Re: [PATCH] fortran: Restore interface to its previous state on error [PR48776]

[Mikael Morin <morin-mikael@orange.fr>]

Le 28/08/2023 à 21:17, Harald Anlauf via Fortran a écrit :
> Hi Mikael,
> 
> On 8/27/23 21:22, Mikael Morin via Gcc-patches wrote:
>> Hello,
>>
>> this fixes an old error-recovery bug.
>> Tested on x86_64-pc-linux-gnu.
>>
>> OK for master?
> 
> I have only a minor comment:
> 
>> +/* Free the leading members of the gfc_interface linked list given in 
>> INTR
>> +   up to the END element (exclusive: the END element is not freed).
>> +   If END is not nullptr, it is assumed that END is in the linked 
>> list starting
>> +   with INTR.  */
>> +
>> +static void
>> +free_interface_elements_until (gfc_interface *intr, gfc_interface *end)
>> +{
>> +  gfc_interface *next;
>> +
>> +  for (; intr != end; intr = next)
> 
> Would it make sense to add a protection for intr == NULL, i.e.:
> 
> +  for (; intr && intr != end; intr = next)
> 
> Just to prevent a NULL pointer dereference in case there
> is a corruption of the chain or something else went wrong.
> 
This would happen in the case END is not a member of the INTR linked 
list.  In that case, the most forgiving would be not freeing any memory 
and just returning.  But it would require walking the list a second time 
to determine before proceeding if END is present, and let's not do work 
that is expected to be useless.

I will just do the change as you suggest it.

> Otherwise it looks good to me.
> 
> It appears that your patch similarly fixes PR107923.  :-)
> 
Good news. :-)
I will double check that none of the testcases there remain unfixed and 
close as duplicate.

I don't know how you manage to make your way through the hundreds of 
open PRs by the way.

Thanks for the review.

> Thanks for the patch!
> 
> Harald
> 
>