Re: This is strange: an RNQ1011 for no apparent reason -- apparent bug in runtime and/or compiler

You must be using I specs to define the record format to use and
failed to define a catch all record identifier. Therefore, the
program finds a record but does not have a definition defined for the
I specs, and thre is no other ways out than to bomb out. The person
who build the program should be talked to to find out why no catch
all was set up.

I agree with this last post. Do you have 'I' specs with record
indicators for the file in question?

No, and no!

For pity's sake, Loek and Paul, IT IS AN EXTERNALLY-DESCRIBED, KEYED PHYSICAL FILE, OPENED IN EXTERNALLY-DESCRIBED, KEYED MODE, WITH NO I-SPECS OTHER THAN THE COMPILER-GENERATED ONES, and I've determined EXACTLY what the problem is: if you CHAIN such a file in an INZSR, and your first CHAIN fails to find a matching record (and, perhaps, there might be other circumstances), the RPG runtime throws a nonsensical and misleading exception.

The fact that it is an externally-described PF makes the exception, BY DEFINITION, nonsensical, since an externally-described PF can only have one format.

The "Record not found" in the dump (even if it didn't make it into the joblog) was the crucial clue, and the only reason I missed it was that I thought there *was* a record for that key (I can only guess that I was looking at a different instance of the file, in another library).

And I think I remember seeing exactly the same misleading exception once before.

If I can come up with a simple testcase to reproduce the problem, I will publish it on the List, and I will follow the off-List suggestion of one of our leading gurus, and open a PMR about it.

For all I know, it might even be something that's fixed in a more current compiler than the one on which this was compiled (I'll look into THAT as well).

--
JHHL