CVS test server
CVS test server
I started one server with bad internet connection running current CVS code. It is on armagetron.kicks-ass.net:4534 and if it appears in the server browser, its name is "CVS Test Server ( aka Crash Me! )". It is running in debug mode in a debugger on a weak machine, so it is not much fun playing there. Jonathan and others already triggered one type of server crash on it
I'll start a twin server on my machine at work with better performance.
Don't be surprised if the bots talk to you. They do so only in debug mode when they think they have trapped you.
I'll start a twin server on my machine at work with better performance.
Don't be surprised if the bots talk to you. They do so only in debug mode when they think they have trapped you.
- Tank Program
- Forum & Project Admin, PhD
- Posts: 6711
- Joined: Thu Dec 18, 2003 7:03 pm
Re: CVS test server
If you look in the devel only section way, way back, I posted a screen shot of this cause it freaked me out .z-man wrote:Don't be surprised if the bots talk to you. They do so only in debug mode when they think they have trapped you.
Yes, but when you tied that on the CVS servers, the debugger hit a breakpoint I removed that breakpoint now, so you can try again and see how you get kicked.Jonathan wrote:I can make bots talk on old servers.
I hope I have fixed the following problems ( on CVS server 2, I can't update server 1 from here ):
- A correcting function of the grid structure had a bug that caused the server to freeze ( hit a breakpoint, it would not have been a crash ) once during testing.
- The server cycle simulation had a bad interference of rubber code and client command interpretion code, causing you to get stuck sometimes kissing a wall until your rubber ran out. If you stayed calm, you survived, but most of the time death was unavoidable.
I added the option to send clients updates of their own cycles' data more frequently and activated it. Updates should now come 10 times per second as opposed to once per second.
I'm afraid I can't fix the following client problems on the server side:
- Rips & "left grid!" crashes ( some settings can make them less frequent, though )
- Position desyncs when you use the rubber feature
- Your cycle temporarily tunneling through all kinds of walls
The more frequent updates should make the last two problems more bearable, however.
- Jonathan
- A Brave Victim
- Posts: 3391
- Joined: Thu Feb 03, 2005 12:50 am
- Location: Not really lurking anymore
Throws me out.z-man wrote:Yes, but when you tied that on the CVS servers, the debugger hit a breakpoint I removed that breakpoint now, so you can try again and see how you get kicked.Jonathan wrote:I can make bots talk on old servers.
Much better, especially when warping.I added the option to send clients updates of their own cycles' data more frequently and activated it. Updates should now come 10 times per second as opposed to once per second.
Buzzsaw effect
i couldn't crash the server but i found an interesting effect (using 2.7)
if one tries to grind too closely the bike (any bike: it happened to bots i trapped as well) get's stuck like a saw in wood, driving on the spot until it explodes. imagine the picture below as being a live capture with only some vibration of the bike (the sparks pattern is also fairly unique).
since i use 2.7 i can't say it's a bug but maybe something to test further?
if one tries to grind too closely the bike (any bike: it happened to bots i trapped as well) get's stuck like a saw in wood, driving on the spot until it explodes. imagine the picture below as being a live capture with only some vibration of the bike (the sparks pattern is also fairly unique).
since i use 2.7 i can't say it's a bug but maybe something to test further?
n54: I though I had fixed one of those bugs, but lately saw indications in the log that it is not completely fixed ( lots of debug message spamming, saying essentially that you're stuck ). I updated the second server to the fixed code at about 9 p.m. GMT, I think your visit was after that. So yes, it is something to investigate, no matter what client version.
the rubber was at 100 when i was there, that'll cause the above fluke ^ and tons of others..
if you wanna test ripping, mebbe like rubber 10 & cycledelay 0 or something..
not that i'm telling the master how to do things, but rubber100 seemed silly
iceman has experimented with LOTS of crazy rubber & other settings
i'd be interested what he has to say about the effect shown in the screenshot.
