ardour/gtk2_ardour/main.cc

458 lines
10 KiB
C++
Raw Normal View History

/*
Copyright (C) 2001-2006 Paul Davis
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 2 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, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
$Id$
*/
#include <sys/types.h>
#include <sys/mman.h>
#include <sys/wait.h>
#include <cerrno>
#include <cstdlib>
#include <signal.h>
#include <unistd.h>
#include <sigc++/bind.h>
#include <gtkmm/settings.h>
#include <pbd/error.h>
#include <pbd/textreceiver.h>
#include <pbd/failed_constructor.h>
#include <pbd/pthread_utils.h>
#include <jack/jack.h>
#include <ardour/version.h>
#include <ardour/ardour.h>
#include <ardour/audioengine.h>
#include <gtkmm/main.h>
#include <gtkmm2ext/popup.h>
#include <gtkmm2ext/utils.h>
#include "version.h"
#include "ardour_ui.h"
#include "opts.h"
#include "i18n.h"
using namespace Gtk;
using namespace GTK_ARDOUR;
using namespace ARDOUR;
Large nasty commit in the form of a 5000 line patch chock-full of completely unecessary changes. (Sorry, doing a "sprint" based thing, this is the end of the first one) Achieved MIDI track and bus creation, associated Jack port and diskstream creation, and minimal GUI stuff for creating them. Should be set to start work on actually recording and playing midi to/from disk now. Relevant (significant) changes: - Creation of a Buffer class. Base class is type agnostic so things can point to a buffer but not care what kind it is (otherwise it'd be a template). Derived into AudioBuffer and MidiBuffer, with a type tag because checking type is necessary in parts of the code where dynamic_cast wouldn't be wise. Originally I considered this a hack, but passing around a type proved to be a very good solution to all the other problems (below). There is a 1:1 mapping between jack port data types and ardour Buffer types (with a conversion function), but that's easily removed if it ever becomes necessary. Having the type scoped in the Buffer class is maybe not the best spot for it, but whatever (this is proof of concept kinda stuff right now...) - IO now has a "default" port type (passed to the constructor and stored as a member), used by ensure_io (and similar) to create n ports. IO::register_***_port has a type argument that defaults to the default type if not passed. Rationale: previous IO API is identical, no changes needed to existing code, but path is paved for multiple port types in one IO, which we will need for eg synth plugin inserts, among other things. This is not quite ideal (best would be to only have the two port register functions and have them take a type), but the alternative is a lot of work (namely destroying the 'ensure' functions and everything that uses them) for very little gain. (I am convinced after quite a few tries at the whiteboard that subclassing IO in any way is not a feasible option, look at it's inheritance diagram in Doxygen and you can see why) - AudioEngine::register_audio_input_port is now register_input_port and takes a type argument. Ditto for output. - (Most significant change) AudioDiskstream abstracted into Distream, and sibling MidiDiskstream created. Very much still a work in progress, but Diskstream is there to switch references over to (most already are), which is the important part. It is still unclear what the MIDI diskstream's relation to channels is, but I'm pretty sure they will be single channel only (so SMF Type 0) since noone can come up with a reason otherwise. - MidiTrack creation. Same thing as AudioTrack but with a different default type basically. No big deal here. - Random cleanups and variable renamings etc. because I have OCD and can't help myself. :) Known broken: Loading of sessions containing MIDI tracks. git-svn-id: svn://localhost/ardour2/branches/midi@641 d708f5d6-7413-0410-9779-e7cbd77b26cf
2006-06-26 12:01:34 -04:00
using namespace PBD;
using namespace sigc;
TextReceiver text_receiver ("ardour");
extern int curvetest (string);
static ARDOUR_UI *ui = 0;
static void
shutdown (int status)
{
char* msg;
if (status) {
msg = _("ardour is killing itself for a clean exit\n");
write (1, msg, strlen (msg));
/* drastic, but perhaps necessary */
kill (-getpgrp(), SIGKILL);
/*NOTREACHED*/
} else {
if (ui) {
msg = _("stopping user interface\n");
write (1, msg, strlen (msg));
ui->kill();
}
pthread_cancel_all ();
}
exit (status);
}
static void
handler (int sig)
{
char buf[64];
int n;
/* XXX its doubtful that snprintf() is async-safe */
n = snprintf (buf, sizeof(buf), _("%d(%d): received signal %d\n"), getpid(), (int) pthread_self(), sig);
write (1, buf, n);
shutdown (1);
}
static void *
signal_thread (void *arg)
{
int sig;
sigset_t blocked;
PBD::ThreadCreated (pthread_self(), X_("Signal"));
pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
/* find out what's blocked right now */
//sigprocmask (SIG_SETMASK, 0, &blocked);
if (pthread_sigmask (SIG_SETMASK, 0, &blocked)) {
cerr << "getting blocked signals failed\n";
}
/* wait for any of the currently blocked signals.
