[fluid-dev] Re: FluidGUI on OS X

[Ebrahim Mayat]

On May 24, 2004, at 6:28 PM, Ken Ellinwood wrote:

> It might not be a simple matter to enable the TCP command port for 
> OS-X.  The command port logic is currently written to assume that all 
> I/O with the client can be accomplished via read() and write() using 
> the file descriptor returned from the socket open call.   This 
> definetly does not work on Windows, where send() and recv() must be 
> used.

The read() function which is defined to read input can be used on OS X. 
It is part of libc and defined in the unistd.h, sys/types.h and 
sys/uio.h headers. The write() API allows you to communicate with other 
users, by copying lines from your terminal to theirs. So yes, the 
read() and write() functions can be used on OS X.

>  The question is, what does the socket API look like on OS-X?  Will 
> read() and write() work to perform I/O over the socket?

The socket() function call on OS X is similar to its BSD 
implementation. So, off the cuff, yes, it should work together with 
read() and write() for I/O. As to the precise application of socket(), 
I haven't had any experience since I am not a programmer but a musician 
who is interested in musical software. :-)

>  If yes, then simply rebuild fluidsynth with the TCP server enabled 
> (in config_macosx_pb.h, comment out the #define WITHOUT_SERVER line).

I'll give this a go.

One other important question: what do you precisely mean by the term 
"TCP command port"? As with Linux, any OS X process can spawn a child 
process and execute any command/program in that subprocess.  This is 
only limited by file system permissions.  Inter-process communication 
occurs via local domain sockets, pipes, or network sockets (e.g., TCP). 
In most cases though, this communication occurs through pipes rather 
than through TCP/IP.

> At one point I had planned on providing a patch for fluidsynth that 
> would introduce the necessary polymorphism required in order to 
> support the command shell on the terminal and over sockets. Its not a 
> small task, and now that Peter has started an effort to provide a JNI 
> interface to fluidsynth, its more likely that a true cross-platform 
> implementation of FluidGUI will rely on the JNI interface and not the 
> TCP command shell interface.

Is JNI some kind of a Java superclass?

Regards.
Ebrahim