DevPac error when compiling?! - Page 2

Galahad/FLT · 22 June 2011, 01:05

Quote:

Originally Posted by StingRay

This can't be the cause of the crash.

Well aware of that, but clearly he needs to give his code more scrutiny, if not the moveq, then code in and around that area, as I assume he gets the system back at least which would suggest his code on exit is to blame.

matthey · 22 June 2011, 02:39

If anyone has or suspects problems with Devpac/GenAm, there is vasm which is very compatible in Devpac compatibility mode and is actively developed...

http://sun.hasenbraten.de/vasm/

Leffmann · 22 June 2011, 04:28

Quote:

Originally Posted by Lonewolf10

I'm pretty sure my code is good and that something Devpac 3.18 is doing is causing this. Does anyone have any pointers?

The are a couple of issues with your startup code, but just like you say it's not your startup code that causes the guru. It gurus because you're not responding to the message that Workbench is sending you.

A program started from the CLI doesn't need to do anything to run properly, but a program started from the Workbench must retrieve a startup message sent by the Workbench before doing anything else, and return this message to the Workbench before quitting.

I can give you my code if you'd like, or you can find instructions on how to do this in the RKM Libraries manual.

Quote:

Originally Posted by matthey

If anyone has or suspects problems with Devpac/GenAm, there is vasm which is very compatible in Devpac compatibility mode and is actively developed...

High five for vasm!

StingRay · 22 June 2011, 09:36

Quote:

Originally Posted by Lonewolf10

Which question? I thought I answered everyones questions/comment.

My question why you are so sure that your good is good and that your problem is caused by Devpac.
Reason I asked that question is simply the fact that it's easy to blame something on the tools you use but in most cases it's your code which is buggy. Trust me on that one.

Quote:

Originally Posted by Lonewolf10

I avoid using libraries as much as possible as they are complicated and I am a newbie when it comes to using libraries, but having reliable system kill/restore libraries is required so that folks can return to workbench afterwards. Personally I'd rather just force them to reset their Amiga's instead.

When it comes to demo coding, you don't need to use many libraries. But those that you use should be used correctly, always check if a library was opened correctly, don't assume anything. Also, rule of thumb, when you have disabled the system (i.e. killed interrupts/DMA) avoid using any system libraries as they expect the system to be intact.

Quote:

Originally Posted by Lonewolf10

The error started happening around the time I started using Devpac 3.18 (Jan 2011?) and I (wrongly?) assumed they were related.

Check which code you have changed since then and remove the new routines one by one and see when the error disappears.

Quote:

Originally Posted by Lonewolf10

I am a little sleepy right now, but I'll take a closer look at it tomorrow.

You want to clear the color registers, but, are you really doing that? Check register a0.

Hewitson · 22 June 2011, 10:26

Quote:

Originally Posted by Lonewolf10

Personally I'd rather just force them to reset their Amiga's instead.

Lonewolf10 · 23 June 2011, 19:13

Quote:

Originally Posted by StingRay

Code:

    lea.l    COLOR00,a0
    moveq    #31-1,d0        ;number of colours
kill_system_colours:
    move.w    #0,a0
    addq    #4,a0
    dbra    d0,kill_system_colours
    rts

When I see this I am pretty sure it's not the assembler which causes your problems. Please check what the code is doing and what you want it to do.

Quite right, I overlooked a couple of stupid things there - namely the size of each colour register (2 bytes not 4) and used direct addressing instead of indirect addressing.

Quote:

Originally Posted by StingRay

My question why you are so sure that your good is good and that your problem is caused by Devpac.
Reason I asked that question is simply the fact that it's easy to blame something on the tools you use but in most cases it's your code which is buggy. Trust me on that one.

You have already proven that I write buggy code occasionally

Quote:

Originally Posted by Leffmann

A program started from the CLI doesn't need to do anything to run properly, but a program started from the Workbench must retrieve a startup message sent by the Workbench before doing anything else, and return this message to the Workbench before quitting.

Ok, I didn't know that. Does that also apply to something started at boot-up?

Quote:

Originally Posted by Leffmann

I can give you my code if you'd like, or you can find instructions on how to do this in the RKM Libraries manual.

Thanks for your kind offer, but I have found the instructions in the RKRM: Libraries manual and will try to write my own code for now.

Regards,
Lonewolf10

Leffmann · 23 June 2011, 22:03

Quote:

Originally Posted by Lonewolf10

Ok, I didn't know that. Does that also apply to something started at boot-up?

