Cycles turning invulnerable

[Manuel Moos]

Sometimes, for no apparent reason, cycles turn invulnerable. At least a clientside recording would be required, better a serverside recording.

[Yann Kaiser (epsy)]

Today on #arma.xclan:
(StickyNoob had that bug some times already)

[23:18] <+epsy>: StickyNoob, ping
[23:19] <StickyNoob>: yo
[23:19] <+epsy>: https://bugs.launchpad.net/armagetronad/0.2.8/+bug/322066
[23:19] <@teabot>: Launchpad bug 322066 in armagetronad/0.2.8 "Cycles turning invulnerable" [High,Incomplete]
[23:19] <+epsy>: I remember having seen you doing such stuff, do you know about any details?
[23:20] <StickyNoob>: not really
[23:20] <StickyNoob>: just a lag issue really
[23:21] <+epsy>: anything particular (except the bug itself) happened at these moments?
[23:21] <StickyNoob>: its generaly after a massive lag spike
[23:21] <+epsy>: what do you see? do you see yourself getting stuck? can you turn?
[23:22] <StickyNoob>: yeah you have full control you just cant die
[23:22] <StickyNoob>: eventualy the server and client seem to sync up and no matter where you are you will die imediately
[23:22] <+epsy>: so you usually notice it then just choose to stick to a wall, to not become a cheat?
[23:24] <StickyNoob>: well as soon as it happens you only a short time until it will kill you
[23:24] <StickyNoob>: and the whole game is so laggy at that point you cant do much useful anyway
[23:24] <+epsy>: a short time?
[23:24] <StickyNoob>: almost anything you see is atleast 2 sec out of date
[23:25] <StickyNoob>: 10-20 sec, 30 maby
[23:25] <+epsy>: ah, right
[23:25] <+epsy>: do you remember in what matches it happened?
[23:26] <StickyNoob>: hmm
[23:26] <StickyNoob>: its happend in atleast one competition but i cant remember which
[23:26] <+epsy>: yeah, maybe you remember the opponent?
[23:26] <StickyNoob>: hmm no
[23:27] <StickyNoob>: thinking CT for some reason
[23:27] <+epsy>: I swear we had you having that against CT
[23:27] <StickyNoob>: im fairly sure it was a final
[23:27] <+epsy>: AFL perhaps?
[23:27] <StickyNoob>: maby
[23:27] <+epsy>: hopefully AFL should have client recordings
[23:28] <+epsy>: mind if I paste this to the bug?
[23:28] <StickyNoob>: could try injecting some artifical lag spikes
[23:28] <StickyNoob>: yeah fine

[Manuel Moos (z-man)]

I don't think that's the same. The invulerability can clearly be seen by other players; what StickyNoob describes sounds like regular lag where the 'you're dead' message just hasn't arrived on his client yet.

[Yann Kaiser (epsy)]

we did see him stuck :)

[Manuel Moos (z-man)]

Looked at the recording in detail. From t=77.896 (game round time 53.2303047), pruAiTeam's cycle ony gets the same snyc message over and over. It's not really invulnerable, it seems, just frozen in time and space. Odd.

[Manuel Moos (z-man)]

Yay! Happened during the preparation for the semifinal on the ladle. Attached is the clientside recording. Around T=430, m_nus, the center of +, goes in a circle and gets stuck.

[Manuel Moos (z-man)]

Found the reason, a premature exit out of the cycle simulation caused by floating point rounding combined with packet loss. It's possible there are still other ways cycle simulation can get stuck, of course. The safety net stays in place.