Dynamic time diff bug

[Nairdaa]

A reminder for our current lazy dev. He knows what to do, but is too busy at the moment.

Steps taken to reproduce: Play, watch the timer suddenly going from for example -0.1 sec to +20 sec and stay +x forever (counting, but wrong).

Replay bots:
bhop_mae
Positon: setpos 1266.043823 423.967499 -29.968689;setang 19.866026 171.646530 0.000000
Style: Scroll

bhop_badges
Starts acting already outside of the startzone
Style: Normal

bhop_mae.zip
bhop_badges.zip

Let me know if you need more. That kind of stuff never happened on log's approach to dynamic time difference.

[Nairdaa]

minusnotrng_plusrng.zip example demo playing live with the dynamic timer diff

bhop_mae

[kidfearless]

I'm unable to reproduce it and have a nagging feeling that you are still running the version that was broken.
https://youtu.be/7pxriZA4jFc
The replay was 2665 frames long at 100 tick giving it a 26 second time. I finished with a 22 second time and a dynamic time difference of around 4 seconds.
Tested on the latest very_good_yes

[Nairdaa]

Downloaded latest, compiled, tried again it's still fucked up. Told you to come and see how to reproduce, but you still don't know how to reproduce that even tho I recorded a video for you. Come and see it live how to cause it.

[Nairdaa]

[ Heroine ] kebenjekajse : now
♫♫ an idiot sandwich : what happened
[ Heroine ] kebenjekajse : said -10s
[ Heroine ] kebenjekajse : when i started
♫♫ an idiot sandwich : k
[ Heroine ] kebenjekajse : it went away now
♫♫ an idiot sandwich : disappeared?
[ Heroine ] kebenjekajse : yea but back again now :D

[Nairdaa]

Also, after updating it's not as fucky as it used to, but still fucks up sometimes.

[Nairdaa]

bhop_mae.zip demo of another odd shit

[Nairdaa]

Also, when you go back to startzone (not with !r) your dynamic time is not back to 0 when you have more than 10 seconds on the timer (you can have even 30 on it, walk back to startzone, walk out and it will show -10 sec on dynamic timer right away).

Your thing is bugged.

[Nairdaa]

sm_r
  StopTimer()
    Shavit_OnStop()
      ClearFrames()
        gA_PlayerFrames.Clear()

callstack of sm_r thats why the shit only resets on !r and not when you enter startzone without !r
fix the prerrun and the dynamic timer will be fixed too

cheers

[kidfearless]

Also, when you go back to startzone (not with !r) your dynamic time is not back to 0 when you have more than 10 seconds on the timer (you can have even 30 on it, walk back to startzone, walk out and it will show -10 sec on dynamic timer right away).

Your thing is bugged.

I told you many times before that that was an issue. And thinking about it now, if you use a set start plugin it can mess with it as well. Or checkpointing inside the start zone.

[rtldg]

That PR fixes a part of this.
Since only shavit_replay_dynamictimedifference_search seconds are searched through the replay frames, the time difference still won't be perfect unless shavit_replay_dynamictimedifference_search is bigger than the replay.
tl;dr: raise shavit_replay_dynamictimedifference_search

I messed around with grabbing the difference for every player every 10th frame (~0.1s on CSS) and then returning that cached but who knows how well servers with many people & a big replay will handle that.
Thoughts me adding that (for testing or whatever) ^ @kidfearless?

[kidfearless]

Unfortunately sorting through large arraylists is pretty heavy on the server. The way I have it implemented now is pretty accurate until it goes outside the search range. But if a player is 10 seconds ahead or behind the replay than that accuracy shouldn't really matter.

As far as having the skipping frames thing goes, as long as there's a convar to control the number of frames skipped then I think that'll be a great idea.

[Nairdaa]

The way that log implemented it worked just fine with 20 ppl on the server and a 10h long replay :/ I still don't understand why it was reworked. It was accurate and really not cpu heavy.