[Patch, Fortran] PR fortran/45776: More variable definition checks

[Daniel Kraft]

Jerry DeLisle wrote:
> On 09/24/2010 02:06 PM, Daniel Kraft wrote:
>> Here's an updated patch with the suggestions and IRC discussion 
>> incorporated --
>> earlier than I expected, but anyways ;)
>>
>> What do you think about the new solution? I will run a fresh regtest 
>> tomorrow,
>> but at least the io_*, write_*, read_* and namelist_* tests seem to 
>> pass. Ok?
>>
> 
> Looking pretty good.  What about this?
> 
> -          /* Free this pointer now so that a warning/error is not 
> triggered
> -         below for the "Extension".  */
> -          gfc_free_expr (dt->extra_comma);
> +          /* Nullify this pointer now so that a warning/error is not
> +         triggered below for the "Extension".  */
>            dt->extra_comma = NULL;
> 
> If the *expr extra_comma is not already NULL then how can it's memory be 
> freed later if it is set to NULL here?

Because extra_comma is always (if it is present) a pointer to dt_io_kind 
-- so dt_io_kind is free'ed always, and extra_comma can just be set to 
NULL if we want the information "that it is not present".  As I saw it, 
extra_comma is (if it is set) always just the same expr as dt_io_kind is 
-- thus I'd like to share it.

Of course, if you like it better, I can allocate a new gfc_expr* for 
both of them and keep the old free'ing logic.

Yours,
Daniel

-- 
http://www.pro-vegan.info/
--
Done:  Arc-Bar-Cav-Kni-Ran-Rog-Sam-Tou-Val-Wiz
To go: Hea-Mon-Pri