exxos's Atari, Amiga & retro forum

Posted: **17 May 2022 21:34**

I honestly cannot remember who wrote this routine...

*-----------------------------------------------------------------------------*
EXXOS_BusErrorTest:
*-----------------------------------------------------------------------------*
	move.l 		bas_params(pc),a0		; a0 = basic variable ptr
	move.l		(a0),a1				; a1 = basic variable (address to test)
	move.l		sp,d2				; d2 = old stack pointer
	move.l		$8,d1				; d1 = old berr handler

	move.w		#.berr-EXXOS_BusErrorTest,d3	; d3 = err handler offset
	lea		EXXOS_BusErrorTest(pc,d3.w),a2	; a2 = err handler absolute address
	move.l		a2,$8				; set new berr handler

	move.l		(a1),d0				; read from address
	move.l		d1,$8				; restore berr handler
	move.l		#1,(a0)				; return 1
	rts
.berr:
	move.l		d2,sp				; restore stack to before berr
	move.l		d1,$8				; restore berr handler
	move.l		#0,(a0)				; return 0
	rts

Maybe someone have some ideas as to what is going on with it. IIRC was working when I would test a RAM location, basically because that location is used to returned a value back to BASIC code.

So I did a test routine where I could return to separate values by doing this..

Code: Select all

	move.l 		bas_params(pc),a0		; a0 = basic variable ptr    6 ?
	move.l		(a0),a1				; a1 = basic variable (address to test) FF8964   
 	move.l		#1234,(a0)			; return 1234 in X&
	move.l		#5678,4(a0)			; return 5678 in Y&
	rts

And this correctly returns the value to BASIC.

Capture.PNG

So I make the assumption that return 1 and return 0 get changed from #0,(a0) TO #0,4(a0) as that is what worked in my test routine.

However something has broken in the bus error handler and I have no idea what. I thought it was down to those code changes but even if I comments them out the handler crashes :(

If I simply put a RTS the start of the code then things don't crash. So definitely wrong with that bus error handler routine somehow :shrug:

Posted: **17 May 2022 22:00**

When I call the routine I have been passing the actual address as a parameter EG:

Code: Select all

 t&=&HFF8964 
  Y&=66666
 call loc testcode&,EXXOS_BusErrorTest,t& ,varptr(Y&)

Which bomns out.

But passing the RAM pointer to the variable doesn't crash, but it returns 0 regardless of it the address is valid or not.

Code: Select all

 t&=&HFF8964 
  Y&=66666
 call loc testcode&,EXXOS_BusErrorTest,varptr(t&) ,varptr(Y&)

EDIT:

Must be something else broken somewhere as if I copy the working code into the BUSERROR routine and just RTS early, it doesn't return #1234 :WTF:

Posted: **17 May 2022 22:48**

FFS!

I used t& as a variable and it NEVER returned a value. I changed it to X& and now its starting to behave.. and t& isn't even used anywhere in the program :WTF: Feking bonkers faults again :pullhair: :headbang:

Posted: **17 May 2022 23:01**

Are we good now then?

Posted: **17 May 2022 23:02**

Sooo..

STEEM - no Flashyclock

1.PNG

Real hardware with FC

IMG_0588.JPG

So no feking idea why hisoft basic didn't like that variable name. I hate computers!

:comp:

and proof..

Capture.PNG

It's not just because its lower case either. Only possible thing is if t& is some internal thing :shrug: eitherway, waste 2 days just working that out :roll: :headbang:

Posted: **17 May 2022 23:11**

Glad you've got it sorted out.

Since you're doubtlessly feeling completely better and not at all sore about it any more, can we now go back to taking the Gipsy Kiss out of you for using a Noddy language?

Too soon..?

;)

BW

Posted: **17 May 2022 23:29**

Badwolf wrote: 17 May 2022 23:11 Since you're doubtlessly feeling completely better and not at all sore about it any more, can we now go back to taking the Gipsy Kiss out of you for using a Noddy language?

:chairsmack:

It's always me which finds bonkers bugs in stuff regardless of what I use :roll: It's not like I even code much of anything :roll:

Posted: **19 May 2022 16:33**

There is still something odd going on :roll:

Would this code work for WORD or BYTE access ? I can peek a byte or word in STOS and I get a number back, but a LONG I get BERR. Fair enough. But this code looks like it deals with longs ? So would that screw up WORD access checks ?!

Code: Select all

	move.l 		bas_params(pc),a0		; a0 = basic variable ptr    6 ?
	move.l		(a0),a1				; a1 = basic variable (address to test) FF8964 
  
	move.l		sp,d2				; d2 = old stack pointer
	move.l		$8,d1				; d1 = old berr handler

	move.w		#.berr-EXXOS_BusErrorTest,d3	; d3 = err handler offset
	lea		EXXOS_BusErrorTest(pc,d3.w),a2	; a2 = err handler absolute address
	move.l		a2,$8				; set new berr handler

	move.l		(a1),d0				; read from address
	move.l		d1,$8				; restore berr handler

	move.l		#1,4(a0)				; return 1
	rts
.berr:
	move.l		d2,sp				; restore stack to before berr
	move.l		d1,$8				; restore berr handler

	move.l		#0,4(a0)				; return 0
	rts

Posted: **19 May 2022 16:39**

OK so doing

Code: Select all

move.w		(a1),d0

Seems to work now..

Posted: **19 May 2022 17:29**

It looks like accessing a1+2 does trigger the error handler then ?

exxos's Atari, Amiga & retro forum

Bus error handling problem

Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem

Re: Bus error handling problem