According to the man page found in linux 2.6 and 2.4, sigwait()
never returns an error. This is incorrect. Checking the man
pages for some other *nix systems makes it clear that
sigwait() can return several error codes, one of which
is EINTR. This happens if the thread receives a signal
which is not in the blocked set.
We do not expect that to happen, and if it did we should generally
exit as planned. However, under 2.6, the ptrace facility used
by gdb seems to also cause sigwait() to return with EINTR
but with a signal that sigwait cannot understand. As a result,
"sig" is set to zero, an impossible signal number.
Handling the EINTR code makes it possible to debug
ardour on a 2.6 kernel.
*/
int swerr;
again:
if ((swerr = sigwait (&blocked, &sig))) {
if (swerr == EINTR) {
goto again;
} else {
cerr << "sigwait failed with " << swerr << endl;
}
}
cerr << "Signal " << sig << " received\n";
if (sig != SIGSEGV) {
/* unblock signals so we can see them during shutdown.
this will help prod developers not to lose sight
of bugs that cause segfaults etc. during shutdown.
*/
sigprocmask (SIG_UNBLOCK, &blocked, 0);
}
shutdown (1);
/*NOTREACHED*/
return 0;
}
int
catch_signals (void)
{
struct sigaction action;
pthread_t signal_thread_id;
sigset_t signals;
// if (setpgid (0,0)) {
if (setsid ()) {
warning << string_compose (_("cannot become new process group leader (%1)"),
strerror (errno))
<< endmsg;
}
sigemptyset (&signals);
sigaddset(&signals, SIGHUP);
sigaddset(&signals, SIGINT);
sigaddset(&signals, SIGQUIT);
sigaddset(&signals, SIGPIPE);
sigaddset(&signals, SIGTERM);
sigaddset(&signals, SIGUSR1);
sigaddset(&signals, SIGUSR2);
/* install a handler because otherwise
pthreads behaviour is undefined when we enter
sigwait.
*/
action.sa_handler = handler;
action.sa_mask = signals;
action.sa_flags = SA_RESTART|SA_RESETHAND;
for (int i = 1; i < 32; i++) {
if (sigismember (&signals, i)) {
if (sigaction (i, &action, 0)) {
cerr << string_compose (_("cannot setup signal handling for %1"), i) << endl;
return -1;
}
}
}
/* this sets the signal mask for this and all
subsequent threads that do not reset it.
