On Mon, 1 Jun 1998, Dennis Plunkett wrote:

> At 07:00 PM 30/05/98 +0100, you wrote:
> >Hi I've been struggling with a problem for a while now so I thought that
> >maybe somebody out there may have seen this one before.
> >
> >Basically I'm seeing my RBIE/RBIF (Int on port b change) interrupt routine
> >being called even though the RBIE (enable flag) is clear. Basically my int
> >handler looks like this.....
> >
> >InterruptVector
> >     push                 ; Always save registers etc
> >CheckInterrupt
> >     btfsc INTCON, RBIF   ; Is it a PORTB int ?
> >     goto ServicePORTB    ; yes so go do it
> >...
> >...
> >ServicePORTB
> >     btfss INTCON, RBIE   ; Test to see if we have ints enabled
> >     nop                  ; This should never be executed!
> >     bcf INTCON, RBIE     ; turn off further ints
> >     movf PORTB, W        ; read port to reset it
> >     bcf INTCON, RBIF     ; say we have seen this change
> >....
> >....
> >
> >EndInterrupt
> >     pop                  ; restore things
> >     retfie               ; return and re-enable ints
> >
> >Ok what I see happening is that if I set a breakpoint on the nop just after
> >the ServicePORTB label. Then From time to time I get a break. This was
> >really screwing up the rest of my interrupt routine (easy to fix just
> >replace the nop with a goto to skip the int routine). What seems to happen
> >is that the int routine gets called once with RBIE enabled - OK, then
> >immediately again with RBIE now disabled - not so good. So my question is -
> >should this happen? If so why? BTW the interrupt gets enabled (RBIE set in
> >the main body of the code
> >
> >The environment I'm using is an ICEPIC emulator with the
> >16C71/16C7110/16C711 module installed. The signal that I'm feeding into a
> >single pin on PORTB is a rather noisy standard Radio Control pulse (+ve
> >pulse of length 1-2mS repeated every 20mS).
> >
> >Thanks
> >
> >Andy
> >
> >
>
>
> Hello, bit of trouble eh?
> Ok, the code looks simple enough and from what I can see you shouldn't get
> into the interrupt routine when RBIE is clear. You indicate that this
> happens from time to time, thus during the normal run period? So when the
> break occurs the RBIE flag is clear, this would indicate that the interrupt
> is disabled:-
>
> 1.      Does any part of your code set this flag other than the main section?
> 2.      Did an interrupt occur that caused this condition?
>
> Those are the simple parts, the next thing is the code itself, can you look
> at the code trace?
> Do you do any modifications to the Programme counter, you know things like a
> look up table where you change the contents of the programme counter. The
> reason for this is simple, often code will vector into an unrequired code
> section if either:-
>
> 1.      Uncontrolled setting of the programme counter (Check mods on PACL?)
> 2.      Stack overflow (Check call level)
>
> Both of these points can be found by code inspection.
>
> Hope I have helped
>
> Dennis
> -=====================================================================-
>
> Dennis Plunkett: Embedded Hardware, Software design
> NEC Australia DRMASS
> Line Interface cards
> TRX software
> ISDN interface card
> ph 03 9264-3867
>
> -=====================================================================-
>
>

Another bogus location:

have you branch around ServicePORTB ? Otherwise you can enter it unwanted
fashion... (xcus for a simple option, but it is still possible).

Imre