[patch, libfortran] Fix matmul result zeroing for empty arguments

[Jerry DeLisle]

On 06/05/2017 02:52 AM, Thomas Koenig wrote:
> Hello world,
> 
> the attached patch moves the zeroing of the result variable before the
> early return. This is done so that the result for zero-sized arguments
> is still valid.
> 
> The bug was only in the library version, but I have also added a test
> case for the inline version to make sure the bug does not suddenly
> appear there.
> 
> OK for trunk?


Yes OK for Trunk and 7

Thanks,

Jerry