*/
if (pthread_sigmask (SIG_SETMASK, &signals, 0)) {
cerr << string_compose (_("cannot set default signal mask (%1)"), strerror (errno)) << endl;
return -1;
}
/* start a thread to wait for signals */
if (pthread_create_and_store ("signal", &signal_thread_id, 0, signal_thread, 0)) {
cerr << "cannot create signal catching thread" << endl;
return -1;
}
pthread_detach (signal_thread_id);
return 0;
}
string
which_ui_rcfile ()
{
string rcfile;
char* env;
if ((env = getenv ("ARDOUR2_UI_RC")) != 0 && strlen (env)) {
rcfile = env;
} else {
rcfile = "ardour2_ui.rc";
}
rcfile = find_config_file (rcfile);
if (rcfile.empty()) {
warning << _("Without a UI style file, ardour will look strange.\n Please set ARDOUR2_UI_RC to point to a valid UI style file") << endmsg;
} else {
cerr << "Loading ui configuration file " << rcfile << endl;
}
return rcfile;
}
gint
show_ui_callback (void *arg)
{
ARDOUR_UI * ui = (ARDOUR_UI *) arg;
ui->hide_splash();
ui->show ();
return FALSE;
}
void
gui_jack_error ()
{
MessageDialog win (_("Ardour could not connect to JACK."),
false,
Gtk::MESSAGE_INFO,
(Gtk::ButtonsType)(Gtk::BUTTONS_NONE));
win.set_secondary_text(_("There are several possible reasons:\n\
\n\
1) JACK is not running.\n\
2) JACK is running as another user, perhaps root.\n\
3) There is already another client called \"ardour\".\n\
\n\
Please consider the possibilities, and perhaps (re)start JACK."));
win.add_button (Stock::QUIT, RESPONSE_CLOSE);
win.set_default_response (RESPONSE_CLOSE);
win.show_all ();
win.set_position (Gtk::WIN_POS_CENTER);
/* we just don't care about the result */
win.run ();
}
static bool
maybe_load_session ()
{
/* If no session name is given: we're not loading a session yet, nor creating a new one */
if (!session_name.length()) {
ui->hide_splash ();
if (!Config->get_no_new_session_dialog()) {
ui->new_session (true);
}
return true;
}
/* Load session or start the new session dialog */
string name, path;
bool isnew;
if (Session::find_session (session_name, path, name, isnew)) {
error << string_compose(_("could not load command line session \"%1\""), session_name) << endmsg;
return false;
}
if (!new_session) {
/* Loading a session, but the session doesn't exist */
if (isnew) {
error << string_compose (_("\n\nNo session named \"%1\" exists.\n\
To create it from the command line, start ardour as \"ardour --new %1"), path) << endmsg;
return false;
}
ui->load_session (path, name);
} else {
/* TODO: This bit of code doesn't work properly yet
Glib::signal_idle().connect (bind (mem_fun (*ui, &ARDOUR_UI::cmdline_new_session), path));
ui->set_will_create_new_session_automatically (true); */
/* Show the NSD */
ui->hide_splash ();
if (!Config->get_no_new_session_dialog()) {
ui->new_session (true);
}
}
return true;
}
Large nasty commit in the form of a 5000 line patch chock-full of completely unecessary changes. (Sorry, doing a "sprint" based thing, this is the end of the first one) Achieved MIDI track and bus creation, associated Jack port and diskstream creation, and minimal GUI stuff for creating them. Should be set to start work on actually recording and playing midi to/from disk now. Relevant (significant) changes: - Creation of a Buffer class. Base class is type agnostic so things can point to a buffer but not care what kind it is (otherwise it'd be a template). Derived into AudioBuffer and MidiBuffer, with a type tag because checking type is necessary in parts of the code where dynamic_cast wouldn't be wise. Originally I considered this a hack, but passing around a type proved to be a very good solution to all the other problems (below). There is a 1:1 mapping between jack port data types and