>> Rexx covers this by guaranteeing that if a single clause contains
(any number of) references to TIME() and DATE() the TOD clock will be
read only once and that value used for all.  The TIME macro should nave no
need to read the TOD clock more than once. 

Paul,

If that is the definition of the midnight pitfall, then DFSORT already 
covers that. If you used DFSORT date formats (Date1 thru Date5 ), DFSORT 
reads the TIME macro once and stores the value. It is not changed in 
between despite how many times you have the references of Date parm in 
your control cards.

Thanks,
Kolusu
DFSORT Development
IBM Corporation



From:   Paul Gilmartin <[email protected]>
To:     [email protected]
Date:   09/13/2016 09:26 AM
Subject:        Re: DFSORT - ICETOOL - Search for text and replace with 
date
Sent by:        IBM Mainframe Discussion List <[email protected]>



On Tue, 13 Sep 2016 09:06:28 -0700, Sri h Kolusu  wrote:

>>> You could consider:  //SYSUT1  DD  *,SYMBOLS= instead of ICETOOL and 
>substitute the dynamic system symbols for &YYMMDD and &HHMMSS, however:
>
>What exactly is the midnight pitfall? If you used the LOCAL system 
symbols 
>then you would get the local time.  DFSORT also has several date formats 
>and it is based on the time the step ran. 
> 
It's probably covered, depending on how "the time the step ran" is 
obtained.

>If you meant that a job that started at 11:59:59 PM and the SORT step ran 

>at 00:00:10 then we would get the next date as SORT would not have any 
>idea about when the Job started. 
> 
The pitfall, however narrow, is that if midnight occurs between the 
evaluations
of &YYMMDD and &HHMMSS the composite value may be in error by (almost)
24 hours.  Rexx covers this by guaranteeing that if a single clause 
contains
(any number of) references to TIME() and DATE() the TOD clock will be
read only once and that value used for all.  The TIME macro should nave no
need to read the TOD clock more than once.  UNIX time() function returns
a single raw RTC value and provides functions to format that.

I suppose that if processors are fast enough that the gap, accumulated 
over
the expected life of all installed z/OS systems, meets a six sigma 
criterion
it's not worth concern.  Has anyone done the calculation?

-- gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN






----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to