> AFAICT, the docs aren't that clear about **all** of the weird interactions
> between IRQs, sleep, and WDT, but looking at Fig 8-14 for the 16C84
> (pg 2-767 in '95 databook) shows that the interrupt inputs have a
> parallel "sneak" pathway that bypasses the normal IRQ channel and goes
> straight to the wakeup from sleep ckt. So doesn't matter whether GIE is
> enabled, but does matter if INTE is.

Yes that's quite true, b0 will generate a wake-up whatever GIE is set to
and if INTE is set a true IRQ will be processed. As I said, I can stop b0
generating an immediate IRQ when s/w is already in an ISR, but the
problem still is that somehow the PIC stores a request, ready to process
it once the current ISR has finished. So far I've found no way to stop
this happening.

If <bcf intf> doesn't prevent it then where the heck is this request being
stored ? INTF would be the obvious cuplrit

If there really is an insurmountable problem when using WDT/SLEEP/
b0 inside an ISR I'd like to see it documented, so that I can abandon
that method and rewrite some code. If b0 can be completely disabled/
ignored I'd rather stick with the code I've got

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.