Date: Tue, 15 Mar 2005 16:08:32 -0500
Reply-To: Peter Crawford <peter.crawford@BLUEYONDER.CO.UK>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Crawford <peter.crawford@BLUEYONDER.CO.UK>
Subject: Re: Error Message of the century is it a bug?
On Tue, 15 Mar 2005 10:34:56 -0500, Charles Harbour
<charles.harbour@PEARSON.COM> wrote:
>Toby,
>
>I ran into a problem on the mainframe where my job wasn't giving me the
>entire log. My job was looping, so I put logic in the loop to only loop a
>couple of times and print out my loop control values--trouble was, when it
>only looped a couple of times, I never got anything to the log. After
>going a couple of rounds with sas support, they suggested writing the log
>to a file, and make sure that bufno=1. The problem was that the default
>number of buffers (yes, there are buffers even for the log) was something
>like 5, and when the job terminated, instead of writing out what was in
the
>buffers, they got flushed.
>
>I'm not an experienced unix user, but maybe the same thing is happening
>here? An example of the host syntax is:
>
> //SASLOG DD SYSOUT=*,DCB=(LRECL=137,BLKSIZE=137,RECFM=FBA,BUFNO=1)
>
>And if this isn't your problem, this can be a lesson to folks that have
had
>their log eaten by the system.
>
>Hope you get it resolved.
>
>CH
>
>On Tue, 15 Mar 2005 01:11:59 +0000, toby dunn <tobydunn@HOTMAIL.COM>
wrote:
>
>>Jim,
>>
>>I am running V9. on a unix box in batch mode. MAcros are involved but
>since
>>I ussually never nest them more than two deep I don;t have a problem with
>>them killing sas. I sent SAS tech support the log and they proptly
>replyied
>>that they had no clue as to what is going on from teh log and since it
>>worked with ten and fiv breaks in the MLF and it dies some where in te
proc
>>tabulate it has to be with how tabulate handles MLF's. Now Paul Choate
>>(what a great guy he is) worked on it some today and found that the
>>preloadfmt option might be the problem as when one takes it out the error
>>goes away but the tabulate stops somewhere around 3/4 of the way through
>the
>>process and spits out blank cells in the output table. Then I get a
warning
>>about there are some formats that are too small, which I looked at what
>proc
>>format was producing and none should be and when applied by hand via a
>data
>>step work just fine. So I have resolved while using proc tabulate with
MLF
>>would be teh best solution if it would work I must try another method.
So
>>unless one of use has a great aphiany or SAS tech support can figure it
out
>>I am left with doing the hard way.
>>
>>
>>
>>Toby Dunn
>>
>>
>>
>>
>>From: Jim Groeneveld <jim1stat@YAHOO.CO.UK>
>>Reply-To: Jim Groeneveld <jim1stat@YAHOO.CO.UK>
>>To: SAS-L@LISTSERV.UGA.EDU
>>Subject: Re: Error Message of the century is it a bug?
>>Date: Mon, 14 Mar 2005 17:53:53 -0500
>>Hi Toby,
>>
>>The error message is very familiar to us all. It's a standard one.
>>But what do you mean with "SAS died"? What SAS version? Interactive or
>>batch? Macros involved? What memory load?
>>
>>I remember a buggy macro nesting other macros about 147 (if I remember it
>>well) levels deep. SAS (batch) silently died, no errors, no warnings,
>>nothing. Reproducing and debugging did not yield any errors, other than
the
>>deep, but not intended nesting. I programmed counting the levels and SAS
>>always stopped at the same level. Later I heard that that indeed was its
>>fatal, acute hart beat failure. Of course avoiding the deep nesting
solved
>>the problem.
>>
>>See also [those nasty two line long URLs]:
>>http://listserv.uga.edu/cgi-bin/wa?A2=ind0212A&L=sas-
>>l&P=R4810&D=1&H=0&O=D&T=1
>>
>>Now, was this piece of information helpful? Yes / No |OK| ;-)
>>
>>Regards - Jim.
>>--
>>Y. (Jim) Groeneveld, MSc., Biostatistician, Vitatron b.v., NL
>>Jim.Groeneveld_AT_Vitatron.com (replace _AT_ by AT sign)
>>http://www.vitatron.com, http://home.hccnet.nl/jim.groeneveld
>>
>>[common disclaimer]
>>
>>
>>
>>On Fri, 11 Mar 2005 21:25:31 +0000, toby dunn <tobydunn@HOTMAIL.COM>
wrote:
>>
>> >I am creating a program that creates weighted ranges tha get feed into
a
>> >proc format which will create a Multi Label Format. These get feed
into
>a
>> >proc tabulate. While I was trying it out I specified that it give me
20
>> >ranges for both the columns and rows. SAS died and gave me this note:
>> >
>> >NOTE: The SAS System stopped processing this step because of errors.
>> >
>> >There were no other error or warning or notes that would help me out.
>> >Mprint Symbolgen and mlogic are no help.
>> >
>> >
>> >Toby Dunn
a long time ago I had trouble with tabulate on a big mainframe.
(well it was big at the time circa 1990) . The problem was the
memory required for the 5 way crossing that generated a need
for something like 23000 M intersection nodes.... and that was
just from 3500 obs.
That time I reduced cardinality (the number of class levels) .
Since then, I'm assured, tabulate statistic gathering has been
improved.
MLF would need some different re-engineering, I expect.
The tabulate intersection that is 20 by 20 doesn't seen too tough!
However 20 is at one of the old tabulate limits.
Perhaps it is not worth trying the depth= option because it seems
to make no difference in my sas9.
What did make a difference is the output destination. My old
problem continues for the listing destination. When I switch to html,
I get an error message at depth of 19 class vars in one table.
ERROR: Limit of 2147483647 potential interactions within a crossing
exceeded for the table statement at line 27334.
You don't report this message, but with different ods destination
..... who knows ?
Good Luck
Peter Crawford
p.s.
to unbuffer your log, use
-unbufflog
then you might see the last messages sent by your sas session to the log.
Peter
| 
