[/eVo/Divinity]

BladeKiller, on 07 October 2012 - 07:27 PM, said:

phd, I connect using the external IP of my server not the internal one. The server sees my internal IP though. This started with a change to auth during the 4.2.002 dev cycle.

If your computer and a server are behind the same firewall and especially on the same subnet, this is expected behavior because the destination is the same as the source.

[LtStriker]

BladeKiller, on 07 October 2012 - 09:50 PM, said:

LtStriker, we haven't had any other reports of server crashes. Have you double checked the MD5 checksums for the 4.2.002 files on your server? Was there anything in the server's logs? What was happening when it crashed? Do you have extra files on the server other than just 4.2.002 like maps, mods, bots? The earlier crash was caused by one bad hit mesh so it only happened if the bad area of the hit mesh was hit.

Hi Bladekiller

We don't have anything extra. No mods or bots. We have a few third party maps.
I downloaded the file from your site myself. Lemmy ran the update which seemed to go fine. Didn't run any checksums.

If the crash was caused by the bad hit mesh, is it possible a remnant of 4.2.001 was left behind and can be causing the problem?

What should we be looking for?

We'll have a look into the linux logs and let you know if we find anything.

Thanks,

Striker

[zyzz]

LtStriker, on 08 October 2012 - 05:52 AM, said:

Hi Bladekiller

We don't have anything extra. No mods or bots. We have a few third party maps.
I downloaded the file from your site myself. Lemmy ran the update which seemed to go fine. Didn't run any checksums.

If the crash was caused by the bad hit mesh, is it possible a remnant of 4.2.001 was left behind and can be causing the problem?

What should we be looking for?

We'll have a look into the linux logs and let you know if we find anything.

Thanks,

Striker

My client wasn't updated correctly and I had some troubles. You could take a look at the pk3 files. *5.pk3 should be updated as well as all executables.

[X3igh7]

BladeKiller, on 06 October 2012 - 07:49 PM, said:

I have tested the hit bug reported in that video and it does not happen when the server and the client are matched. I don't know how they managed to do that video. We tested with male and female crouched and uncrouched, with and without kevlar and there were not unhits If you come across this again please get a server side demo and supply us with full details about the server and the clients so we can debug it. As it is we can not replicate this report.

Thanks for checking that out BK. I'm definitely looking forward to switching FTW over to 4.2 (and very excited for the AUTH system).

[FS Frankie V]

LtStriker, on 08 October 2012 - 05:52 AM, said:

Hi Bladekiller

We don't have anything extra. No mods or bots. We have a few third party maps.
I downloaded the file from your site myself. Lemmy ran the update which seemed to go fine. Didn't run any checksums.

If the crash was caused by the bad hit mesh, is it possible a remnant of 4.2.001 was left behind and can be causing the problem?

What should we be looking for?

We'll have a look into the linux logs and let you know if we find anything.

Thanks,

Striker

Do you recall what map the crash occurred? Is such a crash predictable rather than a random event? Can you supply a log prior to the crash?

The crash in 4.2.001 was caused by a malformed hit mesh that would collapse that only occurred on a few frames and at the moment of collision detection when a player was shot from behind so although it 's improbable it's still possible. We were able to discover this through a log posting of a sever operator that indicated an improper collision report followed up by visual inspection so you could try checking your logs to see if there is naything out of the nornmal occuring.

However.

A different technique for gluing the hitmesh to the primary and LOD models is being used in 4.2.002 so HM collapse is highly unlikely else the level of server crashes reported would still be above acceptable levels.

As for the “badhits” we have done some preliminary investigation and so far the indicators we have on hand, taken from such servers in real time, suggests an improperly installed server package where the animation.cfg file does not match the client.

The test for such a miss match is easy in that if you fire over the head of a crouched player and get torso hits then there is a miss match and you should report to the server operator that they need to reinstall the server client in a clean directory. This is possibly a side effect of installing 4.2 over a 4.1 build.

In the mean time demos are still good but the IP address of the effected server would be better.

[moonq3]

Nice update. We've been running it on CTF East for about a week now. No crashes and great hits. Well done FS.

-- MooN

[{BsTrD}TheRealFubar]

Good Job guys, now we just need some West Coast Servers!! Most of the servers on the list are over 100, which isn't bad, but.....

[LtStriker]

Frankie V, on 08 October 2012 - 06:24 PM, said:

Do you recall what map the crash occurred? Is such a crash predictable rather than a random event? Can you supply a log prior to the crash?

The crash in 4.2.001 was caused by a malformed hit mesh that would collapse that only occurred on a few frames and at the moment of collision detection when a player was shot from behind so although it 's improbable it's still possible. We were able to discover this through a log posting of a sever operator that indicated an improper collision report followed up by visual inspection so you could try checking your logs to see if there is naything out of the nornmal occuring.

However.

A different technique for gluing the hitmesh to the primary and LOD models is being used in 4.2.002 so HM collapse is highly unlikely else the level of server crashes reported would still be above acceptable levels.

As for the “badhits” we have done some preliminary investigation and so far the indicators we have on hand, taken from such servers in real time, suggests an improperly installed server package where the animation.cfg file does not match the client.

The test for such a miss match is easy in that if you fire over the head of a crouched player and get torso hits then there is a miss match and you should report to the server operator that they need to reinstall the server client in a clean directory. This is possibly a side effect of installing 4.2 over a 4.1 build.

In the mean time demos are still good but the IP address of the effected server would be better.

Hey

Crashed on Raiders. Hasn't crashed since. We played Raiders a few times since then.
Haven't had bad hits. Hits are good.

Exactly which log would you like?

Tried the mismatch test you mentioned and no problems there at all. All hits right on, crouching or not.

Lemmy is our Linux Server Admin, and will respond with what he did on the server.


On a different note, on the client side, we are getting "Server is for low pings only" on first join.

Thanks,

Striker

This post has been edited by LtStriker: 10 October 2012 - 11:48 PM

[FS CMM|v3nd3tta]

Got some Images of the "UnnamedPlayer" Bug :)



First I saw him at the end scoreboard at a lazy round...



Then he renamed to "a".



There he was...


asked him if he did it on purpose... but he didn't answer hat one.

The people already know that bug, so debug it ;)

ps: was at this server, 4.2.002.
http://www.urbanterr...14.26.75:27960/

[FS BladeKiller]

V3nd3tta, I notice that this server uses the B3 bot. We will have to try testing with the bot because this doesn't happen on our servers which do not use a server bot. 3rd party apps do add a lot of variables to the mix making it hard to track down some bugs.