I was trying to PCM sound a little bit in Python, but all the packages I've tried were too much or were undocumented or dead, so I decided to make a simple one with libao.
I used as start point this source code by xiph.org that plays 440Hz for 1 second, and I compiled it with gcc -o ao_example ao_example.c -lao -ldl -lm and I run successfully this code, immediately hearing a 440Hz sine for 1 second in both channels.
So far, so good.
So I $ cp ao_exemple.c mySoundAo.c and I edited mySoundAo.c to be compiled as Python module. The full code follows:
#include <math.h>
#include <stdio.h>
#include <string.h>
#include <ao/ao.h>
#include <Python.h>
#define BUF_SIZE 4096
static PyObject* py_soundAo(PyObject* self, PyObject* args)
{
ao_device *device;
ao_sample_format format;
int default_driver;
char *buffer;
int buf_size;
int sample;
float freq = 440.0;
int i;
/* -- Initialize -- */
fprintf(stderr, "libao example program\n");
ao_initialize();
/* -- Setup for default driver -- */
default_driver = ao_default_driver_id();
memset(&format, 0, sizeof(format));
format.bits = 16;
format.channels = 2;
format.rate = 44100;
format.byte_format = AO_FMT_LITTLE;
/* -- Open driver -- */
device = ao_open_live(default_driver, &format, NULL /* no options */);
if (device == NULL) {
fprintf(stderr, "Error opening device.\n");
return Py_BuildValue("", 0);
}
/* -- Play some stuff -- */
buf_size = format.bits/8 * format.channels * format.rate;
buffer = calloc(buf_size,
sizeof(char));
for (i = 0; i < format.rate; i++) {
sample = (int)(0.75 * 32768.0 * sin(2 * M_PI * freq * ((float) i/format.rate)));
/* Put the same stuff in left and right channel */
buffer[4*i] = buffer[4*i+2] = sample & 0xff;
buffer[4*i+1] = buffer[4*i+3] = (sample >> 8) & 0xff;
}
ao_play(device, buffer, buf_size);
/* -- Close and shutdown -- */
ao_close(device);
ao_shutdown();
return Py_BuildValue("", 0);
}
static PyMethodDef mySoundAo_methods[] = {
{"soundAo", py_soundAo, METH_VARARGS},
{NULL, NULL}
};
void initmySoundAo()
{
(void) Py_InitModule("mySoundAo", mySoundAo_methods);
}
So I compiled as gcc -shared -I/usr/include/python2.7/ -o mySoundAo.so mySoundAo.c -lpython2.7 -lm -lsndfile -lao -ldl and I had this warning:
In file included from /usr/include/python2.7/Python.h:8:0,
from mySoundAo.c:5:
/usr/include/python2.7/pyconfig.h:1158:0: warning: "_POSIX_C_SOURCE" redefined [enabled by default]
/usr/include/features.h:214:0: note: this is the location of the previous definition
Doesn't sound much dangerous, so I moved on.
In python, I did the following:
$ python
Python 2.7.2+ (default, Oct 4 2011, 20:03:08)
[GCC 4.6.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import mySoundAo
>>> mySoundAo.soundAo()
libao example program
Error opening device.
>>>
With no sound. Inspecting the code a little bit, I discovered the function ao_initialize(); hangs for about 4 second and the following line default_driver = ao_default_driver_id(); sets this variable to -1 (error).
This behavior is strange, since it's almost the same code.
So, any ideas to make this work?
Thank you!
The warning you get is harmless, simply moving #include <Python.h> to the top should let the standard library correctly recognize the macro is already defined.
The problem is probably caused by a miscompiled /usr/lib/ao/plugins-4/libalsa.so (this file is mentioned if you set debug in ~/.libao.conf). Since the ao's alsa plugin fails to load, ao tries all other options, and exhausts the nas timeout of 4 seconds (that's the cause for the delay).
To check whether the miscompiled (or mislinked) libalsa.so is the problem, run
$ ldd -r /usr/lib/ao/plugins-4/libalsa.so > /dev/null
undefined symbol: ao_is_big_endian (/usr/lib/ao/plugins-4/libalsa.so)
The error in the output should points to a problem with the symbol. You can simply download libao yourself, and patch out the line in libao-*/src/plugins/alsa/ao_alsa.c, or copy the definition from ao_is_big_endian, or fix the linking.