[no subject]

[RMS at MIT-AI (Richard M. Stallman)]

Lots of lossage happens when PACKAGE gets set to a random object.

I just fixed the bug whereby the error handler would get an error.
But there are others.

First of all, if you are in a Lisp listener,
even though the error handler would set the package
to something winning and give you a chance to set
it permanently, you can't get into the error handler.
If the lisp listener is doing a read, then the errors
just print a message and invite you to rub out.
This is nice, except that there is no way to get
a chance to type commands at the error handler to fix anything.
I suppose that typing (1) might get into the error handler.

Warm booting tries to fix up the package, but it loses
because some background process usually manages to run
and halt the machine before LISP-REINITIALIZE fixes it.
It prints "the process such-and-such got an error"
and halts without printing anything more.
Shouldn't scheduling be off until LISP-REINITIALIZE
has got far enough to want it to be on again?
The problem happens with different background processes
on each occasion.  Supposedly the screen-manager, the kbd process,
and the chaos-background process have all been seen to do this.