English Amiga Board

English Amiga Board (http://eab.abime.net/index.php)
-   support.WinUAE (http://eab.abime.net/forumdisplay.php?f=5)
-   -   WinUAE 4.9.0 beta series (Was 4.5.0) (http://eab.abime.net/showthread.php?t=104099)

AMIGASYSTEM 04 November 2020 21:26

Quote:

Originally Posted by hexaae (Post 1438943)
If it's a too fast -> division by zero case, there's a patch on Aminet to solve these kind of errors: http://m68k.aminet.net/search?query=div0

Thanks hexaaem NoMoreDiv0 solves, now Fusion works well even with JIT ;)

Toni Wilen 04 November 2020 21:45

Quote:

Originally Posted by Toni Wilen (Post 1438992)
This is not specific to some JITted instruction but something more weird because if I disable some JIT instruction, it starts working and disabling some others will break it even more "Error 41 in line 8: Invalid expression"

Most likely arexx does something really really weird that triggers some quite hidden bug..

EDIT: found something, there is 64-bit specific optimization related to flags. THE works if it is disabled. More debugging to do..

Fixed. winuae64.7z updated. Was too simple, 64-bit JIT uses different m68k to x64 flag mapping but required define was missed. (I'd have expected much larger problems than rexx interpreter getting confused, so most likely this code was only used when there was some specific JIT to non-JIT or vice versa transition was needed)

hexaae 04 November 2020 22:03

:bowdown

EDIT:
Fixed! Confirmed!

ziosante 05 November 2020 02:35

Quote:

Originally Posted by hexaae (Post 1438943)
If it's a too fast -> division by zero case, there's a patch on Aminet to solve these kind of errors: http://m68k.aminet.net/search?query=div0
For me it solved random crashes of Scout monitor and finally made my emulated WB system rock solid.


Thanks, it worked!!!
Sorry for my mistake, I assumed it was a WinUAE related issue but it's a Fusion fail

Toni Wilen 05 November 2020 19:32

https://download.abime.net/winuae/fi...nuae_4500b9.7z
https://download.abime.net/winuae/fi...ae64_4500b9.7z

Beta 9:

- 64-bit JIT used wrong flag handling mode which broke some programs (b1).
- Improved chipset display mode frame rate timing stability.

Toni Wilen 05 November 2020 19:47

Extra status report: I didn't expect to fix remaining JIT problems that "easily" so it seems usual December official release date is possible. Unless something even worse is found..

All large updates are now done, the rest is bug fixing. I hope..

Viceroy 05 November 2020 20:44

Nice one ;-)

Seiya 05 November 2020 21:49

Thank you Toni

White 06 November 2020 13:57

Toni
thanks for the latest beta and your work,
68k + CVision 3D/64 works very well

Skumball404 06 November 2020 19:46

Bug: The A3000 and A4000 Quickstart configurations do not add the Z2 fast RAM, only Chip RAM.

Toni Wilen 06 November 2020 20:04

Quote:

Originally Posted by Skumball404 (Post 1439428)
Bug: The A3000 and A4000 Quickstart configurations do not add the Z2 fast RAM, only Chip RAM.

No. A3000 and A4000 have 32-bit mainboard RAM (It is under advanced memory options) which is not in autoconfig address space. Z2 (which is 16-bit) would be painfully slow.

Also see first post. Even if it had been a bug, it would have been offtopic in this thread.

alanwall 07 November 2020 19:46

My prior problem with Image Engineer is now gone with this beta !

Toni Wilen 08 November 2020 19:24

https://download.abime.net/winuae/fi...uae_4500b10.7z
https://download.abime.net/winuae/fi...e64_4500b10.7z

Beta 10:

- Added workaround for Windows socket behavior difference compared to Unix/Amiga. Windows return truncated data and error status if bsdsocket.library recv/recvfrom() supplied buffer is smaller than available data. Unix/Amiga sockets will also return truncated data but don't return error status.
- Debug logging $bfff00 write truncated long words to words.
- Log window size was not stored in ini mode.
- ini file is now only saved if contents has been modified.
- Fixed CV64/3D in Z2 mode. (CGX4 hang, Picasso96 wrong colors in some modes)
- Fixed PicassoIV overlay. (Possibly can only happen with some P96 versions, at least some set overlay at VRAM address zero which accidentally worked)
- Fixed Cirrus Logic blitter bug that was introduced few betas ago (for example some CGX4 modes had ghost shell window cursor images)
- uaedbg invalid access message now includes information if it was caused by any chipset DMA channel. Memwatch logging must be enabled (w + return in debugger or any memwatch point enabled)

AZka 08 November 2020 20:01

Thank You!

falken 09 November 2020 00:36

Thank you Toni

AMIGASYSTEM 09 November 2020 15:20

Toni I noticed that with WinUAE 4.5.0 beta 10 I crash all my old OS2.1 based configurations, with WinUAE 4.4.0 no problem.

I attach A2286 XTIDE OS 2.1.uae, Log and .dmp config files

Toni Wilen 09 November 2020 16:33

Quote:

Originally Posted by AMIGASYSTEM (Post 1439976)
Toni I noticed that with WinUAE 4.5.0 beta 10 I crash all my old OS2.1 based configurations, with WinUAE 4.4.0 no problem.

I attach A2286 XTIDE OS 2.1.uae, Log and .dmp config files

This is previously reported compiler issue. Use 64-bit as a workaround. I'll try other compiler version if there is no update in next few weeks..

EDIT: remember to always test and report last working version!

AMIGASYSTEM 09 November 2020 17:11

Toni last version worked is WinUAE 4.4.0 beta8, have the problem instead all beta versions of WinUAE 4.5.0.

I use Win7 32Bit and I can't use the 64Bit version, I could try it on other PC anyway.

Toni Wilen 09 November 2020 20:26

Quote:

Originally Posted by AMIGASYSTEM (Post 1439995)
Toni last version worked is WinUAE 4.4.0 beta8, have the problem instead all beta versions of WinUAE 4.5.0.

I use Win7 32Bit and I can't use the 64Bit version, I could try it on other PC anyway.

It can't be same reason in pre-4.5.0 versions. This was introduced when PCem was updated to v16.

Attach dump from 4.4.0, thanks. (Not really beta related but it still might be same problem after all)

Mad-Matt 09 November 2020 22:14

Small bug that's probably a quick fix.

With CVision3D (Not tested any others) Doublescan is not producing expected behavior with HiColor screens.

With both cgxmode and p96mode selecting a 320x240 or any other low res mode results with doublescan only vertically doubling in 15/16 mode. 8,24,32 bit modes fully double vert/hori as expected.


All times are GMT +2. The time now is 07:53.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, vBulletin Solutions Inc.

Page generated in 0.12109 seconds with 11 queries