NMI causes game crash

[oerg866]

Hi,

I'm writing a replacement for the FM emulation TSR for early VIA chipsets. They assert an NMI whenever a data port write to the OPL ports happens, which a TSR can hook to generate the audio data.

Unfortunately, DOS/32A seems to just crash and burn whenever an NMI happened (the original DOOM does not). Can the extender be configured/linked in a way where NMIs are ignored like in the original DOOM/DOOMII, which both work okay with my TSR?

Best
Eric

[viti95]

I'll take a look, there is an utility to configure DOS/32A but don't know much about NMIs. This same issue happens with the VIAFMTSR utility.

[viti95]

A small update on this, I forgot that the DOS32AWE extender is a fork of DOS32a that is able to fix the NMI problem for the AWE TSR, and also does for VIAFMTSR. Maybe the solution is to create a DOS32a fork with the NMI fixed, and integrate it into FastDoom.

[dougvj]

It may be possible to install a handler for the NMI vector with DPMI and ignore the interrupt with stock DOS32a. I'll investigate

[sneakernets]

I'd like to add that the later Toshiba Librettos (the ones with MMX CPUs, if that helps) do not handle certain NMIs in a robust way as one would expect, and can cause this exact crash when in Windows 9x. the Yamaha Chip is a PCI-based implementation which requires a little voodoo to get DOS programs to see it, which can result in the program bombing if the program itself is trying to do any extra funny business. I wish I could provide more info here, but getting a Libretto for a dev to test on isn't exactly cheap anymore.