on a side note, when i saw rubber at 100, i tried to go thru my own line, unsuccesfully
i can do it locally on 2.6 with rubber above 10 & cycledelay at 0 or very low (by pressing left & right at same time 4 times fast, the second you run into the line) probably its more of a bouncing effect, but, it happens in an instant & looks like you go straight thru the line at a right angle like +
much easier to do locally than on an internet game, but, still possible
if you wanna test ripping, mebbe like rubber 10 & cycledelay 0 or something..
not that i'm telling the master how to do things, but rubber100 seemed silly
iceman has experimented with LOTS of crazy rubber & other settings
i'd be interested what he has to say about the effect shown in the screenshot.
on a side note, when i saw rubber at 100, i tried to go thru my own line, unsuccesfully
i can do it locally on 2.6 with rubber above 10 & cycledelay at 0 or very low (by pressing left & right at same time 4 times fast, the second you run into the line) probably its more of a bouncing effect, but, it happens in an instant & looks like you go straight thru the line at a right angle like +
much easier to do locally than on an internet game, but, still possible
The rubber is at 100 precisely to uncover weird effects and strain the data structures to their max. With sensible rubber settings, n54 would not have been able to tell the difference between getting stuck and dieing right away. Of course, for real gameplay, it is silly. And for fun server settings, I'm no way the master.
Jonathan is, by the way, really productive when it comes to finding bugs. He triggered another one causing the server to hit a breakpoint, there must have been a massive wall bend going on ( must... work... harder... to... eliminate... those... ).
And yeah, tunneling through walls has been made almost impossible. You may have a chance on fresh walls. I'll add a BUG_ variable to enable tunneling again.
Jonathan is, by the way, really productive when it comes to finding bugs. He triggered another one causing the server to hit a breakpoint, there must have been a massive wall bend going on ( must... work... harder... to... eliminate... those... ).
And yeah, tunneling through walls has been made almost impossible. You may have a chance on fresh walls. I'll add a BUG_ variable to enable tunneling again.
I was curious about the difference between wall bending and wall ripping. They would seem to occur from the same actions (tight grind stuff). How does a wall bend manifest? Anyway, just curious.
I found for my own ripping experiments that I really needed some decent brake. I'm using an older slower computer, and just can't do it that well without braking some. Now this brings me to another possible bug that I haven't heard discussed for a long time. -> Andre was able to hit the wall and turn in a way that would stick him into 'slow mode'. He would not accelerate when grinding up against a trail. Being stuck at the lowest speed possible, without acceleration, he could uturn against the outer wall repeatedly until he would rip it. I achieved the 'slow mode' only once. I watched Andre do it repeatedly. Has anyone else experienced this 'slow mode'?
I found for my own ripping experiments that I really needed some decent brake. I'm using an older slower computer, and just can't do it that well without braking some. Now this brings me to another possible bug that I haven't heard discussed for a long time. -> Andre was able to hit the wall and turn in a way that would stick him into 'slow mode'. He would not accelerate when grinding up against a trail. Being stuck at the lowest speed possible, without acceleration, he could uturn against the outer wall repeatedly until he would rip it. I achieved the 'slow mode' only once. I watched Andre do it repeatedly. Has anyone else experienced this 'slow mode'?
That has happened to me a few times on your old server, ish. To start accelerating again, you just pressed v. I've seen this happen on Swampy's once or twice, so the bug was still in 2.7. I've yet to see it happen in a cvs server, but there are very few of them.ishAdmin wrote:I was curious about the difference between wall bending and wall ripping. They would seem to occur from the same actions (tight grind stuff). How does a wall bend manifest? Anyway, just curious.
I found for my own ripping experiments that I really needed some decent brake. I'm using an older slower computer, and just can't do it that well without braking some. Now this brings me to another possible bug that I haven't heard discussed for a long time. -> Andre was able to hit the wall and turn in a way that would stick him into 'slow mode'. He would not accelerate when grinding up against a trail. Being stuck at the lowest speed possible, without acceleration, he could uturn against the outer wall repeatedly until he would rip it. I achieved the 'slow mode' only once. I watched Andre do it repeatedly. Has anyone else experienced this 'slow mode'?