DAXTER crash in softgpu, when frameskip is enabled

[sum2012]

Game or games this happens in

NPUG80329

What area of the game / PPSSPP

After loading rom

What should happen

Don't crash

Logs

Platform

Windows

Mobile phone model or graphics card

Nvidia Geforce 710

PPSSPP version affected

v1.14.4-690-g2dc3a6b05-windows-amd64

Last working version

None

Graphics backend (3D API)

OpenGL / GLES

Checklist

• Test in the latest git build in case it's already fixed.
• Search for other reports of the same issue.
• Try resetting settings or older versions and include if the issue is related.
• Try without any cheats and without loading any save states.
• Include logs or screenshots of issue.

[unknownbrackets]

I can't reproduce in other games - this is a bit strange. Can you try a Solution -> Clean, and then build again? Sometimes after refactoring, things get messed up.

Doe sit happen in any other games?

-[Unknown]

[sum2012]

Yes,I try rebuild same,I forget another line of useful of crash trace.
I tested 20 games,only this game crash

I can't reproduce in other games - this is a bit strange. Can you try a Solution -> Clean, and then build again? Sometimes after refactoring, things get messed up.

Doe sit happen in any other games?

-[Unknown]

1

[sum2012]

v1.14.4-690-g2dc3a6b05 full debug log full debug log:
https://gist.github.com/sum2012/b53fd1476c4239c77c47a4a7e99452ca

[hrydgard]

Hm, weird. I'm not getting this crash.

[sum2012]

@hrydgard
frame skip 1 to 9
frame skip type:both type

[hrydgard]

Yup, that does blow up immediately (tried frameskip 1). Weird!

(Though frameskipping with the software renderer is kind of a weird combination, given how the software renderer is focused on accuracy, and skipping stuff certainly isn't accurate, and can break things... Maybe we shouldn't allow it)

[hrydgard]

Heh, some .. interesting memory corruption is happening.

This is the SoftGPU object viewed in the debugger:

Something overwrote a whole bunch of pointers with 0303030303...

[hrydgard]

Data breakpoints to the rescue. It's the memset in SoftGPU::MarkDirty, for whatever reason it stomps outside the vramDirty_ array:

It's only 2048 bytes long, but it writes 2048 bytes into it at an offset of 560.

No idea why this would only happen when frameskipping...

Aha... the plot thickens, note the SKIPFRAME check:

Alright, seems we are estimating too large a region here.

[sum2012]

Thanks @hrydgard