[patch, fix] Formatted output of IEEE exceptional values

[Jerry DeLisle]

Jerry DeLisle wrote:
> Tobias Schlüter wrote:
> 
>> Jerry DeLisle wrote:
>>
>>> The attached patch fixes the formatting discussed in:
>>>
>>> http://gcc.gnu.org/ml/fortran/2005-07/msg00169.html
>>>
>>> 2005-07-13 Jerry DeLisle  <jvdelisle@verizon.net>
>>>
>>>     * libfortran/write.c: Fix field width checks for printing 'Infinity'
>>>     or 'Inf'.
>>>
>>> Ok to commit to mainline?
>>
>>
>>
>> Ok.  Also for 4.0.
> 
> 
> Commited to 4.1.  I am getting a regression failure on the 4.0 branch 
> with this.  I will investigate and report in a bit.
> 
> Jerry
> 
There is a previous test case in gfortran.fortran-torture/execute that I 
failed to see before.  When I tested I did not see the failure on 4.1 
branch.  (I must have been up too late).  I am bootstrapping the latest 
after my commit today to see what it is doing.

Regardless, it gets down to this.  The test case expects "Infinity" with 
no sign for both positive and negative infinity when the field width is 
8.  This makes no sense, but looking at the related previous pr12839 it 
talks about F2003 standard.  I don't have a way to to look at F2003.  So 
  if there is some one who can or knows, please advise.

In the meantime, I propose changing the test case to expect "    -Inf" 
and "    +Inf" for a field width of 8 following common sense.

Is this OK?

Regards,

Jerry

PS Paul Thomas once asked me if I get any sleep :)