ardour Buffer types (with a conversion function), but that's easily removed if it ever becomes necessary. Having the type scoped in the Buffer class is maybe not the best spot for it, but whatever (this is proof of concept kinda stuff right now...) - IO now has a "default" port type (passed to the constructor and stored as a member), used by ensure_io (and similar) to create n ports. IO::register_***_port has a type argument that defaults to the default type if not passed. Rationale: previous IO API is identical, no changes needed to existing code, but path is paved for multiple port types in one IO, which we will need for eg synth plugin inserts, among other things. This is not quite ideal (best would be to only have the two port register functions and have them take a type), but the alternative is a lot of work (namely destroying the 'ensure' functions and everything that uses them) for very little gain. (I am convinced after quite a few tries at the whiteboard that subclassing IO in any way is not a feasible option, look at it's inheritance diagram in Doxygen and you can see why) - AudioEngine::register_audio_input_port is now register_input_port and takes a type argument. Ditto for output. - (Most significant change) AudioDiskstream abstracted into Distream, and sibling MidiDiskstream created. Very much still a work in progress, but Diskstream is there to switch references over to (most already are), which is the important part. It is still unclear what the MIDI diskstream's relation to channels is, but I'm pretty sure they will be single channel only (so SMF Type 0) since noone can come up with a reason otherwise. - MidiTrack creation. Same thing as AudioTrack but with a different default type basically. No big deal here. - Random cleanups and variable renamings etc. because I have OCD and can't help myself. :) Known broken: Loading of sessions containing MIDI tracks. git-svn-id: svn://localhost/ardour2/branches/midi@641 d708f5d6-7413-0410-9779-e7cbd77b26cf
2006-06-26 12:01:34 -04:00
#ifdef VST_SUPPORT
/* this is called from the entry point of a wine-compiled
executable that is linked against gtk2_ardour built
as a shared library.
*/
extern "C" {
int ardour_main (int argc, char *argv[])
#else
int main (int argc, char *argv[])
#endif
{
ARDOUR::AudioEngine *engine;
vector<Glib::ustring> null_file_list;
// needs a better home.
Glib::thread_init();
gtk_set_locale ();
(void) bindtextdomain (PACKAGE, LOCALEDIR);
(void) textdomain (PACKAGE);
pthread_setcanceltype (PTHREAD_CANCEL_ASYNCHRONOUS, 0);
// catch_signals ();
text_receiver.listen_to (error);
text_receiver.listen_to (info);
text_receiver.listen_to (fatal);
text_receiver.listen_to (warning);
if (parse_opts (argc, argv)) {
exit (1);
}
if (curvetest_file) {
return curvetest (curvetest_file);
}
cout << _("Ardour/GTK ")
<< VERSIONSTRING
<< _("\n (built using ")
<< gtk_ardour_major_version << '.'
<< gtk_ardour_minor_version << '.'
<< gtk_ardour_micro_version
<< _(" with libardour ")
<< libardour_major_version << '.'
<< libardour_minor_version << '.'
<< libardour_micro_version
#ifdef __GNUC__
<< _(" and GCC version ") << __VERSION__
#endif
<< ')'
<< endl;
if (just_version) {
exit (0);
}
if (no_splash) {
cerr << _("Copyright (C) 1999-2006 Paul Davis") << endl
<< _("Some portions Copyright (C) Steve Harris, Ari Johnson, Brett Viren, Joel Baker") << endl
<< endl
<< _("Ardour comes with ABSOLUTELY NO WARRANTY") << endl
<< _("not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.") << endl
<< _("This is free software, and you are welcome to redistribute it ") << endl
<< _("under certain conditions; see the source for copying conditions.")
