James Mastros wrote:
>
> Hey all. I hate to send user-level questions to a development list, but
> given all the sucesses that have been reported, and the lack of anywhere
> else to turn...
>
> I've tried booting off of several different disks: the Windows '98 Emergency
> Boot Disk, the IBM Drive Fitness Test disk, and the Debian 2.1 install disk.
Yes, I think Ramon is right. These would be expected to fail, or
are encountering an instruction which is not emulated yet.
FWIW, right now, I have CPUID virtualized to return a stripped Pentium.
So anything compiled to use otherwise will fail. Will change that
later, but it's important for debugging now.
Please test again for the next version. These reports are helpful.
The next batch of mods I'm working on now, should handle more
CPU mode transitions, in our virtualized context. For example,
one mode is PM->RM where the descriptor caches are left in
a PM state. This is handled differently, then PM->RM where
descriptor caches are loaded with RM compatible values before
the transition. In the 1st case, we have to run guest code
in PM and virtualize accesses to the segment registers. In
the 2nd case, we can run guest code in V86M, and let the guest
code access segment registers normally, using the CPU to
reload them properly.
These new capabilities should fix the bootloader issues, DOS access
to extended memory, and allow more stuff to run.
BTW, I'm currently about 26Million instructions into a Win'95 boot,
using pure emulation mode.
-Kevin
