mjohnson/espeak-ng
1
0
Fork 0
mirror of https://github.com/rhasspy/espeak-ng.git synced 2026-10-31 05:35:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use pcaudiolib for the audio API layer.

Browse Source
This commit is contained in:
Reece H. Dunn
2016-02-07 19:34:08 +00:00
parent bb465a45d8
commit 4f676ed175
9 changed files with 61 additions and 1819 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+5 -2
View File
@@ -20,7 +20,7 @@ build:
* Build the code with a C99 compiler, instead of a C++ compiler.
* Provide a pkg-config file (patch by Luke Yelavich).
* Use -fPIC to support sparc/sparc64 architectures.
* Use the system's portaudio header files.
* Removed the local portaudio header files.
* Use the system's sonic library and header files.
* Output phoneme compilation errors to stderr.
* Generate build failures if building phoneme, intonation or dictionary files
@@ -44,7 +44,10 @@ restructuring:
* Converted the documentation to markdown.
* Group the Windows and POSIX mbrowrap code to provide the `mbrowrap.h`
implementation in a single place.
* Use the `wave_*` APIs in synchronous playback mode.
* Replaced the audio APIs with PCAudioLib to improve portability of the audio
and to share that across different projects.
* Reworked the synchronous audio to share the code paths with asynchronous
audio.
cleanup:
Makefile.am
+8 -31
View File
@@ -102,10 +102,14 @@ espeak_ng_include_HEADERS = \
lib_LTLIBRARIES += src/libespeak-ng.la
src_libespeak_ng_la_LDFLAGS = -version-info $(SHARED_VERSION) -lpthread -lm
src_libespeak_ng_la_LDFLAGS = -version-info $(SHARED_VERSION) -lpthread -lm \
${PCAUDIOLIB_LIBS}
src_libespeak_ng_la_CFLAGS = -Isrc/include -Isrc/include/compat \
-fPIC -fvisibility=hidden -D _BSD_SOURCE -D_DEFAULT_SOURCE -D _POSIX_C_SOURCE=200112L \
-pedantic -fno-exceptions -D PATH_ESPEAK_DATA=\"$(DATADIR)\" -DLIBESPEAK_NG_EXPORT
-pedantic -fno-exceptions -D PATH_ESPEAK_DATA=\"$(DATADIR)\" -DLIBESPEAK_NG_EXPORT \
${PCAUDIOLIB_CFLAGS}
src_libespeak_ng_la_SOURCES = \
src/libespeak-ng/compiledata.c \
src/libespeak-ng/compiledict.c \
@@ -147,33 +151,6 @@ src_libespeak_ng_la_SOURCES += \
src/libespeak-ng/fifo.c
endif
if AUDIO_RUNTIME
src_libespeak_ng_la_LDFLAGS += -lpulse -lpulse-simple -lportaudio
src_libespeak_ng_la_CFLAGS += -DUSE_PULSEAUDIO -DUSE_PORTAUDIO
src_libespeak_ng_la_SOURCES += \
src/libespeak-ng/wave.c \
src/libespeak-ng/wave_pulse.c
else
if AUDIO_PULSEAUDIO
src_libespeak_ng_la_LDFLAGS += -lpulse
src_libespeak_ng_la_CFLAGS += -DUSE_PULSEAUDIO
src_libespeak_ng_la_SOURCES += src/libespeak-ng/wave_pulse.c
else
if AUDIO_PORTAUDIO
src_libespeak_ng_la_LDFLAGS += -lportaudio
src_libespeak_ng_la_CFLAGS += -DUSE_PORTAUDIO
src_libespeak_ng_la_SOURCES += src/libespeak-ng/wave.c
else
if AUDIO_SADA
src_libespeak_ng_la_CFLAGS += -DUSE_SADA
src_libespeak_ng_la_SOURCES += src/libespeak-ng/wave_sada.c
else
src_libespeak_ng_la_SOURCES += src/libespeak-ng/wave.c
endif
endif
endif
endif
bin_PROGRAMS += src/speak-ng
if HAVE_RONN
@@ -181,7 +158,7 @@ man1_MANS += src/speak-ng.1
endif
src_speak_ng_LDADD = src/libespeak-ng.la
src_speak_ng_LDFLAGS = -static -lm
src_speak_ng_LDFLAGS = -static -lm ${PCAUDIOLIB_LIBS}
src_speak_ng_CFLAGS = -Isrc/libespeak-ng -Isrc/include -D _POSIX_C_SOURCE=200112L
src_speak_ng_SOURCES = src/speak-ng.c
@@ -191,7 +168,7 @@ if HAVE_RONN
man1_MANS += src/espeak-ng.1
endif
src_espeak_ng_LDADD = src/libespeak-ng.la
src_espeak_ng_LDADD = src/libespeak-ng.la ${PCAUDIOLIB_LIBS}
src_espeak_ng_CFLAGS = -Isrc/include
src_espeak_ng_SOURCES = src/espeak-ng.c
README.md
+2 -5
View File
@@ -38,8 +38,8 @@ In order to build eSpeak NG, you need:
Optionally, you need:
1. the pulseaudio development library to enable pulseaudio output;
2. the portaudio development library to enable portaudio output;
1. the [pcaudiolib](https://github.com/rhdunn/pcaudiolib) development library
to enable audio output;
3. the [sonic](https://github.com/waywardgeek/sonic) development library to
enable sonic audio speed up support;
4. the `ronn` man-page markdown processor to build the man pages.
@@ -61,9 +61,6 @@ Optional dependencies:
| Dependency | Install |
|--------------|-----------------------------------------|
| pulseaudio | `sudo apt-get install libpulse-dev` |
| portaudio 18 | `sudo apt-get install libportaudio-dev` |
| portaudio 19 | `sudo apt-get install portaudio19-dev` |
| sonic | `sudo apt-get install libsonic-dev` |
| ronn | `sudo apt-get install ruby-ronn` |
configure.ac
+15 -93
View File
@@ -104,104 +104,29 @@ AC_CHECK_FUNCS([strrchr])
AC_CHECK_FUNCS([strstr])
dnl ================================================================
dnl PulseAudio checks.
dnl PCAudioLib checks.
dnl ================================================================
AC_ARG_WITH([pulseaudio],
[AS_HELP_STRING([--with-pulseaudio], [use the pulseaudio library for audio output @<:@default=yes@:>@])],
AC_ARG_WITH([pcaudiolib],
[AS_HELP_STRING([--with-pcaudiolib], [use the pcaudiolib library for audio output @<:@default=yes@:>@])],
[])
if test "$with_pulseaudio" = "no"; then
echo "Disabling pulseaudio output support via pulseaudio"
have_pulseaudio=no
if test "$with_pcaudiolib" = "no"; then
echo "Disabling audio output support via pcaudiolib"
have_pcaudiolib=no
else
PKG_CHECK_MODULES(PULSEAUDIO, [libpulse >= 0.9],
AC_CHECK_HEADERS([pcaudiolib/audio.h],
[
have_pulseaudio=yes
have_pcaudiolib=yes
PCAUDIOLIB_CFLAGS=
PCAUDIOLIB_LIBS=-lpcaudio
],[
have_pulseaudio=no
have_pcaudiolib=no
])
fi
dnl ================================================================
dnl PortAudio checks.
