Waiting For Game Info Hangs

[Norfleet]

What's the story behind this new 2.11 feature that causes games to regularly hang on "waiting for game info"? According to the patch notes, something was apparently changed in the netcode in 2.11, and now it's much, much worse!

[Gandalf Parker]

Ive been waiting for someone to reply saying they have the same problem. So far it seems to be something that only affects certain copies.

[Norfleet]

It seems to afflict a number of Windoze servers, and ONLY Windoze servers. It also seems to only kick in after the game has been running for many hours, and doesn't occur all at once, but instead, gradually, as the "waiting for game info" period can become interminably long....before this phase starts taking longer than anyone reasonably has patience for, and kills the process to try reconnecting again. Sometimes, this banishes the problem for a few connection attempts, but ultimately, it exacerbates the problem for any future connect attempts, and the situation deterioriates: Players who are ALREADY connected to the server often notice nothing unusual, however.

Killing the process and restarting it always banishes the problem completely for at least a few hours.

It's definitely not related to "specific copies", because the kludgemush.com cluster had 3 different Windoze boxen that all suffered this problem: We moved all the games to a Linux box yesterday, and now the games are as stable as the ones on Mosehansen's, which is also a Linux box: No more WFGI hangs. Argitoth's server, a Windoze box, and Xavier's, another Windoze box, also seem to suffer from this on occasion, although it's hard to tell as I don't have too many games there and elsewhere. Their boxes also nagot constantly on turn generations, something that I am unable to reproduce. Not too many people can host 24h servers.

This strongly suggests that it's a Windoze-specific problem, as Windoze networking code and networking code for unices are completely different: Everything else would largely be the same, except for the network code, which I currently suspect is the culprit. It should also be noted that, according to the changelog, the networking code was altered in 2.11, and this problem did not occur in 2.08.

[Cainehill]

Originally posted by Gandalf Parker:
Ive been waiting for someone to reply saying they have the same problem. So far it seems to be something that only affects certain copies.

I thought it was too obvious this was a problem to warrant replying. If a game is hosted on a Windows box, it gets slower and slower and slower, until it seems like the game has completely hung.

Sometimes it takes 5 minutes. Sometimes it takes longer - I'll get up, go to the kitchen, make a snack, come back and it's ready. Sometimes it takes so much longer I kill the dom2 process and do something else entirely.

[Gandalf Parker]

Sounds like its worth testing then. Thanks for bringing it up.

[alexti]

Originally posted by Norfleet:
This strongly suggests that it's a Windoze-specific problem, as Windoze networking code and networking code for unices are completely different: Everything else would largely be the same, except for the network code, which I currently suspect is the culprit.

If you're talking about the applications network code, you can write pretty much the same code. Either OS supports berkeley sockets API, so unless Illwinter intentionally implements networking differently, it should be the same.

TCP transport implementation (that comes with OS) is a different thing though. But then it's not clear why only Dom2 suffers from the problems. Though Windows is not as good at networking, the degree of problems is unreasonable even for Windows. I'm puzzled about what it might be. Maybe some setsockopt option is different by default? SO_LINGER for example?