genic.f miscompiles with -O2 or greater

[Tobias Burnus]

On 07/26/2010 09:07 AM, Tobias Schlüter wrote:
> I think -fwhole-file might be able to look through the pointers, and
> then gets utterly confused.

Difficult to tell without a complete test case, but I think it might be
the middle end which you confuse by thinking that two objects do not
alias as they have a different type.  Does -fno-strict-aliasing work?

But as written, without knowing at which optimization level it occurs
and without a full test case, we can do little. (Except of trying to
create a test case to reproduce it, but the bottom up method is much ore
time consuming than the top-down method. Besides there are still enough
-fwhole-file gfortran bug - thus I am too lazy to try finding a test case.)

Tobias