Mon, 23 Dec 2002 15:28:36 -0500
Tony Clark <firstname.lastname@example.org> wrote:
> If we need a watchdog then we had all better go back to win95, there it would
> be useful. Seriously though this type of application shouldn't need one. I
> use watchdogs in embedded stuff but there are quite different animals. They
> shouldn't need them either but some of the embedded guys I work with don't
> realise that :)
When your really serious about your embedded device, you don't want it
to not be operational. Its almost impossible these days to test every
single test case of every input/output possiblity and all states that
the device can get into. Consequently, its a nice safeguard to throw
in the watchdog.
Granted, it should't crash. But its nice when it does (because it will)
to have automatic recovery.
How about having the watchdog write a log message, and have Myth present
the log on launch if the last launch resulted in a crash. Similiar
to Windows knowing you did a good shutdown vs. a bad shutdown and
forcing fdisk to run on startup. :) If the last run crashed, Myth
shows a message. If it didn't, no message.
This might be able to turn into a nice way to send feedback to the
developer. If you have gdb installed and built Myth with debugging
symbols on you can get a stack trace from the core, email it out,
and delete the core - all from the comfort of the couch and the remote
control. Just throw a button "Send Crash Info to Developers" in that
message box to kick off a shell script doing this work.