/**************************************************************************** ** ** Copyright (C) 2009 Nokia Corporation and/or its subsidiary(-ies). ** All rights reserved. ** Contact: Nokia Corporation (qt-info@nokia.com) ** ** This file is part of the Qt3Support module of the Qt Toolkit. ** ** $QT_BEGIN_LICENSE:LGPL$ ** No Commercial Usage ** This file contains pre-release code and may not be distributed. ** You may use this file in accordance with the terms and conditions ** contained in the Technology Preview License Agreement accompanying ** this package. ** ** GNU Lesser General Public License Usage ** Alternatively, this file may be used under the terms of the GNU Lesser ** General Public License version 2.1 as published by the Free Software ** Foundation and appearing in the file LICENSE.LGPL included in the ** packaging of this file. Please review the following information to ** ensure the GNU Lesser General Public License version 2.1 requirements ** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html. ** ** In addition, as a special exception, Nokia gives you certain additional ** rights. These rights are described in the Nokia Qt LGPL Exception ** version 1.1, included in the file LGPL_EXCEPTION.txt in this package. ** ** If you have questions regarding the use of this file, please contact ** Nokia at qt-info@nokia.com. ** ** ** ** ** ** ** ** ** $QT_END_LICENSE$ ** ****************************************************************************/ #include "q3process.h" #ifndef QT_NO_PROCESS #include "qapplication.h" #include "private/q3membuf_p.h" #include #include QT_BEGIN_NAMESPACE //#define QT_Q3PROCESS_DEBUG /*! \class Q3Process \brief The Q3Process class is used to start external programs and to communicate with them. \compat You can write to the started program's standard input, and can read the program's standard output and standard error. You can pass command line arguments to the program either in the constructor or with setArguments() or addArgument(). The program's working directory can be set with setWorkingDirectory(). If you need to set up environment variables pass them to the start() or launch() functions (see below). The processExited() signal is emitted if the program exits. The program's exit status is available from exitStatus(), although you could simply call normalExit() to see if the program terminated normally. There are two different ways to start a process. If you just want to run a program, optionally passing data to its standard input at the beginning, use one of the launch() functions. If you want full control of the program's standard input (especially if you don't know all the data you want to send to standard input at the beginning), use the start() function. If you use start() you can write to the program's standard input using writeToStdin() and you can close the standard input with closeStdin(). The wroteToStdin() signal is emitted if the data sent to standard input has been written. You can read from the program's standard output using readStdout() or readLineStdout(). These functions return an empty QByteArray if there is no data to read. The readyReadStdout() signal is emitted when there is data available to be read from standard output. Standard error has a set of functions that correspond to the standard output functions, i.e. readStderr(), readLineStderr() and readyReadStderr(). If you use one of the launch() functions the data you pass will be sent to the program's standard input which will be closed once all the data has been written. You should \e not use writeToStdin() or closeStdin() if you use launch(). If you need to send data to the program's standard input after it has started running use start() instead of launch(). Both start() and launch() can accept a string list of strings each of which has the format, key=value, where the keys are the names of environment variables. You can test to see if a program is running with isRunning(). The program's process identifier is available from processIdentifier(). If you want to terminate a running program use tryTerminate(), but note that the program may ignore this. If you \e really want to terminate the program, without it having any chance to clean up, you can use kill(). Although you may need quotes for a file named on the command line (e.g. if it contains spaces) you shouldn't use extra quotes for arguments passed to addArgument() or setArguments(). The readyReadStdout() signal is emitted when there is new data on standard output. This happens asynchronously: you don't know if more data will arrive later. In the above example you could connect the processExited() signal to the slot UicManager::readFromStdout() instead. If you do so, you will be certain that all the data is available when the slot is called. On the other hand, you must wait until the process has finished before doing any processing. Note that if you are expecting a lot of output from the process, you may hit platform-dependent limits to the pipe buffer size. The solution is to make sure you connect to the output, e.g. the readyReadStdout() and readyReadStderr() signals and read the data as soon as it becomes available. Please note that Q3Process does not emulate a shell. This means that Q3Process does not do any expansion of arguments: a '*' is passed as a '*' to the program and is \e not replaced by all the files, a '$HOME' is also passed literally and is \e not replaced by the environment variable HOME and the special characters for IO redirection ('>', '|', etc.) are also passed literally and do \e not have the special meaning as they have in a shell. Also note that Q3Process does not emulate a terminal. This means that certain programs which need direct terminal control, do not work as expected with Q3Process. Such programs include console email programs (like pine and mutt) but also programs which require the user to enter a password (like su and ssh). \section1 Notes for Windows users Some Windows commands, for example, \c dir, are