What do you want to see in Armagetron soon? Any new feature ideas? Let's ponder these ground breaking ideas...
12 posts • Page 1 of 1
So I know it's been mentioned a few times for CTF servers that people using the 4.0 client screen's would go black and they would have to reconnect. For a while I think a lot of people thought it had to do with the amount of respawns in a round, I think I've figured out what triggers it though. Once or twice I've made a macro to mess around and I'll end up disconnecting from the server, so I'm thinking that too many turns in one round will crash your connection with the server. Which would explain ctf with all the long rounds, respawns and multibinding. Occasionally people disconnect in fort too, and I'm guessing it happens after really long rounds, I've never actually paid attention enough to notice it but it's my best guess for why a bunch of us get disconnected. Not sure if this helps any of the developers figure out why it's happening or how to fix it, just something I've noticed.
It only happens to v0.4 clients as far as I can tell. For instance I'm using v0.4 on Windows.me from an earlier thread (ctf-specific) wrote:At the start of the round, the camera gets placed into a rather small square arena, no cycles or zones. Chat and console output is visible, the scoreboard does not get updated when players join though (not sure if scores are, but entering players are definitely not updated, leaving players are though).
This happens as player and as spectator.
The log output we get on the server (at least on mine; quite recent sty+ct) is:
I have noticed it happens on servers that do a lot of respawns (CTF, as well as deathmatch servers with instant respawns). I've had a quick look at sty+ct code, and the respawn code does include a "new gCycle()" which does likely mean new netobjects. Maybe 0.4 can't handle many new ones?
Code: Select all
 User 2 does not know about netobject 14289 ( gCycle nr. 14289 owned by ePla$  User 2 does not know about netobject 14288 ( gCycle nr. 14288 owned by ePla$  User 2 does not know about netobject 14287 ( gCycle nr. 14287 owned by ePla$  User 1 does not know about netobject 14289 ( gCycle nr. 14289 owned by ePla$  User 1 does not know about netobject 14288 ( gCycle nr. 14288 owned by ePla$  User 1 does not know about netobject 14287 ( gCycle nr. 14287 owned by ePla$  User 1 does not know about netobject 14286 ( Nameless NetObject nr. 14286 ).  User 1 does not know about netobject 14285 ( Nameless NetObject nr. 14285 ).  User 1 does not know about netobject 14284 ( Nameless NetObject nr. 14284 ).
Don't know if there is an exact way to reproduce it; but the deathmatch-instant-respawn servers do this almost every round. (Durf's "Hunt the Wabbit", my "Bounty Hunt").
Maybe it's something with those respawned cycles dying / being created at the end of rounds? (speculation, not yet tested)
Shall I try building my server in debug mode or something? I would need a quick tutorial on that then. It runs on debian and if command line (ssh) access is enough for it then that would be preferred.
Last edited by vov on Sat Apr 11, 2015 12:56 pm, edited 1 time in total.
I've seen it happen to people in 2.8 as well vov, it's not as common but it does happen every now and then, same bug happens to me in fort sometimes, so that's why I was saying I'm not sure if it actually is respawns or not, since it happens in HR which doesn't have respawns either, same black screen.
All that speculation has its use, but what we would really need is a recording from a 0.4 client where this happens. I'm trying to produce one, but so far, no luck.
How did you make the recording? If via the Windows start menu entry, it should be on your Desktop.Magi wrote:I tried making a recording with .4 and it successfully crashed but I can't find where the file went o.o
Ah, that might do it. I don't get a completely black screen, but all the cycles are missing. The client debug output says they are never created. And the server log from earlier said the clients did not know about the cycles. What I need to find out now from the recording is whether the cycle creation messages are sent by the server or not. If they are sent, I should be able to find out why they are ignored.