(j3.2006) 09-119

[Robert Corbett]

John Reid wrote:
> 
> Van Snyder wrote:
> 
>>09-119 remarks that 9.5.4p4 prohibits a unit from being connected to
>>more than one file, and file from being connected to more than one unit.
>>It proposes an exception is needed for OUTPUT_UNIT and ERROR_UNIT, in
>>that OUTPUT_UNIT is preconnected to the same file on every image, and
>>similarly for ERROR_UNIT.
>>
>>9.5.1p3 says that an external unit identified by a particular value is
>>the same external unit in all program units of a program.  Is it the
>>same external unit on all images?  This is the place to say, one way or
>>another.
> 
> 
> In think 2.3.2 is pretty clear:
> 
> "Each image has its own execution state, floating-point status (14.7), and set 
> of data objects, input/output units, and procedure pointers."
> 
> Cheers,
> 
> John.

Yes, that certainly does look like a problem.

Your proposed solution to that problem raises the issue of how output
to those units is to be synchronized.  Must the user arrange for the
synchronization in his code or is the Fortran processor required to
do the synchroniztion?  Is the synchonization to be done at the record
level?  If so, how is nonadvancing I/O to be handled?

Bob Corbett