not provided by separate applications, but by the command interpreter. If you attempt to use Q3Process to execute these commands directly it won't work. One possible solution is to execute the command interpreter itself (\c cmd.exe on some Windows systems), and ask the interpreter to execute the desired command. Under Windows there are certain problems starting 16-bit applications and capturing their output. Microsoft recommends using an intermediate application to start 16-bit applications. \sa Q3Socket */ /*! \enum Q3Process::Communication This enum type defines the communication channels connected to the process. \value Stdin Data can be written to the process's standard input. \value Stdout Data can be read from the process's standard output. \value Stderr Data can be read from the process's standard error. \value DupStderr Both the process's standard error output \e and its standard output are written to its standard output. (Like Unix's dup2().) This means that nothing is sent to the standard error output. This is especially useful if your application requires that the output on standard output and on standard error must be read in the same order that they are produced. This is a flag, so to activate it you must pass \c{Stdout|Stderr|DupStderr}, or \c{Stdin|Stdout|Stderr|DupStderr} if you want to provide input, to the setCommunication() call. \sa setCommunication() communication() */ /*! Constructs a Q3Process object. The \a parent and \a name parameters are passed to the QObject constructor. \sa setArguments() addArgument() start() */ Q3Process::Q3Process( QObject *parent, const char *name ) : QObject( parent, name ), ioRedirection( false ), notifyOnExit( false ), wroteToStdinConnected( false ), readStdoutCalled( false ), readStderrCalled( false ), comms( Stdin|Stdout|Stderr ) { init(); } /*! Constructs a Q3Process with \a arg0 as the command to be executed. The \a parent and \a name parameters are passed to the QObject constructor. The process is not started. You must call start() or launch() to start the process. \sa setArguments() addArgument() start() */ Q3Process::Q3Process( const QString& arg0, QObject *parent, const char *name ) : QObject( parent, name ), ioRedirection( false ), notifyOnExit( false ), wroteToStdinConnected( false ), readStdoutCalled( false ), readStderrCalled( false ), comms( Stdin|Stdout|Stderr ) { init(); addArgument( arg0 ); } /*! Constructs a Q3Process with \a args as the arguments of the process. The first element in the list is the command to be executed. The other elements in the list are the arguments to this command. The \a parent and \a name parameters are passed to the QObject constructor. The process is not started. You must call start() or launch() to start the process. \sa setArguments() addArgument() start() */ Q3Process::Q3Process( const QStringList& args, QObject *parent, const char *name ) : QObject( parent, name ), ioRedirection( false ), notifyOnExit( false ), wroteToStdinConnected( false ), readStdoutCalled( false ), readStderrCalled( false ), comms( Stdin|Stdout|Stderr ) { init(); setArguments( args ); } /*! \fn Q3Process::~Q3Process() Destroys the instance. If the process is running, it is not terminated! The standard input, standard output and standard error of the process are closed. You can connect the destroyed() signal to the kill() slot, if you want the process to be terminated automatically when the instance is destroyed. \sa tryTerminate() kill() */ /*! Returns the list of arguments that are set for the process. Arguments can be specified with the constructor or with the functions setArguments() and addArgument(). Note that if you want to iterate over the list, you should iterate over a copy, e.g. \snippet doc/src/snippets/code/src_qt3support_other_q3process.cpp 0 \sa setArguments() addArgument() */ QStringList Q3Process::arguments() const { return _arguments; } /*! Clears the list of arguments that are set for the process. \sa setArguments() addArgument() */ void Q3Process::clearArguments() { _arguments.clear(); } /*! Sets \a args as the arguments for the process. The first element in the list is the command to be executed. The other elements in the list are the arguments to the command. Any previous arguments are deleted. Q3Process does not perform argument substitutions; for example, if you specify "*" or "$DISPLAY", these values are passed to the process literally. If you want to have the same behavior as the shell provides, you must do the substitutions yourself; i.e. instead of specifying a "*" you must specify the list of all the filenames in the current directory, and instead of "$DISPLAY" you must specify the value of the environment variable \c DISPLAY. Note for Windows users. The standard Windows shells, e.g. \c command.com and \c cmd.exe, do not perform file globbing, i.e. they do not convert a "*" on the command line into a list of files in the current directory. For this reason most Windows applications implement their own file globbing, and as a result of this, specifying an argument of "*" for a Windows application is likely to result in the application performing a file glob and ending up with a list of filenames. \sa arguments() addArgument() */ void Q3Process::setArguments( const QStringList& args ) { _arguments = args; } /*! Adds \a arg to the end of the list of arguments. The first element in the list of arguments is the command to be executed; the following elements are the command's arguments. \sa arguments() setArguments() */ void Q3Process::addArgument( const QString& arg ) { _arguments.append( arg ); } #ifndef QT_NO_DIR /*! Returns the working directory that was set with setWorkingDirectory(), or the current directory if none has been explicitly set. \sa setWorkingDirectory() QDir::current() */ QDir Q3Process::workingDirectory() const { return workingDir; } /*! Sets \a dir as the working directory for processes. This does not affect running processes; only processes that are started afterwards are affected. Setting the working