Re: [PATCH V4 5/6] Resolve dynamic target types of pointers.

[Bernhard Heckel <bernhard.heckel@intel.com>]

On 17/10/2016 13:24, Bernhard Heckel wrote:
> On 17/10/2016 12:32, Yao Qi wrote:
>> Bernhard Heckel <bernhard.heckel@intel.com> writes:
>>
>>>>> +set test "print intap, not associated"
>>>>> +gdb_test_multiple "print intap" $test {
>>>>> + -re " = \\(PTR TO -> \\( $int \\(:,:\\)\\)\\) <not
>>>>> associated>\r\n$gdb_prompt $" {
>>>>> +    pass $test
>>>>> +  }
>>>>> +  -re " = <not associated>\r\n$gdb_prompt $" {
>>>>> +    pass $test
>>>>> +  }
>>>>> +}
>>>> Why do we need to different regex here?  Need to handle the 
>>>> differences
>>>> of different compilers?  IMO, the former looks reasonable to me.
>>> A pointer to an array (especially to Variable Length Array, VLA) is
>>> described in Gfortran's  compiled debug info
>>> like normal array types + associated property of an pointer.
>> Could you show the debug information gfortran and ifort generates
>> respectively?
>
> see attachment
>
>>
>>> Not sure if this is a bug in Gfortran or not? I was also thinking to
>>> make the
>>> output equal, and check if an array has pointer attributes but ..... I
>>> don't really like this approach. This would
>>> cause a lot of "if .. else" construct in the code for example:
>>> De-reference this pointer, Address operator, ...
>> If gfortran people confirm it is a bug, we can xfail the test.

I just looked again in the specification. It is not a bug.
FORTRAN specification (http://www.j3-fortran.org/doc/year/04/04-007.pdf) 
mention this as "array pointer"

Intel Deutschland GmbH
Registered Address: Am Campeon 10-12, 85579 Neubiberg, Germany
Tel: +49 89 99 8853-0, www.intel.de
Managing Directors: Christin Eisenschmid, Christian Lamprechter
Chairperson of the Supervisory Board: Nicole Lau
Registered Office: Munich
Commercial Register: Amtsgericht Muenchen HRB 186928