The game is either mis-detecting the S3 chipset, or not using it in a way that is compatible with DOSBox. However, machine=svga_et4000 or machine=svga_paradise works OK. If using machine=svga_s3, the game works if you force the use of VESA by adding a command-line parameter of "vesa".

The issue is very alike to Wooden Ships and Iron Men Won't start
Wouldn't surprise me if both games share a common SVGA graphics library.

I tried all the machines, nothing, same thing... the screan just flickers the text from DOSBox but the game is running, hearing everything, but can't see anything. I also tried forcing vesa from the post you linked, not sure if I did it right. Never done it before. The game's exe = scream.bat

So how would I type it to force vesa, they typed " wsim /v"

/v is force vesa?

You do realize you have to change the machine type in the conf file and restart DOSBox? just typing "machine=svga_et4000" at the prompt doesn't actually switch to the other machine type.

If I didn't spell it out clearly enough, type "scream vesa" to run with forced VESA.

original game? hehe.

Fixed!! Forcing VESA worked and I can see! Thanks Ripsaw8080!

Lightmaster, yes, original game. I heard there were install bugs, but I guess I had no problems installing the game.

The game works without forcing VESA in 0.72 (which does have S3 emulation). This CRTC detection code is used:

1   mov   dx,03d4h
2   mov   ax,4838h
3   out   dx,ax      ; write 0x48 to index 0x38 (Reg lock 1)
4   mov   ax,0a539h
5   out   dx,ax      ; write 0xa5 to index 0x39 (Reg lock 2)
6   mov   al,47h
7   out   dx,ax      ; write 0xa5 to index 0x47 (HGC orgX)
8   inc   dx
9   in    al,dx      ; read index 0x47
10   mov   ah,al
11   mov   al,0ffh
12   out   dx,al      ; write 0xff to index 0x47
13   in    al,dx      ; read index 0x47

In the last port read, DOSBox 0.72 always reads 0, but 0.73 reads back the 0xff that was written. The game also misbehaves in 0.72 if index 0x47 returns what is written to it. This is not to say the current register behavior is wrong, only that it appears to be a hinge point for the issue.

Does it work if you mask off bits 11-15?

err 3-7

It works when any value other than 0xff is read back after being written, so masking any bits at all; but just to be sure I changed it as you specified, and the game works without forcing VESA:

1   case 0x47:  /*  HGC orgX */
2-     return vga.s3.hgc.originx&0xff;
3+     return vga.s3.hgc.originx&0x07;

The other game (WSIM) works as well with any bits masked on 0x47.

Unfortunately that is not how real hardware behaves. Might be used to ID some older S3 chips.

I see a special video mode has been added in the svn for this

Yes, the discussion was through PMs.

ripsaw8080 wrote:

So in 0.72 the games *failed* to detect S3 and fell through to using VESA; however in 0.73 the S3 was detected because of improved emulation accuracy of the regs, but then there was no support for mode 0x69. I see you called it a regression in the change log, and I guess it is in a sense; but the truth of the issue is a bit more twisted. Happy

Thanks for clearing that up.