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)

Toni Wilen 28 September 2021 18:46

https://download.abime.net/winuae/fi...uae_4900b35.7z
https://download.abime.net/winuae/fi...e64_4900b35.7z

Beta 35:

- Adjusted 68000 IPL sampling timing. (Spectre Party / Phenomena, F1GP by MicroValue. Not Microprose!)
- Blizzard PPC: if RESET m68k instruction executed, apparently board logic also forces external CPU reset.
- Fixed GDI handle leak (GUI panel change always leaked 2 font handles)
- Updated GUI tab order.
- If emulator was started by selecting statefile and GUI was opened after starting emulation, every time GUI Misc panel was opened, selected statefile was set to reload when GUI exited. (b31?)

Viceroy 29 September 2021 10:46

Just noticed a possible bug with the latest 35 beta.

Let's say I had "Fastest possible" set, and also for some silly reason "Cycle-exact (Full)" set.

If I then untick "Cycle-exact (Full)", then "Fastest possible" now changes to "Approximate".

Toni Wilen 29 September 2021 20:49

Quote:

Originally Posted by Viceroy (Post 1509011)
Just noticed a possible bug with the latest 35 beta.

Let's say I had "Fastest possible" set, and also for some silly reason "Cycle-exact (Full)" set.

If I then untick "Cycle-exact (Full)", then "Fastest possible" now changes to "Approximate".

It is side-effect of "silly" changes. It is ok :)

Viceroy 29 September 2021 21:25

Quote:

Originally Posted by Toni Wilen (Post 1509071)
It is side-effect of "silly" changes. It is ok :)

Thank you ;-)

amilo3438 30 September 2021 18:59

Quote:

Originally Posted by Toni Wilen (Post 1508958)
- Adjusted 68000 IPL sampling timing. (Spectre Party / Phenomena, F1GP by MicroValue. Not Microprose!)

Made In Croatia by Binary -> there now is a problem at start of the demo in 4900b35! (old CPU ucode IPL problem mentioned in the "Problematic Demos" thread)

(Quickstart A500 1.3 OCS/ECS 512 + 512 config.)


EDIT: If untick "CIA TOD Bug" in Advanced chipset it works again!
http://eab.abime.net/showpost.php?p=...&postcount=699 --> "untick "CIA TOD Bug" in Advanced chipset as a workaround"
http://eab.abime.net/showpost.php?p=...&postcount=700 --> "CIA TOD bug is the only option that is a sort of a hack fix"
http://eab.abime.net/showpost.php?p=...&postcount=706 --> "disk timing sensitive"
http://eab.abime.net/showpost.php?p=...&postcount=710 --> Properly fixed!
http://eab.abime.net/showpost.php?p=...&postcount=711 --> Explanation!

Zarnal 30 September 2021 19:11

Quote:

Originally Posted by amilo3438 (Post 1509211)
Made In Croatia by Binary -> there is problem at start of the demo in 4900b35!

(Quickstart A500 1.3 OCS/ECS 512 + 512 config.)


Yes ( original OCS/ECS version ). The old problem is back ( crash and Guru ).



http://ftp.amigascne.org/pub/amiga/G...eInCroatia.adf

Toni Wilen 01 October 2021 08:48

It does stupid interrupt setup (=IPL timing dependent. No one is surprised anymore.), CIA bug change is only a workaround, stupid setup enables CIA-B interrupt which state depends on CIA TOD bug.

This can be worked around easily but I'll need do some testing first.

Toni Wilen 01 October 2021 20:50

Another 68000 IPL timing update. Made in Croatia works again and F1GP and Spectreparty still works :)

saimo 02 October 2021 23:49

@Toni

Starting from b33, floppy drive emulation is broken: if cycle-exact emulation is not selected, writing to floppy images either lasts forever or even causes the system to freeze (but WinUAE responds to F12 and is able to reset). Floppy emulation speed does not matter (i.e. 100% speed doesn't solve the problem if cycle-exact emulation is off and turbo speed works fine if cycle-exact emulation is on).
Initially I stumbled upon this problem when writing to an ADF on a USB pen drive and I thought that the device was broken, but after several tests and reformatting, I came to the conclusion that the problem might be elsewhere; so I created a couple of ADFs on the desktop and the result was the same.
For testing, you can use the last configuration file I sent you (if you don't have it anymore and you need it, just let me know).

Rotareneg 03 October 2021 02:39

I couldn't duplicate that with an A500 quickstart without cycle exact, but with an A1200 non-cycle exact quickstart Workbench would freeze if I used the turbo key while formatting a floppy disk, worked fine if I left it to format at the normal speed (with floppy emulation set to either 100% or turbo, didn't matter.)

saimo 03 October 2021 13:24

I did more tests and found something interesting.

First off, something I forgot to mention:
* FFS;
* ROMs: 3.1
* OS: 3.9

The problem happens with files of at least 17170 bytes (yes, such an odd and exact figure).
The tool used for writing the files does not matter - I have tried the Workbench standard drap&drop function, the Copy command from shell and ClassAction.

Then, I tried with Copy after booting without startup-sequence: it works fine with files of any size. Eventually, I found that the cause is the NoClick option in the trackdisk prefs of MCP: once disabled, writing works fine again.

