On 7/15/06, Stéphane Magnenat <[EMAIL PROTECTED]> wrote:
Hi,
Thanks for the great work Bradley ! :-)
For the checksum bug, are we sure it is a problem related with order
reception ? If so, we can simulate packet drop (by not sending some packets)
on two glob2 running on the same machine to see if it triggers the bug.
Well there are a few ways it could have happenned. The first one is if
a NullOrder or some other type of order wasn't recieved, in which case
the default checksum value is 0, which would cause the crash, and it
would also explain why this bug only crashes one machine, not all of
the machines (as normal desyncs would).
Another possibility is that the computer sent a checksum of 0,
however, this doesn't explain why only one computer crashes, and not
all of them. I can't think of another possibility, although I don't
know the network code as well as you do.
For solving it, I'll commit some code that prints everytime a new
checksum is recieved (some 10000 times in an average length game),
along with the uStep is was recieved in.
Next, people should play lots of netgames untill they get crashes like
the one in the other email.
Thirdly, when they do crash, they'll mail us a reply on this email
with the console output on each machine untill the point of crashing.
This will either validate or invalidate my theorie, hopefully allowing
me to fix the bug.
For the alpha20, I fully agree. I still think we need to do a 1.0 soon, so
perhaps this alpha20 should be beta1 ? Do you feel the code is good enough
for this ?
The code is stable enough for a beta 1. Single player games very
rarely crash, its only multiplayer games thats a cause for
instability.
-Bradley Arsenault
_______________________________________________
glob2-devel mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/glob2-devel
