Tales of Phantasia with Chinese patch black screen.

[daniel229]

log information

49:24:864 user_main    W[MM]: MemmapFunctions.cpp:130 WriteToHardware: Invalid address 0005fd36
49:24:864 user_main    W[MM]: MemmapFunctions.cpp:93 ReadFromHardware: Invalid address 0005fd38
49:24:864 user_main    W[MM]: MemmapFunctions.cpp:130 WriteToHardware: Invalid address 0005fd38
49:24:864 user_main    E[KERNEL]: HLE\sceKernelThread.cpp:668 ThreadQueueList should not be empty.
49:24:864 user_main    E[*]: MsgHandler.cpp:54 Warning: ThreadQueueList should not be empty.

JPCSP fixed it in jpcsp/jpcsp@d458e1f

[hrydgard]

We must have some different issue than the one JPCSP fixed, I checked that we already parse right in 1e109a3 .

[unknownbrackets]

ThreadQueueList should never become empty. That's basically brain death for the HLE.

Does this do anything?
hrydgard:master...unknownbrackets:hle-minor

Most importantly, if you hit Yes, what does it log?

-[Unknown]

[daniel229]

Still got tha error,

39:48:211 user_main    W[MM]: MemmapFunctions.cpp:93 ReadFromHardware: Invalid address 0005fd38
39:48:211 user_main    W[MM]: MemmapFunctions.cpp:130 WriteToHardware: Invalid address 0005fd38
39:48:211 user_main    E[KERNEL]: ppsspp\Core/HLE/ThreadQueueList.h:85 ThreadQueueList should not be empty.
39:48:211 user_main    E[*]: MsgHandler.cpp:54 Warning: ThreadQueueList should not be empty.

After clicking "Yes" button,then a lot of Invalid addresses going on ,then pressing the circle button and wait a few seconds,it skips the language selection screen and go ingame.

41:55:013 user_main    W[MM]: MemmapFunctions.cpp:93 ReadFromHardware: Invalid address 0005fd36
41:55:013 user_main    W[MM]: MemmapFunctions.cpp:130 WriteToHardware: Invalid address 0005fd36
41:55:013 user_main    W[MM]: MemmapFunctions.cpp:93 ReadFromHardware: Invalid address 0005fd38
41:55:013 user_main    W[MM]: MemmapFunctions.cpp:130 WriteToHardware: Invalid address 0005fd38
41:55:013 user_main    E[HLE]: HLE\HLE.cpp:550 Unimplemented HLE function SysMemForKernel_c9805775
41:55:016 user_main    E[KERNEL]: HLE\sceKernelThread.cpp:2671 sceKernelResumeThread(273): thread not suspended
41:55:016 user_main    E[KERNEL]: HLE\sceKernelThread.cpp:2661 sceKernelResumeThread(276): cannot suspend current thread

[unknownbrackets]

Hmm, I was hoping for a Failed to reschedule: out of threads on queue log line...

Sounds like it's using kernel memory management functions and that's why it's got the invalid addresses.

-[Unknown]

[sum2012]

I told gid15 that try to fix this game in jpcsp.He did it.
Anyway we need Implement SysMemForKernel_1c1fbfe7 and SysMemForKernel_636c953b
Jpcsp source reference:
jpcsp/jpcsp@3c25717
full log: https://gist.github.com/sum2012/3b36c375c9e211d5f924

[sum2012]

We also need fix ModuleMgrForKernel_a1a78c58 (sceKernelLoadModuleDisc).
Jpcsp source reference:
jpcsp/jpcsp@fccb747

[sum2012]

@daniel229 can you help to do "replace frame download" of https://cloud.githubusercontent.com/assets/2177532/9947780/761c3376-5dcf-11e5-9899-82e9f79a83f1.jpg
? Thanks.

You need merge #7965 first (Use soft gpu to see that picture)

[daniel229]

Great,that pr works pretty fine.
@unknownbrackets
The hashes don't save to knownfuncs.ini since switch to VS 2015,the last working build is v1.0.1-891-g1c67de0

I catch these function and it's not like framebuffer download things.

[daniel229]

That screen works fine in D3D9.