latest status report: I think it's fixed now. I finally tracked down the obscure allocation-related problem: it was double-freeing something, but it was an STL element, so it was impossible to debug (valgrind couldn't even find it) until I found the GLIBCPP_FORCE_NEW environment variable.

I'm running the server under valgrind throughout the rest of the night and tomorrow morning. Around the afternoon I'll make sure it's clean and restart it normally. At that point, I hope to declare the server stable, but I'm still open to the possibility that it is not.

There's a minor client-side bug in the network stuff, so we'll release a new client this week most likely. Bugfixes only.

Are we going to post each time? Let me save time:

it's up

it's down

it's up

it's down

it's up

it's down

Man, it CRASHED!

Hey man, this is why we're all here, and this is what makes it fun! I love this stuff. Good luck O intrepid devs! Vanquish the bugs and charge on to glory! One day you guys are gonna have a round of drinks on me.

Ewwwwwww!
I hope you shower before we start drinking off of you. ;)

Oooooh, the GLIBCPP_FORCE_NEW! I see! :P
Anyway.. good luck with the server, mate.

If it's any help, the small crash (the one AFTER the "cap crash" was preceeded by the bots going at fantastic speeds. I'm assuming that it was me lagging but the bots were fluidly moving...not jumpy. Hope this helps.

Okay. I officially declare the server stable (knock on wood).

OW! That was my head a1k0n! And, you're supposed to use knuckles, NOT a 2X4!

No, no, he was using your head to knock on the 2x4.

good use for a clue-by-four nonetheless

From Vendetta News "...The issues had nothing to do with any specific user activity (flag caps, mines, or anything else)...."

You know full well that "I" figured it all out and you really don't want to give credit to anything except your precious valgrind! I see how this test works now...

:-) Let's find the next bug!

*ahem* it's _almost_ stable anyway. Took a few days. That particular bug is corrected, I think. (It was a slightly overzealous assert() for a condition that turns out to be valid).

I have never understood why developers use assert() when they really don't want things to crash ;) Isn't it better to just test the condition and append a warning to an error log?

because Igrok, asserts are meant for conditions that should NEVER occur. For example, you COULD do a assert(0 != 0). NOW... as far as we know, 0 should ALWAYS equal 0. :) IF NOT... that means something seriously wrong has occured. It's a good bug tracking mechanism.

If I assert on something that isn't supposed to ever happen, then I can take a look at the resulting core dump and figure out how and why it did happen. If I print a warning, I just get thoroughly confused (and then inevitably, since something that isn't supposed to happen happened, the program quickly goes down hill and crashes in a vastly confusing way.) Better to crash on a known condition than crashing as a result of long periods of execution in a dubious state.

Bells and Whistles Server status:


The server is: DOWN.

EDIT: 2 seconds later it is up. Lol.

Second edit: I checked, it was down, everyone else was kicked off.