John is right. I was not clear in my first note, although the second
one spells out "accessing and updating the callers variables". Most of
the *vars* in CALLPIPEs use the numeric operand that accesses a higher
level caller's variables. As I said elsewhere, the problem seems to be
more limited than I thought: It happens when we use CALLPIPE CMS XXX
and XXX uses PIPE ... *VAR* 1 ... . That is not a big enough problem
that we can't work around it.
There are no RITA EXECs or PIPE EXECs.
I meant what I said about the "pipe" global variable:
Pipe = "rita" /* actually set from a config file, ... or not. */
Pipe "(listerr ... )",
We don't use ADDRESS COMMAND. We exploit and depend on ADDRESS CMS.
Sorry for any confusion.
Would anyone care to answer my questions about the ALL stage?
-----Original Message-----
From: CMSTSO Pipelines Discussion List
[mailto:[EMAIL PROTECTED] On Behalf Of John P. Hartmann
Sent: Thursday, May 29, 2008 2:51 PM
To: [email protected]
Subject: Re: [CMS-PIPELINES] FW: Rita+*VAR*+CALLPIPE
>>> PIPE is "pipe", and not when it's "rita"
You didn't mean that, right? You meant that pipes are issued with
ADDRESS
COMMAND PIPE and Rita is fired up with ADDRESS COMMAND RITA, right?
If not I'll bet you half a shiny bit that you have a RITA EXEC around
(or
possibly a PIPE one, but that would be even more mindboggling).
j.
2008/5/29 John P. Hartmann <[EMAIL PROTECTED]>:
> I suspect Hobart is not telling us the whole truth about his
experiments.
> In particular we might be held in the dark as to which other operands
he is
> using on the REXX variable access stages.
>
> /*
> /*
> Signal on novalue
> pipe='rexxvars|take 1|console'
> Address COMMAND
> 'PIPE' pipe
> 'RITA' pipe
> Exit RC
>
> Gets me this response (as it should):
>
> s CMS COMMAND TR EXEC A1 TR
> CMS
> s CMS COMMAND TR EXEC A1 TR
> CMS
> CPU Utilization by Pipeline Specification 29 May 2008
> 20:35:34
>
>
> 0.099 ( 0.099) ms total in "NoName001" (1
> invocation)
>
>
> 2.015 ms
> total.
>
>
> Detailed output from Rita in UNNAMED RITA007.
>
> Adding MAIN to the rexxvar stage gets the same response (as it
should).
>
> Yes, RITA does run a REXX filter to process the output of RUNPIPE.
And
> yes, if you go out of your way you might be able to interfere with it,
> though I cannot think of how to do so off the top of my head.
>
>
> j.
>
> 2008/5/29 SPITZ, HOBART CTR DFAS <[EMAIL PROTECTED]>:
>
> That's exactly what it sounds like, but, per p.3 of "Streamlining Your
>> Pipes", we are using RITA MODULE. (I don't think we actually had a
>> choice. :-) )
>>
>> We use a configuration variable PIPE that we set to RITA when we want
to
>> get performance info., so we know that the programs haven't changed.
>> Our pipe statements look like:
>>
>> Pipe "(listerr end ? name" UsedName"."ExecType"."LineNo()")",
>> ...
>>
>> A number of programs all show the same problem. Since these programs
>> sub-filters all receive and set the callers variables correctly when
>> PIPE is "pipe", and not when it's "rita", it would seem to be
something
>> external to the programs.
>>
>> Any ideas?
>>
>> - Hobart
>>
>> -----Original Message-----
>> From: CMSTSO Pipelines Discussion List
>> [mailto:[EMAIL PROTECTED] On Behalf Of Rob van der Heij
>> Sent: Wednesday, May 28, 2008 5:14 PM
>> To: [email protected]
>> Subject: Re: [CMS-PIPELINES] FW: Rita+*VAR*+CALLPIPE
>>
>> On Wed, May 28, 2008 at 4:08 PM, SPITZ, HOBART CTR DFAS
>> <[EMAIL PROTECTED]> wrote:
>>
>> > When we run RITA and we have a filter, any *VAR* stages (REXXVARS,
>> VAR,
>> > VARLOAD, VARSET, etc.), in a CALLPIPE, seem to not be able to get
the
>> > variables that they get when we run without RITA.
>> >
>> > Are we doing something wrong?
>>
>> Could it be you have a PIPE or RITA EXEC involved. That would
>> introduce another level of execcomm and mess up things.
>>
>> Rob
>>
>
>
