Different results from CFF (EBCIDIC) stage between envs

[fridge]

Hi - dont know if anyone can help but have a strange issue that cant understand

Have a simple job that writes to a CFF stage (PX) as follows (EBCIDIC file on AIX env)

FIELD A - DISPLAY NUMERIC (4)
FIELD B - CHAR (3)
FIELD C - CHAR (10)

On one enviorment (Dev) it creates the file as expected e.g.

1234ABCXXXXXXXXXX

However on the other env (UAT) it appears as

123}ABCXXXXXXXXXX

The display numeric is presumably coming though as signed on the incorrect one

For info am running on UNIX (AIX) and getting the above results by using the dd conv=ascii command to check the results.

Originally had the issue on the dev env - but was solved by a recompile - presumed the binaries had got out of sync somehow with the Job - however tried the same on UAT and still get the incorrect format


I have checked both jobs and even bought back the jobs to same enviornment and done a compare - datastage says exactly the same - and checked the parallel layout on the CFF stage - again the same.

Any input gratefully recieved.

[chulett]

I really have no idea what might be going on but wanted to point out that "123}" isn't any kind of proper signed conversion of "1234"... which I'm sure you know. That "}" is a "negative zero" whereas you should have a "D" there.

[fridge]

Thanks - sorry was using it as an example - as the actual fle is a bit more complicated datawise - also I typed it wrong should have been 123{

However this was the output after conversion via dd comand

[chulett]

Ok, so a positive zero then.

[ArndW]

Steve, if you turn on $OSH_PRINT_SCHEMAS can you check whether the column datatype going into your CFF stage in the "bad" environment is signed or not - that would narrow down the search for a cause.

[fridge]

Actually its Ed (you know the one you worked with whom taught you all you know :D )

Actually resolved this finally - but had to
copy the original jobs
delete the originals
rename the jobs back
compile and run

So they get assigned new job numbers , repository objects etc

And came out as proper display numeric

Going to have to try to replicate the issue - but suspect that the format was originally Signed Numeric, got changed in the CFF stage recompiled and worked on DEv , but when moved to UAT and recompiled didnt

How mad is that

[ArndW]

Sorry Ed, I must have equated the userid with size Since you soaked up all my knowledge, can I have some of it back please? A couple of hit-points would do me well right now.

Glad to hear that you've solved it, even though the solution was non-trivial.