Do you mean in the bootblock or the startup-sequence? Anything run from the startup-sequence is run just like from the CLI.

There are no methods to start program files from the bootblock, so any code you load yourself from the diskdrive can just be run as is. If you want your bootblock to enter AmigaDOS then there's a small procedure that needs to be followed, and it also needs to comply with the AmigaOS ABI and preserve registers D2-D7 and A2-A6.

Lonewolf10 · 23 June 2011, 22:19

Quote:

Originally Posted by Leffmann

Do you mean in the bootblock or the startup-sequence?

I currently load my demo via the startup-sequence, but it can also be loaded via workbench.

Any future projects will (hopefully) be loaded via the bootblock.

Thanks for answering for both ways anyway

Regards,
Lonewolf10

h0ffman · 23 June 2011, 23:58

Quote:

Originally Posted by Lonewolf10

*bump!*

So hows the demo going hoffman?

Edit:

Back on topic, my demo works fine (in workbench atleast), but when it exits I get an error:

Program failed (error 87000004)

I'm pretty sure my code is good and that something Devpac 3.18 is doing is causing this. Does anyone have any pointers?

Regards,
Lonewolf10

You lot are hijackin my thread!!

To answer Lonewolf, the coding was going very well. So far i've managed to do

scroller
scroller cookie cut over a background
protracker note scrolling display
bitplane sine wave shifting
protracker / theplayer scope (orignally ripped from an old PT source but well an truly my own code now!)
a module deconstruction and rebuild system

I think my proudest effect so far is what I call the banksy effect. spray paints over the screen, which then peels away to reveal a logo, much like stencil art.

Also I put together an entire 3 disk long play music disk using most of the stuff listed above, however, it still has a nasty bug involving the CIA replay and the loader... I'm sure Stingray could fill you in on the fun thats been!!

BTW - that original block of code in this thread?? absolute bag of shite! Thankfully I got a good system startup block from Stingray.

Next on the Agenda....

Finish music disk

Start proper demo!

pmc · 24 June 2011, 08:43

Quote:

Originally Posted by h0ffman

Next on the Agenda....

Finish music disk

Start proper demo!

\o/

Lonewolf10 · 25 June 2011, 22:37

I have re-written my kill system and restore system routines (see below), but now whilst my demo works as soon as it tries to return to workbench all I get is a black screen

I also tried saving register values on start-up and restoring on exit, but as soon as I started my demo I got an error (80000004) so I have commented them out for now.

What am I missing, or doing wrong?

New additions are in bold

Code:

;Library offsets taken from Amiga Machine Language, pages 240-243

ExecBase	equ	4

;Exec.Library calls
*******************
CloseLib	equ	-414
FindPort	equ	-390
FindTask	equ	-294
Forbid		equ	-132
GetMsg		equ	-372
OldOpenLib	equ	-408
Permit		equ	-138
ReplyMsg	equ	-378
WaitPort	equ	-384


;Graphics.Library calls
***********************
DisownBlitter	equ	-462
OwnBlitter	equ	-456
LoadView	equ	-222
WaitBlit	equ	-228
WaitTOF		equ	-270


;Intuition.Library calls
************************
RemakeDisplay	equ	-384


kill_system:				;**** Kill System (cleanly?) ****

*	movem.l	d0-d7/a0-a6,-(sp)	;store current register values
	move.l	ExecBase,a6
	jsr	Forbid(a6)		;multitasking off

***					;new, 24th june 2011

	lea.l	demo_name,a1
	jsr	FindPort(a6)		
	cmp	#0,d0
	beq.s	killsys_error
	move.l	d0,msg_port		;save msg port address

	lea.l	msg_port,a0
	jsr	WaitPort(a6)
	lea.l	msg_port,a0
	jsr	GetMsg(a6)	
	move.l	d0,system_msg

