Has the bug occured in previous versions? If yes, please specify version(s): Not sure.
Description of the bug: When testing how XM interacts with Offset commands in Compat and MPT mode, i discovered what i think is unintended behavior. While in Compat mode, Offset commands whose value is larger than the sample size were outright ignored. That's expected. In MPT mode, those same values were processed by playing the note at the end of the sample (i could tell because the channel VU would light up but not play the note).
However, if a normal sample loop was set (with the loop end the same value as the last sample number), the sample loop would not activate -- shouldn't it? With a Bidi loop though, the looping was activated from the END of the Bidi loop -- even if you set the Loop End to NOT the last sample. Here's the kicker: it did this no matter what the 9xx value was (900, 9FF, or anything in between).
How often does it happen?: Always.
How to reproduce (step by step description):
I set up a sample and an instrument in an XM track and tried all tests with different loop settings, between each mode (Compat and MPT). I also called the Offset command with and without the note on the same row, and with different values.