Sorry, Sourceforge is blocked for me.

I noticed that when using [poly 1 0] (note stealing off)

when I would smash multi-note chords and release notes,

sometimes I would be left pressing a note,

yet the output of [poly] is silence.

I thought this might be a bug with the Windows MIDI driver,

but it is definitely a bug with [poly], because with higher

values of N, [poly N 0] knows the correct notes

remaining.

See attached ZIP file.
There is an HTML+image included which explains

how to run the patch to demonstrate the issue.

The problem is there is a race condition.

Lets say you smash the following chord:

("smash" means play all notes as simultaneously as possible)

24 26 28 29 31 33 [C-D-E-F-G-A]

Now the value which will be latched by [poly 1 0] is random.

Let's say it is 28.

Now we release 33, 31,29 .... ALL GOOD!

Now release 29.
NOTE 24, 26, 28 are still depressed!

[poly] doesn't know what to do here, and releases 28, ALL NOTES OFF. Yet I am still holding down 3 notes.

The problem here is how do deal with this condition is ambiguous.

Should [poly] have "highest note value" priority? (and jump to 26)
Or "lowest note value" priority? (and jump to 24)
Or "order received" priority? (and jump to to the latest received of [24,26])
Or maybe "reverse order received" priority? (and jump to the earliest received of [24,26])

I would request a feature enhancement to provide some way to set

the mode of poly (new inlet, message to inlet1, and/or starting parameter) to chose one of these four modes.

In the meantime, I think I can code my own version of what I need as an abstraction.

Thanks Miller and everyone who contributes here for such an awesome tool/toy!

--
May you, and all beings
be happy and free from suffering :)
-- ancient Buddhist Prayer (Metta)