WinUAE 3.6.0 beta series

[ross]

Hi Toni, various notes about debugger (some beta related, some not), numbered list.

1) in latest beta 'w' command is a bit unstable..
Do not trigger in some situation and vice versa trigger with no reason.

Attached an ADF that explain.
The executed code is this simple snippet:

	machine mc68030

	st	$100.w
	movea.l	$4.w,a6
	lea	super(pc),a5
	jsr	-30(a6)
	move.b	$f80000,d0
	rts
super	move.l	#$2111,d0
	movec	d0,CACR
	rte

No config attached because every machine with a CACR register is affected (so a quickstart a1200 suffice).

Start the machine and before inserting the ADF go to debugger and write:

w 0 f80000

w 1 100 1 w ff

Now insert, wait code loading and the memwatch at $100 trigger, that one at $f80000 do not...
(but there are various memory zone that do not work, even chipset registers);


2) in debugger (with a machine with 8MB z2 memory) if you type

s "something"

the response is bad:

s "something"
Searching from 00000000 to 00A00000..
Scanning.. 00000000 - 00200000 (Chip memory)
Scanning.. 00000000 - 00a00000 (Chip memory)

3) in machines with cache enabled if you use the 'a' command in debugger the code inserted is unusable if memory location used is in some cache
yes, I know, this is some sort of self modify code, but we need an escape:
a new command in debugger to flush cache (i've tried setting CACR manually but do not works) or simply auto-flush after last 'a' instruction inserted


4) 'fs' command do not work for me (it may be that I do not use it well..)
from help:

fs <lines to wait> | <vpos> <hpos> Wait n scanlines/position.

fs n scanline wait works, a defined position don't.

- Disassembler now also shows data in memory, makes step by step tracing less annoying, for example "CMP.W (A4)+ [2f09],D2". Not shown if address is not plain RAM or ROM.

Very useful!

Thanks, regards,
ross

[Toni Wilen]

Originally Posted by ross

command is a bit unstable..
Do not trigger in some situation and vice versa trigger with no reason.

Maybe fixed again. (or something else gets broken)

2) in debugger (with a machine with 8MB z2 memory) if you type

s "something"

the response is bad:

This goes to category "unsupported". May or may not be fixed someday..

3) in machines with cache enabled if you use the 'a' command in debugger the code inserted is unusable if memory location used is in some cache
yes, I know, this is some sort of self modify code, but we need an escape:
a new command in debugger to flush cache (i've tried setting CACR manually but do not works) or simply auto-flush after last 'a' instruction inserted

It isn't that simple. 68020+ has long prefetch, it needs reload too if you modify instruction close to current PC and that can't be automatic. (or timing would change each time debugger is exited)

I haven't yet decided how to handle this. Toggleable automatic mode sounds nice but it may be too annoying to implement..

4) 'fs' command do not work for me (it may be that I do not use it well..)
from help:

fs <lines to wait> | <vpos> <hpos> Wait n scanlines/position.

fs n scanline wait works, a defined position don't.

Perhaps it works better now but note that it can't be accurate if CPU is blocked, it needs to wait until CPU gets clock cycles.

[Toni Wilen]

Originally Posted by ross

Very useful!

I think this also should check data cache contents first and mark it with some extra character if it was cached and also if memory contents are different. (EDIT: done!)

Hmm.. Maybe some other commands should also check caches, for example "m".

This is getting complex..

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3600b4.7z
http://www.winuae.net/files/b/winuae64_3600b4.7z

Beta 4:

- If DirectDraw mode surface allocation fails, retry it with smaller size.
- Smaller than 512 block size and writing to HDF block zero was always detected as failed.
- SCSI MODE SENSE format parameters/rigid drive geometry and uaehf.device TD_GETGEOMETRY didn't return manual physical geometry if it was set.
- If config file has sound driver name that can't be found, try to find default device of same type before defaulting to first in list.
- 68030 MMU + prefetch mode delayed bus error fix.
- Added Cubo CD32 config dip switches (Cubo Switch #1 to #8), event activation = switch state toggled. State is only logged, perhaps some kind of on-screen message will be added later.
- Added Cubo CD32 option in Quickstart panel. Currently only enables serial port fake touch screen used by some later games, responds "ok" to all touch screen commands.
- Expansion panel string datatype was unimplemented (forgotten..). NIC MAC option is now visible.
- Disassembler [xxxx] data values (b3) now support 68030+ data caches. If data is located in cache and data in memory is identical, it is prefixed with *, if data is cached and data in memory is different: [value in cache:value in memory]
- Ignore DMS unpacking errors if all data is already unpacked (for example if dms has extra data at the end etc)

[ross]

Originally Posted by Toni Wilen

- Disassembler [xxxx] data values (b3) now support 68030+ data caches. If data is located in cache and data in memory is identical, it is prefixed with *, if data is cached and data in memory is different: [value in cache:value in memory]

Noticed in previous pre-beta

Thanks for your work!
ross

[Toni Wilen]

http://www.winuae.net/files/b/winuae_3600b5.7z
http://www.winuae.net/files/b/winuae64_3600b5.7z

