Puae2021 core crashes when loading save states

Libretro Cores
#1

Hi everyone, When using the latest puae2021 core (2022-12-13 22:59) (via retroarch via launchbox) I save a state (F2), reload it (F4) and the emulation either crashes to desktop or shows a green flickering screen in the still running emulation. I can still successfully load save states from previous versions of puae2021, only new states crash.

Retroarch log content:

  • [INFO] [Input]: Game Focus => ON
  • [INFO] [Input]: Game Focus => OFF
  • [INFO] [Input]: Game Focus => ON
  • [INFO] [Joypad]: Found joypad driver: “xinput”.
  • [INFO] [Autoconf]: Xbox 360 Wireless Controller configured in port 1.
  • [libretro INFO] waiting_blits detected PC=00000d64
  • [libretro DEBUG] Output audio buffer capacity set to 3675
  • [libretro INFO] waiting_blits detected PC=00000d72
  • [libretro INFO] waiting_blits detected PC=00000d80
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [libretro INFO] waiting_blits detected PC=000075ec
  • [INFO] [Input]: Game Focus => OFF
  • [INFO] [State]: State slot: 140
  • [INFO] [State]: Saving state “X:\Poison\Retro_Central\RetroArch\states\Turrican (2.0_0092).state141”, 11025672 bytes.
  • [INFO] [State]: Loading state “X:\Poison\Retro_Central\RetroArch\states\Turrican (2.0_0092).state141”, 11025672 bytes.
  • [INFO] [State]: Saving state “RAM”, 11025672 bytes.
  • [libretro INFO] CPU: 68EC020, PC=00000F0E
  • [libretro INFO] unknown chunk ‘’ size 25 bytes
  • [libretro INFO] Reset at 00000F14. Chipset mask = 00000007
  • [libretro INFO] SNDRATE 313.0227.050.000000=3552550.000000
  • [libretro INFO] PAL mode V=50.0000Hz H=15625.0879Hz (227x312+1) IDX=10 (PAL) D=0 RTG=0/0
  • [libretro INFO] CPU=68020 Chipset=AGA PAL
  • [libretro INFO] State restored
  • [libretro INFO] Building CPU table for configuration: 68020 prefetch 24-bit
  • [libretro INFO] 1883 CPU functions
  • [libretro INFO] Building CPU, 45954 opcodes (2 1 1)
  • [libretro INFO] CPU=68020, FPU=0, MMU=0, JIT=CPU=0.
  • [libretro INFO] SNDRATE 313.0227.050.000000=3552550.000000
  • [libretro INFO] PAL mode V=50.0000Hz H=15625.0879Hz (227x312+1) IDX=10 (PAL) D=0 RTG=0/0
  • [libretro INFO] Illegal instruction: dfff at 00000F0E -> 00000000
  • [libretro INFO] B-Trap ffff at 0 (000000000729b072)
  • [libretro INFO] B-Trap ffff at 0 (000000000729b072)
  • [libretro INFO] Illegal instruction: bfff at 00000000 -> 00000000
  • [libretro INFO] B-Trap fef7 at 0 (000000000729b072)
  • [libretro INFO] B-Trap ff7f at 0 (000000000729b072)
  • [libretro INFO] B-Trap ffed at 0 (000000000729b072)
  • [libretro INFO] Illegal instruction: bfff at 00000000 -> 00000000
  • [libretro INFO] B-Trap ffff at 0 (000000000729b072)
  • [libretro INFO] B-Trap ff7f at 0 (000000000729b072)
  • [libretro INFO] B-Trap fdff at 0 (000000000729b072)
  • [libretro INFO] B-Trap fdff at 0 (000000000729b072)
  • [libretro INFO] B-Trap ffff at 0 (000000000729b072)
  • [libretro INFO] B-Trap ffff at 0 (000000000729b072)
  • [libretro INFO] Illegal instruction: bfff at 00000000 -> 00000000
  • [libretro INFO] B-Trap ffbb at 0 (000000000729b072)
  • [libretro INFO] Exception 20 3 (0) at 0 -> 0!
  • …last line repeated multiple times…

Amiga emulation has a high emotional value for me, so I’m horrified :wink: As you can see, this s my Turrican save #140, which objectively is cheating :grimacing: No, really, please please fix this (or tell me what to do differently)

#2

Thanks for notifying. It was a botched and rather needless audio backport operation on my part. Fix pushed. All new saved states were affected, so just save again from a working state.

