Re: The condition was caused by the program output data containing below hex 40 ...

Not sure how the clauses "have no idea how many" and "a few dozens of records" are ensured to be congruous... but I offer anyhow:

The following SQL requests would suffice; implemented with full table-scan:

DELETE FROM The_File
WHERE
locate(
x'3F'
, translate(
the_bad_column
, x'3F'
, /* next two lines must be a contiguous hex constant */
x'000102030405060708090A0B0C0D0E0F101112131415161718191A1B1C1D1E1F
202122232425262728292A2B2C2D2E2F303132333435363738393A3B3C3D3E3F'
, x'3F'
)
) > 0

DELETE FROM The_File
WHERE
translate(
the_bad_column
, x'3F'
, /* next two lines must be a contiguous hex constant */
x'000102030405060708090A0B0C0D0E0F101112131415161718191A1B1C1D1E1F
202122232425262728292A2B2C2D2E2F303132333435363738393A3B3C3D3E3F'
, x'3F'
) LIKE ('%' concat x'3F' concat '%')

Note: if the character 0xFF is similarly problematic, despite the OP implying characters with a hex code point of less than 0x40, then add the x'FF' to the end of the from-string in the TRANSLATE scalar.

Regards, Chuck

On 03-Dec-2013 13:32 -0800, Booth Martin wrote:

In this instance the data is is only for stress testing. Losing a
few dozens of records won't impact what I am doing. and in a live
situation this same error would have already ground things to a
halt.

On 12/3/2013 3:05 PM, Jon Paris wrote:

Many - but can you be a bot more specific? After all just deleting
records is rarely a cure (or at least it is one where the cure can
be as bad as the disease).

Definition of fields? How were the records created, etc. etc.

On 2013-12-03, at 3:52 PM, Booth Martin wrote:

The file has about 2.5 million records. Some of those records, I
have no idea how many, give an error and my RPG program fails.
Is there a way to process the file and delete those troublesome
records?