Beta 5:

- Do not fallback to standard D3D9 from D3D9Ex if -nod3dshader command line is used, removed also other now mostly useless fallbacks.
- Reduce CPU Idle when DMA operation is active (directory filesystem/UAE HDF or true DMA SCSI controller).
- If Wait for blitter was enabled, blit used all available cycles and statefile was saved when blit was active: CPU state was saved before PC was increased. Restoring statefile would incorrectly restart the blit.
- Action Replay III CIA read/write internal NMI triggers need small delay (NMI must be delayed until following instruction completes). Fixes Virus Test "Boot" option hang.
- 341b2 mouse/joystick button release when input is uncaptured didn't fully set the new state.
- Include "JIT direct switched off: reason" message in error log.
- Relative path mode ignored Filesys path.
- Host FPU mode: set host FPU round to nearest mode temporarily (if needed) before calling any native C-library trigonometric functions, other rounding modes can return (very) wrong results.
- AVI recording file splitting (2G limit) flushed sound buffers causing random sound glitches. Added -max_avi_size <bytes> -command line parameter to set split size.

[bladecgn]

Is it just me but holding left + right mousebutton doesn't let me enter the early startup menu any more.

[Retro-Nerd]

Works fine for me. Maybe load and resave your config.

[Toni Wilen]

Originally Posted by bladecgn

Is it just me but holding left + right mousebutton doesn't let me enter the early stratup menu any more.

What does "any more" mean? Did beta 4 work? (required info!)

[Snake79]

Confirmed (winuae_3600b5) Amiga Early Startup Control works fine for me also.

[bladecgn]

It works for me very rarely if I go directly (just before first boot) to it.

If I am booted up, keep holding both mousebuttons and press CTRL+A+A the menu is not entered again.

Beta 4 is fine!

[thomas]

Originally Posted by bladecgn

If I am booted up, keep holding both mousebuttons and press CTRL+A+A the menu is not entered

I can confirm that. If you hold down the mouse buttons before you press Ctrl-A-A in RTG fullscreen, the buttons are not recognised after the reset.

If you first press Ctrl-A-A and then the mouse buttons, it works ok.

[AnnaWu]

Dear Toni,

I get a Boot Disk failure.

Bridgeboard
A2088T Color
A1060 A2088 MSDOS 1987.img (360 KB)

No problems with WinUAE 3.5.1

P.S. It seems, the raw disk image format (*.IMG/360KB/PC Bridge 40) is no more supported.
Tested with A1060 Sidecar/A2088/A2088T settings.
Boot disks (*.IMG/1.2MB/1.44MB/PC Bridge 80) are working.
Tested with A2286/A2386SX settings.

If you need a disk image file (PC side) please tell me.

[Toni Wilen]

Originally Posted by AnnaWu

Dear Toni,

I get a Boot Disk failure.

Bridgeboard
A2088T Color
A1060 A2088 MSDOS 1987.img (360 KB)

No problems with WinUAE 3.5.1

P.S. It seems, the raw disk image format (*.IMG/360KB/PC Bridge 40) is no more supported.
Tested with A1060 Sidecar/A2088/A2088T settings.
Boot disks (*.IMG/1.2MB/1.44MB/PC Bridge 80) are working.
Tested with A2286/A2386SX settings.

If you need a disk image file (PC side) please tell me.

3.5.1? Which is last working beta? (Too many unknown variables without knowing exact beta)

[mark_k]

At a guess, could it be something to do with this change in 3.6.0b1?
- Don't support 360k floppy images in Amiga drive unless it is configured as 5.25" drive. (Fixes 360k DS vs 720k SS misdetection)

[AnnaWu]

Originally Posted by Toni Wilen

3.5.1? Which is last working beta? (Too many unknown variables without knowing exact beta)

Since the WinUAE 3.6.0 beta series (winuae_3600b1-winuae_3600b5) I have this kind of problem.
The last working beta version was winuae_3500b14 and of course the official winuae_3500.
and WinUAE 3.5.1 (Public Beta 0, 2017.07.17)

[AnnaWu]

Originally Posted by mark_k

At a guess, could it be something to do with this change in 3.6.0b1?
- Don't support 360k floppy images in Amiga drive unless it is configured as 5.25" drive. (Fixes 360k DS vs 720k SS misdetection)

You are right, mark_k.

[AnnaWu]

Originally Posted by mark_k

At a guess, could it be something to do with this change in 3.6.0b1?
- Don't support 360k floppy images in Amiga drive unless it is configured as 5.25" drive. (Fixes 360k DS vs 720k SS misdetection)

mark_k, it seems only 360KB disk images have a problem with the WinUAE 3.6.0 beta series.
(*.IMG/720KB/PC Bridge 40) are working.

Toni, if you need a bootable 360KB disk image pls tell me.

[AnnaWu]

Toni you changed something?

www.winuae.net/files/b/winuae.7z
02.10.2017 20:13:44

(*.IMG/360KB/PC Bridge 40) seems to work.

Thank you!

[Toni Wilen]

It was quick fix, I am not yet sure if it is 100% correct fix.