(j3.2006) when does an error condition occur in executing EXECUTE_COMMAND_LINE

Michael Ingrassia michaeli
Fri Jan 16 16:09:28 EST 2009


>What is there to standardize?

You're making a good case for eliminating EXITSTAT, CMDSTAT, and
CMDMSG from the description of the intrinsic!
But the committee thought that having them added value.  I am not advocating
adding EXITSTAT for assignment statements!  But if someone is, I want them
to spell out how it works and what guarantees are possible.  

I'm pointing out that while EXITSTAT/CMDSTAT/CMDMSG 
do seem to be in the spirit of the error conditions
that the standard specifies for I/O and for ALLOCATE, they don't 
seem to work the same way. 

I think that a processor that segfaults in an assignment statement
in a standard conforming program does not conform to the standard.

I think the current draft guarantees that a command is executed when
EXECUTE_COMMAND_LINE is called.

I think that if EXECUTE_COMMAND_LINE is called with a single argument
and the command isn't executed, then the processor does not conform to the
draft standard.

I think that if a vendor actually detects that a command wasn't executed,
and actually returns a non-zero code that indicates that, then 
the processor still does not conform to the draft.  Because the draft
says that the command will be executed.  This is the part I want to change.
I would like a processor that detects an error condition to have a legitimate
excuse which keeps them standard conforming.

Clearer? 

	--Michael I.




 	--Michael I.



More information about the J3 mailing list