dnl ================================================================
AC_ARG_WITH([portaudio],
[AS_HELP_STRING([--with-portaudio], [use the portaudio library for audio output @<:@default=yes@:>@])],
[])
if test "$with_portaudio" = "no"; then
echo "Disabling portaudio output support via portaudio"
have_portaudio=no
else
AC_CHECK_HEADERS([portaudio.h],
[
# Check the portaudio library.
AC_CHECK_LIB([portaudio], [Pa_IsStreamActive]) # portaudio 19
AC_CHECK_LIB([portaudio], [Pa_StreamActive]) # portaudio 18
# Then use the headers to determine the portaudio version.
# This is because on some systems with both libportaudio0 and
# libportaudio2 installed, portaudio.h and -lportaudio refer
# to different versions.
AC_CHECK_FUNC([Pa_IsStreamActive],
[
have_portaudio=19
],[
AC_CHECK_FUNC([Pa_StreamActive],
[
have_portaudio=18
],[
have_portaudio=no
])
])
],[
have_portaudio=no
])
fi
dnl ================================================================
dnl Audio checks.
dnl ================================================================
AC_ARG_WITH([sada],
[AS_HELP_STRING([--with-sada], [use the Solaris SADA audio API @<:@default=no@:>@])],
[])
if test "$with_sada" = "yes" ; then
have_sada=yes
else
have_sada=no
fi
if test "$have_portaudio" = 18 -o "$have_portaudio" = 19 ; then
if test "$have_pulseaudio" = yes ; then
PKG_CHECK_MODULES(PULSEAUDIO_SIMPLE, [libpulse-simple >= 0.9],
[
have_pulseaudio=yes
AUDIO=runtime
],[
have_pulseaudio=no
AUDIO=portaudio
])
else
AUDIO=portaudio
fi
elif test "$have_pulseaudio" = yes ; then
AUDIO=pulseaudio
elif test "$have_sada" = yes ; then
AUDIO=sada
else
AUDIO=disabled
fi
AC_SUBST(AUDIO)
AM_CONDITIONAL(AUDIO_RUNTIME, [test x"$AUDIO" = xruntime])
AM_CONDITIONAL(AUDIO_PULSEAUDIO, [test x"$AUDIO" = xpulseaudio])
AM_CONDITIONAL(AUDIO_PORTAUDIO, [test x"$AUDIO" = xportaudio])
AM_CONDITIONAL(AUDIO_SADA, [test x"$AUDIO" = xsada])
AC_SUBST(PCAUDIOLIB_CFLAGS)
AC_SUBST(PCAUDIOLIB_LIBS)
dnl ================================================================
dnl Optional compilation checks.
@@ -319,14 +244,11 @@ AC_MSG_NOTICE([
C99 Compiler: ${CC}
C99 Compiler flags: ${CFLAGS}
pulseaudio: ${have_pulseaudio}
portaudio: ${have_portaudio}
sada: ${have_sada}
audio configuration: ${AUDIO}
Sonic: ${have_sonic}
PCAudioLib: ${have_pcaudiolib}
Klatt: ${have_klatt}
MBROLA: ${have_mbrola}
Sonic: ${have_sonic}
Async: ${have_async}
Extended Dictionaries:
src/libespeak-ng/speech.c
+31 -12
View File
@@ -33,6 +33,10 @@
#include <unistd.h>
#include <wchar.h>
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
#include <pcaudiolib/audio.h>
#endif
#if defined(_WIN32) || defined(_WIN64)
#include <fcntl.h>
#include <io.h>
@@ -51,7 +55,6 @@
#include "espeak_command.h"
#include "fifo.h"
#include "event.h"
#include "wave.h"
#ifndef S_ISDIR
#define S_ISDIR(mode) (((mode) & S_IFMT) == S_IFDIR)
@@ -63,7 +66,9 @@ espeak_EVENT *event_list = NULL;
int event_list_ix = 0;
int n_event_list;
long count_samples;
void *my_audio = NULL;
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
struct audio_object *my_audio = NULL;
#endif
static const char *option_device = NULL;
static unsigned int my_unique_identifier = 0;
@@ -107,17 +112,21 @@ static int dispatch_audio(short *outbuf, int length, espeak_EVENT *event)
voice_samplerate = event->id.number;
if (out_samplerate != voice_samplerate) {
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if (out_samplerate != 0) {
// sound was previously open with a different sample rate
wave_close(my_audio);
audio_object_close(my_audio);
sleep(1);
}
#endif
out_samplerate = voice_samplerate;
my_audio = wave_open(voice_samplerate, option_device);
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
audio_object_open(my_audio, AUDIO_OBJECT_FORMAT_S16LE, voice_samplerate, 1);
if (!my_audio) {
err = ENS_AUDIO_ERROR;
return -1;
}
#endif
#ifdef USE_ASYNC
if ((my_mode & ENOUTPUT_MODE_SYNCHRONOUS) == 0)
event_init();
@@ -125,9 +134,10 @@ static int dispatch_audio(short *outbuf, int length, espeak_EVENT *event)
}
}
if (outbuf && length && a_wave_can_be_played) {
wave_write(my_audio, (char *)outbuf, 2*length);
}
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if (outbuf && length && a_wave_can_be_played)
audio_object_write(my_audio, (char *)outbuf, 2*length);
#endif
#ifdef USE_ASYNC
while (event && a_wave_can_be_played) {
@@ -218,9 +228,12 @@ ESPEAK_NG_API espeak_ng_STATUS espeak_ng_InitializeOutput(espeak_ng_OUTPUT_MODE
{
option_device = device;
my_mode = output_mode;
my_audio = NULL;
out_samplerate = 0;
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
my_audio = create_audio_device_object(device, "eSpeak", "Text-to-Speech");
#endif
// buflength is in mS, allocate 2 bytes per sample
if ((buffer_length == 0) || (output_mode & ENOUTPUT_MODE_SPEAK_AUDIO))
buffer_length = 200;
@@ -486,8 +499,10 @@ espeak_ng_STATUS sync_espeak_Synth(unsigned int unique_identifier, const void *t
end_character_position = end_position;
espeak_ng_STATUS aStatus = Synthesize(unique_identifier, text, flags);
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if ((my_mode & ENOUTPUT_MODE_SPEAK_AUDIO) == ENOUTPUT_MODE_SPEAK_AUDIO)
wave_flush(my_audio);
audio_object_drain(my_audio);
#endif
return aStatus;
}
@@ -790,8 +805,10 @@ ESPEAK_NG_API espeak_ng_STATUS espeak_ng_Cancel(void)
event_clear_all();
#endif
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
if ((my_mode & ENOUTPUT_MODE_SPEAK_AUDIO) == ENOUTPUT_MODE_SPEAK_AUDIO)
wave_close(my_audio);
audio_object_close(my_audio);
#endif
embedded_value[EMBED_T] = 0; // reset echo for pronunciation announcements
for (int i = 0; i < N_SPEECH_PARAM; i++)
@@ -832,8 +849,10 @@ ESPEAK_NG_API espeak_ng_STATUS espeak_ng_Terminate(void)
#endif
if ((my_mode & ENOUTPUT_MODE_SPEAK_AUDIO) == ENOUTPUT_MODE_SPEAK_AUDIO) {
wave_close(my_audio);
wave_terminate();
#ifdef HAVE_PCAUDIOLIB_AUDIO_H
audio_object_close(my_audio);
audio_object_destroy(my_audio);
#endif
out_samplerate = 0;
}
src/libespeak-ng/wave.c
-719
View File
@@ -1,719 +0,0 @@
/*
* Copyright (C) 2007, Gilles Casse <gcasse@oralux.org>
* Copyright (C) 2015-2016 Reece H. Dunn
* based on AudioIO.cc (Audacity-1.2.4b) and wavegen.cpp
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see: <http://www.gnu.org/licenses/>.
