SQL to Excel crashing since version 6.2.20230927.0

[petersioufi]

All our excel reports are now experiencing MCH check "Pointer not set for location referenced" in vvOut_execSQLtoSS. We traced with debug to a null pointer error not checked properly. See attached jpeg. We checked with later build, 6.2.20231026.0 and we have same error.

If we respond to error message with "G", it works second time. Not sure if that is for all cases.

This is critical for us because all our nightly reporting crashes and holds up processing.

We have dumps if that helps, Error message sent to QSYSOPR is RNQ0202.

[robert.swanson]

Hi Peter,

It would be helpful to see the call stack of the job when this occurs. If these Excel download jobs are running in batch (meaning in a non-CGI job) then I suspect some sort of exception logic is being triggered that fails when executed outside of CGI. A look at the call stack would help confirm that.

[petersioufi]

This is the call stack at the point of crash with "Pointer not set for location referenced."
NB: This is a batch RPG program PSR900 (not part of Valence CGI jobs) doing a stand-alone call to vvOut_execSQLtoSS

Type Program Statement Procedure
QCAPCMD QSYS /016E
PSR900 APPOBJ _QRNP_PEP_PSR900
PSR900 APPOBJ 9100 PSR900
VVSRVPGM VALENCE6P 206000 VVOUT_EXECSQLTOSS
VVSRVPGM VALENCE6P 504500 VVOUT_INITIALIZE
VVSRVPGM VALENCE6P 374300 VVUTILITY_GETVALENCESETTING
VVSRVPGM VALENCE6P 389909 VVUTILITY_REPLACESUBS
VVSRVPGM VALENCE6P 12800 VVIN_CHAR
VVSRVPGM VALENCE6P 133800 PULLANDVALIDATEQUERYSTRING
QMHPDEH QSYS /012E
QMHUNMSG QSYS /0600
QTEMDEBP QSYS /01FB
QTESTOPH QSYS _C_pep
QTESTOPH QSYS 200 main
QTENPTS QSYS 20 qtescren
QTENPTS QSYS 36 ProgramSource
QUIDSPP QSYS /00E4
QUIMGFLW QSYS /03C3
QUIEXFMT QSYS /0099
QUIINMGR QSYS /0582
QWSGET QSYS /0667

We have the same problem on other Excel calls. Look at this case:

// Initialise the spreadsheet
if vvOut_SSOpenWorkbook(vvOut) and
vvOut_SSOpenWorksheet(vvOut: %addr(Col): %elem(Col));

Type Program Statement Procedure
1 QCMD QSYS /01C8
ORR110B APPOBJ _QRNP_PEP_ORR110B
ORR110B APPOBJ 12300 ORR110B
VVSRVPGM VALENCE6P 214500 VVOUT_SSOPENWORKBOOK
VVSRVPGM VALENCE6P 504500 VVOUT_INITIALIZE
VVSRVPGM VALENCE6P 374300 VVUTILITY_GETVALENCESETTING
VVSRVPGM VALENCE6P 389909 VVUTILITY_REPLACESUBS
VVSRVPGM VALENCE6P 12800 VVIN_CHAR
VVSRVPGM VALENCE6P 133800 PULLANDVALIDATEQUERYSTRING
QRNXIE QSYS 1 _QRNX_CALL_FC_H
QRNXIE QSYS 2 _QRNX_DFT_ERROR
QRNXIE QSYS 18 _QRNX_INQ
QRNXIE QSYS 10 InqMsg

It also holds a lock on VVFUSION *DTAARA causing subsequent processes to hang waiting or crash.

[robert.swanson]

Thanks, that stack points me to the root cause. The vvUtility_replaceSubs procedure was modified recently to check session variables in addition to settings. Retrieving a session variable requires a session ID (sid), and of course there won't be a session ID in a non-Valence CGI job. So it's blowing up on the vvIn_char call where it's attempting to retrieve a sid.

I will email you a patch shortly that should resolve this problem. Assuming it works we'll include it in the next build.