Re: [MiNT] Invalid S?_MAGIC

[Peter Slegg <p.slegg@scubadivers.co.uk>]

On Sat, 15 Oct 2011 12:14:21 , Peter Persson <pep.fishmoose@gmail.com> wrote:
>
>
> 15 okt 2011 kl. 00:34 skrev Peter Slegg:
> >> Did you really isolate it to be caused by Thing? Because like I said
> -
> >> the kernel will kill Thing even if the problem is caused by some
> other
> >> process.
> >
> > If the kernel catches the memory violation and kills Thing, could it
> not
> > ask the user first ?
>
> Somehow this seems like a very bad idea. The purpose of memory
> protection is to prevent bad things from happening; you can't know for
> sure *why* a memory violation happened (even though it's generally a AES
> msg protocol vs. non-global memory mishap in this case). If you allow
> the access to take place, you could end up with a crash. If you ignore
> the access, you could also end up with a crash. It doesn't make sense.
>
> 1. Configure applications (binary header) properly.
> 2. Don't use ill-behaved applications.
> 3. Celebrate.
>
> -- PeP

We know there misbehaving apps that everybody uses that don't actually
do any harm.

In the situation where, for instance, Texel does something that causes
Thing to be killed, it seems faily pointless.

The app that misbehaved is still running and the app that was killed is
going to be restarted again.

If a warning pops-up and says:

Thing is trying to access protected memory belonging to Texel.
It could offer me the choice of killing Texel, killing Thing or
allowing the access.

Nothing is really gained by automatically killing an unsuspecting
app other than forcing people to turn MP off.

In an ideal world none of this would happen but we have a far from
ideal world with lots of legacy apps written for single tasking or
making bad assumptions about the OS. A user friendly solution is needed.

Regards,
Peter