1 Like
#3

Hi sonninnos, That was quick, thank you very much! I used the auto updater in RA and when this didn’t work, I got the core here: https://buildbot.libretro.com/nightly/windows/x86_64/latest/ Unfortunately, the symptoms are still the same.

[INFO] [Input]: Game Focus => ON
[INFO] [Input]: Game Focus => OFF
[INFO] [State]: Saving state "X:\Poison\Retro_Central\RetroArch\states\Turrican2 (1.4_CDTV).state4", 11025672 bytes.
[INFO] [State]: Loading state "X:\Poison\Retro_Central\RetroArch\states\Turrican2 (1.4_CDTV).state4", 11025672 bytes.
[INFO] [State]: Saving state "RAM", 11025672 bytes.
[libretro INFO] CPU: 68EC020, PC=00004BFE
[libretro INFO] unknown chunk '' size 25 bytes
[libretro INFO] Reset at 00004C04. Chipset mask = 00000007
[libretro INFO] SNDRATE 313.0*227.0*50.000000=3552550.000000
[libretro INFO] PAL mode V=50.0000Hz H=15625.0879Hz (227x312+1) IDX=10 (PAL) D=0 RTG=0/0
[libretro INFO] CPU=68020 Chipset=AGA PAL
[libretro INFO] State restored
[libretro INFO] Building CPU table for configuration: 68020 prefetch 24-bit
[libretro INFO] 1883 CPU functions
[libretro INFO] Building CPU, 45954 opcodes (2 1 1)
[libretro INFO] CPU=68020, FPU=0, MMU=0, JIT=CPU=0.
[libretro INFO] SNDRATE 313.0*227.0*50.000000=3552550.000000
[libretro INFO] PAL mode V=50.0000Hz H=15625.0879Hz (227x312+1) IDX=10 (PAL) D=0 RTG=0/0
[libretro INFO] Illegal instruction: dfff at 00004BFE -> 00000000
[libretro INFO] B-Trap ffff at 0 (0000000006f0d062)
[libretro INFO] B-Trap ffff at 0 (0000000006f0d062)
[libretro INFO] Illegal instruction: bfff at 00000000 -> 00000000
[libretro INFO] B-Trap fef7 at 0 (0000000006f0d062)
[libretro INFO] B-Trap ff7f at 0 (0000000006f0d062)
[libretro INFO] B-Trap ffed at 0 (0000000006f0d062)
[libretro INFO] Illegal instruction: bfff at 00000000 -> 00000000
[libretro INFO] B-Trap ffff at 0 (0000000006f0d062)
[libretro INFO] B-Trap ff7f at 0 (0000000006f0d062)
[libretro INFO] B-Trap fdff at 0 (0000000006f0d062)
[libretro INFO] B-Trap fdff at 0 (0000000006f0d062)
[libretro INFO] B-Trap ffff at 0 (0000000006f0d062)
[libretro INFO] B-Trap ffff at 0 (0000000006f0d062)
[libretro INFO] Illegal instruction: bfff at 00000000 -> 00000000
[libretro INFO] B-Trap ffbb at 0 (0000000006f0d062)
[libretro INFO] Exception 20 3 (0) at 0 -> 0!
[...]

Did I fetch the right core? I tested with version 2.6.1.3a75159 but don’t know which version I used before, sorry.

#4

The buildbot hasn’t updated the core yet. That is the old version hash.

Also hot tip for the future: Wrap the output with three backticks when posting logs.

Like so.
And so on.
1 Like
#5

Thank you very much. I will be patient now.

