Memory Layout

["Jon Paris" <Jon.Paris@xxxxxxxxxxxxxx>]

 >> Anyone got any "way out there in left field" message stories from wrong
parm lengths? That they want to share so others may be looking out for them.

The hardest one to diagnose I ever saw was in a COBOL program (although
there is nothing unique about COBOL in this respect).  Customer complained
that he did a simple re-compile and now the program didn't work.  All he had
changed was an interest rate - nothing else.  The strange thing was that in
an A calls B calls C situation it blew up on the second program call from A
to B with an invalid parameter pointer error.  This was some 10+ minutes
into the run (the corruption it turned out happened in the first minute or
so).  Turned out the corruption had been going on for 7 years (moving blanks
to a field defined as 10 long in the A, 10 long in B, but 80 long in the
utility routine C ). Previously the errant spaces had always landed in a
print buffer, which was overlaid immediately afterwards with the next print
line.  However, in the 7 years since this program was last compiled, the
compiler had changed.  In order to improve efficiency it now pre-resolved
addresses for any fields being used as parms, and stored the pointer after
the DS in which the field resided.  It was this pointer that was being nuked
since the 10 char field was at the end of the DS and the extra 70 blanks had
to go somewhere!!

Jon Paris
Partner400