Linux Bug in CVS source

[minorgod]

Ever since the bug fix for the chat changes in cs1.6, I have a major problem on my Linux server. I'm getting MAJOR lag. It only becomes noticable once there are 4-5 players (bots) on the server. I connect and my ping is like 450+ over my LAN. The more bots that join, the higher the ping gets. Anyway, the game is unplayable with the latest source. It was working great before the update, so I'm pretty sure it's not a problem with my machine or network. I'm running Mandrake Linux on a 1Ghz P3. Has anyone else seen this problem on Linux?

Originally Posted by minorgod

Has anyone else seen this problem on Linux?

Humm it may well be so...???

I'm running the last version (Friday 18th if not mistaken), and there is some more lag. I was assuming it was a plain network lag and not the server lagging itself. It is about 100-200 compared to the usual 50-60... Strange but it could be the network.

There are a lot of tracing/debugging message specially when running on a map without nodes with the current CVS version. This may well be the explanation as well.

-eric

[stefanhendriks]

could be, i don't know about that. As i said tons of times the CVS version is not for real usage , so its hard for me to take 'bugs' like this seriously. It is worth a look though because if its REALLY a problem.... well we got a problem then.

[stefanhendriks]

Okay; what i have done:

- deleted any rbn/rbx file of cs_estate
- rebuild the entire source (deleted dll, etc)
- i ran the map
- spawned 9 bots (10 players in total)
- i ran around, the more i ran around the first time, the more laggy it became (not so unlogical, because the bots can now calculate paths... out of lots of nodes, the more i move, the more connections get added, etc)
- i die at T spawn
- some t's there move the path back to the ct spawn where i began, + CT's there move to the T's (not all CT's).
- 2 t's and 3 ct's remain
- i ended round
- next round, there was NO lag at all (no fps drop at me, steady at 60 fps (which is my max btw))

i don't know what the real problem is, i can think of log() commands being given to frequently somewhere?

[minorgod]

I'll try running the server with no bots to see if it seems more laggy than it should. It may be impossible to get enough real users for a real test though. I have recompiled with debugging enabled, but where does the log file get written? The map I'm testing is de_torn. I did try deleting all my .rbn and .rbx files, but didn't notice a difference. I'll try testing on some more maps and see if I can find any pattern. Let me know if there's any info that would help you help me.

[stefanhendriks]

the log file is written in your hl ded server directory (so main dir of HL).

I have updated the source, it has less comments and logging, but i can't asure you it fixed your problem though.

[minorgod]

The update introduced a compilation bug on Linux...compiler said:


NodeMachine.cpp: In member function `void cNodeMachine::path_think(cBot*, float)':
NodeMachine.cpp:3257: error: name lookup of `iGn' changed for new ISO `for' scoping
NodeMachine.cpp:3018: error: using obsolete binding at `iGn'
make: *** [NodeMachine.o] Error 1

So I went to line 3018 and changed this:

for (int iGn=0; iGn < MAX_GOALS; iGn++)

to this:

int iGn=0;
for (iGn=0; iGn < MAX_GOALS; iGn++)

And now it compiles. I'm now going to try to track down the lag again.

[minorgod]

Just tried to post my realbot log file, but it's too big and caused a fatal "out of memory" error on this forum. Should I e-mail it to you or something? Also, did the metamod code change after this latest steam update? I never got any new metamod source when I updated from CVS, so maybe the problem is there?

[stefanhendriks]

hmm, be sure you have the latest metamod source, the Getstarted.zip file contains a metamod source, which i believe is the latest. But to be sure you can download it from the metamod site. Anyway, i use the exact same source and it runs fine.

I applied the linux compile fix. Thx.

[minorgod]

I think the linux lag issue I was experiencing several weeks ago was either fixed by a CVS update or by a Steam update. Either way I now only have intermittent lag spikes over 200 with 14-16 bots and 2 humans playing. Installed mods are Metamod, Realbot WIP, AMX, WC3FT, AdminMod and StatsMe running on a 1Ghz P3. Also running MySQL for saving AMX/WC3 info. Lagspikes will happen every couple of minutes, but only last a few seconds (5-10). I thought it might only happen when a new map starts and all the bots are alive and moving around at the same time, but the number of live bots doesn't seem to have any effect. It will sometimes lagspike when there's just a couple of players left alive. Still, it's very playable now and the lag is usually so short it goes unnoticed. Pings rarely jump over 250 now when they do spike and I'm pretty convinced now that it's just the slow processor maxing out. Something is definitely taking more CPU cycles than it used to though.

In all, I'm very happy to finally have these various mods working this well together on a low-end P3. RealBot WIP is really starting to kick some booty! My best friend and I have been playing against the various realbot versions for over a year and it seems like the WIP is almost as good as the old non-WIP version ever was. Bots are showing more interesting behaviors and working together better. I'm still seeing some problems with bots taking the same path every time and some getting stuck and blocking each other in hallways, etc. From the WIP source code, it looks like there's some work going on to allow automatic generation of .rbn files from .bsp files. That would be really cool if there was a way to render some higher quality nodes from the bsps. Maybe a program that would run overnight and do things that would be too cpu intensive to do during the game, such as shoot tracelines in all directions from every location in the game and use the data to fix existing node files. Just some random speculation of course.

Now if only I could figure out which mod has control of my mapcycle!