Weird record lock in 36EE

["M. Lazarus" <mlazarus@xxxxxxxx>]

Hi all,

 I've just migrated some RPG II programs from the *M36 to the 36EE (under 
v5r2) and I'm running into a weird record lock situation.  It looks like a 
deadly embrace situation, which didn't exist on the "real" S/36.

 The msg is: RPG9998 [which IBM tech support couldn't even find!], Options 
(  23F ) An input or output error occurred in file ABC.

 Cause: RPG program MYPGM detected an I/O error CPF5032 at statement 
00000000. [This is odd because observability has not been 
removed.]  Message text for error CPF5032 is Record 1 member M040310 
already locked to this job..

 Here's some pertinent sample code:

// LOAD MYPGM
// FILE NAME-ABC1,LABEL-ABC,DISP-SHR,BYPASS-PRF
// FILE NAME-ABC2,LABEL-ABC,DISP-SHR,BYPASS-PRF
// FILE NAME-ABC3,LABEL-ABC,DISP-SHR,BYPASS-PRF
// RUN

 Note that BYPASS-PRF (to explicitly turn off ODP sharing) was added at 
the suggestion of the IBM'er, but didn't help, as I suspected it wouldn't, 
since the 36EE is already set not to automatically share ODPs.

 The 3 files are defined as UP or US.  After some processing (based on the 
input indicator) there is an O spec that attempts to delete the 
record.  That is when the pgm bombs.

IABC1     NS   01    1 CA
I                             1   80 REC
I        XX
IABC2     NS   02    1 CB
I                             1   80 REC
I        XX
IABC3     NS   03    1 CC
I                             1   80 REC
I        XX
 ....
OABC1     DDEL        01
OABC2     DDEL        02
OABC3     DDEL        03


 This pgm has been running for years on the /36.  PTFs are fairly current 
(2nd to last cume.) Any ideas?  TIA.

 -mark