key not detected for some tracks

I tried making a new database with 2.2 and analyzing it. All tracks that had a valid detected BPM also had a detected key, so this is a regression in master. Opening the database with analyses done with 2.2 with master works fine, so this is an issue with the analysis in master. I presume it is related to the recent multithreaded analysis changes.

Unfortunately I'm not able to confirm this. Both the ad-hoc analysis of the PlayerManager as well as the multi-threaded batch analysis reliably detect both bpm and key of all analyzed tracks without exceptions.

A previous bug was already fixed, I will update the status: https://bugs.launchpad.net/mixxx/+bug/1813413

The same tracks are affected when analyzing different databases, so I am doubtful this is because of a race condition.

For a few track no key is detected, no matter how often or when I analyze them. I think that's expected. The settings of the QM key analyzer may affect this behavior.

Key analysis is still single-threaded and synchronized with a mutex.

That is not expected. All tracks that were able to be read and weren't blank filler tracks had a key detected with 2.2.

Enabling the "Re-analyze keys when settings change or 3rd-party keys are present" preference option and reloading the affected tracks does not result in their keys being detected.

The multithreaded analysis branch ( https://github.com/mixxxdj/mixxx/pull/1624 ) was merged before the VAMP removal branch ( https://github.com/mixxxdj/mixxx/pull/926 ). The merge commit for the multithreaded analysis branch (b9c9b7f) correctly analyzes these tracks' keys, but the merge commit for the VAMP removal branch (18e9314) does not. I did a git bisect to find which commit introduced the regression, but I could not identify that because none of the commits in the VAMP removal branch build before 1fcbf31c07314edfe1c83b355554182b0fbb300e

To clarify, the bug is present in commit 1fcbf31c07314edfe1c83b355554182b0fbb300e

I will check the DownmixAndOverlapHelper (including some refactoring and code style fixes).

A simple logging statement gives some insights:
Warning [AnalyzerThread 0 #1]: No key detected in analyzed window: 25

Seems to be a rounding error in GetKeyMode::process()

https://github.com/mixxxdj/mixxx/pull/2107

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/9655