CachingReader: Invalid sample position from engine

[Uwe Klotz]

Branch: master / 2.1

Deck 2 was playing and I was adjusting a cue point on deck 1. When hitting cue on deck 1 suddenly the music stopped. My first guess was a wrong midi mapping, but than I noticed that a debug assertion stopped Mixxx. Only a test run with controller debugging enabled ;)

Please see the attached stack trace.

[Uwe Klotz (uklotzde-deactivatedaccount)]

This has now happened multiple times when setting a cue point. Only one deck is affected, it doesn't matter if another deck is playing or not. The engine code for calculating the sample offser when setting cue points is definitely broken!

[Daniel Schürmann (daschuer)]

This is probably caused be the late latest changes, allowing fractional sample cue points.
We need to allow fractional hints for the caching reader as well.

[Uwe Klotz (uklotzde-deactivatedaccount)]

I disagree. The caching reader can only read from frame boundaries. Sub-frame calculations must be done in the engine.

[Uwe Klotz (uklotzde-deactivatedaccount)]

By switching to frame positions this would work as expected. It's just a missing adjustment of sample positions to even numbers during sub-frame calculations.

[Daniel Schürmann (daschuer)]

Yes, the caching reader can only read full frames, and the engine can request fractional frames.
This bug is just an issue where we round to full frames. I have no strong opinion about the right place right now.

[Uwe Klotz (uklotzde-deactivatedaccount)]

The sub-frame calculations (and interpolation) must be done once, but not twice. If resampling between the sampling rate provided by decoding and the internal rate of the engine is needed, that could be the right place.

[Be (be.ing)]

https://github.com/mixxxdj/mixxx/pull/1223

[Swiftb0y (swiftb0y)]

Mixxx now uses GitHub for bug tracking. This bug has been migrated to:
https://github.com/mixxxdj/mixxx/issues/8812