Subject: | |
From: | |
Reply To: | |
Date: | Mon, 22 May 1995 17:26:50 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Dan writes:
> Does anyone know why a program would crash when its SL was octcomp'ed, but
> runs fine when the SL was not octcomp'ed?
Yes.
Remember...OCTCOMP is a compiler, and compilers have bugs. The bugs in
OCTCOMP vary from release to release, and I suspect you got bit by one.
One that I can think of:
FOR loops (TBA & MTBA instructions, for instance) make a *usually*
valid assumption that the programmer won't be changing the FOR-loop
limit or counter or variable address in the stack. So, if the
programmer *does* do something that nasty, the OCT'd version won't
work like the non-OCT'ed ("emulated") version.
That problem's been around forever, and I reported it to HP around 86 or so.
You could try OCT'ing the program a segment at a time, to see if
you can locate which segment is causing the problem.
Or, try a different version of OCTCOMP...I've seen problems like you
describe come & go & come & go over four releases.
Stan
|
|
|