Re: [openMSX-devel] Disk emulation works for me now

[Joost Y. D. <j....@qa...>]

'Maarten ter Huurne' wrote about '[openMSX-devel] Disk emulation works for me now' - Mon, Dec 31, 2001 at 04:14:36PM CET
> Hi,
> 
> The reason disk emulation wouldn't work yet was this:
>   // TODO, currently this just assumes RAM is in PS:3/SS:3
>   // this is of course quite wrong
> and NMS8250 has RAM in slot 3-2. After hardcoding 0xAA instead of 0xFF for 
> the secundary slot select register it works.
> A more reasonable guess for the RAM slot to load in is to take the slot 
> selected at #C000..#FFFF. This allows loading only in the primary mapper, 
> which is not the way DOS1 does it, but in a single mapper system or under 
> DOS2 it works just fine.

fixed. Please give this a swing ;-)

> I also discovered a small bug in MSXDiskRomPatch:
>   for (int sector = sector_number; num_sectors--; sector++)
>   {
>     ...
>   }
>   regs.BC.B.h--;
>
> The decrease of B should be inside the loop ofcourse, after all B contains 
> the number of sectors not yet loaded. Actually, you could use B as the loop 
> guard. This code occurs twice, once for read, once for write.

fixed.

> I have some doubts about the usefulness of this statement:
>   if (transfer_address + num_sectors*MSXDiskRomPatch::sector_size > 0x10000)
>   {
>     // read would overflow memory, adapt:
>     num_sectors = (0x10000 - transfer_address) / MSXDiskRomPatch::sector_size;
>   }
> A read like that would crash a real MSX as well, no need to protect against 
> that. A write like that should be allowed. Since writeMem has an address 
> parameter of type "word", there is no risk openMSX would read or write 
> out-of-bounds.

I'm not sure... perhaps the diskrom does some checking also.
I'd like to investigate later how the real diskrom handles it
without patching, to have some comparison.

> I hope you're not tired of screen shots yet... see attachment.

Never :)

> Undeadline doesn't work well, though. During fade outs, the game gets very 
> slow. This happens for example halfway the intro or when you start a new 
> game. Also, in the game itself the screen split is very wrong making the game 
> almost unplayable. I think I have to do some serious reviewing on the line 
> interrupt code.

Joost