[PATCH, gfortran] Re: Cray Pointers

[Asher Langton]

At 10:10 AM -0400 8/15/05, Jack Howarth wrote:
>Asher,
>    The discussion of cray pointers being incompatible with modules

Cray pointers aren't incompatible with modules, though there was a 
bug which was fixed in last Friday's patch.

>reminds me of an oddity that I saw testing xplor-nih when built with
>gfortran on MacOS X 10.4. I found that one of the tests core dumped
>on exiting...
>
>http://gcc.gnu.org/ml/fortran/2005-08/msg00135.html
>This only occurs when I build xplor-nih with -fcray-pointer and seems
>to occur when using one of the xplor-nih c++ modules that calls
>fortran modules recursively. This is despite the fortran only using
>LOC calls and no explicit cray pointers. When I drop the -fcray-pointer
>flag the segfault goes away

That sounds like a problem.  If there are no Cray pointers in a code, 
toggling the fcray-pointer flag should have no effect.  I will look 
through the patch today for any place where this could be an issue. 
We were not able to reproduce that bug on our end, though.

>. Is it possible that the incompatibility of
>cray pointers and modules extends to the LOC call itself?

The loc intrinsic implementation and the Cray pointer implementation 
are not related.  There is no common code.


-Asher