[no subject]

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

Whenever I spend a few hours editing on a Lisp machine, I run into a
few mysterious errors.  I don't define any new commands and I haven't
been hacking ZWEI, so I'm sure this must happen to everyone.  I don't
usually remember enough about what the ZWEI data structure is supposed
to look like to investigate, so I just ^Z them.  I developed this
habit for two reasons:

1) It usually works
2) I usually don't know enough to figure out anything else
3) The error usually is not related to the particular command
I had just typed.  Usually it is due to something else that went wrong
earlier.  Most of the errors are in redisplay.
4) The errors are often nonreproduceable (if it is reproduceable, I
have something to investigate).

Nevertheless, when you have a program with nonreproduceable bugs that
strike fairly often, you have to do your best to look for them.  It's
no use saying "I won't do anything until I get a report I like".  This
attitude just leads to bugs not getting fixed.  If the only way ZWEI
can be made reliable (and, in my opinion, it is not) is to spend a lot
of time looking over the code and to put in debugging checks, then do
it!

I sometimes complain to people when they leave out information from a
bug report if they have reproduceable problems, but that is because I
know the person can easily determine the information and send a better
report.  When the problem is not reproduceable it is purposeless to
complain.  You have the information you have.  The question is, do
you try find the bug or you decide to leave it in?

I also sometimes complain to people who fail to dump unreproduceable
crashes.  Sadly, there is no way to dump a crashed ZWEI, unless I feel
like clobbering one of the bands.  I wouldn't normally expect that
there would be a band that I could obviously clobber.  If there were a
crash analysis program that would check over all the ZWEI data
structures, you could ask people to run that.  Until then, you cannot
expect anything but sketchy bug reports and you ought to do your best
with them since you will never get anything better to work with.