#6

It works! Thank you again, for the fix and the unbelievable response time :slightly_smiling_face:

#7

Excellent! My bad for breaking it for over a month…

#8

Hello everyone, I have a similar probleme with PUAE 4.9.1 d9a8dfb, Dec 2 2022 22:55:01 on linux retroarch 1.14.0 (Git 932b4ad3db). Amiga games run perfectly, but when I load a save state, the screen displays a mess. Program seems still working, but nothing is readable on the screen. Retroarch still works during the problem. I dont have this problem with UAE4ARM, neither for other systems. end of logs: [libretro INFO] FS: Flashback_v3.2_1163 (flags=00000004,E=0,ED=0,EF=1,native=’/roms2/amiga/Flashback_v3.2_1163.lha’) sta rting… [libretro INFO] Illegal instruction: 4e7a at 00239C1C -> 00239C2C [libretro INFO] B-Trap F200 at 00239C46 -> 00239C6A [libretro INFO] B-Trap F017 at 00239C92 -> 00239CC4 [libretro INFO] AUD0: forced idle state PER=2048 PC=009483b2 [libretro INFO] AUD1: forced idle state PER=2048 PC=009483be [libretro INFO] AUD2: forced idle state PER=2048 PC=009483b2 [libretro INFO] AUD3: forced idle state PER=2048 PC=009483be [INFO] [State]: Saving state “/roms2/amiga/Flashback_v3.2_1163.state”, 11027688 bytes. [INFO] [State]: File already exists. Saving to backup buffer … [INFO] [State]: Loading state “/roms2/amiga/Flashback_v3.2_1163.state”, 11027688 bytes. [INFO] [State]: Loading state “/roms2/amiga/Flashback_v3.2_1163.state”, 11027688 bytes. [INFO] [State]: Saving state “RAM”, 11027688 bytes. [libretro INFO] Saved with: ‘UAE 4.9.1’, description: ‘libretro’ [libretro INFO] CPU: 68EC020, PC=009FEBDC [libretro INFO] blitter state restore error [libretro INFO] Chunk ‘BLTX’ total size 90 bytes but read 87 bytes! [libretro INFO] unknown chunk ‘’ size 25 bytes [libretro INFO] Building CPU, 45912 opcodes (2 0 1) [libretro INFO] CPU=68020, FPU=0, MMU=0, JIT=0. 24-bit [libretro INFO] Reset at 009FEBE2. Chipset mask = 00000007 [libretro INFO] SNDRATE 313.0227.050.000000=3552550.000000 [libretro INFO] PAL mode V=50.0000Hz H=15625.0879Hz (227x312+1) IDX=10 (PAL) D=0 RTG=0/0 [libretro INFO] SERIAL: period=372, baud=9600, hsyncs=14, bits=8, PC=9febe2 [libretro INFO] CPU=68020 Chipset=AGA PAL [libretro INFO] State restored [libretro INFO] Your Amiga program just did something terribly stupid 009FEBDC PC=009FEBE2 [libretro INFO] 009FEBBC [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] [libretro INFO] 009FEBCC … … [libretro INFO] 00000000 2048K/0 = 2048K ID C32 Chip memory [libretro INFO] 00200000 8704K/0 = 8704K – F32 [libretro INFO] 00A80000 1024K/2 = 512K ID F32 Kickstart ROM (1483A091) [libretro INFO] =KS ROM v3.1 (A1200) rev 40.68 (512k) [libretro INFO] 00B80000 448K/0 = 448K – F32 [libretro INFO] 00BF0000 64K/0 = 64K – CIA CIA [libretro INFO] 00C00000 1536K/0 = 1536K – C32 [libretro INFO] 00D80000 256K/0 = 256K – C16 Gayle (low) [libretro INFO] 00DC0000 64K/0 = 64K – C16 Battery backed up clock (MSM6242B) [libretro INFO] 00DD0000 64K/0 = 64K – C16 Gayle (low) [libretro INFO] 00DE0000 64K/0 = 64K – C16 Gayle (high) [libretro INFO] 00DF0000 64K/0 = 64K – C16 Custom chipset [libretro INFO] 00E00000 512K/1 = 512K ID F32 Kickstart ROM (1483A091) [libretro INFO] =KS ROM v3.1 (A1200) rev 40.68 (512k) [libretro INFO] 00E80000 64K/0 = 64K – F32 Autoconfig [reset] [libretro INFO] 00E90000 448K/0 = 448K – F32 [libretro INFO] 00F00000 64K/1 = 64K – F32 UAE Boot ROM [libretro INFO] 00F10000 448K/0 = 448K – F32 [libretro INFO] 00F80000 512K/1 = 512K ID F32 Kickstart ROM (1483A091) [libretro INFO] =KS ROM v3.1 (A1200) rev 40.68 (512k) [libretro INFO] CPU halted: reason = 3 PC=009febe2 [libretro INFO] CPU cycleunit: 256 (0.500) [libretro INFO] 1872 CPU functions [libretro INFO] Your Amiga program just did something terribly stupid 009FEBDC PC=009FEBDC [libretro INFO] 009FEBBC [libretro INFO] 0000 [libretro INFO] 0000 [libretro INFO] 0000 … …

