Re: Weird problem a customer is having with a file (James H H Lampert)

I'm here, but I don't have an answer to the problem.

Can you send me a copy of the file?  Maybe just the record that gets the
error, with any sensitive info redacted.

--Dave

James H H Lampert wrote:

George Lemen <georgelemen@xxxxxxxxxx> wrote:

You may have encountered a known problem . . . where if

a physical file created by SQL has a logical view/file created over it, something like the "last use" date gets corrupted and the object ends up (eventually)

being damaged.

. . .

 Don't know if that is what you are looking at, . . . .

Well, I've also got some people at IBM looking at the problem file, and as it turns out, they could tell right away that it didn't come from SQL. When I asked the customer how it was created, I found out it's an output file from Query. And I don't think it ever had any logicals defined over it.

Just about the only things I know about it are:

1. There's a null-capable date field in it,

2. Reading a record with that field nulled-out causes a data mapping error,

3. The record with the nulled-out field still does make it into the low-level input buffer, BUT the null-map for the record has a 2, instead of a 1, for the nulled-out field,

4. The records without nulled-out fields can be updated, and new records can be added, but any attempt to null-out the field in a new or updated record gets ignored, and

5. The record with the nulled-out field rejects all attempts to update it, regardless of whether I plug a valid date into the field, or I plug either a 1 or a 2 into the field's position in the null-map, throwing a data-mapping error every time.

I haven't a clue what could be causing the problem, but I'm convinced that the "2" in the null-map is a key clue.

Leif Svalgaard, Gene Gaunt, Dave McKenzie, any other MI gurus, WHERE ARE YOU?

--
JHHL