Bug fix in sysex handling and incorporation of various minor changes …

…for header guards and compiler warnings. Also, fixed mm (midi monitor) ability to filter messages and control output. Details: The interface between system-dependent and system-independent code was originally based on the idea that the system did a lot of work to parse out messages, so it should be simple to push these messages to the PortMidi input queue. This should have avoided re-parsing MIDI bytes and enabled fast word-at-a-time handling of sysex data. However, MacOS MIDI handling is largely undocumented and has many complexities: packets can contain multiple messages or partial (sysex) messages, messages can have embedded 1-byte real-time messages. To handle these special cases, code became messy and to add to the complexity, the mess was split between system-dependent (mmmacosxcm.c) and system-independent (portmidi.c) code. Even after years, a new bug was found by te-johan, and there may be other problems with embedded real-time messages. Rather than continuing to patch the code, this commit greatly simplifies code by completely reparsing MIDI in system-independent code, but retains the option of passing in fully formed MIDI short messages.

CMakeLists.txt

@@ -65,8 +65,6 @@ endif()
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR})
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR})
 
-option(BUILD_PMDEFAULTS
-       "build the legacy Java application PmDefaults" OFF)
 option(BUILD_JAVA_NATIVE_INTERFACE
        "build the Java PortMidi interface library" OFF)
 
@@ -92,11 +90,6 @@ if(BUILD_JAVA_NATIVE_INTERFACE)
   set(PMJNI_IF_EXISTS "pmjni")  # used by INSTALL below
 else(BUILD_JAVA_NATIVE_INTERFACE)
   set(PMJNI_IF_EXISTS "")       # used by INSTALL below
-  if(BUILD_PMDEFAULTS)
-    message(FATAL_ERROR
-      "Cannot build PM_Defaults program (BUILD_PM_DEFAULTS) without option "
-      "BUILD_JAVA_NATIVE_INTERFACE")
-  endif(BUILD_PMDEFAULTS)
 endif(BUILD_JAVA_NATIVE_INTERFACE)
 
 

