search

ppy / osu

rhythm is just a *click* away!
https://osu.ppy.sh
MIT License
15.44k stars 2.3k forks source link

Missing a hold note release causes random misses in inverse maps #9608

Closed Joppe27 closed 4 years ago

Joppe27 commented 4 years ago

When missing a hold note release, specifically on inverse maps or maps with a lot of LNs, there will be random misses throughout the rest of the map. I think it misses every perfect release in the specific column you missed on before, until you get a great, good, etc. (not a perfect/marvelous) on a release in that same column. It's hard to explain, please watch this video where I intentionally miss the first LN if it isn't clear: https://youtu.be/he3tHxYS7eU

map: https://osu.ppy.sh/beatmapsets/502203#mania/1111137 version: 2020.717.0 no logs

bdach commented 4 years ago

The hypothesis given above seems incorrect (as in I cannot reproduce anything like getting misses on one column after missing a release). I also am not sure what I'm looking at in the video.

I don't think we can do any investigation into this until a replay is provided (which there is no convenient way to access right now). Or someone who can play mania good enough to be able to play the attached map looks into it. I'm certainly not that sort of person.

Joppe27 commented 4 years ago

The hypothesis given above seems incorrect (as in I cannot reproduce anything like getting misses on one column after missing a release). I also am not sure what I'm looking at in the video.

I don't think we can do any investigation into this until a replay is provided (which there is no convenient way to access right now). Or someone who can play mania good enough to be able to play the attached map looks into it. I'm certainly not that sort of person.

This bug only occurs on pretty dense LN maps like the one I attached, not on every mania map. In the video, you can see osu! displaying misses every time there is a gap in the left most lane (which is why i suspect it's missing the releases), while I'm not missing (i only missed the very first hold note). Let me know if you need any other info/video

bdach commented 4 years ago

I suppose if you provided your client.db file we should be able to fish out the replay from the depths of the files folder. (Or if we add the functionality to export replays, I guess - that's actually something that might be useful going forward...)

Joppe27 commented 4 years ago

Should I include my files folder too? This is client.db: https://www1.zippyshare.com/v/KbDoRxup/file.html

bdach commented 4 years ago

You can just include files/3/35/3571c14bdefbf8e278049542568e587056db7bc5ac10fc1305766cbc470cd757. It should be the .osr for the play on the video if I'm not mistaken.

Joppe27 commented 4 years ago

https://www95.zippyshare.com/v/6Pw1gtkM/file.html I get a no replay data message for this one, is that normal?

bdach commented 4 years ago

Thanks, can reproduce. Haven't looked into it in detail (will require a proper slowdown) but definitely something seems to be afoot. While it's not exactly the play captured on the video, some weird misses on the first column from the left are visible about ~45 secs in.

Joppe27 commented 4 years ago

This bug has returned in the newest version 2020.820.0, can this issue be reopened or do I create a new one?

bdach commented 4 years ago

The fix applied previously is unlikely to have regressed as it's covered with automated tests. However as noted in the PR description this can still occur due to note lock which will be addressed separately (and is tracked at #2854).