events with timestamps after replay.frames

[dsjoerg]

In WoL, we could rely on the fact that replay.frames was the length of the replay (in frames). And there would certainly not be any events with a frame timestamp after replay.frames.

Well not anymore.

import sc2reader replay = sc2reader.load_replay("http://dl.dropbox.com/sh/k9nl1w4lox9smsu/K26MovC2fE/sc2replays/2.0.0.23925/Korhal%20City%20%287%29.SC2Replay?dl=1") print replay.events[-1].frame 26073 print replay.events[-1] 27.09 Wraithan PlayerLeaveEvent

print replay.frames 25561

Is the parser mis-parsing? Is there a second frame timestamp somewhere?

If this is not a mis-parse, then we may be required to do some hackery to detect the "true" game length and not rely on replay.frames any more.

[dsjoerg]

This replay was release_string '2.0.0.23925'

So maybe it was a bug that has been fixed in later releases. I'll survey the replays and see which releases this has been observed to happen in.

[dsjoerg]

Oh wow. This problem happens with 100% of the 2.0.0.23925 replays.

And with 0% of the 2.0.0.24247 and 2.0.3.24764 replays. We have ~460 replays from 2.0.0.24247, so I'm confident the issue was resolved in that build. We only have 6 replays from 2.0.3.24764, but I think it's quite likely the bug stayed fixed.

Given that the bug has been fixed, I dont think it's terribly important to do anything about this, at least not for ggtracker. Closing, feel free to re-open if someone wants to do something about this.

[GraylinKim]

Great, thanks for investigating @dsjoerg. Its called beta for a reason. I am perfectly happy reporting bad data for beta replays that actually had bad data in them.