Re: MiNT 1.15.1.beta-1 problems

[Michael Schwingen <rincewind@tubul.dascon.de>]

On Tue, Nov 03, 1998 at 10:44:28PM +1200, Mario Becroft wrote:
> > If MiNT takes control of the Reset vector, then it is highly 
> > possible that parts of the hardware would be prevented from
> > re-initializing after a crash, if a register is held locked.
> 
> I'm sorry, but I think any computer technician would tell you that this is
> nonsense.

Not necessary.

MiNT installs its own reset vector, which restores a lot of registers and
system variables before continuing with the boot process.

If *anything* fails there, the machine will completely lock up - there is no
recovery from bus errors etc. in that stage of the TOS initialization.

Now the point is that most of what MiNT does in the reset vector does not
seem to be necessary, so when removing this stuff, we could reduce the risk
of a hard lockup (and be it only because some crashing program overwrote
part of memory, be it MiNTs reset code or the data needed there - without a
reset handler, TOS would do a clean reboot when pressing RESET).

> > That would definitely explain why the hardware freezes and
> > video goes bezerk after an AES crash, when I try pressing 
> > Reset to reboot.
> 
> I think there could be any number of reasons why that would happen. I
> don't think any of them have anything to do with a register being "held
> locked" (what ever that is supposed to mean).

The only thing which can cause the machine to completely lock up so that
even the RESET button does not work is a reset vector which is pointing to
some code which crashes.

> Everyone seems to be talking about the reset vector, I have not looked at
> how this relates to MiNT, but it's certainly possible that if this was
> pointing to garbage, exactly this problem could happen (I believe this is
> what other people have in fact observed when the reset vector points to
> fast RAM which is no longer valid after a reset). But don't quote me on
> that, I may be rather missing the point, as I haven't been folloing that
> discussion closely.

It seems so. Now, what was your point actually?

cu
Michael
-- 
Michael Schwingen, Ahornstrasse 36, 52074 Aachen