Crossfire Mailing List Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Server crash in /city/anthony/prison



On Sun, Jun 18, 2000 at 09:04:32PM +0200, Jan Echternach wrote:
> On Sun, Jun 18, 2000 at 02:18:32AM +0200, Neil Muller wrote:
> > I see that currently there are 4 log-levels specified. Probably whats needed
> There are virtually no error messages when running the server, unless
> you hit a real problem.  I don't think different *_error variants are
> worth implementing.

There are two obvious advantages to implementing different error levels.
Firstly, the ability to turn errors off that aren't interesting while 
searching for a problem and the second is it would allow us to implement
some consitent prefixing of error messages. For instance, given the command
LOG(...,MAP_ERROR, "message") we can ensure its prefixed by
MAP_ERROR when dispalyed, which should make it easier to use commands like
grep, etc. on the log-file.

OK, not all errors fit into easily definable categories, but it easy enough
to create a general server class or something of that nature. 

> What could be useful, however, is separating errors with a (probably)
> safe workaround from those errors that are likely to crash the server
> or make it otherwise unusable, even if detected.  
Agreed. Also, since it should not be possible to disable displaying of these
types of errrors, I'll buy the argument that added some sort of category
definition may not be nessecary, although the automatic prefixing might still
be useful.

>   llevCritical
Never off, increments nroferrors, needs not be split into subsections
>   llevError
Switchable. (Incrementing nroferrors for these might be a
commandline option.) I still think subsectioning would be useful.
>   llevInfo
Info strikes me as additional information provided on the running of the
server, either statistics or information on how the code is running. 
Definately should not increment nroferrors. 

Question: I far as I can see, the server doesn't generate any statistics at
the moment. Would this be a useful feature for anybody? 

-- 
Neil Muller