Even if the problem is triggered by a patch, I've been using that forever, on both my real A1200 and on UAE (whose system is basically a clone of my real Amiga plus RTG), and, as reported, the problem only surfaced with b33, so maybe it's worth investigating.

Toni Wilen 03 October 2021 14:30

It is not directly floppy related but trackdisk.device use of blitter queue (MFM encoding) when writing to the disk. Blitter interrupts seem to have something wrong in "less compatible" modes.

Toni Wilen 03 October 2021 17:19

https://download.abime.net/winuae/fi...uae_4900b36.7z
https://download.abime.net/winuae/fi...e64_4900b36.7z

Beta 36:

- Lores Display panel resolution mode: horizontal DIW values had wrong masking, bitplane left/right edge missed 1 pixel column or had 1 pixel column too much. (b29)
- 68000 IPL timing adjustment, IPL copy seems to be done when CPU is going to do read/write data during second part of memory cycle if followed by prefetch memory access. Previously it was after memory access which seems to be too late. (Made In Croatia / Binary)
- Horizontal mid screen HAM mode enable/disable combined with resolution change didn't anymore work correctly (Runaway / 2000AD, "Use joystick to move scroller!!" part)
- Fixed "box art" window image filehandle leak.
- Warp mode was much slower in Direct3D 9 compared to Direct3D 11.
- Don't emulate chipset emulation interrupt delays if not approximate/more compatible CPU speed. Delays are far too large in fast modes where chipset/chip ram accesses have unrestricted speeds. Fixes random hangs when formatting/writing to floppies.

saimo 03 October 2021 21:49

Quote:

Originally Posted by Toni Wilen (Post 1509606)
- Don't emulate chipset emulation interrupt delays if not approximate/more compatible CPU speed. Delays are far too large in fast modes where chipset/chip ram accesses have unrestricted speeds. Fixes random hangs when formatting/writing to floppies.

Confirmed to work fine now. Thanks for the quick fix!

Viceroy 03 October 2021 23:06

Hi Toni,

I'm happy to say that scrolling through "covers" from start to finish worked perfectly with DamienD's collection ;-)

...but then I decided to exit WinUAE, start again, and then scroll through title / screenshots. Crashed about 65% through...

Must be a memory leak here also.

Toni Wilen 05 October 2021 18:02

Quote:

Originally Posted by Viceroy (Post 1509669)
Hi Toni,

I'm happy to say that scrolling through "covers" from start to finish worked perfectly with DamienD's collection ;-)

...but then I decided to exit WinUAE, start again, and then scroll through title / screenshots. Crashed about 65% through...

Must be a memory leak here also.

I found another (old) bug related to this window. But does it happen randomly or always when you select specific config file?

Always do the test 3 times to see if it is random or not :)

jbenam 05 October 2021 19:17

2 Attachment(s)
Hi Toni,

Just got a crash with b36 while roaming thru the options. DMP attached.

Also I can't get RTG to auto-switch anymore - the main screen becomes black and I can't do anything. I have been able to work around the issue by specifying two separate monitor. I'm sure it must be something weird I have done in the config, but I can't find the correct option to make it work again :nuts

Config attached just to be sure.

Thanks!


EDIT: Actually I haven't got RTG enabled at boot on my install - it should start in High Res, I usually switch to RTG later. If I disable the Voodoo 3, then I don't get the black screen issue :nuts

Viceroy 05 October 2021 19:38

Quote:

Originally Posted by Toni Wilen (Post 1509910)
I found another (old) bug related to this window. But does it happen randomly or always when you select specific config file?

Always do the test 3 times to see if it is random or not :)

Not loading any config, just scrolling through the configurations list from start to finish.

Usually occurs around 65% percent but sometimes later. ;-)

Toni Wilen 06 October 2021 18:49

Quote:

Originally Posted by jbenam (Post 1509923)
Also I can't get RTG to auto-switch anymore - the main screen becomes black and I can't do anything. I have been able to work around the issue by specifying two separate monitor. I'm sure it must be something weird I have done in the config, but I can't find the correct option to make it work again :nuts

Does END+F9 still switch normally? Native (usually grey if there is nothing to be seen) and Voodoo (black). In that case it switches correctly, there simply isn't anything to see.

EDIT: HDF needed to debug this more.

jbenam 10 October 2021 21:34

Quote:

Originally Posted by Toni Wilen (Post 1510099)
Does END+F9 still switch normally? Native (usually grey if there is nothing to be seen) and Voodoo (black). In that case it switches correctly, there simply isn't anything to see.

EDIT: HDF needed to debug this more.

Yep, it does. So WinUAE actually autoswitches to RTG even if it's not active, as soon as you enable an RTG card?

If that's intended behaviour, then consider my "bug report" closed :)
If you still need me to upload the HDF, just let me know and I will gladly do so.

If I can make a suggestion, maybe an CHIPSET/RTG indicator in the status bar might help troubleshoot what some users might seem like a bug?

Example: I would probably enable the RTG card before even installing drivers, if the same thing as above happens, I would end up with a non-booting system, spending the next few hours troubleshooting a non-issue, or even thinking that maybe the RTG card I have selected in WinUAE is not working properly :laughing Anyway, these are my two cents, thanks as always for your incredible job :great


All times are GMT +2. The time now is 13:50.

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

Page generated in 0.12347 seconds with 11 queries