why endless loop when CICS GETMAIN is executed for 0 length?

dejunzhu · Active User Joined: 08 May 2008 Posts: 390 Location: China

in my subprogram, I have below CICS statement:

dick scherrer · Posted: Thu May 10, 2012 8:31 am

Hello,

Suggest you post enough code for someone to be able to help (NOT the entire program).

From what little you've posted, it appears you have some kind of programming error. Possibly because the code returns to the call over and over. . .

Bill O'Boyle · Posted: Thu May 10, 2012 8:53 am

Add the NOHANDLE keyword to the GETMAIN API. After the GETMAIN, check EIBRESP for an error; DFHRESP(LENGERR) which is 22 or DFHRESP(NOSTG) which is 42. Most likely, it will be 22.

The EIBRCODE will equal X'E10000000000 (when EIBRESP is 22) or X'E20000000000 (when EIBRESP is 42).

The EIBFN for a GETMAIN API is X'0C02'.

Please post the Picture Clause for CMLSEADR-DB-BUFFER-LEN.

Try debugging the transaction using CEDF. After you execute the GETMAIN API, check the above EIB fields for the values posted. A value of other than ZERO for EIBRESP and other than LOW-VALUES for EIBRCODE is considered an error in this API.

Peter cobolskolan · Posted: Thu May 10, 2012 10:27 am

I assume this is the same program as in your previous post. The code you showed contained a Handle Abend. I asked you why, but no answer.
I ask the same question again, because Handle Abend tells CICS that "I have a better way to handle the error than you!".
This might be the cause of the loop, as a Getmain with zero length ALWAYS casues an error.
By the way: Why zero length?
If this might occur, you should check for zero and avoid the Getmain.
Solve the problem with the zero length befor you go on, this is probably the real error.

dejunzhu · Active User Joined: 08 May 2008 Posts: 390 Location: China

hi, BIll,

there is no NOHANDLE keyword in GETMAIN grammar:

Peter cobolskolan · Posted: Thu May 10, 2012 11:53 am

I recommend you read about Resp Option for CICS Commands, which would clarify the Nohandle This outdates the Handle Condition, which I saw in your pervious post. This was used pre -85, and the preferred method is the Resp Option, when you write a new program.
I assume you havent specified the Handle Condition for Lengerr, neither you have the Resp Option. When an error occurs, the CICS default is to abend the task. But with my assumption that you have the Handle Abend in your program, this ovverrides CICS actions for handling the error, and your routine gets control.
As you really dont know how to handle CICS raised conditions, you might end up in illogic and eventully a loop.

Bill O'Boyle · Posted: Thu May 10, 2012 6:17 pm

Just reviewed your previous post and CMLSEADR-DB-BUFFER-LEN is in fact, a fullword. An error can easily be raised (such as a zero-length) when the FLENGTH keyword is used with a halfword (9999 COMP) and this why the Picture Clause was requested. So, you're OK with this field.

You need to review the NOHANDLE and RESP options of a given CICS API, introduced in the mid/late 1980's with CICS/MVS 1.7.

I agree with Peter that a HANDLE ABEND is worse than just letting the transaction ABEND on its own.

When a HANDLE CONDITION is coded, an internal GETMAIN is issued and a subsequent FREEMAIN is issued either at enclave termination or at task termination.

Using NOHANDLE or RESP eliminates these internal GETMAIN's/FREEMAIN's.

dejunzhu · Active User Joined: 08 May 2008 Posts: 390 Location: China

thanks for your suggestion.

I have coded input length check before GETMAIN.