new gfortran failures on darwin8

[Andreas Tobler]

Jerry DeLisle wrote:
> Jack Howarth wrote:
>> Jerry,
>>    We are still seeing these additional failures on
>> powerpc-apple-darwin8/9 with current gcc trunk.
>>           Jack
>>
>> On Tue, Jun 26, 2007 at 10:35:18AM -0700, Jerry DeLisle wrote:
>>> My recent patch to change the error format to include the file name 
>>> and unit number may be the culprit.  This touched on gfortran.h so 
>>> you may need to bootstrap and see if that fixes it.
>>>
>>> If not, I will work with you to sort this out.
>>>
>>> error_format is a new test case that went with the patch.  read_comma 
>>> I never encountered a problem with.
> 
> Please try this simple patch first.  If it does not work, let me know 
> and I will work on Plan B. :)  The patch is to unit.c in libgfortran/io/
> 
> I see that find_unit sets the private lock of the unit and requires 
> unlock_unit.  Why it works on some systems without the unlock I do not 
> know.
> 
> Thanks for reporting and testing.

Here it doesn't seem to work. (Only rebuilt libgfortran on ppc, x86 will 
follow just for convenience). And sorry for the delay. Lots of non gcc 
stuff.

Regarding the cpu cycles, here the tests just hangs, no cpu cycles wasted.

Thanks,
Andreas