Re: [Win95]

From: Jeremy Elson (jelson@circlemud.org)
Date: 02/08/97


> The reason the logs are not sent out to the "syslog" file is simple.
> CircleMud writes to stderr.  The autorun script for UNIX redirects
> stderr to the syslog file.  Simple solution is to do the same for
> Win95 (if it's possible?).  As for syslog.CRASH...

Yes, Circle just writes to stderr and under Win95's DOS box there's no way
to redirect stderr.  pl12 will write logs to a file instead of stderr
(although, the default "file" will be stderr anyway, for compatibility),
so that windows users can see their logs.

> Can we all agree that "syslog.CRASH" is the worst-named file of
> 1996?  And at this rate, maybe it'll repeat it's victory in 1997.
> The syslog.CRASH file does not tell you what happened that made
> the mud crash.  It's the last 50 (more or less) lines of the syslog
> file.  This is made by using the UNIX 'tail' utility.  Since the
> mud can't really log problems in the code that make it crash, the
> syslog.CRASH file will only show errors which Circle recognized.
> Or in other words, probably world file problems.  When the mud is
> actually running, there is very few logs sent in comparison to when
> it's just starting up, and there's really nothing to error check.
> So syslog.CRASH for all practical purposes is worthless in tracking
> down crashes.  Perhaps renaming to syslog.REBOOT in the stock
> distribution would be better.  Or completely eliminating it since,
> IMO, it's worthless and misleading.

Well, let me explain why it's there before you call it worthless.  I agree
that for *you*, specifically, it is worthless -- but then, you probably
understand how the code works better than 95% of the other people who use
it, so syslog.CRASH may not be worthless for everyone.

First, I have to explain that the one thing that annoys me most about
CircleMUD's existance is all the mail I get from clueless MUD-imp
wannabees.  In other words, people who know *nothing* about UNIX, C,
programming in general, or even how to use "tar", but want to run a MUD
anyway because they're pissed at the imp of their old MUD or just think it
would be a cool power trip to have their own MUD.  Clueless MUD imps
monopolize huge amounts of the time I devote to responding to MUD email.
(To be civil, I won't name names, but I could put together a *highly*
amusing collection of email I've gotten over the past few years ;-)).

A lot of Circle's recent new features (well, I'm using "recent" rather
liberally here since there hasn't been a new patchlevel in a few months) 
are there solely to reduce the amount of email I get from clueless MUD imp
wannabees.  For example, the cool "configure" script was originally added
because I was replying to half-a-dozen letters a week from people who
didn't know how to configure the Makefile for their system.  They would
either ignore the instructions in the README, or they'd read the
instructions but when they got to the part that said "If you're using
Solaris..." they wouldn't know the name of their operating system so
they'd just skip it.  "configure" was originally added to set up Circle
automatically for people who didn't know how.

Anyway, the point I'm rambling to is that the "autorun" script can lead to
very confusing behavior for people who don't understand how CircleMUD
works.  This was an accident and I didn't realize what the confusing
behavior was until I finally was able to find the common thread tying
together a bunch of email I got from some very confused imps over the
course of several months.

Consider the case of a newbie UNIX user who is following the instructions
in the README file to get Circle up for the first time.  Now, imagine
there is some error that prevents Circle from booting -- for example, a
Linux user trying to run the MUD without TCP/IP compiled into the kernel,
or someone trying to start another copy of Circle on a machine where
Circle is already running, etc.  They type 'autorun', just like the
instructions say.  The MUD tries to come up and then dies 1 second later. 
The autorun script then immediately moves the syslog to the 'logs'
directory and goes to sleep for 60 seconds.  Then it starts all over
again... the MUD tries to come up, dies, and the logs are removed.  The
confused user is left waiting for his mud to come up, and it never does. 
It appears that autorun is working, but nothing is coming up in the syslog
(it only exists for a second before the script gets rid of it!) and the
MUD never lets him connect.

Sorry, I didn't mean for this mail to be so long.. the point is just that
although *you*, an expert user, know how to use "tail" and know that
autorun stashes logs in the 'log' directory, a newbie user does not know
these things.  For them, seeing a file appear called "syslog.CRASH" that
says "bind: port already in use", or something can be a big help.

And, the best test of success - I've stopped getting confused mail from
people who can't figure out if autorun is working or not. ;-)

-Jeremy



This archive was generated by hypermail 2b30 : 12/18/00 PST