***
	lea.l	graphics_lib,a1		;open Graphics.Library
	moveq	#0,d0
	jsr	OldOpenLib(a6)		;returns library address in d0
*****
	cmp	#0,d0
	beq.s	killsys_error		;new <<<--- 24 june 2011
*****
	move.l	d0,graphics_base	;store address
	move.l	d0,a6			;graphics.lib base -> a6
*	beq.s	allsquare_exit
	jsr	OwnBlitter(a6)		;no input/output for both calls
	jsr	WaitBlit(a6)
	move.l	#0,a1
	jsr	LoadView(a6)		;no output
	jsr	WaitTOF(a6)		;no input/output (call twice?)
	jsr	WaitTOF(a6)

	move.w	INTENAR,old_INTENAR	;store current settings
	move.w	INTREQR,old_INTREQR
	move.w	DMACONR,old_DMACONR

	move.w	#$7FFF,INTENA		;disable everything
	move.w	#$7FFF,INTREQ
	move.w	#$07FF,DMACON

	or.b	#%10000000,$BFD100	;as interrupts are off, we manually
	and.b	#%10000111,$BFD100	;switch off disk drive motor

	lea.l	COLOR00,a0
	moveq	#31-1,d0		;number of colours
kill_system_colours:
	move.w	#0,(a0)			;was a0
	addq	#2,a0
	dbra	d0,kill_system_colours
	rts

demo_name:
	dc.b	"DD_AllSquare",0
	even

taskCB:
	dc.l	0

msg_port:
	dc.l	0

system_msg:
	dc.l	0

graphics_lib:
	dc.b	"graphics.library",0
	even

graphics_base:
	dc.l	0

intuition_lib:
	dc.b	"intuition.library",0
	even

intuition_base:
	dc.l	0

old_INTENAR:
	dc.w	0

old_INTREQR:
	dc.w	0
	
old_DMACONR:
	dc.w	0


restore_system:
	move.w	#$7FFF,INTENA		;*** disable everything ***
	move.w	#$7FFF,INTREQ
	move.w	#$07FF,DMACON

	lea.l	intuition_lib,a1	;*** Open Intuition.Library ***
	move.l	ExecBase,a6
	moveq	#0,d0
	jsr	OldOpenLib(a6)		;returns library address in d0
	move.l	d0,intuition_base	;store address
	move.l	d0,a6			;intuition.lib base -> a6
	jsr	RemakeDisplay(a6)

	move.l	ExecBase,a6
	move.l	156(a6),a6
	move.l	38(a6),COP1LCH		;restore workbench copperlist
	move.w	#0,COPJMP1		;write any value to COPJMP1

	move.w	old_INTENAR,d0		;*** restore system settings ***
	bset	#15,d0
	move.w	d0,INTENA
	move.w	old_INTREQR,d0
	bset	#15,d0
	move.w	d0,INTREQR
	move.w	old_DMACONR,d0
	bset	#15,d0
	move.w	d0,DMACON

	move.l	graphics_base,a6
	jsr	WaitTOF(a6)
	jsr	DisownBlitter(a6)

	move.l	ExecBase,a6

***				;new, 24th june 2011

	move.l	system_msg,a1
	jsr	ReplyMsg(a6)
***

	jsr	Permit(a6)
*	movem.l	(sp)+,d0-d7/a0-a6	;restore original register values
	rts

Regards,
Lonewolf10

hitchhikr · 25 June 2011, 23:37

Code:

;Library offsets taken from Amiga Machine Language, pages 240-243

DMACONR		equ	$dff002
INTENAR		equ	$dff01c
COP1LCH		equ	$dff080
COPJMP1		equ	$dff088
DMACON		equ	$dff096
INTENA		equ	$dff09a
COLOR00		equ	$dff180

ExecBase	equ	4

;Exec.Library calls
*******************
CloseLib	equ	-414
FindPort	equ	-390
FindTask	equ	-294
Forbid		equ	-132
GetMsg		equ	-372
OldOpenLib	equ	-408
Permit		equ	-138
ReplyMsg	equ	-378
WaitPort	equ	-384


;Graphics.Library calls
***********************
DisownBlitter	equ	-462
OwnBlitter	equ	-456
LoadView	equ	-222
WaitBlit	equ	-228
WaitTOF		equ	-270


;Intuition.Library calls
************************
RemakeDisplay	equ	-384


kill_system:				;**** Kill System (cleanly?) ****

*	movem.l	d0-d7/a0-a6,-(sp)	;store current register values
	move.l	ExecBase,a6
	jsr	Forbid(a6)		;multitasking off