[INFO] [Core]: Content ran for a total of: 00 hours, 00 minutes, 28 seconds. [INFO] [Core]: No content, starting dummy core. [INFO] [Core]: Content ran for a total of: 00 hours, 00 minutes, 00 seconds. [INFO] [Core]: Unloading game… [INFO] [Core]: Unloading core… [INFO] [Core]: Unloading core symbols… [INFO] [Core]: Saved core options file to “/home/ark/.config/retroarch/retroarch-core-options.cfg”. [INFO] Threaded video stats: Frames pushed: 1712, Frames dropped: 12. [INFO] RetroArch 1.14.0 (Git 932b4ad3db) [INFO] === Build ======================================= [INFO] Capabilities: ASIMD [INFO] Built: Dec 17 2022 [INFO] Version: 1.14.0 [INFO] Git: 932b4ad3db [INFO] =================================================

Amiga emulation has a high emotional value for me too ! please help me

#9

States are working fine here even with that same game, so maybe it is game specific or a random encounter, so try with version 3.4 of Flashback slave and other games.

And PLEASE wrap log output with three backticks because that is completely unreasonable reading…

like
this
#10

Sorry for the poor log. I tried with all my amiga games and there is the same problem for all. it was working well in a recent past, and I dont remember after what update it went wrong (arkos update for instance) I reset all core options: nothing

#11

If the states are taken with the same version, I have no idea, since they are working here…

#12

I will use retroarch32 and uae4ARM, not the same experience but it is working. Thanks for your help

#13

Sorry to bump but I get a similar issue with save states, though only with “The Addams Family” (‘Addams Family, The (1992)(Ocean).adf’ - crc: 60F5D2BB) using the most recent PUAE core and RA 1.19.0,

Loading a state results in a messed glitchy screen and messed up audio (if any was playing at the time of the save):

1
1857×607 17.6 KB
2
2861×609 42.6 KB

No crashes though. This also occurs in PUAE 2021 and previous versions of the PUAE core that I’ve tested (such as 4.9). I haven’t experienced this problem with any other game so far when loading save states.

For the purpose of the test every core option was left to default.

It looks like a long-standing issue it seems. Curious if @sonninnos can reproduce the bug and why it seems to occur only with this game.

#14

Same result happens with same or any version of WinUAE, so nothing to be done here. Haven’t tested the newer ones than the core version yet though.

1 Like
#15

Indeed, I also have the same bugs on the Puae core. I’ve checked Winuae 5.0 and 5.3 (64 bit) standalone versions and I can’t reproduce this errors. It may be a core bug of Puae or Retroarch.

#16

Be more specific about what you mean by “same bugs” and “this errors”…

Just tried Addams Family with WinUAE 5.3 and indeed it works. But with 5.0 it does the exact same thing as the core.

#17

On the Puae 5.0 core in Retroarch I have the same problems as the author of the post. When I load a save state, the graphics and music are completely broken. I tested Winuae 5.0 and 5.3 and the save state works ok with Adams Family. In Winuae I used the A500 quick start settings, the game image in IPF format.

#18

States are working fine here in every other situation than Addams Family.