Re: Two internal compiler errors

[Janus Weil <janus at gcc dot gnu dot org>]

> it is bug #47728.

Thanks.


> As the other one seems to be already covered, I have not made a PR out of that.
> (Unless you think I should)

Well, I'd say it would not hurt. We now have at least three
independent reports of this issue, and these things tend to get lost a
bit if they just appear as a side-note in some other PR. It's up to
you.

Cheers,
Janus




> 2011/2/14 Arjen Markus <arjen.markus895@gmail.com>:
>> Hi Janus,
>>
>> I will open a PR for both.
>>
>> Regards,
>>
>> Arjen
>>
>> 2011/2/13 Janus Weil <janus@gcc.gnu.org>:
>>> Hi Arjen,
>>>
>>>> while experimenting with the object-oriented features of Fortran 2003
>>>> I ran into a few ICEs.
>>>
>>> thanks for the test drive :)
>>>
>>>
>>>> The compiler stops with a message that there was a segmentation fault,
>>>> the reported line is line 56.
>>>> (No syntax errors or other messages of any kind are reported)
>>>
>>> Ok, I can confirm this with the current trunk and the following reduction:
>>>
>>> program test_objects
>>>
>>> ? implicit none
>>>
>>> ? type, abstract :: shape
>>> ? end type
>>>
>>> ? type, extends(shape) :: rectangle
>>> ? ? ? real :: width, height
>>> ? end type
>>>
>>> ? class(shape), dimension(2) :: object
>>>
>>> ? object(1) = rectangle( 1.0, 2.0 )
>>>
>>> end program test_objects
>>>
>>>
>>> There are at least two problems with this:
>>> 1) Your "object" is missing one of the attributes POINTER or
>>> ALLOCATABLE. Declaring "plain" class objects like this is forbidden.
>>> 2) gfortran currently does not support dimensionful class variables.
>>> This of course is not a flaw of your program, but of gfortran ;)
>>>
>>> However, it seems the respective error messages are being swallowed by
>>> the ICE (which is caused by the last line). Could you open a PR for
>>> this? I will take care of fixing it ...
>>>
>>> Cheers,
>>> Janus
>>>
>>
>