***					;new, 24th june 2011

	lea.l	demo_name,a1
	jsr	FindPort(a6)		
	move.l	d0,msg_port		;save msg port address
	beq.b	no_msg

	lea.l	msg_port,a0
	jsr	WaitPort(a6)
	lea.l	msg_port,a0
	jsr	GetMsg(a6)	
	move.l	d0,system_msg
no_msg:
***
	lea.l	graphics_lib,a1		;open Graphics.Library
	moveq	#0,d0
	jsr	OldOpenLib(a6)		;returns library address in d0
*****
	move.l	d0,graphics_base	;store address
	beq.w	killsys_error		;new <<<--- 24 june 2011
*****
	move.l	d0,a6			;graphics.lib base -> a6
*	beq.s	allsquare_exit
	jsr	OwnBlitter(a6)		;no input/output for both calls
	jsr	WaitBlit(a6)
	move.l	#0,a1
	jsr	LoadView(a6)		;no output
	jsr	WaitTOF(a6)		;no input/output (call twice?)
	jsr	WaitTOF(a6)

	move.w	INTENAR,old_INTENAR	;store current settings
	move.w	DMACONR,old_DMACONR

	move.w	#$7FFF,INTENA		;disable everything
	move.w	#$07FF,DMACON

	or.b	#%10000000,$BFD100	;as interrupts are off, we manually
	and.b	#%10000111,$BFD100	;switch off disk drive motor

	lea.l	COLOR00,a0
	moveq	#32-1,d0		;number of colours
kill_system_colours:
	move.w	#0,(a0)			;was a0
	addq	#2,a0
	dbra	d0,kill_system_colours

loop:	btst	#6,$bfe001		;left mouse button
	bne.b	loop
	bra.w	restore_system

demo_name:
	dc.b	"DD_AllSquare",0
	even

taskCB:
	dc.l	0

msg_port:
	dc.l	0

system_msg:
	dc.l	0

graphics_lib:
	dc.b	"graphics.library",0
	even

graphics_base:
	dc.l	0

intuition_lib:
	dc.b	"intuition.library",0
	even

intuition_base:
	dc.l	0

old_INTENAR:
	dc.w	0

old_DMACONR:
	dc.w	0

restore_system:
	move.w	#$7FFF,INTENA		;*** disable everything ***
	move.w	#$07FF,DMACON

	lea.l	intuition_lib,a1	;*** Open Intuition.Library ***
	move.l	ExecBase,a6
	moveq	#0,d0
	jsr	OldOpenLib(a6)		;returns library address in d0
	move.l	d0,intuition_base	;store address
	move.l	d0,a6			;intuition.lib base -> a6
	jsr	RemakeDisplay(a6)

	move.l	ExecBase,a6
	move.l	156(a6),a6
	move.l	38(a6),COP1LCH		;restore workbench copperlist
	move.w	#0,COPJMP1		;write any value to COPJMP1

	move.w	old_INTENAR,d0		;*** restore system settings ***
	or.w	#$c000,d0
	move.w	d0,INTENA
	move.w	old_DMACONR,d0
	bset	#15,d0
	move.w	d0,DMACON

	move.l	graphics_base,a6
	jsr	WaitTOF(a6)
	jsr	DisownBlitter(a6)

killsys_error:

	move.l	ExecBase,a6

***				;new, 24th june 2011

	move.l	system_msg,d0
	beq.b	no_replymsg
	move.l	d0,a1
	jsr	ReplyMsg(a6)
no_replymsg:
***

	jsr	Permit(a6)
*	movem.l	(sp)+,d0-d7/a0-a6	;restore original register values
	moveq	#0,d0			;return code
	rts

Lonewolf10 · 26 June 2011, 01:58

Thanks for trying to help, hitchhikr, but I already have those registers defined - I just forgot to list them in the above code.

Regards,
Lonewolf10

hitchhikr · 26 June 2011, 02:11

Check the code i modified more carefully.

Leffmann · 26 June 2011, 16:11

That code will still not work. You have to call FindTask(0) and then find the message port structure at offset pr_MsgPort. You also need to have multitasking disabled before you call ReplyMsg and leave it disabled, or your program memory might get freed up and reused before you've quit. AmigaOS will enable the multitasking again after your program has quit.

The way you find the system's Copper program through the exec base is probably not safe either. You can f.ex enable interrupts first before calling RethinkDisplay or RemakeDisplay and then enable DMA, and the system will write the Copper pointer itself.

Here's a minimal WBStartup you can use like this:

jsr WBStartup
(run demo)
moveq #0, d0
rts

phx · 27 June 2011, 10:06

As Leffmann already assumed, this is causing your black screen by starting a bad copper list.

