Going through rim walls/rips --> Lag? Bug?

TiTnAsS · Post by **TiTnAsS** » Sun Feb 20, 2005 6:01 pm

It happens on both fresh and old walls...it happens in network games (only happens on the cvs server...)...and ask lucifer for the settings...

NOTE: Ser is also one of these ripping "experts"...he might be nice to help you...i'll also get focus ramen and you talking on aim when your both on.

Post by **Lucifer** » Sun Feb 20, 2005 6:22 pm

TiTnAsS wrote:It happens on both fresh and old walls...it happens in network games (only happens on the cvs server...)...and ask lucifer for the settings...

Attached.

Website · Post by **Z-Man** » Sun Feb 20, 2005 6:50 pm

Thanks, TiTnAsS and Lucifer! I could already reproduce it. Chronology of the screenshot: Don't know about the line that enters from the left, but from the lines at the bottom, I was first racing up the leftmost of them, then down the rightmost ( zigzaging a bit ) and then up the middle wall. The gap is real, I did not get killed .1 seconds later.

Post by **Lucifer** » Sun Feb 20, 2005 6:54 pm

Is that the gap TnA's complaining about? I consider that a feature and use it to build slingshots....

Post by **Lucifer** » Sun Feb 20, 2005 7:08 pm

z-man wrote: n54:
Are you using an older client ( not CVS )? It may in fact be that the new server side code will make ripping older clients easier. But i'm afraid I could only stop this if I disabled wall grinds ( you would need a min distance to walls to survive ).

So, a human solution is required then. How about we curb our tendency to freak out over rips? (I wonder how many people i"ve pissed off on my own server by telling people to rip the grid, heh) How about we think "Oh, that's probably only ripped for existing clients, but the next version won't be ripped" instead of "You bastard, you ripped the grid! The server's going to crash, on no!" Can we do that? And then, please, drive into the rip. I put Hell on cvs code precisely so the developers would have a Breakfast server out there for testing for the next release, so beat up on it! Break it, please! And then come here and talk about it, and I'll produce logs/etc as needed to find these bugs and squash them into oblivion (or at least into Norway).

Website · Post by **Z-Man** » Sun Feb 20, 2005 7:46 pm

Lucifer wrote:Is that the gap TnA's complaining about? I consider that a feature and use it to build slingshots....

I think it is. I tried to drive closer to my wall, and instead I got farther away and left a gap. You consider that a feature?

Oh dear. Another BUG_ variable I'll have to introduce.
BTW, it happens on local games, too.

About the ripping of older clients: I found a way toat least reduce the chances for that without changing the gameplay for new clients; I've just made the wall detection code for the network synchronization and rubber code consistent. It's good for the code quality, too

There'll be config variables to fine tune the ability to do tight grinds vs the possibility of ripping. And yes, we should not freak out just because some people have more fun with the bugs in AA than with the features.

ishAdmin · Post by **ishAdmin** » Sun Feb 20, 2005 7:56 pm

I don't mind the ripping. I mind only the crashing when ripping.

I thought that ripping was an easter egg you left on purpose actually. I was up late one night reading the docs, one too many hot chocolates in me, and well, I read stuff between the lines that you just didn't put there.

TiTnAsS · Post by **TiTnAsS** » Sun Feb 20, 2005 7:57 pm

It is the gap, but myn adjusted me the wrong way to...

Post by **Lucifer** » Sun Feb 20, 2005 8:10 pm

z-man wrote:
Lucifer wrote:Is that the gap TnA's complaining about? I consider that a feature and use it to build slingshots....
I think it is. I tried to drive closer to my wall, and instead I got farther away and left a gap. You consider that a feature? Oh dear. Another BUG_ variable I'll have to introduce.
BTW, it happens on local games, too.

Bouncing at the wrong time and at seemingly inconsistent times is annoying, but the gap that's left by the bounce can be used. So, I'd like to see bouncing happen more consistently and predictably to reduce the number of bounces at the wrong time, but not taken out completely.

I guess it's technically a bug because of the break in the wall, and I can live without it, so you don't have to make another BUG_ variable to accomodate me.

Website · Post by **Z-Man** » Sun Feb 20, 2005 10:38 pm

First, I'll have to find out what is causing this bouncing

What disturbs me is that I've only seen it on the windows client on my normal PC and not the linux client on my laptop. However, the gap is with absolute certainty a server bug. I've added a bug entry so it's not forgotten about.

Website · Post by **Z-Man** » Mon Feb 21, 2005 12:03 pm

The reason for the windows/linux difference apparently was that I ran some newer code on linux where the bouncing bug seems to be fixed ( I disabled one instance of possible backward simulation ). I'm not sure about it yet, however. The new code is in gCycle.cpp version 1.11 if you want to test it.

Bouncing seems to be caused by an odd combination of the code handling cycle_delay, cycle_rubber and low framerates. I'm afraid it could not be made reliable as it is now, it would have to be added as a real feature. And a complicated one since it would need full networking support ( when it happens as a bug, you often lag-slide from the correct position to the bounced position ). If your heart hangs on it, submit an "Anti-Bug-Defamation-League" feature request, and we'll see what can be done.

Website · Post by **Z-Man** » Mon Feb 21, 2005 2:00 pm

About the wall bending from the last demo video:
Nice demonstration! The wall seems to be clearly bent the same way on the server and the client. I hope wall bending is fixed in CVS ( eGrid.cpp version 1.16 ), but as with the rips, clients will still be affected even if the server is fixed.

Post by **dlh** » Mon Feb 21, 2005 8:28 pm

z-man wrote:About the wall bending from the last demo video:
Nice demonstration! The wall seems to be clearly bent the same way on the server and the client. I hope wall bending is fixed in CVS ( eGrid.cpp version 1.16 ), but as with the rips, clients will still be affected even if the server is fixed.

Just got a video clip of a wall bend on the cvs server. I can't drive around in this one, so its probably not bent on the server.

Post by **Lucifer** » Tue Feb 22, 2005 2:27 am

I'll try to do another checkout and build in the near future. I just put a rack'n'pinion in my car and boy am I tired.

z-man: I can give you shell access and you could build it if you wanted. I realize if you did all that you could just as easily run your own test server...

/dev/null · Post by **/dev/null** » Tue Feb 22, 2005 3:47 am

The linux 2.6 client bounces at high framerates also, its just much harder to do. If you do two fast turns you will die, you have to slow the time between the turns a little.