IBM Mainframe Forum

how we can check the syntax errors in jcl other than typerun=scan? please help me...

What is your problem in running the "scan"?

Unless there is some fancy software package, the only other thing I can think of is "desk-checking". This means looking at the stuff yourself very hard. Often employed, many years ago, when your compile/run was waiting on an input queue at 10am but unlikely to be processed before 5pm. Mostly looking for logic errors, but you'd tend to check everything. Very effective way of learning how to "look" at something to start working out how it is wrong. I suspect very under-used these days.

Terminology is critical in IT,where similar terms may mean very different thigns. There is no such thing on a job card as "TYPERUN=SCAN" as your post put it. There is, however, a "TYPRUN=SCAN" (note the missing E).

Contact your site support group and find out if CA-JCLCHECK (or any similar product) is installed at your site. If so, you can use that package. If not, as BillyBoyo said, you can desk check the JCL. If you're looking for something else, you'll have to provide more information than you have so far.

Another thing you can do is code, as the first step before the proc, a non-existent program. If this gets to run then the JCL is, most likely, ok (other errors can still exist). E.G.

[quote="NicC"]Another thing you can do is code, as the first step before the proc, a non-existent program. [quote]

Second time I've seen you mention this. Neat, but what advantage does it have over TYPRUN=SCAN, or another way, why do you prefer to use it? I've never used the TYPRUN=SCAN (couldn't see the point). Does the SCAN sit waiting in a queue rather than bombing straight away?

Scan will pick up syntax errors - there are other errors that are found when the job actually initiates. Don't ask me which - I don't make misteaks! Long time since I did - at least with JCL - well, writing procs.

Maybe the scan picks up syntax errors, the non existent program data sets that do not exist? But this is easy to test and play with, something for the person that started the thread to test and post his findings.

I have previously used NicC's method. Correct JCL, typo in the program name.

Writing JCL, once you have a number of samples kicking around, was really just an "editor" task. For "batch" anyway (if that does not seem too obvious, I mean without DBs). I used to just type "submit" (which we called "run"). If no "notify" in a couple of seconds, the JCL was basically OK. Then, after about eight minutes, notify says "S806". No big biggie.