PRs 28971 and 28993

[Paul Thomas]

I propose to clear the two above PRs tomorrow morning, unless anybody 
objects; both are marked as "severe" but both test cases work with the 
most recent gfortran.  I have to confess that one (28971) was a 
regression in that it is OK with GCC 4.0.3 20051106, did not work with 
4.2.0 in July but is OK now.  Neither Steve nor I can see what fixed it 
and I do not have the time or the inclination to find out by reverting 
my tree to that time.  Maybe we should put the testscase in the 
testsuite to guard against a further regression?  The other is filed 
against gcc-4.0.3 but does compile fine 4.0.3, dated as above, and any 
subsequent version.

Thank you both for the PRs.  It would be very helpful if you would both 
update to the latest gfortran and would continue to post such problems 
as you find to Bugzilla.

Paul