*/
#include "config.h"
#include <assert.h>
#include <math.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#include <time.h>
#include <unistd.h>
#ifdef PLATFORM_WINDOWS
#include <windows.h>
#endif
#include <espeak-ng/espeak_ng.h>
#include "speech.h"
#include "wave.h"
#ifdef USE_PORTAUDIO
#include "portaudio.h"
#undef USE_PORTAUDIO
// determine portaudio version by looking for a #define which is not in V18
#ifdef paNeverDropInput
#define USE_PORTAUDIO 19
#else
#define USE_PORTAUDIO 18
#endif
#ifdef USE_PULSEAUDIO
// create some wrappers for runtime detection
// checked on wave_open
static int pulse_running;
// wave.cpp (this file)
void *wave_port_open(int, const char *);
size_t wave_port_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize);
int wave_port_close(void *theHandler);
void wave_port_terminate();
void wave_port_flush(void *theHandler);
void *wave_port_test_get_write_buffer();
// wave_pulse.cpp
int is_pulse_running();
void *wave_pulse_open(int, const char *);
size_t wave_pulse_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize);
int wave_pulse_close(void *theHandler);
void wave_pulse_terminate();
void wave_pulse_flush(void *theHandler);
void *wave_pulse_test_get_write_buffer();
// wrappers
void *wave_open(int srate, const char *device)
{
pulse_running = is_pulse_running();
if (pulse_running)
return wave_pulse_open(srate, device);
else
return wave_port_open(srate, device);
}
size_t wave_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize)
{
if (pulse_running)
return wave_pulse_write(theHandler, theMono16BitsWaveBuffer, theSize);
else
return wave_port_write(theHandler, theMono16BitsWaveBuffer, theSize);
}
int wave_close(void *theHandler)
{
if (pulse_running)
return wave_pulse_close(theHandler);
else
return wave_port_close(theHandler);
}
void wave_terminate()
{
if (pulse_running)
wave_pulse_terminate();
else
wave_port_terminate();
}
void wave_flush(void *theHandler)
{
if (pulse_running)
wave_pulse_flush(theHandler);
else
wave_port_flush(theHandler);
}
// rename functions to be wrapped
#define wave_open wave_port_open
#define wave_write wave_port_write
#define wave_close wave_port_close
#define wave_terminate wave_port_terminate
#define wave_flush wave_port_flush
#endif
#define MAX_SAMPLE_RATE 22050
#define FRAMES_PER_BUFFER 512
#define BUFFER_LENGTH (MAX_SAMPLE_RATE*2*sizeof(uint16_t))
static char myBuffer[BUFFER_LENGTH];
static char *myRead = NULL;
static char *myWrite = NULL;
static int out_channels = 1;
static int my_stream_could_start = 0;
static int wave_samplerate;
static int mInCallbackFinishedState = false;
#if (USE_PORTAUDIO == 18)
static PaDeviceID myOutputDevice = 0;
static PortAudioStream *pa_stream = NULL;
#endif
#if (USE_PORTAUDIO == 19)
static struct PaStreamParameters myOutputParameters;
static PaStream *pa_stream = NULL;
#endif
static int userdata[4];
static PaError pa_init_err = 0;
// time measurement
// The read and write position audio stream in the audio stream are measured in ms.
//
// * When the stream is opened, myReadPosition and myWritePosition are cleared.
// * myWritePosition is updated in wave_write.
// * myReadPosition is updated in pa_callback (+ sample delay).
static uint32_t myReadPosition = 0; // in ms
static uint32_t myWritePosition = 0;
static void init_buffer()
{
myWrite = myBuffer;
myRead = myBuffer;
memset(myBuffer, 0, BUFFER_LENGTH);
myReadPosition = myWritePosition = 0;
}
static unsigned int get_used_mem()
{
char *aRead = myRead;
char *aWrite = myWrite;
unsigned int used = 0;
assert((aRead >= myBuffer)
&& (aRead <= myBuffer + BUFFER_LENGTH)
&& (aWrite >= myBuffer)
&& (aWrite <= myBuffer + BUFFER_LENGTH));
if (aRead < aWrite)
used = aWrite - aRead;
else
used = aWrite + BUFFER_LENGTH - aRead;
return used;
}
static PaError start_stream()
{
PaError err;
my_stream_could_start = 0;
mInCallbackFinishedState = false;
err = Pa_StartStream(pa_stream);
#if USE_PORTAUDIO == 19
if (err == paStreamIsNotStopped) {
// not sure why we need this, but PA v19 seems to need it
err = Pa_StopStream(pa_stream);
err = Pa_StartStream(pa_stream);
}
#endif
return err;
}
/* This routine will be called by the PortAudio engine when audio is needed.
