  Found by GCC -fanalyzer.
> ---
>  src/xselect.c | 20 ++++++++------------
>  1 file changed, 8 insertions(+), 12 deletions(-)
>
> diff --git a/src/xselect.c b/src/xselect.c
> index ae15fecccc..3f35842daa 100644
> --- a/src/xselect.c
> +++ b/src/xselect.c
> @@ -2454,9 +2454,6 @@ If the value is 0 or the atom is not known, return the empty string.  */)
>    (Lisp_Object value, Lisp_Object frame)
>  {
>    struct frame *f = decode_window_system_frame (frame);
> -  char *name = 0;
> -  char empty[] = "";
> -  Lisp_Object ret = Qnil;
>    Display *dpy = FRAME_X_DISPLAY (f);
>    struct x_display_info *dpyinfo;
>    Atom atom;
> @@ -2468,17 +2465,16 @@ If the value is 0 or the atom is not known, return the empty string.  */)
>  
>    block_input ();
>    x_catch_errors (dpy);
> -  name = (atom ? x_get_atom_name (dpyinfo, atom,
> -				  &need_sync) : empty);
> +  char *name = atom ? x_get_atom_name (dpyinfo, atom, &need_sync) : NULL;
>    had_errors_p = need_sync && x_had_errors_p (dpy);
>    x_uncatch_errors_after_check ();
> -
> -  if (!had_errors_p)
> -    ret = build_string (name);
> -
> -  if (atom && name) xfree (name);
> -  if (NILP (ret)) ret = empty_unibyte_string;
> -
> +  Lisp_Object ret = empty_unibyte_string;
> +  if (name)
> +    {
> +      if (!had_errors_p)
> +	ret = build_string (name);
> +      xfree (name);
> +    }
>    unblock_input ();
>  
>    return ret;

How can name be NULL if had_errors_p is false?  x_get_atom_name always
returns something if need_sync is false, and XGetAtomName always returns
a string unless an error occured.

Re: master eb37e4814e 3/3: Fix unlikely null pointer dereference

[Paul Eggert]

On 5/27/22 17:46, Po Lu wrote:

> How can name be NULL if had_errors_p is false?  x_get_atom_name always
> returns something if need_sync is false, and XGetAtomName always returns
> a string unless an error occured.

Oh, you're right - I hadn't noticed that. So my recent patch didn't fix 
any runtime bugs, though it did pacify GCC and I hope made the confusing 
code a bit easier to follow for humans too.

I see you made further improvements in the area; thanks.

Re: master eb37e4814e 3/3: Fix unlikely null pointer dereference

[Po Lu]

Paul Eggert <eggert@cs.ucla.edu> writes:

> Oh, you're right - I hadn't noticed that. So my recent patch didn't
> fix any runtime bugs, though it did pacify GCC and I hope made the
> confusing code a bit easier to follow for humans too.

Thanks.