<< endl;
}
try {
ui = new ARDOUR_UI (&argc, &argv, which_ui_rcfile());
}
catch (failed_constructor& err) {
error << _("could not create ARDOUR GUI") << endmsg;
exit (1);
}
if (!no_splash) {
ui->show_splash ();
if (session_name.length()) {
gtk_timeout_add (4000, show_ui_callback, ui);
}
}
try {
engine = new ARDOUR::AudioEngine (jack_client_name);
Large nasty commit in the form of a 5000 line patch chock-full of completely unecessary changes. (Sorry, doing a "sprint" based thing, this is the end of the first one) Achieved MIDI track and bus creation, associated Jack port and diskstream creation, and minimal GUI stuff for creating them. Should be set to start work on actually recording and playing midi to/from disk now. Relevant (significant) changes: - Creation of a Buffer class. Base class is type agnostic so things can point to a buffer but not care what kind it is (otherwise it'd be a template). Derived into AudioBuffer and MidiBuffer, with a type tag because checking type is necessary in parts of the code where dynamic_cast wouldn't be wise. Originally I considered this a hack, but passing around a type proved to be a very good solution to all the other problems (below). There is a 1:1 mapping between jack port data types and ardour Buffer types (with a conversion function), but that's easily removed if it ever becomes necessary. Having the type scoped in the Buffer class is maybe not the best spot for it, but whatever (this is proof of concept kinda stuff right now...) - IO now has a "default" port type (passed to the constructor and stored as a member), used by ensure_io (and similar) to create n ports. IO::register_***_port has a type argument that defaults to the default type if not passed. Rationale: previous IO API is identical, no changes needed to existing code, but path is paved for multiple port types in one IO, which we will need for eg synth plugin inserts, among other things. This is not quite ideal (best would be to only have the two port register functions and have them take a type), but the alternative is a lot of work (namely destroying the 'ensure' functions and everything that uses them) for very little gain. (I am convinced after quite a few tries at the whiteboard that subclassing IO in any way is not a feasible option, look at it's inheritance diagram in Doxygen and you can see why) - AudioEngine::register_audio_input_port is now register_input_port and takes a type argument. Ditto for output. - (Most significant change) AudioDiskstream abstracted into Distream, and sibling MidiDiskstream created. Very much still a work in progress, but Diskstream is there to switch references over to (most already are), which is the important part. It is still unclear what the MIDI diskstream's relation to channels is, but I'm pretty sure they will be single channel only (so SMF Type 0) since noone can come up with a reason otherwise. - MidiTrack creation. Same thing as AudioTrack but with a different default type basically. No big deal here. - Random cleanups and variable renamings etc. because I have OCD and can't help myself. :) Known broken: Loading of sessions containing MIDI tracks. git-svn-id: svn://localhost/ardour2/branches/midi@641 d708f5d6-7413-0410-9779-e7cbd77b26cf
2006-06-26 12:01:34 -04:00
ARDOUR::init (*engine, use_vst, try_hw_optimization);
ui->set_engine (*engine);
} catch (AudioEngine::NoBackendAvailable& err) {
gui_jack_error ();
error << string_compose (_("Could not connect to JACK server as \"%1\""), jack_client_name) << endmsg;
return -1;
} catch (failed_constructor& err) {
error << _("could not initialize Ardour.") << endmsg;
return -1;
}
if (maybe_load_session ()) {
ui->run (text_receiver);
ui = 0;
}
delete engine;
ARDOUR::cleanup ();
shutdown (0);
Large nasty commit in the form of a 5000 line patch chock-full of completely unecessary changes. (Sorry, doing a "sprint" based thing, this is the end of the first one) Achieved MIDI track and bus creation, associated Jack port and diskstream creation, and minimal GUI stuff for creating them. Should be set to start work on actually recording and playing midi to/from disk now. Relevant (significant) changes: - Creation of a Buffer class. Base class is type agnostic so things can point to a buffer but not care what kind it is (otherwise it'd be a template). Derived into AudioBuffer and MidiBuffer, with a type tag because checking type is necessary in parts of the code where dynamic_cast wouldn't be wise. Originally I considered this a hack, but passing around a type proved to be a very good solution to all the other problems (below). There is a 1:1 mapping between jack port data types and ardour Buffer types (with a conversion function), but that's easily removed if it ever becomes necessary. Having the type scoped in the Buffer class is maybe not the best spot for