Pd installation instructions outdated?

[jarmitage]

Trying to install the OS X version of the lib in this release and trying to follow this instruction:

For Pd, add ml.*.pd_* to your Pd search path and add the "ml" library to your startup flags by opening up the application Preferences and adding "-lib ml" to the Startup Flags.

• Am I right in thinking that add ml.*.pd_* to your Pd search path means just having the folder of *.pd_darwin files in the search path (if I do this the objects do instantiate correctly)? If so this instruction could be simplified a bit.

• Is it necessary to add the startup flag? If I do this I still get the post message ml can't load library:

[jamiebullock]

Hi @jarmitage. It's not so much the instructions that are outdated as the fact there was file (ml.pd_darwin) missing from the release.

You technically don't need the -lib ml flag. This just allows you to run ml-lib as a Pd library, so you just have one external (ml.pd_darwin) that contains everything rather than lots of different externals. It should work fine without -lib ml in which cans you just need the directory containing the ml-lib externals in your Pd path.

I've now updated the release to contain ml.pd_darwin in case you want to try it.

[jarmitage]

Hi @jamiebullock, I tried the new release and added the -lib ml flag again. The library appears to load, but I cannot create an ml object. Also getting the error maximum object loading depth 1000 reached:

[jamiebullock]

There's no ml object as such. The fact it states:

ml-lib - machine learning library for Max and Pure Data version 0.18.0 (c) 2013 Carnegie Mellon University

Shows the library has loaded correctly.

After that you have successfully created an instance of ml.mlp. Everything is working!

[jarmitage]

Ah great!

Should I worry about that error?

[jamiebullock]

No, you're only getting the error because you're trying to load a library ml as if it were an object, which isn't supported.

[jarmitage]

@jamiebullock is the Linux build also missing a ml.pd_linux object?

[jarmitage]

@jamiebullock this release titled "Bela Build" doesn't have the ARM32 build: https://github.com/cmuartfab/ml-lib/releases/tag/0.18.0

Any chance you can add that? Can use 0.17.0 for now!

[jamiebullock]

Hi @jarmitage it's possible the ml.pd_linux is missing from the Linux release. Do you need it?

Re: ARM32, yes I can do another generic ARM build, however I'd be interested to know if the Bela build works on other ARM platforms. Might be worth trying!

[jarmitage]

We are doing good on Bela with 0.17.0 ARM32 right now!

Are we missing anything substantial not having 0.18.0 + ml.pd_linux?

[jamiebullock]

Ah... oops! I made the Bela release for 0.18 without actually attaching the Bela build. I've now done that, so if you go to Releases, there is a Bela-optimised build there.

Interesting that the generic ARM32 build works on Bela. Would be useful to benchmark that against the Bela optimised one.

There are a few other fixes in 0.18, but if you have things working, there's no massive advantage to upgrading.

ml.pd_linux doesn't add any functionality, it's just a convenience if you want to distribute a project with only one external that contains "everything"

[jarmitage]

Awesome thank you. Will compare over the weekend and let you know!