** It may called at interrupt level on some machines so don't do anything
** that could mess up the system like calling malloc() or free().
*/
#if USE_PORTAUDIO == 18
static int pa_callback(void *inputBuffer, void *outputBuffer,
unsigned long framesPerBuffer, PaTimestamp outTime, void *userData)
#else
static int pa_callback(const void *inputBuffer, void *outputBuffer,
long unsigned int framesPerBuffer, const PaStreamCallbackTimeInfo *outTime,
PaStreamCallbackFlags flags, void *userData)
#endif
{
(void)inputBuffer; // unused
(void)outTime; // unused
(void)userData; // unused
int aResult = 0; // paContinue
char *aWrite = myWrite;
size_t n = out_channels*sizeof(uint16_t)*framesPerBuffer;
myReadPosition += framesPerBuffer;
if (aWrite >= myRead) {
if ((size_t)(aWrite - myRead) >= n) {
memcpy(outputBuffer, myRead, n);
myRead += n;
} else {
// underflow
aResult = 1; // paComplete;
mInCallbackFinishedState = true;
size_t aUsedMem = 0;
aUsedMem = (size_t)(aWrite - myRead);
if (aUsedMem)
memcpy(outputBuffer, myRead, aUsedMem);
char *p = (char *)outputBuffer + aUsedMem;
memset(p, 0, n - aUsedMem);
myRead = aWrite;
}
} else {
if ((size_t)(myBuffer + BUFFER_LENGTH - myRead) >= n) {
memcpy(outputBuffer, myRead, n);
myRead += n;
} else if ((size_t)(aWrite + BUFFER_LENGTH - myRead) >= n) {
int aTopMem = myBuffer + BUFFER_LENGTH - myRead;
if (aTopMem)
memcpy(outputBuffer, myRead, aTopMem);
int aRest = n - aTopMem;
if (aRest) {
char *p = (char *)outputBuffer + aTopMem;
memcpy(p, myBuffer, aRest);
}
myRead = myBuffer + aRest;
} else {
// underflow
aResult = 1; // paComplete;
int aTopMem = myBuffer + BUFFER_LENGTH - myRead;
if (aTopMem)
memcpy(outputBuffer, myRead, aTopMem);
int aRest = aWrite - myBuffer;
if (aRest) {
char *p = (char *)outputBuffer + aTopMem;
memcpy(p, myBuffer, aRest);
}
size_t aUsedMem = aTopMem + aRest;
char *p = (char *)outputBuffer + aUsedMem;
memset(p, 0, n - aUsedMem);
myRead = aWrite;
}
}
#ifdef ARCH_BIG
// BIG-ENDIAN, swap the order of bytes in each sound sample in the portaudio buffer
int c;
unsigned char *out_ptr;
unsigned char *out_end;
out_ptr = (unsigned char *)outputBuffer;
out_end = out_ptr + framesPerBuffer*2 * out_channels;
while (out_ptr < out_end) {
c = out_ptr[0];
out_ptr[0] = out_ptr[1];
out_ptr[1] = c;
out_ptr += 2;
}
#endif
return aResult;
}
void wave_flush(void *theHandler)
{
(void)theHandler; // unused
if (my_stream_could_start)
start_stream();
}
static int wave_open_sound()
{
PaError err = paNoError;
PaError active;
#if USE_PORTAUDIO == 18
active = Pa_StreamActive(pa_stream);
#else
active = Pa_IsStreamActive(pa_stream);
#endif
if (active == 1)
return 0;
if (active < 0) {
out_channels = 1;
#if USE_PORTAUDIO == 18
err = Pa_OpenStream(&pa_stream,
// capture parameters
paNoDevice,
0,
paInt16,
NULL,
// playback parameters
myOutputDevice,
out_channels,
paInt16,
NULL,
// general parameters
wave_samplerate, FRAMES_PER_BUFFER, 0,
paNoFlag,
pa_callback, (void *)userdata);
if (err == paInvalidChannelCount) {
// failed to open with mono, try stereo
out_channels = 2;
err = Pa_OpenStream(&pa_stream,
// capture parameters
paNoDevice,
0,
paInt16,
NULL,
// playback parameters
myOutputDevice,
out_channels,
paInt16,
NULL,
// general parameters
wave_samplerate, FRAMES_PER_BUFFER, 0,
paNoFlag,
pa_callback, (void *)userdata);
}
mInCallbackFinishedState = false; // v18 only
#else
myOutputParameters.channelCount = out_channels;
unsigned long framesPerBuffer = paFramesPerBufferUnspecified;
err = Pa_OpenStream(&pa_stream,
NULL, // no input
&myOutputParameters,
wave_samplerate,
framesPerBuffer,
paNoFlag,
pa_callback,
(void *)userdata);
if ((err != paNoError)
&& (err != paInvalidChannelCount)) {
fprintf(stderr, "wave_open_sound > Pa_OpenStream : err=%d (%s)\n", err, Pa_GetErrorText(err));
framesPerBuffer = FRAMES_PER_BUFFER;
err = Pa_OpenStream(&pa_stream,
NULL, // no input
&myOutputParameters,
wave_samplerate,
framesPerBuffer,
paNoFlag,
pa_callback,
(void *)userdata);
}
if (err == paInvalidChannelCount) {
// failed to open with mono, try stereo
out_channels = 2;
myOutputParameters.channelCount = out_channels;
err = Pa_OpenStream(&pa_stream,
NULL, // no input
&myOutputParameters,
wave_samplerate,
framesPerBuffer,
paNoFlag,
pa_callback,
(void *)userdata);
}
mInCallbackFinishedState = false;
#endif
}
return err != paNoError;
}
static void update_output_parameters(int selectedDevice, const PaDeviceInfo *deviceInfo)
{
#if (USE_PORTAUDIO == 19)
myOutputParameters.device = selectedDevice;
myOutputParameters.channelCount = 1;
myOutputParameters.sampleFormat = paInt16;
// Latency greater than 100ms for avoiding glitches
// (e.g. when moving a window in a graphical desktop)
// deviceInfo = Pa_GetDeviceInfo(selectedDevice);
if (deviceInfo) {
double aLatency = deviceInfo->defaultLowOutputLatency;
myOutputParameters.suggestedLatency = aLatency; // for faster response ?
} else
myOutputParameters.suggestedLatency = (double)0.1; // 100ms
myOutputParameters.hostApiSpecificStreamInfo = NULL;
#else
myOutputDevice = selectedDevice;
#endif
}
static const PaDeviceInfo *select_device(const char *device)
{
#if (USE_PORTAUDIO == 19)
int numDevices = Pa_GetDeviceCount();
#else
int numDevices = Pa_CountDevices();
#endif
if (numDevices < 0)
return NULL;
#if (USE_PORTAUDIO == 19)
PaDeviceIndex i = 0, selectedIndex = 0;
#else
PaDeviceID i = 0, selectedIndex = 0;
#endif
const PaDeviceInfo *deviceInfo = NULL;
const PaDeviceInfo *selectedDeviceInfo = NULL;
if (device == NULL) {
#if (USE_PORTAUDIO == 19)
selectedIndex = Pa_GetDefaultOutputDevice();
#else
selectedIndex = Pa_GetDefaultOutputDeviceID();
#endif
selectedDeviceInfo = Pa_GetDeviceInfo(selectedIndex);
}
if (selectedDeviceInfo == NULL) {
for (i = 0; i < numDevices; i++) {
deviceInfo = Pa_GetDeviceInfo(i);
if (deviceInfo != NULL && !strcmp(device, deviceInfo->name)) {
selectedIndex = i;
selectedDeviceInfo = deviceInfo;
}
}
}
if (selectedDeviceInfo)
update_output_parameters(selectedIndex, selectedDeviceInfo);
return selectedDeviceInfo;
}
void *wave_open(int srate, const char *device)
{
PaError err;
pa_stream = NULL;
wave_samplerate = srate;
mInCallbackFinishedState = false;
init_buffer();
// PortAudio sound output library
err = Pa_Initialize();
pa_init_err = err;
if (err != paNoError)
return NULL;
static int once = 0;
if (!once) {
if (!select_device(device))
return NULL;
once = 1;
}
return (void *)1;
}
static size_t copyBuffer(char *dest, char *src, const size_t theSizeInBytes)
{
size_t bytes_written = 0;
unsigned int i = 0;
uint16_t *a_dest = NULL;
uint16_t *a_src = NULL;
if ((src != NULL) && dest != NULL) {
// copy for one channel (mono)?
if (out_channels == 1) {
memcpy(dest, src, theSizeInBytes);
bytes_written = theSizeInBytes;
} else { // copy for 2 channels (stereo)
a_dest = (uint16_t *)dest;
a_src = (uint16_t *)src;
for (i = 0; i < theSizeInBytes/2; i++) {
a_dest[2*i] = a_src[i];
a_dest[2*i+1] = a_src[i];
}
bytes_written = 2*theSizeInBytes;
}
}
return bytes_written;
}
size_t wave_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize)
{
(void)theHandler; // unused
size_t bytes_written = 0;
// space in ringbuffer for the sample needed: 1x mono channel but 2x for 1 stereo channel
size_t bytes_to_write = (out_channels == 1) ? theSize : theSize*2;
my_stream_could_start = 0;
if (pa_stream == NULL) {
if (0 != wave_open_sound())
return 0;
my_stream_could_start = 1;
}
assert(BUFFER_LENGTH >= bytes_to_write);
if (myWrite >= myBuffer + BUFFER_LENGTH)
myWrite = myBuffer;
size_t aTotalFreeMem = 0;
char *aRead;
while (1) {
aRead = myRead;
// write pointer is before read pointer?
