2 / 4 in qseq3 means half the speed, like two ticks are played in the time of four.

2 / 4 in unpunch means double the speed, each tick as 2/4 the length of a regular tick.

Which is the correct one?

The hard-coded value can also sound bad. Adjustable 'humanization' value would be good.

When changing the length, latch3 sends multiple messages on each value change. Also, changing length causes audio drop-outs.

Even better: Support loop ranges. Even betterer: Support additional loop range that follows the current loop, so that loop ranges can be switched seamlessly. Even bettererer: Do it graphically with datastructures.

In certain areas of unpunch's GUI it happens that a "background" scalar is clicked, which messes up the appearance. By switching pages the correct view can be restored, but it would be much nicer if it didn't happen in the first place.

This.

by using new bus order feature of netpd.

They are unnecessary at loading time. This seems to be the reason for this issue: reduzent/netpd#30

The lack of automation is still a huge disadvantage compared to unstep.

How about moving all patches directly in netpd2-patches, that way one could submodule (optionally for development) netpd2-patches into netpd2 by:

git clone git://github.com/reduzent/netpd2.git
cd netpd2
rm patches/.gitignore
rm patches
git submodule git://github.com/reduzent/netpd2-patches.git patches
git submodule update --init --recursive

• this would require that abs are moved in netpd2 into netpd2/patches/abs
• maybe netpd2/.gitignore could contain patches/* to not track patches+abs (if this works when submoduled)

The offset detection of master triggers errors, when being offline:

error: unpack: type mismatch

Before git was common, I once had quite a lot of the TR-808 sounds synthesized in Pd. However, due to some stupid mistake I lost them all but kick and clap. Would be nice to have them back.

There is yet no instrument that generates LFO waveforms for controlling parameters similar to latch3. This should be implemented.

The number box for the delay time allows for negative values, although the actual minimum delay is 0.

Pd 0.51 introduced a new object [pdcontrol] which allows to create clickable links inside patches. Since many instruments already are more or less documented with their dedicated page on netpd.org, they should provide a link to it.

write a wiki page on netpd.org

CPU load caused by the instruments evil and evilst is significant. Judging from the experiences with tpf-client, this could probably be done more efficiently.

The name smells...

When synching a session from a peer, the buttons in the qseq3 main window reflect the correct set states, but all sets are playing regardless. When switching set states manually, they behave as expected.

Recently I have had lots of error messages around using e-del, along the lines of...

'warning: insert/1/e-del_dp.1-r: multiply defined'

I think this also produces a kind of 'leaky' effect with the wrong e-del effect picking up audio streams.

This.

When an effect is loaded while evil or evilst are streaming, the streaming terminates immediately.

Thanks to beem for reporting it.

It's not totally clear when, but probably when instruments or aux channels are added, the list of sources gets increasingly more duplicate entries.

Steps to reproduce:

1. Create a new set
2. Create a sequence into that set
3. Clear the set (with 'X' button in main qseq3 window)
4. Create a sequence again

All the sequences in the previously cleared set do not show a cursor going forward while master is running.

Currently, there is no way to control volume per sample separately.

When saving a state while master is stopped, it starts running without the start/stop toggle indicating it properly. It seems this is a regression that was introduced with the re-ordering of messages for allowing nested effects. master shouldn't change its own state when session is saved (or transferred to peer).

With ripple's waveform window, cue points can be set very precisely. Still it would be handy, if it there would be an option to make cue points snap exactly onto zero-crossings for quicker editing.

Sometimes when using shuffle from master and state is saved as preset, the shuffle clock "loses" it second tick and only even numbered steps are output. Touching the 'shuffle' sliders restores it.

When loading an effect into an insert, the path of the effect parameters is different from the path of the loading itself:

[..b] /1/mx/mx/insert/1/load rphase
[..b] /1/mx/insert/1/rphase/freq 0.0909091

This is an example of an effect in an insert in the main section. The same applies also to the inserts of the instruments:

[..b] /2/sine/mx/insert/1/load rphase
[..b] /2/sine/insert/1/rphase/freq 0.0909091

The loading probably should rather use:

/2/sine/mx/insert/1/load rphase

A metaseq set assigned to a sequence in qseq is lost when the netpd set is saved and reopened.

And sometimes even after the /dump/finished message. Why are these so frivolous?

Sometimes playback starts only after the second loop. However, it can be forced to play by hitting pause and play again.

Only after unsolo-ing and solo-ing a segment, the new length takes effect. This was reported by sqgl, thanks for that.

It seems there is somewhere a slider object with an empty label that causes the scrollbar to appear.

Somehow gate is still missing.

Only the second click actually creates the first aux channel.

People report that selecting effects in mx is finicky. Eni proposes a different approach to scrolling/dragging:

• Click pops up a list of effects
• Click on effect name selects effect

I think this could even be refined by using the first click to show only available libraries and click on library name shows list of effects.

State syncing from peer works OK, but when restoring from preset, too many sets appear, or wrong assignment of instruments to sequences.

An instrument then could have many [i2mx $1 instr] instance. For instance

• [i2mx $1 instr/1]
• [i2mx $1 instr/2]
• etc.

or with additional argument:

• [i2mx $1 instr 1]
• [i2mx $1 instr 2]
• etc.

People reported that latch3 does not work for MIDI controlled parameters. However, it would make totally sense to be able to record movement from a physical controller as opposed to mouse movements.

So that a MIDI keyboard can be used to control instruments using built-in [unpunch] sequencer. Also support for being able to record MIDI notes into [unpunch] would be nice.

It would be nice to be able to apply effects on different sounds separately. 6 mx channel strips per mx might be overkill. Maybe add them dynamically when needed and add a simple router in ripple?

Pd asks if the user wants to save changes to 'mx-dropdown' when trying to quit at the end of a session - as this is for each instance of the patch it can produce a lot of messages. Workaround is to click 'unpatch' at the end of a session but this won't be obvious to many users.

It would help to get a quick overview of what is going on if mx would display the a level meter for each individual instrument.

Here is a strange little bug - if I launch 'unpatch' and create a simple setup with drum-delta or something and qseq3 and give it a polyrhythm of 3/4 with a 12:3 layout, qseq3 will mute every third cell (bom,bom,pause,bom,bom,pause) but as soon as I change the tempo from the default 120 it fixes itself and plays everything properly. I can move the tempo back to 120 and it works ok.
On MacOSX 10.13.6 Pd 0.51-4 netpd 2.3.0

In lilacid one note at the bottom of the unstep piano roll (Fsharp?) will not sound and gives this error message:
tabosc4~: NBLSQUARE119: no such array

When no [unstep] based instrument is loaded and the 'selector' in the midi2unstep bridge is used, the following error is printed:

error: $1: argument number out of range

This can be worked-around by loading another instrument with [i2mx] and closing it again. Sounds like an order problem.

When loading ripple, the following error messages appear in the Pd console:

error: 1440.window-frame: no such template
error: 1440.zero-line: no such template
error: 1440.cue-marker: no such template
error: 1440.cue-dtl: no such template
error: 1440.cue-marker: no such template
error: 1440.cue-dtl: no such template
error: 1440.cue-marker: no such template
error: 1440.cue-dtl: no such template