Does WinUAE clear [Chip]RAM from Ctrl-Amiga-Amiga?

[Photon]

It certainly doesn't on a real A500 and it's in the way of testing right now, so is there a way to turn it off?

If it doesn't, please say so and I can error search elsewhere.

[ross]

Originally Posted by Photon

It certainly doesn't on a real A500 and it's in the way of testing right now, so is there a way to turn it off?

If it doesn't, please say so and I can error search elsewhere.

It depends.

With soft reset (CTRL-A-A or via software), it doesn't clear the memory.
With the GUI and Reset button it clear the memory.

[Photon]

I tested it and Ctrl-Amiga-Amiga definitely clears e.g. $30000, $50000, $7e000, and $c40000. Same with softreset in code. 2.0+ is fine.

I've looked but found no setting for it. I fear this might be a compatibility fix(?) but I've definitely rescued sources and data under 1.3 and this is no longer possible in WinUAE 4.4.0.

[Toni Wilen]

KS 1.3 clears Chip RAM if Chip RAM size is larger than 512k (KS bug)
setpatch -r should fix it.

(and not enough information)

[Photon]

Originally Posted by Toni Wilen

KS 1.3 clears Chip RAM if Chip RAM size is larger than 512k (KS bug)
setpatch -r should fix it.

(and not enough information)

That would explain the differences to 2.0+! Wow, that was a tricky/sneaky bug. Will test at once. Cheers Toni

[ross]

Originally Posted by Photon

That would explain the differences to 2.0+! Wow, that was a tricky/sneaky bug. Will test at once. Cheers Toni

http://eab.abime.net/showpost.php?p=...8&postcount=11
You can also solve the problem of KS1.x and 1MB chip-ram

[Photon]

Ross: Come on, I don't know every bug of every configuration. I leave that to those who have to deal with it like a pro (Toni and you ) Just trying to write correct code here. Sometimes it can't be done it seems!

Tested: Yes, data at $7e000 stays from Ctrl-Amiga-Amiga and code softreset, if <=512KB Chip RAM.

So. Old code is solid for 1.3. Can't do anything about the clear bug.

(Side note: Would be kinda handy to trigger it for a hardreset on 2.0+ tho, since ColdReboot() doesn't )

Anyway thx again for info Toni

[ross]

This is the relevant code.
I skip all the code that setup ColdCapture vector and check for KS1.x.

ColdCapture_code
	;fix 1MB chip on Ks1.3
	bchg	#1,$bfe001		;cosmetic
	movea.l	$3e(a6),a3		;read MaxLocMem
	jmp	$1aa-$184(a5)		;and use it

Basically skip ROM code that setup the wrong $80000 as top of chip-mem and use the full chip-memory available.

No idea what setpatch -r does, but this tricky code works
(I simply dissected the KS to see what the problem was)

[Photon]

Cool, is this your fix for 1.3, or what ColdCapture does on 2.0+?

[ross]

Originally Posted by Photon

Cool, is this your fix for 1.3, or what ColdCapture does on 2.0+?

My fix for 1.3.
You need to execute it very early on KS startup (you only can do it at ColdCapture time because after is to late..)

From 2.0+ >=1MB chip-ram and reset-proof code is properly handled.

[Photon]

OK, just what I was looking for. "We're gonna need a bigger Boot" (c) ...

What is a3/a5/a6 (mostly a5 if a6 is execbase)?

[ross]

Example code to use it:

		moveq	#0,d2
		cmpi.w	#KS13,lib_Version(a6)
		bne.b	.skip
		move.l	#ColdCapture_code,ColdCapture(a6)
		move.w	d2,LowMemChkSum(a6)
		lea	SoftVer(a6),a0
		moveq	#$18,d0
.loop	        add.w	(a0)+,d2
		dbf	d0,.loop
		not.w	d2
		move.w	d2,LowMemChkSum(a6)	;checksum

This ColdCapture code is a one-shot because from what I remember ROM code clear ColdCapture vector after use.

Actually if I remember correctly the KS1.2 also suffers from the exact same bug, but Amigas with 1MB of chips-RAM and KS1.2 never existed...

[ross]

Originally Posted by Photon

OK, just what I was looking for. "We're gonna need a bigger Boot" (c) ...

What is a3/a5/a6 (mostly a5 if a6 is execbase)?

a6= execbase
a3= right top of chip-ram (and not the wrong $80000 hardcoded one)
a5= return code in ROM (the two offset define the faulty code to skip, and I re-enter with a correct a3).

Yes, is a dirty trick

[Photon]

Ah. I thought the first few lines (and even the bchg can be removed) was the fix.

Too experimental for a correct bootblock I feel. If there's no other way, 1.3 users with >512KB chip must accept the KS bug.

And sorry for posting in this forum. I just switched kickstart to test software and got the bug, and knew 1.3 didn't on my real Amiga. Emulator is correct.

[ross]

Originally Posted by Photon

Too experimental for a correct bootblock I feel. If there's no other way, 1.3 users with >512KB chip m

I bet setpatch -r doesn't do much differently than that.
Maybe it will use a lot of overbloated checks but I don't think there are many ways to fix a similar bug that happens in the very initial stages of the kickstart startup.