Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

https://github.com/bgribble/mfp

Music For Programmers, a graphical patching language
https://github.com/bgribble/mfp

jack midi music puredata python

Last synced: 16 days ago
JSON representation

Music For Programmers, a graphical patching language

Lists

README

        

BUILDING AND INSTALLING MFP
===================================

Post version 0.07, 2024-01-15

TL;DR
---------------------

$ ./waf configure --python=python3 --virtualenv --prefix ~/mfpvirt build install install_deps
[...]
$ . ~/mfpvirt/bin/activate
(mfpvirt)$ mfp -v

THINGS TO KNOW
---------------------

OS platform: I have only tested MFP on 32-bit (raspi) and 64-bit
Linux.

python: MFP requires Python 3.x at this time.

waf: MFP uses 'waf', which you can think of as a replacement for
autotools and make. You do not need to install waf; the correct
version is supplied in the MFP source tree. That's the way waf
is designed to work; more information can be found at
https://code.google.com/p/waf/. MFP's "Makefile" or
"configure.in" equivalent is the file "wscript", which is a plain
Python file. The entry points used by waf are the functions
"configure" and "build", defined near the bottom of the file.

virtualenv: At this phase of MFP's development you should
probably build it in a "virtualenv" sandbox. The waf scripts
support this, but you have to install the "virtualenv" tool
(packaged as 'python-virtualenv' on Debian and Ubuntu). Then you
configure with --virtualenv as in the TL;DR above and the build
will go into the virtual environment. When you specify
--virtualenv, the install prefix (--prefix) is the path to the
virtual environment, so you shouldn't let this default to
/usr/local.

setuptools: The build process depends on Python setuptools. If
you have virtualenv installed, you have setuptools already, but
if you aren't using virtualenv you may need to install setuptools
separately. It's packaged as "python-setuptools" on Debian and
Ubuntu.

pip: The modern replacement for "easy_install" to install Python
packages is called pip. If (1) you are using Debian or Ubuntu, (2)
you do not want to use a --virtualenv install, and (3) you want to use
the './waf install_deps' feature to install some Python dependencies,
you need the Debian package 'python-pip', or your distro's equivalent,
installed.

gcc: The C components of mfp (mfpdsp, pluginfo, and testext) are
straightforward C but include some c99-isms, mainly due to LV2. I
have been using versions from gcc-4.8.3 to 9.3.0 with no trouble.

CONFIGURING
-----------------------

The basic form:

$ ./waf configure [options]

A full list of options is shown by ./waf --help, under
"Configuration options". --virtualenv is the only "custom"
option.

LIBRARIES
-----------------------

The "mfpdsp" C library/program needs a few C libraries with their
development headers and pkg-config files installed:

* ladspa.h (LADSPA plugins)
* glib-2.0 (GLib/GObject)
* json-glib-2.0 (JSON serializing/deserializing)
* serd-0 (LV2 TTL parsing)
* jack (duh)
* liblo (OSC library, needed by pyliblo)
* lv2 headers (LV2 load/save)

You'll get errors in the "./waf configure" process if these aren't installed.

Python runtime dependencies are checked for in "configure" and many of
them can be automagically installed by "./waf install_deps" if you
wish.

install_deps must be run after "./waf install". If you
are using a virtualenv, the installed Python libraries will also
be installed into the virtualenv.

Note that the Python language bindings for Gtk+/GLib and Clutter libraries
CANNOT be automatically installed by this build process. You need to
install your OS package manager's package for the "GIR" (GObject
Introspection Repository) data for Clutter, Glib, Gtk+, Gdk,
GtkClutter, and Pango, and the Python infrastructure to use that data.
In Debian, these are packaged as:

gir1.2-clutter-1.0
gir1.2-gtkclutter-1.0
gir1.2-pango-1.0
gir1.2-glib-2.0
gir1.2-gtk-3.0
python-gi

BUILDING
-----------------------

$ ./waf build

This will build the project into ./wafbuild (or other directory as
specified to "configure"). There may be some temporary stuff placed
into ./build/ and ./dist/ by setuptools, so I wouldn't use those as
names for the waf build directory.

INSTALLING
--------------------------

$ ./waf install

The installation root defaults to /usr/local and can be modified
by passing in the --prefix argument to "./waf configure".

If you get errors on the install step to the effect of os.symlink
failing to make a link because the file already exists -- see GitHub
issue #220 and please add a comment indicating your Linux distribution
and that you are having this problem.

Note that the instances of #220 I have seen are failing at the
very tail end of the install process, and are trying to install
files that are already there... which means that you can probably
ignore this error. Try activating the virtualenv and running
mfp. If it launches, you are good to go.

DESKTOP FILE
--------------------------

The template desktop file `mfp.desktop` will be populated with
paths and installed to ${PREFIX}/share/mfp/mfp.desktop. You will
need to copy it to wherever such things go on your system.

The icon path is what I needed to get icons working on my Gnome
system. You may need to change it. If you need to do something
manual, the actual icon image is the one in mfp.svg. I can't
express how crabby I am about how messed up icons are on the
Linux desktop.

For what it's worth, on my system the desktop file should be
~/.local/share/applications/mfp.desktop

RUNNING TESTS
--------------------------

There are some tests, though not as many as I would like. What's
there basically comes in two flavors, all of which can be run
with various incantations of the `nosetests` test runner:

* Python test cases in mfp/test: run with plain `nosetests -v mfp/test/`

* C test cases in mfpdsp/test_*: run with the `testext` plugin
to nosetests, `nosetests -v --with-testext --exe wafbuild/libmfpdsp.so`

Note on nosetests: I started using it because, 12+ years ago, it
was the best test runner. Now other tools like pytest have won
more mindshare. I spent some effort on the plugins to run C tests
for mfpdsp so I am sticking with nosetests.

HOWEVER, the original nosetests is abandoned and doesn't work
with current Python versions, and the "new" version of nosetests
(nose2) changed the way plugins worked, and doesn't support the
--exe option needed to point nose to the libmfpdsp.so file. So I
switched over to "pynose", which is a fork of the original
nosetests, just updated to work with contemporary Python. It gets
installed by install_deps.