🚀 go-pugleaf

In article <1992Sep14.150531.10762@magnus.acs.ohio-state.edu>, szatezal@magnus.acs.ohio-state.edu (Shane M Zatezalo) writes:
|> 
|> If the cpu does hang, and we try rebooting a few more times (via
|> command-command-~) we can successfully boot the machine. (after
|> a few tries)
|> 
|> When we bootup with the monitor screen up, we've had two things happen:
|> 
|> 1. the error message "unexpected kernal page fault failure" 
|> 2. panic: (Cpu 0) MMU invalid descriptor during table walk

I encountered a similar problem, but that one couldn't be the same since I figured out
when it happens:

I once installed a program someone posted the source of that switches between 
beeps every two seconds and rescans the directory every 100 seconds. It is called
'BeepMaster'. When you use that, it might happen (unpredictably) that your NeXT
panics with the invalid descriptor during table walk MMU error. That only happens
if you get a beep 'at the wrong microsecond'. There are times when you only have
to hit backspace in a shell window and let the autorepeat do the panic job, but
sometimes this doesn't work. Since I removed BeepMaster, everything is fine again.
Now the question: Why is BeepMaster doing that? It does the following:

- create another thread.

thread 1:
  - wait 2 seconds
  - unlink RandomBeep.snd in the appropriate directory
  - create a new hard link for RandomBeep.snd
  - loop to the wait

thread 2:
  - wait 100 seconds
  - scan directory for snd-files
  - build new table for thread 1
  - loop again

the two threads are semaphored for not disturbing each other, so: What's on?
I assume that happenes when thread 1 has deleted the RandomBeep and the beep
should come exactly then. But: If you delete a sound file that is the active
beep, guess what happenes: exactly - nothing!

What's up?

Uli