JDEDB_COMMIT_MANUAL resultset limitation?

[BOster]

I am researching on Oracle support, but I thought I would ask this here since someone might know the answer and save me some time.

We rarely use JDEDB_COMMIT_MANUAL option in C BSFNs when doing any type of transaction processing but instead let the caller (APPL, UBE, etc.) determine if TP is enabled and set boundaries, etc. - the way it appears Oracle kind of wants you to do it. However we do occasionally handle TP within a BSFN and we haven't really had any problems, until just recently.

We have an issue in some code that has been in place for years - not sure if the issue was always there or if it is new. We are reading through a result set on one table handle (hRequestRead) and updating the same table using a different table handle (hRequestUpdate). Both table handles are opened on the same HUSER handle intialized with:

<font class="small">Code:</font><hr /><pre>
JDB_InitUser(lpBhvrCom->hEnv, &hUser, (JCHAR *) NULL, JDEDB_COMMIT_MANUAL)
</pre><hr />

We found that the result set on the table doing the reading was only returning the first 100 rows when it should have been returning 600+ rows. If I change the code to open the hRequestRead table using a seperate HUSER handle initialized with

<font class="small">Code:</font><hr /><pre>
JDB_InitBhvr(lpBhvrCom, &hUserAuto, (JCHAR *) NULL, JDEDB_COMMIT_AUTO)
</pre><hr />

hRequestRead returns all the rows that it should in its result set.

Any ideas?

 

[craig_welton]

Pretty strange behaviour Brian. This probably has no impact, but it looks like you are using two different APIs to setup the user handle (JDB_InitUser vs. JDB_InitBhvr). I would expect the two to work the same, but might be worth a shot.

Craig

 

[BOster]

When you use JDEDB_COMMIT_MANUAL, I have found you have to use:

<font class="small">Code:</font><hr /><pre>
JDB_InitUser(lpBhvrCom->hEnv, &hUser, (JCHAR *) NULL, JDEDB_COMMIT_MANUAL);

JDB_FreeUser(hUser);
</pre><hr />

instead of

<font class="small">Code:</font><hr /><pre>
JDB_InitBhvr
JDB_FreeBhvr
</pre><hr />

I never fully understood why but I think that may actually be documented some place. I opened a case with Oracle because I can't find anything that would explain the behavior I am seeing. This code was put in to place in 2010, I am not sure if this has allways been an issue and we never discovered it before. The particular process where we discovered this problem could actually hide this problem - the way the process works it will eventually kind of self correct. My bigger fear is all the other places where we use JDEDB_COMMIT_MANUAL and we do both reads and writes using an HUSER handle initialized with the JDEDB_COMMIT_MANUAL option.

 

[craig_welton]

Cool. Thanks for the info. Keep us posted if you can.

Craig