Code:

        move.l     ExecBase,a6
        move.l     156(a6),a6
        move.l     38(a6),COP1LCH        ;restore workbench copperlist
        move.w     #0,COPJMP1        ;write any value to COPJMP1

156(a6) is ExecBase.IntVects[6]. But you need the GfxBase to restore the copper list from offset 38.

StingRay · 27 June 2011, 12:07

Quote:

Originally Posted by phx

As Leffmann already assumed, this is causing your black screen by starting a bad copper list.

Code:

        move.l     ExecBase,a6
        move.l     156(a6),a6
        move.l     38(a6),COP1LCH        ;restore workbench copperlist
        move.w     #0,COPJMP1        ;write any value to COPJMP1

156(a6) is ExecBase.IntVects[6]. But you need the GfxBase to restore the copper list from offset 38.

You're wrong. This is an old trick to get gfxbase, check my code here for example. I only use this trick when the code needs to be extremely size optimised. It isn't causing the black screen.

I don't know why he uses code like that though (probably copy/pasted from somewhere) when he opens the graphics.lib in a "clean" way in his init code.

As for the black screen, didn't test the code but I see that he still uses RemakeDisplay when the sytem is disabled. That might be one of the reason for the problems.

phx · 27 June 2011, 12:17

Quote:

Originally Posted by StingRay

You're wrong. This is an old trick to get gfxbase, check my code here for example.

Oh, indeed, you're right! This works because IntVects[6] is used for the Blitter interrupts, and graphics.library always installs its base in the iv_Data field to get it passed. Might be safe enough, on classic Amigas. Nice trick.

Lonewolf10 · 27 June 2011, 23:00

Quote:

Originally Posted by hitchhikr

Check the code i modified more carefully.

I did and Leffmann was right, it didn't fix my problem.

Quote:

Originally Posted by phx

As Leffmann already assumed, this is causing your black screen by starting a bad copper list.

Code:

        move.l     ExecBase,a6
        move.l     156(a6),a6
        move.l     38(a6),COP1LCH        ;restore workbench copperlist
        move.w     #0,COPJMP1        ;write any value to COPJMP1

156(a6) is ExecBase.IntVects[6]. But you need the GfxBase to restore the copper list from offset 38.

Err.. no. Everything was working perfectly ok, including returning to workbench, until I tried to fix the problem with the error message I recieved from workbench and ended up making things worse!

Quote:

Originally Posted by Leffmann

Here's a minimal WBStartup...

Many thanks, it's all working as desired and no error messages when returning back to workbench

Quote:

Originally Posted by h0ffman

I think my proudest effect so far is what I call the banksy effect. spray paints over the screen, which then peels away to reveal a logo, much like stencil art.

Sounds very cool

Regards,
Lonewolf10

27 June 2011, 10:06	#36
phx Natteravn Join Date: Nov 2009 Location: Herford / Germany Posts: 2,496	As Leffmann already assumed, this is causing your black screen by starting a bad copper list. Code: move.l ExecBase,a6 move.l 156(a6),a6 move.l 38(a6),COP1LCH ;restore workbench copperlist move.w #0,COPJMP1 ;write any value to COPJMP1 156(a6) is ExecBase.IntVects[6]. But you need the GfxBase to restore the copper list from offset 38.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Compiling WinUAE 2.1.0	rotacak	support.WinUAE	8	06 May 2010 17:09
Help compiling winuaesrc1530	AAG	support.WinUAE	2	08 January 2009 11:45
Compiling BlizKick	Toni Wilen	Coders. General	2	26 December 2007 17:06
Compiling 1.3.4	rotacak	support.WinUAE	70	22 January 2007 08:13
Slowly compiling	stainy	Nostalgia & memories	4	10 September 2002 01:28

22 June 2011, 02:39	#22
matthey Banned Join Date: Jan 2010 Location: Kansas Posts: 1,284	If anyone has or suspects problems with Devpac/GenAm, there is vasm which is very compatible in Devpac compatibility mode and is actively developed... http://sun.hasenbraten.de/vasm/

26 June 2011, 01:58	#33
Lonewolf10 AMOS Extensions Developer Join Date: Jun 2007 Location: near Cambridge, UK Age: 44 Posts: 1,924	Thanks for trying to help, hitchhikr, but I already have those registers defined - I just forgot to list them in the above code. Regards, Lonewolf10

26 June 2011, 02:11	#34
hitchhikr Registered User Join Date: Jun 2008 Location: somewhere else Posts: 511	Check the code i modified more carefully.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)