SP23_I1/8.94 on iSeries - beware

[bmarxhau]

This notice was attached to 8.94.C1 on 1/28 .

Clients running EnterpriseOne iSeries Enterprise Server code compiled on
V5R3 machine without the Hyper PTF/PTFs specified in the IBM info APAR and
running on V5R2 or V5R3 level may experience the following Kernel crashes
with MCH0601 error in the joblog.
From module . . . . . . . . : MUL_1
From procedure . . . . . . : __mpn_mul_1 Statement . . . . . . . . . : 7 To
module . . . . . . . . . : MUL_1 To procedure . . . . . . . : __mpn_mul_1
Statement . . . . . . . . . : 7 Message . . . . : Space offset X'00000000'
or X'0000007F06CDA640' is outside current limit for object JDENET_K ONEWORLD
295085

Client can perform a DSPSRVPGM GNUMP and see build information in the GNUMP
*SRVPGM created between 12/17/04 and 1/27/05. The information will appears
as follows (for example):

Release service program created on . . . . . . . : V5R3M0 Release service
program created for . . . . . . : V5R2M0 Earliest release service program
can run . . . . : V5R2M0



PeopleSoft claimed that it would be resolved with 8.94.C2. However, 8.94.C2
was not packaged correctly, and does NOT resolve the issue. I have two
customers who are experiencing this when running pre-payroll on the iSeries.
One is V5R2 one is V5R3, both are 8.94.C2 customers. The job logs contain
the above errors, the jdedebug logs end with 'Entering MathDivide' This is
not limited to the JDENET_K, but also any UBE that uses BSFNs that utilize
MathDivide functionality. Compling the code with V5R2 flags does NOT
resolve the issue. As stated above, the issue lies with the version of the
GNUMP *SRVPGM that was delivered with 8.94. Using the GNUMP from your 8.93
backup library with the rest of the 8.94 tools release seems to be a good
work around. No final word on when it will be corrected. Seems to also be
present in SP23_I1.

Brent Marxhausen
DBT, Inc.
Independent CNC
EMail: [email protected]

 

[sashton]

We hit this same problem when we installed SP23_I1 on 1/13/05 on ERP 8. We figured it out on our own that you had to swap the GNUMP file from your backup. PeopleSoft claims this is resolved with I2 however, I downloaded I2, stripped out the GNUMP file, swapped it, and we had the same problem. Swapped it back with the old one and we were fine again. I still believe something else is wrong with that program.

 

[rab85020]

We had this same problem (MathDivide/GNUMP) on 8.93.Q1. We are current are
V5R2 with current PTFs.



We loaded R1 and the problem is now fixed. Response Line was under the
impression that the problem was fixed in Q1.

 

[arim_lakha]

Thanks for posting this - we have come across this very same issue and using an old GNUMP (from SP23 lib) resolved it.
Questions for you if you don't mind:
a) Did you hear back from Denver re:ETA for fix?
b) Did this workaround have an inadvertent effect on other applications?

 

[bmarxhau]

It was fixed with 8.94_D1, then broke again with 8.94_F1. Not sure where it
stands now, we just stuck with the older version of GNUMP.

It did affect other applications. We had some issues in A/P processing that
were related to this. Just watch the jdedebug.log on the server to end with
'Entering jde_MathDivide', and a corresponding job log created with
reference to GNUMP.


Brent Marxhausen





_____

From: [email protected] [mailto:[email protected]] On
Behalf Of arim_lakha
Sent: Monday, May 16, 2005 12:42 PM
To: [email protected]
Subject: Re: SP23_I1/8.94 on iSeries - beware


Thanks for posting this - we have come across this very same issue and using
an old GNUMP (from SP23 lib) resolved it.
Questions for you if you don't mind:
a) Did you hear back from Denver re:ETA for fix?
b) Did this workaround have an inadvertent effect on other applications?


_____


The entire <http://www.jdelist.com/ubb/showflat.php?Cat=&Board=> JDELIST
thread is available for viewing.


Looking for a job? Check out the Job


This is the JDELIST EnterpriseOne Mailing List.
The instructions on how to unsubscribe from any JDELIST mailing list are
available here <http://www.jdelist.com/unsubscr.shtml> .
JDELIST is not affiliated with JDEdwardsR.