BUG cries on Bugfarm Fortress and Sumo

For things that have to do with those crazy test servers... and yeah. By request of z-man, and, of course, you gotta obey...

Moderator: Z-Man

Post Reply
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Code: Select all

[3] ~pAul~: ?
[3] ~pAul~: wow
[4] TBL: BUG PAUL EXPLODED FROM NOTHING
You and paul drove head on towards each other, passed each other and paul turned towards you, against your wall. The clients, especially paul's, may have mispredicted that.

Code: Select all

[0] °°rain core dumped maras for 2 points.
[11] °°rain: no hole?
[11] °°rain: BUG: z-man
That was an interesting one :) Turns out there was a huge messup when converting between different coordinate systems while handling walls. Not only was the code wrong, it was far more complicated than it should have been. I wonder why it didn't get uncovered earlier. Some of the "I crashed right in the middle of nowhere" bug cries may have been the result of this, too; it would have produced wrong time distance readings for wall collision points.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Code: Select all

[10] °°rain: omg
[10] °°rain: I passed a wall
The server says it was a clientside effect. It cot a sync of yours from the other side of a wall and lag slided you into place.

Code: Select all

[5] LøÐ|AÐ¥-Lµ¢îfer: BUG on 0.3 dedicated server isn accel if u are betweenn ur
wall and enemy's wal
There is a setting that allows that, yes, but it's not active by default. BUG cries are not the best way to report these issues, by the way. SF bug entries or posts here on the forum are. BUG cries are for immediate bugs on the server you're currently playing on.

For those coming in late, there's a bigger post on the last page you may have missed :)
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Code: Select all

[2] ~*SP*~ ØMÊgÀ: ZMAN BUG Y IS THIS GUY HERE
Referring to Keenan, I presume. He is forced to bear this name so you always recognize him and kick him yourself when he misbehaves. He's on static IP, so the autobanning system should work nicely.

Code: Select all

[10] dookatee |DpG|: BUG: Legit's deaths and kills were all suicides
(From Legit's chat, I assume this means that the system rated them as suicides, but shouldn't have done so) No, only the last couple of times. There were quite a few real kills accounted for.

Code: Select all

[0] !LOL! Legit core dumped SL(Camouflage) for 2 points.
[0] Removing future walls of the cylce that just got killed mercilessly.
[0] !LOL! Legit core dumped ¡LØL!·Durkå for 2 points.
[2] !LOL! Legit: lol
[9] ¡LØL!·Durkå: how
[9] ¡LØL!·Durkå: BUG: I split and was on the other side of a player on opposite team
Legit's cycle was simulated with a time delay because of his higher ping; I assume he passed through Cam's wall (which is legal if his clock was behind Cam's) and directly got you afterwards. If he was stopped by cam's wall on your client, that's a clientside prediction error I just fixed.

Code: Select all

[10] dookatee |DpG|: BUG: wall fade lag, i did drive past the wall after it faded and then diee
I need clientside recordings for this type of bug.

Code: Select all

[11] dookatee |DpG|: BUG: I stopped movin
For this one, too. I already have one I didn't have time to look at yet, so don't rush.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Turns out the previously announced wrong tail end extrapolation code was right after all, I reverted the change.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Mkzelda: sorry, your recordings don't play back far enough :( I tried to get your exact versions (now listed in the recording, haha), but apparently, the lazy version updating hits and your actual code was some versions behind that. Gotta fix that for the future.

Other than that, could you enable VSYNC when you record? My machine isn't fast enough to play them back in real time at 300 FPS...

Vanhayes: The lags I saw were all of the type that was fixed, recognizable by the fact that the cycles make two turns at once, but it's probable that I did not spot all. Playback always has that extra stuttering.
Infamous
Average Program
Posts: 96
Joined: Sun Jul 16, 2006 2:27 pm

Post by Infamous »

z-man wrote:
[1] SL(Infamous): BUG
(On Sumo, fighting the AI, without further explanation)
ahaha if you had scrolled down you would see the bug, not worth it tho :twisted:
meriton
Round Winner
Posts: 256
Joined: Sun Nov 20, 2005 3:33 am

Freeze

Post by meriton »

Shouldn't armagetron say the connection was lost when no server updates have arrived for 30 seconds? I just experienced it merrily extrapolating during that time (after that, I got tired of waiting and cut the "connection" to the server). Unfortunately, I wasn't recording. This happend a couple of minutes ago.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

The timeout mechanism only looks for missing ACK packets, and they only get sent if the client sends something. Which doesn't happen unless you say something or turn your bike. Yeah, perhaps an addition would be in order.
meriton
Round Winner
Posts: 256
Joined: Sun Nov 20, 2005 3:33 am

Post by meriton »

I turned several times during that period.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Hmm, strange things can happen sometimes. On one of the rare occasions I wasn't recording, I got something on my client that looked like a timeout, nothing moved, my chats got swallowed. But when checking the server logs, the chats came through to the server. So either the connection was lost only in one direction, or there was a client bug that threw away the server messages.
User avatar
Rain
Round Winner
Posts: 300
Joined: Sat Apr 15, 2006 2:59 pm
Location: a random empty server playing with bots

Post by Rain »

Walls crossed in Bugfarm fortress.

Happened 5 times in the same week when I was playing.
Attachments
Walls crossed
Walls crossed
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

Rain: any hint of when this screenshot is taken? Did you cry BUG then? It may be a clientside effect, like a wrongly set cycle distance variable (I fixed one or two of those), so, you guessed it, a clientside recording would help.

Code: Select all

[2] Zop: BUG. undwanted restarts when someone joins
Whoops. Fixed. Umm, I can't act on bug cries faster if you repeat them, by the way.
User avatar
Rain
Round Winner
Posts: 300
Joined: Sat Apr 15, 2006 2:59 pm
Location: a random empty server playing with bots

Post by Rain »

Z-Man: that was an example of what happen. I took it sunday 16 july at around 13.00 GMT.
I dont remember if I cryed the BUG. But all players that was there saw that crossing.

Today happened again. I cryed the bug. I was playing with ct-play, all saw it.

It is hard to take a recording because I cannot know when it is going to happen.

I'll try anyway.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

I'll try to locate it. Why don't you just *always* record when playing bugfarm? :) That's the safest way to have a recording when it turns out I needed one. The performance cost should be moderate, it's just dumping data to a file.
User avatar
Z-Man
God & Project Admin
Posts: 11585
Joined: Sun Jan 23, 2005 6:01 pm
Location: Cologne
Contact:

Post by Z-Man »

I didn't get that particular incident, but there were plenty of others, and after confirming that it was indeed a serverside error (ok, I had to see it happen to myself to believe that, always the fastest way to get me out of the denial phase), the change that caused it was easy to identify. Consider all "BUG: I/he/she went through my/his/her wall" cries handled.

There was an incident where a new match was started after one dummy short round; the reason was a score event at a time where the round end was already decided, but not yet executed. That loophole has been eliminated. I won't update the code Bugfarm runs today because the Ladle will be played there and it seems to run mostly fine now.
Post Reply