I second the motion for platform-specific MIDI files. It could be done as
follows. This is a bit sketchy, but obviously something like this would (a)
work, and (b) be very much simpler than the existing code.

An abstract class (A C class, not a C++ class) for input and output drivers
with self-contained Csound input or output port function. The job of the
abstract driver is to encapsulate the business of reading a platform input
port and writing to a Csound input port, or reading a Csound output port and
writing to a platform output port.

typedef struct AbstractDriver
{
    AbstractDriver *(*open)(void (*csoundPort)(void *csoundData));
    int (*close)(struct AbstractDriver *this);
    int(*sizeAvailable)(struct AbstractDriver *this)();
    int (*read)(struct AbstractDriver *this);
    int (*write)(struct AbstractDriver *this);
    int(*control)(struct AbstractDriver *this, int command, void *data);
};

Then each platform would have a derived class (again a C class, not a C++
class), for example on Linux the MIDI input driver could be something like:

typedef struct LinuxMidiInputDriver
{
    AbstractDriver baseDriver;
    unsigned char buffer[4];
    void (*midiInsert)(MIDIEVENT *event));
} ;

AbstractDriver *LinuxMidiInputOpen(void *buffer, int bufferSize, void
(*csoundPort)(MIDIEVENT *event))
{
    LinuxMidiInputDriver *linuxDriver = (LinuxMidiInputDriver
*)malloc(sizeof(LinxMidiInputDriver), );
    linuxDriver.baseDriver->buffer = buffer;
    midiInsert = csoundPort;
    (AbstractDriver *)linuxDriver;

Then in Csound itself would be tables of drivers,each terminated by a null
pointer:

AbstractDriver *inputDrivers[INPUT_DRIVER_COUNT + 1];
AbstractDriver *outputDrivers[OUTPUT_DRIVER_COUNT + 1];
AbstractDriver *audioInputDriver;
AbstractDriver *audioOutputDriver;

And on each platform there would be a driver.c file that fills in the
tables:

On Linux (note that the Csound midiInsert function gets plugged into the
driver):

inputDrivers[MIDI_INPUT_DRIVER] = LinuxMidiInputOpen(midiInsert);
inputDrivers[LINE_INPUT_DRIVER] = LinuxLineInDriverOpen(midiInsert);
On Windows:

inputDrivers[MIDI_INPUT_DRIVER] = WindowsMidiInputOpen(midiInsert);
inputDrivers[LINE_INPUT_DRIVER] = WindowsLineInDriverOpen(midiInsert);

Then kperf would look like this:

long kperf(long controlSampleCount)
{
 extern struct AbstractDriver inputDrivers[];
 extern struct AbstractDriver *outputDrivers[];
 extern struct AbstractDriver *audioInputDrivers[];
 extern struct AbstractDriver *audioOutputDriver[]s;
 for(controlSampleIndex = 0; controlSampleIndex < controlSampleCount;
controlSampleIndex++)
 {
  // Read any pending control inputs.
  int inputsRead = 0;
  for(AbstractDriver *inputDriver = inputDrivers; inputDriver;
inputDrivers++)
  {
   if((*controlInput->available)(controlInput)
   {
    inputsRead = (*controlInput->read)(controlInput);
   }
  }
  if(inputsRead)
  {
   return controlSampleIndex;
  }
  // Initialize the primary audio buffers.
  controlSampleCounter++;
  memset(inputBuffer, 0, inputBufferSize);
  memset(outputBuffer, 0, outputBufferSize);
  // Read any pending audio inputs.
  for(AbstractDriver *audioInput = audioInputDrivers; audioInput;
audioInput++)
  {
   (*audioInput->read)(audioInput);
  }
  // For each active instrument...
  for(ip = &actanchor; ip = ip->nxtact; )
  {
   // ...run each of its opcodes.
   for(pds = (OPDS *)ip; pds = pds->nxtp; )
   {
    (*pds->opadr)(pds);
   }
  }
  // Write any audio outputs.
  for(AbstractDriver *audioOutput = audioOutputDrivers; audioOutput;
audioOutput++)
  {
   audioOutput = (*audioOutput->write)(audioOutput);
  }
  // Write any control outputs.
  for(AbstractDriver *outputDriver = outputDrivers; outputDriver;
outputDriver++)
  {
   if(outputDriver->available(outputDriver))
   {
    (*outputDriver->write)(outputDriver);
   }
  }
 }
 return controlSampleIndex;
}




-----Original Message-----
From: Larry Troxler 
To: Dave Phillips 
Cc: Csound Mail ; Adam Zygmunt
; Jason ? 
Date: Monday, August 09, 1999 3:33 PM
Subject: Re: compiling Linux Csound (official)


>On Mon, 9 Aug 1999, Dave Phillips wrote:
>>
>> However, I don't actually know where the problem lies in midirecv.c,
>> since the #ifdefs are rather numerous and somewhat confusing. If I had a
>> little more time I could narrow it down, but perhaps some other reader
>> can continue this exercise (Larry T, are you there ?). Perhaps it's time
>> to consider platform-specific files for the MIDI i/o ?
>>
>Actually, although midirecv.c definitely isn't correct, as it turns out, I
>have a MIDI driver problem now after a recent upgrade. So once I fix that,
>I'll go back and have a look.
>
>Actually, I do have the unnoficial version working, and the main reason
>I'm using the official one is because it seems to have more of the recent
>opcodes. Hmm, it probably would have been easier just to add the opcodes
>to the unnoficial one.
>
>Larry Troxler
>