[patch,fix, PING] PR21875 Internal Array IO and NIST final fix

[Jerry DeLisle]

Steve Kargl wrote:

> 
> If you think that further work on gfortran may
> accidently lead to a situation of len < 0, then
> you can change it to gcc_assert(*len >= 0).
> 

I see that gcc_assert() does not exist on the libgfortran side and is not used 
anywhere there.  I will add this to my todo list for a future addition.

I confirmed on all my test cases that len is not going negative so I have 
deleted that bit of code in question.  Regression tested again with all OK.  I 
have cleaned up the formatting issues.

I will commit this patch to 4.1 branch later today.  Will wait for 4.0.3 on that 
branch.

Regards, and thank you for everyones support and attention.

Jerry