if (myWrite >= aRead)
aTotalFreeMem = aRead + BUFFER_LENGTH - myWrite;
else // read pointer is before write pointer!
aTotalFreeMem = aRead - myWrite;
if (aTotalFreeMem > 1) {
// -1 because myWrite must be different of aRead
// otherwise buffer would be considered as empty
aTotalFreeMem -= 1;
}
if (aTotalFreeMem >= bytes_to_write)
break;
usleep(10000);
}
aRead = myRead;
// write pointer is ahead the read pointer?
if (myWrite >= aRead) {
// determine remaining free memory to the end of the ringbuffer
size_t aFreeMem = myBuffer + BUFFER_LENGTH - myWrite;
// is enough linear space available (regardless 1 or 2 channels)?
if (aFreeMem >= bytes_to_write) {
// copy direct - no wrap around at end of ringbuffer needed
myWrite += copyBuffer(myWrite, theMono16BitsWaveBuffer, theSize);
} else { // not enough linear space available
// 2 channels (stereo)?
if (out_channels == 2) {
// copy with wrap around at the end of ringbuffer
copyBuffer(myWrite, theMono16BitsWaveBuffer, aFreeMem/2);
myWrite = myBuffer;
myWrite += copyBuffer(myWrite, theMono16BitsWaveBuffer+aFreeMem/2, theSize - aFreeMem/2);
} else { // 1 channel (mono)
// copy with wrap around at the end of ringbuffer
copyBuffer(myWrite, theMono16BitsWaveBuffer, aFreeMem);
myWrite = myBuffer;
myWrite += copyBuffer(myWrite, theMono16BitsWaveBuffer+aFreeMem, theSize - aFreeMem);
}
}
} else // read pointer is ahead the write pointer
myWrite += copyBuffer(myWrite, theMono16BitsWaveBuffer, theSize);
bytes_written = bytes_to_write;
myWritePosition += theSize/sizeof(uint16_t); // add number of samples
if (my_stream_could_start && (get_used_mem() >= out_channels * sizeof(uint16_t) * FRAMES_PER_BUFFER))
start_stream();
return bytes_written;
}
int wave_close(void *theHandler)
{
(void)theHandler; // unused
static int aStopStreamCount = 0;
#if (USE_PORTAUDIO == 19)
if (pa_stream == NULL)
return 0;
if (Pa_IsStreamStopped(pa_stream))
return 0;
#else
if (pa_stream == NULL)
return 0;
if (Pa_StreamActive(pa_stream) == false && mInCallbackFinishedState == false)
return 0;
#endif
// Avoid race condition by making sure this function only
// gets called once at a time
aStopStreamCount++;
if (aStopStreamCount != 1)
return 0;
// Comment from Audacity-1.2.4b adapted to the eSpeak context.
//
// We got here in one of two ways:
//
// 1. The calling program calls the espeak_Cancel function and we
// therefore want to stop as quickly as possible.
// So we use AbortStream(). If this is
// the case the portaudio stream is still in the Running state
// (see PortAudio state machine docs).
//
// 2. The callback told PortAudio to stop the stream since it had
// reached the end of the selection.
// The event polling thread discovered this by noticing that
// wave_is_busy() returned false.
// wave_is_busy() (which calls Pa_GetStreamActive()) will not return
// false until all buffers have finished playing, so we can call
// AbortStream without losing any samples. If this is the case
// we are in the "callback finished state" (see PortAudio state
// machine docs).
//
// The moral of the story: We can call AbortStream safely, without
// losing samples.
//
// DMM: This doesn't seem to be true; it seems to be necessary to
// call StopStream if the callback brought us here, and AbortStream
// if the user brought us here.
#if (USE_PORTAUDIO == 19)
if (pa_stream) {
Pa_AbortStream(pa_stream);
Pa_CloseStream(pa_stream);
pa_stream = NULL;
mInCallbackFinishedState = false;
}
#else
if (pa_stream) {
if (mInCallbackFinishedState) {
Pa_StopStream(pa_stream);
} else {
Pa_AbortStream(pa_stream);
}
Pa_CloseStream(pa_stream);
pa_stream = NULL;
mInCallbackFinishedState = false;
}
#endif
init_buffer();
aStopStreamCount = 0; // last action
return 0;
}
void wave_terminate()
{
Pa_Terminate();
}
#else
void *wave_open(int srate, const char *device)
{
(void)srate; // unused
(void)device; // unused
return (void *)1;
}
size_t wave_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize)
{
(void)theHandler; // unused
(void)theMono16BitsWaveBuffer; // unused
return theSize;
}
int wave_close(void *theHandler)
{
(void)theHandler; // unused
return 0;
}
void wave_terminate()
{
}
void wave_flush(void *theHandler)
{
(void)theHandler; // unused
}
#endif
src/libespeak-ng/wave.h
-38
View File
@@ -1,38 +0,0 @@
/*
* Copyright (C) 2007, Gilles Casse <gcasse@oralux.org>
* Copyright (C) 2015-2016 Reece H. Dunn
* based on AudioIO.cc (Audacity-1.2.4b) and wavegen.cpp
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see: <http://www.gnu.org/licenses/>.
*/
#ifndef WAVE_H
#define WAVE_H
#ifdef __cplusplus
extern "C"
{
#endif
extern void *wave_open(int samplerate, const char *device);
extern size_t wave_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize);
extern int wave_close(void *theHandler);
extern void wave_flush(void *theHandler);
extern void wave_terminate();
#ifdef __cplusplus
}
#endif
#endif
src/libespeak-ng/wave_pulse.c
-587
View File
@@ -1,587 +0,0 @@
/*
* Copyright (C) 2007, Gilles Casse <gcasse@oralux.org>
* Copyright (C) 2015-2016 Reece H. Dunn
* eSpeak driver for PulseAudio
* based on the XMMS PulseAudio Plugin
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see: <http://www.gnu.org/licenses/>.
*/
// TBD:
// * ARCH_BIG
// * uint64? a_timing_info.read_index
// * prebuf,... size?
// * 0.9.6: pb pulse_free using tlength=8820 (max size never returned -> tlength=10000 ok, but higher drain).
