genic.f miscompiles with -O2 or greater

[Tobias Burnus]

Jack Howarth wrote:
> On Sun, Jul 25, 2010 at 10:17:36PM +0200, Tobias Burnus wrote:
>   
>> Jack Howarth wrote:
>>     
>>> A reduced example of this is...
>>> [...]
>>>   
>>>       
>> That is overly reduced. Unless, you have a complete, compilable run-time
>> example, it is difficult to find out what goes wrong
> I think the testcase captures the core of the problem.

Well, it might illustrate a problematic part, but it is not a (runable)
test case.

> [...] The problem appears to be that
>       character*4 iac(*)
> when passed to...
>       SUBROUTINE cmpric(iac)
> [...]
>   

Does it work if you compile with -O0? Does it work if you add TARGET?

I refuse to think or look at the problem in more detail without a
runable test case (i.e. one which fails due to this problem) - it takes
simply too much time, I would like to use for other things.

> doesn't consider iac to be a pointer in
> the argument list since it is declared as
> iac(*) within the subroutine. What do other
> fortran compilers do with this code?

In (g)Fortran, all arguments (except when using %VAL/VALUE) are
pointers. And I have no idea without a small test case what other
compilers do.

Tobias