it, but whatever (this is proof of concept kinda stuff right now...) - IO now has a "default" port type (passed to the constructor and stored as a member), used by ensure_io (and similar) to create n ports. IO::register_***_port has a type argument that defaults to the default type if not passed. Rationale: previous IO API is identical, no changes needed to existing code, but path is paved for multiple port types in one IO, which we will need for eg synth plugin inserts, among other things. This is not quite ideal (best would be to only have the two port register functions and have them take a type), but the alternative is a lot of work (namely destroying the 'ensure' functions and everything that uses them) for very little gain. (I am convinced after quite a few tries at the whiteboard that subclassing IO in any way is not a feasible option, look at it's inheritance diagram in Doxygen and you can see why) - AudioEngine::register_audio_input_port is now register_input_port and takes a type argument. Ditto for output. - (Most significant change) AudioDiskstream abstracted into Distream, and sibling MidiDiskstream created. Very much still a work in progress, but Diskstream is there to switch references over to (most already are), which is the important part. It is still unclear what the MIDI diskstream's relation to channels is, but I'm pretty sure they will be single channel only (so SMF Type 0) since noone can come up with a reason otherwise. - MidiTrack creation. Same thing as AudioTrack but with a different default type basically. No big deal here. - Random cleanups and variable renamings etc. because I have OCD and can't help myself. :) Known broken: Loading of sessions containing MIDI tracks. git-svn-id: svn://localhost/ardour2/branches/midi@641 d708f5d6-7413-0410-9779-e7cbd77b26cf
2006-06-26 12:01:34 -04:00
return 0;
}
Large nasty commit in the form of a 5000 line patch chock-full of completely unecessary changes. (Sorry, doing a "sprint" based thing, this is the end of the first one) Achieved MIDI track and bus creation, associated Jack port and diskstream creation, and minimal GUI stuff for creating them. Should be set to start work on actually recording and playing midi to/from disk now. Relevant (significant) changes: - Creation of a Buffer class. Base class is type agnostic so things can point to a buffer but not care what kind it is (otherwise it'd be a template). Derived into AudioBuffer and MidiBuffer, with a type tag because checking type is necessary in parts of the code where dynamic_cast wouldn't be wise. Originally I considered this a hack, but passing around a type proved to be a very good solution to all the other problems (below). There is a 1:1 mapping between jack port data types and ardour Buffer types (with a conversion function), but that's easily removed if it ever becomes necessary. Having the type scoped in the Buffer class is maybe not the best spot for it, but whatever (this is proof of concept kinda stuff right now...) - IO now has a "default" port type (passed to the constructor and stored as a member), used by ensure_io (and similar) to create n ports. IO::register_***_port has a type argument that defaults to the default type if not passed. Rationale: previous IO API is identical, no changes needed to existing code, but path is paved for multiple port types in one IO, which we will need for eg synth plugin inserts, among other things. This is not quite ideal (best would be to only have the two port register functions and have them take a type), but the alternative is a lot of work (namely destroying the 'ensure' functions and everything that uses them) for very little gain. (I am convinced after quite a few tries at the whiteboard that subclassing IO in any way is not a feasible option, look at it's inheritance diagram in Doxygen and you can see why) - AudioEngine::register_audio_input_port is now register_input_port and takes a type argument. Ditto for output. - (Most significant change) AudioDiskstream abstracted into Distream, and sibling MidiDiskstream created. Very much still a work in progress, but Diskstream is there to switch references over to (most already are), which is the important part. It is still unclear what the MIDI diskstream's relation to channels is, but I'm pretty sure they will be single channel only (so SMF Type 0) since noone can come up with a reason otherwise. - MidiTrack creation. Same thing as AudioTrack but with a different default type basically. No big deal here. - Random cleanups and variable renamings etc. because I have OCD and can't help myself. :) Known broken: Loading of sessions containing MIDI tracks. git-svn-id: svn://localhost/ardour2/branches/midi@641 d708f5d6-7413-0410-9779-e7cbd77b26cf
2006-06-26 12:01:34 -04:00
#ifdef VST_SUPPORT
} // end of extern C block
#endif