README.md

@@ -24,82 +24,104 @@ make         # compile sources and build PortMidi library
 sudo make install  # if you want to install to your system
 ```
 
-**PmDefaults** is a Java-based program for setting default MIDI
-devices. It is necessary if you use `Pm_DefaultInputDeviceID()` or
-`Pm_DefaultOutputDeviceID()` to avoid implementing your own device
-browsing, selection and preferences in your applications. Enable
-`BUILD_PMDEFAULTS` and `BUILD_JAVA_NATIVE_INTERFACE` in ccmake, and
-see `pm_java/README.txt` for more information.
-
-See also notes in `pm_mac/README_MAC.txt`, `pm_win/README_WIN.txt` and
-`pm_linux/README_LINUX.txt`.
-
-## What's New?
-
-PortMidi has some fixes for Apple M1 cpus as of May 23, 2022. This has not yet
-been formally released to allow for further testing, but please use the latest
-code if you want to run on an M1.
-
-PortMidi has some changes in 2021:
-
- - added Pm_CreateVirtualInput() and Pm_CreateVirtualOutput() functions that allow
-   applications to create named ports analogous to devices.
-
- - improvements for macOS CoreMIDI include higher data rates for devices, better
-   handling of Unicode interface names in addition to virtual device creation.
-
- - the notion of default devices, Pm_GetDefaultInputDeviceID(), 
-   Pm_GetDefaultOutputDeviceID and the PmDefaults program have fallen into disuse
-   and are now deprecated.
-
- - Native Interfaces for Python, Java, Go, Rust, Lua and more seem best left
-   to individual repos, so support within this repo has been dropped. A Java
-   interface is still here and probably usable -- let me know if you need it
-   and/or would like to help bring it up to date. I am happy to help with, 
-   link to, or collaborate in supporting PortMidi for other languages. 
-
-For up-to-date PortMidi for languages other than C/C++, check with
-developers. As of 27 Sep 2021, this (and SourceForge) is the only repo with
-the features described above.
+## Installation
+
+My advice is to build PortMidi and statically link it to your
+application. This gives you more control over versions. However,
+installing PortMidi to your system is preferred by some, and the
+following should do it:
+```
+cmake .
+make
+sudo make install
+```
+
+## Language Bindings
+
+Here is a guide to some other projects using PortMidi. There is not
+much coordination, so let us know if there are better or alternative
+bindings for these and other languages:
+
+- Python: Various libraries and packages exist; search and ye shall
+  find.
+- [SML](https://github.com/jh-midi/portmidi-sml2)
+- [OCaml](https://ocaml.org/p/portmidi/0.1) 
+- [Haskell](https://hackage.haskell.org/package/PortMidi) 
+- [Erlang](https://hexdocs.pm/portmidi/PortMidi.html) 
+- [C#](https://github.com/net-core-audio/portmidi)
+- [Rust](https://musitdev.github.io/portmidi-rs/)
+- [Go](https://github.com/rakyll/portmidi)
+- [Odin](https://pkg.odin-lang.org/vendor/portmidi/)
+- [Serpent](https://sourceforge.net/projects/serpent/) - a real-time
+  Python-like language has PortMidi built-in, a MIDI-timestamp-aware
+  scheduler, and GUI support for device selection.
+- [Pd (Pure Data)](https://puredata.info/) uses PortMidi.
+
+
+## What's New? 
+
+(Not so new, but significant:) Support for the **PmDefaults** program,
+which enabled a graphical interface to select default MIDI devices,
+has been removed for lack of interest. This allowed us to also remove
+C code to read and parse Java preference files on various systems,
+simplifying the library. **PmDefaults** allowed simple command-line
+programs to use `Pm_DefaultInputDeviceID()` and
+`Pm_DefaultOutputDeviceID()` rather than creating device selection
+interfaces. Now, you should either pass devices on the command line or
+create your own selection interface when building a GUI
+application. (See pm_tests for examples.)  `Pm_DefaultInputDeviceID()`
+and `Pm_DefaultOutputDeviceID()` now return a valid device if
+possible, but they may not actually reflect any user preference.
+
+Haiku support in a minimal implementation. See TODO's in source.
+
+sndio is also minimally supported, allowing basic PortMidi functions
+in OpenBSD, FreeBSD, and NetBSD by setting USE_SNDIO for CMake, but
+not delayed/timestamped output and virtual devices.
 
 # Other Repositories
 
-PortMidi used to be part of the PortMedia suite, but this repo has been reduced to
-mostly just C/C++ code for PortMidi. You will find some other repositories in this PortMidi project
-set up for language bindings (volunteers and contributors are invited!). Other code removed from
-previous releases of PortMedia include:
+PortMidi used to be part of the PortMedia suite, but this repo has
+been reduced to mostly just C/C++ code for PortMidi. You will find
+some other repositories in this PortMidi project set up for language
+bindings (volunteers and contributors are invited!). Other code
+removed from previous releases of PortMedia include:
 
 ## PortSMF
 
-A Standard MIDI File (SMF) (and more) library is in the [portsmf repository](https://github.com/PortMidi/portsmf).
+A Standard MIDI File (SMF) (and more) library is in the [portsmf
+repository](https://github.com/PortMidi/portsmf).
 
-PortSMF is a library for reading/writing/editing Standard MIDI Files. It is
-actually much more, with a general representation of events and updates with
-properties consisting of attributes and typed values. Familiar properties of
