Mail Archives: djgpp-workers/1997/10/15/00:41:23
On 14 Oct 97 at 22:52, Charles Sandmann wrote:
> > How would I go about testing [VDS]? I don't think I have any 16 bit dma
> > devices at home, and 8 bit dma doesn't seem to work above 1 M (You'ld
> > think it would on an AT, the page registers didn't HAVE to stay 4
> > bit), or am I missing something?.
>
> Testing is one of the issues. I want a test program which uses many
> of the features, be able to test it on W95, make sure the test
> program seems sound, then make sure it does the right things with
> the CWSDPMI implementation. I have a sound card which supports
> 16bit DMA, but I am really too busy to spend much time on the
> testing.
I'll see what I can do (I've got the VDS somewhere, and if not, I
imagin it's on x2ftp).
>
> For example, VDS on W95 seems to always return a 63.5K length DMA
> transfer buffer - we ought to do the same (we currently return 16K
> or 32K). So my obsession with getting it right may delay getting it
> officially into a release. With my primary platform anymore (NT)
> not even supporting VDS, I'll probably do more supervising than
> doing...
Is the size of the buffer really an issue? Anyway, I've written some
code that tries (and usually succeeds) in allocating a 64.0000k,64k
aligned memory buffer in dos memory, and not one single undocumented
feature is used (other than the implicit non-move of the resize
function).
>
> Then the real question comes up ... should the effort be put into
> MWDPMI instead? I actually think so, since the compiler is much
> higher quality and more available to everyone wanting to hack...
>
Yes, I think mwdpmi is more important (another DPMI 1.0 server!!) and
the VDS effort can go into that (though cwsdpmi could currently be
used as a testbed for VDS). BTW, ootb, I got mwdpmi to run my
programs, but it randomly resets my computer (tripple faults I
assume, possibly during mode switching (my guess)).
Bill
--
Leave others their otherness.
- Raw text -