[CF-Devel] Forgotten child

Sat Oct 4 01:23:14 CDT 2003

  Few quick notes:

  One reason for locking CVS access (in selected areas) is simply the fact that 
there are different modules.  There are some number of cases where person was 
originally granted access in one area, and then unexpectedly moved to another 
area which wasn't initially envisioned.

  I care most about the server code simply because that in a sense is the most 
visible piece - if some code is checked in that breaks functionality, or just 
makes the server crash, that is much more a problem than most any other issues 
(image not quite right, etc).

  I'd also say that there are very few commits which didn't follow procedure.  I 
said in my previous message, 90% are perfectly fine in procedure and practive 
(number might be higher than that).  But one bad one can be quite a pain to fix 
down the road.

  And while the idea of 'having the person who introduced the bug fix it' is 
nice in theory, it doesn't really work for a few reasons:
1) If the bug is serious (frequent server crashes for example), waiting for that 
person to fix the bug may just not be a viable option.
2) In the process of investigating the bug and laying blame, often times by the 
time that is done, the fix is equally as obvious.

  Now for bugs which don't cause crashes/aren't serious, then having the 
original person who put the code in fix it is the ideal behaviour.  But it is 
sometimes hard to do that.

  Yanking the code isn't necessarily any better.  Point #2 still applies in that 
case (what code is responsible).  But there is also the potential issue of 
commits done after the offending code was put in.  There isn't handy way to pull 
out just the offending code and leave those later bits in.

  Plus, there are some number of longstanding bugs or RFE's that aren't related 
to new code, but would still be nice for those to get fixed.

  Re metalforge and clients:  Metalforge for a long time has been known as a 
'runs latest code' server.  In fact, I think the metaserver even says it is CVS 
code.

  Now in this particular case, metalforge probably got hit a bit worse in that 
it hadn't had an update for a while before I did the last one, so I think a 
bunch of things basically changed at once.  If it has been done more 
incrementally, probably not as much stuff would have been hit.

The reason metalforge runs CVS code is that's the only way to find bugs.  As 
mentioned elsewhere, the bug related to exp wrapping had been around for 
_months_ before metalforge was updated, and apparantly no one else ran into it.

_______________________________________________
crossfire-devel mailing list

     crossfire-devel at lists.real-time.com

     https://mailman.real-time.com/mailman/listinfo/crossfire-devel