Kaeon - Horizontal Scrolling Corruption

[JWatersMeet]

Describe the bug

When running the shoot-em-up Kaeon, the graphics are corrupt in game when the playfield scrolls horizontally. See the video below.

https://github.com/joncampbell123/dosbox-x/assets/136824455/af061dce-eb08-43be-9063-d0f2a82e8c1e

Steps to reproduce the behaviour

Start Kaeon, Start New Game, Notice Corruption.

Expected behavior

The playfield should scroll horizontally smoothly.

What operating system(s) this bug have occurred on?

Windows 11

What version(s) of DOSBox-X have this bug?

All Known

Used configuration

No response

Output log

No response

Additional information

I have tried all kinds of options to fix, including manually adjusting CPU cycles to get the timing right, but no luck. The game renders correctly in x86Box.

Have you checked that no similar bug report(s) exist?

• [X] I have searched and didn't find any similar bug report.

Code of Conduct & Contributing Guidelines

• [X] I agree to follow the code of conduct and the contributing guidelines.

[rderooy]

Have you tried different VGA adapter options, in particular machine=vgaonly?

[JWatersMeet]

Thanks for the suggestion, but no luck unfortunately, including vgaonly.

[grapeli]

This is a regression that came with some patch backported from dosbox-svn. I briefly checked under dosbox 0.74-2 it works fine. Under dosbox-svn and dosbox-staging incorrectly. Probably, and under one of the older versions of dosbox-x, it also worked correctly.

[MX9000]

@grapeli I don't know if this information would be helpful in identifying the regression cause, but the game guide refers to a particular screen refresh technique that is used only if a certain range of RAM addresses are not allocated to high memory.

[JWatersMeet]

Thanks all - I've tracked the regression range down by a manual bisection - the rendering broke in the Jan 11th 2016 build of DOSBox-X.

https://github.com/joncampbell123/dosbox-x/releases/tag/dosbox-x-windows-v0.801.gusandbootfixes.b4

Gravis ultrasound DMA, IRQ, and DMA streaming emulation fixes. Boot stage reorganization, to shift to hardware -> BIOS -> DOS bootup process. S3 graphics emulation fixes. Linear framebuffer fixes, Windows no longer crashes with S3 emulation. VESA BIOS hacks were added for games that have modelist reading problems.

As such, I assume the regression is in the changes at yhe link below. https://github.com/joncampbell123/dosbox-x/compare/dosbox-x-windows-v0.801.newyears.b3...dosbox-x-windows-v0.801.gusandbootfixes.b4

[grapeli]

I ran git bisect. With default settings it was working fine until before this commit c0e07de. With this change, you need to change the default s3 card to et4000.

Works correctly. dosbox-x -set machine=svga_et4000

[JWatersMeet]

Setting it to an ET4000 fixes it - many thanks!

Though, I definitely see it break under the default S3 SVGA option and VGA Only options in the regression range i posted above - is that becuase this would also break on a real S3 card and the changes in the regression rane actually make that behaviour accurate?

[grapeli]

@JWatersMeet I just tentatively assumed it might be a regression. It's hard to say what it should look like (I'm not a programmer). Operation in dosbox-x is the same as in dosbox-staging and dosbox-svn, for correct operation use machine=svga_et4000. It's not a big hassle or complication.

Just out of curiosity, I checked the operation of this game under Qemu and Dosemu2, in both cases the graphics are displayed correctly, but in both cases the game runs much, much too fast.

[JWatersMeet]

I'll mark as closed given there's a fix. Yes, it runs too fast as well, but at least it runs now.