//
#include "config.h"
#include <assert.h>
#include <math.h>
#include <pthread.h>
#include <pulse/pulseaudio.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#include <time.h>
#include <unistd.h>
#include <espeak-ng/espeak_ng.h>
#include "speech.h"
#include "wave.h"
enum {
/* return value */
PULSE_OK = 0,
PULSE_ERROR = -1,
PULSE_NO_CONNECTION = -2
};
#ifdef USE_PULSEAUDIO
#define SAMPLE_RATE 22050
#define ESPEAK_FORMAT PA_SAMPLE_S16LE
#define ESPEAK_CHANNEL 1
#define MAXLENGTH 132300
#define TLENGTH 4410
#define PREBUF 2200
#define MINREQ 880
#ifdef USE_PORTAUDIO
// rename functions to be wrapped
#define wave_open wave_pulse_open
#define wave_write wave_pulse_write
#define wave_close wave_pulse_close
#define wave_terminate wave_pulse_terminate
#define wave_flush wave_pulse_flush
// check whether we can connect to PulseAudio
#include <pulse/simple.h>
int is_pulse_running()
{
pa_sample_spec ss;
ss.format = ESPEAK_FORMAT;
ss.rate = SAMPLE_RATE;
ss.channels = ESPEAK_CHANNEL;
pa_simple *s = pa_simple_new(NULL, "eSpeak", PA_STREAM_PLAYBACK, NULL, "is_pulse_running", &ss, NULL, NULL, NULL);
if (s) {
pa_simple_free(s);
return 1;
} else
return 0;
}
#endif
static pthread_mutex_t pulse_mutex;
static pa_context *context = NULL;
static pa_stream *stream = NULL;
static pa_threaded_mainloop *mainloop = NULL;
static int do_trigger = 0;
static uint64_t written = 0;
static int time_offset_msec = 0;
static int just_flushed = 0;
static int connected = 0;
static int wave_samplerate;
#define CHECK_DEAD_GOTO(label, warn) do { \
if (!mainloop || \
!context || pa_context_get_state(context) != PA_CONTEXT_READY || \
!stream || pa_stream_get_state(stream) != PA_STREAM_READY) { \
if (warn) \
fprintf(stderr, "Connection died: %s\n", context ? pa_strerror(pa_context_errno(context)) : "NULL"); \
goto label; \
} \
} while (0);
#define CHECK_CONNECTED(retval) \
do { \
if (!connected) return retval; \
} while (0);
#define CHECK_CONNECTED_NO_RETVAL(id) \
do { \
if (!connected) { return; } \
} while (0);
static void subscribe_cb(struct pa_context *c, enum pa_subscription_event_type t, uint32_t index, void *userdata)
{
(void)userdata; // unused
assert(c);
if (!stream ||
index != pa_stream_get_index(stream) ||
(t != (PA_SUBSCRIPTION_EVENT_SINK_INPUT|PA_SUBSCRIPTION_EVENT_CHANGE) &&
t != (PA_SUBSCRIPTION_EVENT_SINK_INPUT|PA_SUBSCRIPTION_EVENT_NEW)))
return;
}
static void context_state_cb(pa_context *c, void *userdata)
{
(void)userdata; // unused
assert(c);
switch (pa_context_get_state(c))
{
case PA_CONTEXT_READY:
case PA_CONTEXT_TERMINATED:
case PA_CONTEXT_FAILED:
pa_threaded_mainloop_signal(mainloop, 0);
break;
case PA_CONTEXT_UNCONNECTED:
case PA_CONTEXT_CONNECTING:
case PA_CONTEXT_AUTHORIZING:
case PA_CONTEXT_SETTING_NAME:
break;
}
}
static void stream_state_cb(pa_stream *s, void *userdata)
{
(void)userdata; // unused
assert(s);
switch (pa_stream_get_state(s))
{
case PA_STREAM_READY:
case PA_STREAM_FAILED:
case PA_STREAM_TERMINATED:
pa_threaded_mainloop_signal(mainloop, 0);
break;
case PA_STREAM_UNCONNECTED:
case PA_STREAM_CREATING:
break;
}
}
static void stream_success_cb(pa_stream *s, int success, void *userdata)
{
assert(s);
if (userdata)
*(int *)userdata = success;
pa_threaded_mainloop_signal(mainloop, 0);
}
static void context_success_cb(pa_context *c, int success, void *userdata)
{
assert(c);
if (userdata)
*(int *)userdata = success;
pa_threaded_mainloop_signal(mainloop, 0);
}
static void stream_request_cb(pa_stream *s, size_t length, void *userdata)
{
(void)length; // unused
(void)userdata; // unused
assert(s);
pa_threaded_mainloop_signal(mainloop, 0);
}
static void stream_latency_update_cb(pa_stream *s, void *userdata)
{
(void)userdata; // unused
assert(s);
pa_threaded_mainloop_signal(mainloop, 0);
}
static int pulse_free(void)
{
size_t l = 0;
pa_operation *o = NULL;
CHECK_CONNECTED(0);
pa_threaded_mainloop_lock(mainloop);
CHECK_DEAD_GOTO(fail, 1);
if ((l = pa_stream_writable_size(stream)) == (size_t)-1) {
fprintf(stderr, "pa_stream_writable_size() failed: %s", pa_strerror(pa_context_errno(context)));
l = 0;
goto fail;
}
/* If this function is called twice with no pulse_write() call in
* between this means we should trigger the playback */
if (do_trigger) {
int success = 0;
if (!(o = pa_stream_trigger(stream, stream_success_cb, &success))) {
fprintf(stderr, "pa_stream_trigger() failed: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
while (pa_operation_get_state(o) != PA_OPERATION_DONE) {
CHECK_DEAD_GOTO(fail, 1);
pa_threaded_mainloop_wait(mainloop);
}
if (!success)
fprintf(stderr, "pa_stream_trigger() failed: %s", pa_strerror(pa_context_errno(context)));
}
fail:
if (o)
pa_operation_unref(o);
pa_threaded_mainloop_unlock(mainloop);
do_trigger = !!l;
return (int)l;
}
static int pulse_playing(const pa_timing_info *the_timing_info)
{
int r = 0;
const pa_timing_info *i;
assert(the_timing_info);
CHECK_CONNECTED(0);
pa_threaded_mainloop_lock(mainloop);
for (;;) {
CHECK_DEAD_GOTO(fail, 1);
if ((i = pa_stream_get_timing_info(stream)))
break;
if (pa_context_errno(context) != PA_ERR_NODATA) {
fprintf(stderr, "pa_stream_get_timing_info() failed: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
pa_threaded_mainloop_wait(mainloop);
}
r = i->playing;
memcpy((void *)the_timing_info, (void *)i, sizeof(pa_timing_info));
fail:
pa_threaded_mainloop_unlock(mainloop);
return r;
}
static void pulse_write(void *ptr, int length)
{
CHECK_CONNECTED_NO_RETVAL();
pa_threaded_mainloop_lock(mainloop);
CHECK_DEAD_GOTO(fail, 1);
if (pa_stream_write(stream, ptr, length, NULL, PA_SEEK_RELATIVE, (pa_seek_mode_t)0) < 0) {
fprintf(stderr, "pa_stream_write() failed: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
do_trigger = 0;
written += length;
fail:
pa_threaded_mainloop_unlock(mainloop);
}
static int drain(void)
{
pa_operation *o = NULL;
int success = 0;
int ret = PULSE_ERROR;
CHECK_CONNECTED(ret);
pa_threaded_mainloop_lock(mainloop);
CHECK_DEAD_GOTO(fail, 0);
if (!(o = pa_stream_drain(stream, stream_success_cb, &success))) {
fprintf(stderr, "pa_stream_drain() failed: %s\n", pa_strerror(pa_context_errno(context)));
goto fail;
}
while (pa_operation_get_state(o) != PA_OPERATION_DONE) {
CHECK_DEAD_GOTO(fail, 1);
pa_threaded_mainloop_wait(mainloop);
}
if (!success)
fprintf(stderr, "pa_stream_drain() failed: %s\n", pa_strerror(pa_context_errno(context)));
