Misc MOL questions

[Samuel Rydh]

On Tue, Mar 11, 2003 at 11:59:07PM -0800, Dara Hazeghi wrote:
> Now on to the questions...
> 
> So regarding MacOS 7.6 (and 8.0). It seems that the
> AWACS check is the cause of my troubles. Question is
> how can I work around it. The hang is early in the
> bootup sequence, even before extensions have begun
> loading, so I doubt MOLAudio is the answer. I've tried
> disabling audio (and startupboing) and that doesn't
> work. Any other suggestions? Obviously it's not a
> _big_ deal, but...

Well, I think it is of some importance to keep MOL
compatible with these really old MacOS versions.

> As to MOL networking in MacOS 8.1, I'm pretty much
> baffled. I configure it to use DHCP, tell it to use
> Ethernet Slot C1, and as soon as I make any network
> request, everything save screen updates hang, for a
> real long time.

I just did some tests, and somethings does not work
right with networkings in the oldworld settings.
I'll investigate. I would guess it is the
interrut assignment stuff. In oldworld mode, only
three interrupts are available to MOL so they have to
be shared.

> Nor for that matter, does anything
> ever make it through the network. For 4-5 minutes,
> MacOS appears to hang, and then regains interactivity,
> as if nothing had happened. In the process, I get a
> number of "Warning: Ethernet packet dropped." Any
> ideas on how to debug this?

The 'ethernet package dropped' message is normal.
MacOS dynamically increases the receive buffer
when a package is dropped so a couple of dropped
packages is normal.

> Final question is regarding VirtualPC. I gather that
> it's not expected to work.

Well, I believe VirtualPC does something really low level.
I don't have VirtualPC here, so I can't debug it.

> On MacOS X, indeed, I alway get "Unhandled PowerPC
> Exception: 0x300 data access fault, in VirtualPC, 
> but MacOS manages fine.

You mean, OS X native.

> But on MacOS 9, it also crashes the emulator. Under debugging
> mode, I get: Exiting due to crashed thread, and the
> debugger itself is hung.
>
> Not in debug mode, I simply get a segmentation fault.

That is really bad. The MOL process should never crash, regardsless
of what the hosted OS does. I'm very interested in the
stacktrace...

> Not that I also receive some
> screen corruption (upper parts of the screen turn
> black), so it could be an X bug.

Screen corruption? The MOL window or the X window?

> P.S. It's cool to see that booting OS X on MOL is now
> dramatically faster than native.

I'm actually planning further improvements in this area.
By asynchronous filling the disk cache in advance during
the startup, it is possible to almost completely eliminate
the delays due to disk access.

/Samuel