[patch, fortran] Fix PR38709, ice-on-invalid-code

Paul Richard Thomas paul.richard.thomas@gmail.com
Fri Apr 10 14:00:00 GMT 2009


On Fri, Apr 10, 2009 at 3:33 PM, Daniel Franke <franke.daniel@gmail.com> wrote:
> On Friday 10 April 2009 15:19:34 you wrote:
>> > On Saturday 04 April 2009 18:31:44 Daniel Franke wrote:
>> >> Attached patch fixes an ICE-on-invalid with zero-sized arrays in
>> >> initialization expressions.
>>
>> On the face of it this looks OK for trunk, up to a question:
>>
>> Is it not worthwhile verifying that the size really is zero by
>> counting the elements rather than checking the shape?  The reason that
>> I ask is that the shape can come fairly late in the game, IRC.  Mind
>> you, for it to be useful, you would then have to go off and determine
>> the shape...
>
> Paul,
>
> please note the assert on the presence of the shape that has been removed.
> Further, the shape is used just below the change. So, generally we have a
> shape here but in the presence of an empty array?!

Duuuh!  OK, then:-)

Paul



More information about the Gcc-patches mailing list