else
ret = PULSE_OK;
fail:
if (o)
pa_operation_unref(o);
pa_threaded_mainloop_unlock(mainloop);
return ret;
}
static void pulse_close(void)
{
drain();
connected = 0;
if (mainloop)
pa_threaded_mainloop_stop(mainloop);
connected = 0;
if (context) {
pa_context_disconnect(context);
pa_context_unref(context);
context = NULL;
}
if (mainloop) {
pa_threaded_mainloop_free(mainloop);
mainloop = NULL;
}
}
static int pulse_open(const char *device)
{
pa_sample_spec ss;
pa_operation *o = NULL;
int success;
int ret = PULSE_ERROR;
assert(!mainloop);
assert(!context);
assert(!stream);
assert(!connected);
pthread_mutex_init(&pulse_mutex, (const pthread_mutexattr_t *)NULL);
ss.format = ESPEAK_FORMAT;
ss.rate = wave_samplerate;
ss.channels = ESPEAK_CHANNEL;
if (!pa_sample_spec_valid(&ss))
return false;
if (!(mainloop = pa_threaded_mainloop_new()))
goto fail;
pa_threaded_mainloop_lock(mainloop);
if (!(context = pa_context_new(pa_threaded_mainloop_get_api(mainloop), "eSpeak")))
goto fail;
pa_context_set_state_callback(context, context_state_cb, NULL);
pa_context_set_subscribe_callback(context, subscribe_cb, NULL);
if (pa_context_connect(context, NULL, (pa_context_flags_t)0, NULL) < 0) {
fprintf(stderr, "Failed to connect to server: %s", pa_strerror(pa_context_errno(context)));
ret = PULSE_NO_CONNECTION;
goto fail;
}
if (pa_threaded_mainloop_start(mainloop) < 0)
goto fail;
// Wait until the context is ready
pa_threaded_mainloop_wait(mainloop);
if (pa_context_get_state(context) != PA_CONTEXT_READY) {
fprintf(stderr, "Failed to connect to server: %s", pa_strerror(pa_context_errno(context)));
ret = PULSE_NO_CONNECTION;
if (mainloop)
pa_threaded_mainloop_stop(mainloop);
goto fail;
}
if (!(stream = pa_stream_new(context, "unknown", &ss, NULL))) {
fprintf(stderr, "Failed to create stream: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
pa_stream_set_state_callback(stream, stream_state_cb, NULL);
pa_stream_set_write_callback(stream, stream_request_cb, NULL);
pa_stream_set_latency_update_callback(stream, stream_latency_update_cb, NULL);
pa_buffer_attr a_attr;
a_attr.maxlength = MAXLENGTH;
a_attr.tlength = TLENGTH;
a_attr.prebuf = PREBUF;
a_attr.minreq = MINREQ;
a_attr.fragsize = 0;
if (pa_stream_connect_playback(stream, device, &a_attr, (pa_stream_flags_t)(PA_STREAM_INTERPOLATE_TIMING|PA_STREAM_AUTO_TIMING_UPDATE), NULL, NULL) < 0) {
fprintf(stderr, "Failed to connect stream: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
// Wait until the stream is ready
pa_threaded_mainloop_wait(mainloop);
if (pa_stream_get_state(stream) != PA_STREAM_READY) {
fprintf(stderr, "Failed to connect stream: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
// Now subscribe to events
if (!(o = pa_context_subscribe(context, PA_SUBSCRIPTION_MASK_SINK_INPUT, context_success_cb, &success))) {
fprintf(stderr, "pa_context_subscribe() failed: %s", pa_strerror(pa_context_errno(context)));
goto fail;
}
while (pa_operation_get_state(o) != PA_OPERATION_DONE) {
CHECK_DEAD_GOTO(fail, 1);
pa_threaded_mainloop_wait(mainloop);
}
pa_operation_unref(o);
do_trigger = 0;
written = 0;
time_offset_msec = 0;
just_flushed = 0;
connected = 1;
pa_threaded_mainloop_unlock(mainloop);
return PULSE_OK;
fail:
if (mainloop)
pa_threaded_mainloop_unlock(mainloop);
if (ret == PULSE_NO_CONNECTION) {
if (context) {
pa_context_disconnect(context);
pa_context_unref(context);
context = NULL;
}
if (mainloop) {
pa_threaded_mainloop_free(mainloop);
mainloop = NULL;
}
} else
pulse_close();
return ret;
}
void wave_flush(void *theHandler)
{
(void)theHandler; // unused
}
void *wave_open(int srate, const char *device)
{
stream = NULL;
wave_samplerate = srate;
if (pulse_open(device) != PULSE_OK)
return NULL;
return (void *)1;
}
size_t wave_write(void *theHandler, char *theMono16BitsWaveBuffer, size_t theSize)
{
(void)theHandler; // unused
size_t bytes_to_write = theSize;
char *aBuffer = theMono16BitsWaveBuffer;
assert(stream);
size_t aTotalFreeMem = 0;
pthread_mutex_lock(&pulse_mutex);
while (1) {
aTotalFreeMem = pulse_free();
if (aTotalFreeMem >= bytes_to_write)
break;
// TBD: check if really helpful
if (aTotalFreeMem >= MAXLENGTH*2)
aTotalFreeMem = MAXLENGTH*2;
// 500: threshold for avoiding too many calls to pulse_write
if (aTotalFreeMem > 500) {
pulse_write(aBuffer, aTotalFreeMem);
bytes_to_write -= aTotalFreeMem;
aBuffer += aTotalFreeMem;
}
usleep(10000);
}
pulse_write(aBuffer, bytes_to_write);
terminate:
pthread_mutex_unlock(&pulse_mutex);
return theSize;
}
int wave_close(void *theHandler)
{
(void)theHandler; // unused
static int aStopStreamCount = 0;
// Avoid race condition by making sure this function only
// gets called once at a time
aStopStreamCount++;
if (aStopStreamCount != 1)
return 0;
int a_status = pthread_mutex_lock(&pulse_mutex);
if (a_status) {
aStopStreamCount = 0; // last action
return PULSE_ERROR;
}
drain();
pthread_mutex_unlock(&pulse_mutex);
aStopStreamCount = 0; // last action
return PULSE_OK;
}
void wave_terminate()
{
pthread_mutex_t *a_mutex = NULL;
a_mutex = &pulse_mutex;
pthread_mutex_lock(a_mutex);
pulse_close();
pthread_mutex_unlock(a_mutex);
pthread_mutex_destroy(a_mutex);
}
#endif
src/libespeak-ng/wave_sada.c
-332
View File
@@ -1,332 +0,0 @@
/*
* Copyright (C) 2008, Sun Microsystems, Inc.
* Copyright (C) 2015-2016 Reece H. Dunn
* eSpeak driver for Solaris Audio Device Architecture (SADA)
* Written by Willie Walker, based on the eSpeak PulseAudio driver
* from Gilles Casse
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see: <http://www.gnu.org/licenses/>.
*/
#include "config.h"
#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <stdlib.h>
#include <string.h>
#include <stropts.h>
#include <sys/audioio.h>
#include <unistd.h>
#include <espeak-ng/espeak_ng.h>
#include "speech.h"
#include "wave.h"
#define SAMPLE_RATE 22050
#define SAMPLE_SIZE 16
#ifdef USE_SADA
static const char *sun_audio_device = "/dev/audio";
static int sun_audio_fd = -1;
// The total number of 16-bit samples sent to be played via the
// wave_write method.
//
static uint32_t total_samples_sent;
// The total number of samples sent to be played via the wave_write
// method, but which were never played because of a call to
// wave_close.
//
static uint32_t total_samples_skipped;
// The last known playing index after a call to wave_close.
