[gfortran] Fix PR 24643

[Tobias Schlüter]

Paul Thomas wrote:
>>Coming to think of it, I'll not submit the testcase, as it would be testing
>>exactly the same thing.
> 
> It would not hurt to modify your new testcase to include the if 
> statement and to reference the PR, would it?

I don't think it will be of any benefit either.  The code takes exactly the
same route, and its exactly the same condition for the bug to occur, both
times we're parsing an expression.  As for adding the PR to the testcase, if
someone really wants to do some archaeology he can look at PR 24643 and look
at its duplicates.

If you think that it has any benefits, I will gladly add this, I just can't
think of any.

- Tobi