-pitch, time, duration, and channel are built into events and updates to make
-them faster to access and more compact.
+PortSMF is a library for reading/writing/editing Standard MIDI
+Files. It is actually much more, with a general representation of
+events and updates with properties consisting of attributes and typed
+values. Familiar properties of pitch, time, duration, and channel are
+built into events and updates to make them faster to access and more
+compact.
 
-To my knowledge, PortSMF has the most complete and useful handling of MIDI
-tempo tracks. E.g., you can edit notes according to either beat or time, and
-you can edit tempo tracks, for example, flattening the tempo while preserving
-the beat alignment, preserving the real time while changing the tempo or 
-stretching the tempo over some interval.
+To my knowledge, PortSMF has the most complete and useful handling of
+MIDI tempo tracks. E.g., you can edit notes according to either beat
+or time, and you can edit tempo tracks, for example, flattening the
+tempo while preserving the beat alignment, preserving the real time
+while changing the tempo or stretching the tempo over some interval.
 
-In addition to Standard MIDI Files, PortSMF supports an ASCII representation
-called Allegro. PortSMF and Allegro are used for Audacity Note Tracks.
+In addition to Standard MIDI Files, PortSMF supports an ASCII
+representation called Allegro. PortSMF and Allegro are used for
+Audacity Note Tracks.
 
 ## scorealign
 
-Scorealign used to be part of the PortMedia suite. It is now at the [scorealign repository](https://github.com/rbdannenberg/scorealign).
-
-Scorealign aligns
-audio-to-audio, audio-to-MIDI or MIDI-to-MIDI using dynamic time warping (DTW)
-of a computed chromagram representation. There are some added smoothing tricks
-to improve performance. This library is written in C and runs substantially 
-faster than most other implementations, especially those written in MATLAB,
-due to the core DTW algorithm. Users should be warned that while chromagrams
-are robust features for alignment, they achieve robustness by operating at 
-fairly high granularity, e.g., durations of around 100ms, which limits 
-time precision. Other more recent algorithms can doubtless do better, but
-be cautious of claims, since it all depends on what assumptions you can 
-make about the music.
+Scorealign used to be part of the PortMedia suite. It is now at the
+[scorealign repository](https://github.com/rbdannenberg/scorealign).
+
+Scorealign aligns audio-to-audio, audio-to-MIDI or MIDI-to-MIDI using
+dynamic time warping (DTW) of a computed chromagram
+representation. There are some added smoothing tricks to improve
+performance. This library is written in C and runs substantially
+faster than most other implementations, especially those written in
+MATLAB, due to the core DTW algorithm. Users should be warned that
+while chromagrams are robust features for alignment, they achieve
+robustness by operating at fairly high granularity, e.g., durations of
+around 100ms, which limits time precision. Other more recent
+algorithms can doubtless do better, but be cautious of claims, since
+it all depends on what assumptions you can make about the music.

pm_common/CMakeLists.txt

@@ -79,11 +79,13 @@ if(NOT WIN32)
 endif()
 
 # Check for sndio
-include (FindPackageHandleStandardArgs)
-find_path(SNDIO_INCLUDE_DIRS NAMES sndio.h)
-find_library(SNDIO_LIBRARY sndio)
-find_package_handle_standard_args(Sndio
-    REQUIRED_VARS SNDIO_LIBRARY SNDIO_INCLUDE_DIRS)
+if(USE_SNDIO)
+  include (FindPackageHandleStandardArgs)
+  find_path(SNDIO_INCLUDE_DIRS NAMES sndio.h)
+  find_library(SNDIO_LIBRARY sndio)
+  find_package_handle_standard_args(Sndio
+      REQUIRED_VARS SNDIO_LIBRARY SNDIO_INCLUDE_DIRS)
+endif(USE_SNDIO)
 
 # first include the appropriate system-dependent file:
 if(SNDIO_FOUND AND USE_SNDIO)

pm_common/pminternal.h

@@ -126,9 +126,10 @@ typedef struct pm_internal_struct {
         * sending data from the middle of a sysex message. If a sysex
         * message is filtered, sysex_in_progress is false, causing the
         * message to be dropped. */
-    PmMessage sysex_message; /* buffer for 4 bytes of sysex data */
-    int sysex_message_count; /* how many bytes in sysex_message so far */
-
+    PmMessage message; /* buffer for 4 bytes of sysex data */
+    int message_count; /* how many bytes in sysex_message so far */
+    int short_message_count; /* how many bytes are expected in short message */
+    unsigned char running_status; /* running status byte or zero if none */
     int32_t filters; /* flags that filter incoming message classes */
     int32_t channel_mask; /* filter incoming messages based on channel */
     PmTimestamp last_msg_time; /* timestamp of last message */
@@ -148,6 +149,8 @@ typedef struct pm_internal_struct {
     uint32_t fill_length; /* how many sysex bytes to write */
 } PmInternal;
 
+/* what is the length of this short message? */
+int pm_midi_length(PmMessage msg);
 
 /* defined by system specific implementation, e.g. pmwinmm, used by PortMidi */
 void pm_init(void); 

pm_common/pmutil.h

@@ -9,6 +9,9 @@
     available for other uses.
  */
 
+#ifndef PORTMIDI_PMUTIL_H
+#define PORTMIDI_PMUTIL_H
+
 #ifdef __cplusplus
 extern "C" {
 #endif /* __cplusplus */
@@ -177,3 +180,5 @@ PMEXPORT PmError Pm_SetOverflow(PmQueue *queue);
 #ifdef __cplusplus
 }
 #endif /* __cplusplus */
+
+#endif // PORTMIDI_PMUTIL_H