//
static uint32_t last_play_position = 0;
static uint32_t wave_samplerate;
static int wave_get_remaining_time(uint32_t sample, uint32_t *time);
// wave_open
//
// DESCRIPTION:
//
// initializes the audio subsytem.
//
// GLOBALS USED/MODIFIED:
//
// sun_audio_fd: modified to hold the file descriptor of the opened
// audio device.
//
void *wave_open(int srate, const char *device)
{
if (device == NULL)
device = sun_audio_device;
audio_info_t ainfo;
wave_samplerate = srate;
if ((sun_audio_fd = open(device, O_WRONLY)) < 0) {
fprintf(stderr, "wave_open() could not open: %s (%d)\n", device, sun_audio_fd);
return NULL;
}
ioctl(sun_audio_fd, AUDIO_GETINFO, &ainfo);
ainfo.play.encoding = AUDIO_ENCODING_LINEAR;
ainfo.play.channels = 1;
ainfo.play.sample_rate = wave_samplerate;
ainfo.play.precision = SAMPLE_SIZE;
if (ioctl(sun_audio_fd, AUDIO_SETINFO, &ainfo) == -1) {
fprintf(stderr, "wave_open() failed to set audio params: %s\n", strerror(errno));
close(sun_audio_fd);
return NULL;
}
return (void *)sun_audio_fd;
}
// wave_write
//
// DESCRIPTION:
//
// Meant to be asynchronous, it supplies the wave sample to the lower
// audio layer and returns. The sample is played later on. [[[WDW -
// we purposely do not open the audio device as non-blocking because
// managing that would be a pain. So, we rely a lot upon fifo.cpp and
// event.cpp to not overload us, allowing us to get away with a
// blocking write. event.cpp:polling_thread in particular appears to
// use get_remaining_time to prevent flooding.]]]
//
// PARAMETERS:
//
// theHandler: the audio device file descriptor
// theMono16BitsWaveBuffer: the audio data
// theSize: the number of bytes (not 16-bit samples)
//
// GLOBALS USED/MODIFIED:
//
// total_samples_sent: modified based upon 16-bit samples sent
//
// RETURNS:
//
// the number of bytes (not 16-bit samples) sent
//
size_t wave_write(void *theHandler,
char *theMono16BitsWaveBuffer,
size_t theSize)
{
size_t num;
#if defined(BYTE_ORDER) && BYTE_ORDER == BIG_ENDIAN
// BIG-ENDIAN, swap the order of bytes in each sound sample
int c;
char *out_ptr;
char *out_end;
out_ptr = (char *)theMono16BitsWaveBuffer;
out_end = out_ptr + theSize;
while (out_ptr < out_end) {
c = out_ptr[0];
out_ptr[0] = out_ptr[1];
out_ptr[1] = c;
out_ptr += 2;
}
#endif
num = write((int)theHandler, theMono16BitsWaveBuffer, theSize);
// Keep track of the total number of samples sent -- we use this in
// wave_get_read_position and also use it to help calculate the
// total_samples_skipped in wave_close.
//
total_samples_sent += num / 2;
return num;
}
// wave_close
//
// DESCRIPTION:
//
// Does what SADA normally would call a flush, which means to cease
// all audio production in progress and throw any remaining audio
// away. [[[WDW - see comment in wave_flush.]]]
//
// PARAMETERS:
//
// theHandler: the audio device file descriptor
//
// GLOBALS USED/MODIFIED:
//
// last_play_position: modified to reflect play position the last time
// this method was called
// total_samples_sent: used to help calculate total_samples_skipped
// total_samples_skipped: modified to hold the total number of 16-bit
// samples sent to wave_write, but which were
// never played
// sun_audio_fd: used because some calls to wave_close seem to
// pass a NULL for theHandler for some odd reason
//
// RETURNS:
//
// The result of the ioctl call (non-0 means failure)
//
int wave_close(void *theHandler)
{
int ret;
audio_info_t ainfo;
int audio_fd = (int)theHandler;
if (!audio_fd)
audio_fd = sun_audio_fd;
// [[[WDW: maybe do a pause/resume ioctl???]]]
ret = ioctl(audio_fd, I_FLUSH, FLUSHRW);
ioctl(audio_fd, AUDIO_GETINFO, &ainfo);
// Calculate the number of samples that won't get
// played. We also keep track of the last_play_position
// because wave_close can be called multiple times
// before another call to wave_write.
//
if (last_play_position != ainfo.play.samples) {
last_play_position = ainfo.play.samples;
total_samples_skipped = total_samples_sent - last_play_position;
}
return ret;
}
// wave_is_busy
//
// DESCRIPTION:
//
// Returns a non-0 value if audio is being played.
//
// PARAMETERS:
//
// theHandler: the audio device file descriptor
//
// GLOBALS USED/MODIFIED:
//
// sun_audio_fd: used because some calls to wave_is_busy seem to
// pass a NULL for theHandler for some odd reason
//
// RETURNS:
//
// A non-0 value if audio is being played
//
int wave_is_busy(void *theHandler)
{
(void)theHandler; // unused
uint32_t time;
if (total_samples_sent >= 1)
wave_get_remaining_time(total_samples_sent - 1, &time);
else
time = 0;
return time != 0;
}
// wave_terminate
//
// DESCRIPTION:
//
// Used to end our session with eSpeak.
//
// GLOBALS USED/MODIFIED:
//
// sun_audio_fd: modified - closed and set to -1
//
void wave_terminate()
{
close(sun_audio_fd);
sun_audio_fd = -1;
}
// wave_flush
//
// DESCRIPTION:
//
// Appears to want to tell the audio subsystem to make sure it plays
// the audio. In our case, the system is already doing this, so this
// is basically a no-op. [[[WDW - if you do a drain, you block, so
// don't do that. In addition the typical SADA notion of flush is
// currently handled by wave_close. I think this is most likely just
// terminology conflict between eSpeak and SADA.]]]
//
// PARAMETERS:
//
// theHandler: the audio device file descriptor
//
void wave_flush(void *theHandler)
{
(void)theHandler; // unused
}
// wave_get_remaining_time
//
// DESCRIPTION:
//
// Returns the remaining time (in ms) before the sample is played.
// The sample in this case is a return value from a previous call to
// wave_get_write_position.
//
// PARAMETERS:
//
// sample: an index returned from wave_get_write_position representing
// an index into the long continuous stream of 16-bit samples
// time: a return value representing the delay in milliseconds until
// sample is played. A value of 0 means the sample is either
// currently being played or it has already been played.
//
// GLOBALS USED/MODIFIED:
//
// sun_audio_fd: used to determine total number of samples played by
// the audio system
// total_samples_skipped: used in remaining time calculation
//
// RETURNS:
//
// Time in milliseconds before the sample is played or 0 if the sample
// is currently playing or has already been played.
//
int wave_get_remaining_time(uint32_t sample, uint32_t *time)
{
uint32_t a_time = 0;
uint32_t actual_index;
audio_info_t ainfo;
if (!time)
return -1;
ioctl(sun_audio_fd, AUDIO_GETINFO, &ainfo);
// See if this sample has already been played or is currently
// playing.
//
actual_index = sample - total_samples_skipped;
if ((sample < total_samples_skipped) ||
(actual_index <= ainfo.play.samples))
*time = 0;
else {
a_time = ((actual_index - ainfo.play.samples) * 1000) / wave_samplerate;
*time = (uint32_t)a_time;
}
return 0;
}
#endif