RE: fatal error handling

["Howard Chu" <chu@platinum.com>]

Y'know, there's another way... What you're talking about here is much like
how a Unix kernel handles a system panic - the kernel does a core dump, it
doesn't just output a text log somewhere. Also, the dump isn't written via a
filesystem, it's written by dedicated code to a reserved portion of the
disk. If you really want a safe way of recording status after a fatal error
in the MiNT kernel, you're going to want to do something similar.

In particular, the Unix kernel has a dedicated disk partition reserved for
the virtual memory swap space - its accessed by raw blocks, not as a
filesystem. The dump should contain the entire data & bss segment of the
kernel, as well as any heap and stack, and the register contents. This will
be a tons more useful than any text error log, because it will allow a
reconstruction of the failure condition (using an appropriate debugger
tool).

> -----Original Message-----
> From: Martin-Eric Racine [mailto:q-funk@megacom.net]
> Sent: Friday, June 12, 1998 4:52 PM
> To: MiNT-List
> Subject: Re: fatal error handling
>
>
>  >> Further to this, it would be nice if MiNT wrote an error log,
>  >> to help track the cause, after rebooting, IMHO.
>  >
>  >Would it be safe to do this after a *fatal* error?
>  >What if it's the filesystem for the disk where you're writing
>  >the log that went *KAZOOM*...?
>
> Boot partition is always TOSFS, so it could be written there.
>
>
> ----------------------------------------------------------------
> Martin-Eric Racine         # The Atari TT030 Homepage with FAQ #
> FUNKYWARE CREATIONS inc.   http://www.megacom.net/~q-funk/TT030/
> ----------------------------------------------------------------
>