directory is especially useful for processes that try to access files with relative paths. \sa workingDirectory() start() */ void Q3Process::setWorkingDirectory( const QDir& dir ) { workingDir = dir; } #endif //QT_NO_DIR /*! Returns the communication required with the process, i.e. some combination of the \c Communication flags. \sa setCommunication() */ int Q3Process::communication() const { return comms; } /*! Sets \a commFlags as the communication required with the process. \a commFlags is a bitwise OR of the flags defined by the \c Communication enum. The default is \c{Stdin|Stdout|Stderr}. \sa communication() */ void Q3Process::setCommunication( int commFlags ) { comms = commFlags; } /*! Returns true if the process has exited normally; otherwise returns false. This implies that this function returns false if the process is still running. \sa isRunning() exitStatus() processExited() */ bool Q3Process::normalExit() const { // isRunning() has the side effect that it determines the exit status! if ( isRunning() ) return false; else return exitNormal; } /*! Returns the exit status of the process or 0 if the process is still running. This function returns immediately and does not wait until the process is finished. If normalExit() is false (e.g. if the program was killed or crashed), this function returns 0, so you should check the return value of normalExit() before relying on this value. \sa normalExit() processExited() */ int Q3Process::exitStatus() const { // isRunning() has the side effect that it determines the exit status! if ( isRunning() ) return 0; else return exitStat; } /*! Reads the data that the process has written to standard output. When new data is written to standard output, the class emits the signal readyReadStdout(). If there is no data to read, this function returns a QByteArray of size 0: it does not wait until there is something to read. \sa readyReadStdout() readLineStdout() readStderr() writeToStdin() */ QByteArray Q3Process::readStdout() { if ( readStdoutCalled ) { return QByteArray(); } readStdoutCalled = true; Q3Membuf *buf = membufStdout(); readStdoutCalled = false; return buf->readAll(); } /*! Reads the data that the process has written to standard error. When new data is written to standard error, the class emits the signal readyReadStderr(). If there is no data to read, this function returns a QByteArray of size 0: it does not wait until there is something to read. \sa readyReadStderr() readLineStderr() readStdout() writeToStdin() */ QByteArray Q3Process::readStderr() { if ( readStderrCalled ) { return QByteArray(); } readStderrCalled = true; Q3Membuf *buf = membufStderr(); readStderrCalled = false; return buf->readAll(); } /*! Reads a line of text from standard output, excluding any trailing newline or carriage return characters, and returns it. Returns an empty string if canReadLineStdout() returns false. By default, the text is interpreted to be in Latin-1 encoding. If you need other codecs, you can set a different codec with QTextCodec::setCodecForCStrings(). \sa canReadLineStdout() readyReadStdout() readStdout() readLineStderr() */ QString Q3Process::readLineStdout() { QByteArray a( 256 ); Q3Membuf *buf = membufStdout(); if ( !buf->scanNewline( &a ) ) { if ( !canReadLineStdout() ) return QString(); if ( !buf->scanNewline( &a ) ) return QString( QLatin1String(buf->readAll()) ); } uint size = a.size(); buf->consumeBytes( size, 0 ); // get rid of terminating \n or \r\n if ( size>0 && a.at( size - 1 ) == '\n' ) { if ( size>1 && a.at( size - 2 ) == '\r' ) a.chop(2); else a.chop(1); } return QString(QString::fromLatin1(a.constData())); } /*! Reads a line of text from standard error, excluding any trailing newline or carriage return characters and returns it. Returns an empty string if canReadLineStderr() returns false. By default, the text is interpreted to be in Latin-1 encoding. If you need other codecs, you can set a different codec with QTextCodec::setCodecForCStrings(). \sa canReadLineStderr() readyReadStderr() readStderr() readLineStdout() */ QString Q3Process::readLineStderr() { QByteArray a( 256 ); Q3Membuf *buf = membufStderr(); if ( !buf->scanNewline( &a ) ) { if ( !canReadLineStderr() ) return QString(); if ( !buf->scanNewline( &a ) ) return QString( QString::fromLatin1( buf->readAll().constData() ) ); } uint size = a.size(); buf->consumeBytes( size, 0 ); // get rid of terminating \n or \r\n if ( size>0 && a.at( size - 1 ) == '\n' ) { if ( size>1 && a.at( size - 2 ) == '\r' ) a.chop(2); else a.chop(1); } return QString( QString::fromLatin1( a.constData() ) ); } /*! \fn bool Q3Process::start( QStringList *env ) Tries to run a process for the command and arguments that were specified with setArguments(), addArgument() or that were specified in the constructor. The command is searched for in the path for executable programs; you can also use an absolute path in the command itself. If \a env is null, then the process is started with the same environment as the starting process. If \a env is non-null, then the values in the stringlist are interpreted as environment setttings of the form \c {key=value} and the process is started in these environment settings. For convenience, there is a small exception to this rule: under Unix, if \a env does not contain any settings for the environment variable \c LD_LIBRARY_PATH, then this variable is inherited from the starting process; under Windows the same applies for the environment variable \c PATH. Returns true if the process could be started; otherwise returns false. You can write data to the process's standard input with writeToStdin(). You can close standard input with closeStdin() and you can terminate the process with tryTerminate(), or with kill(). You can call this function even if you've used this instance to create a another process which is still running. In such cases, Q3Process closes the old process's standard input and deletes pending data, i.e., you lose all control over the old process, but the old process is not terminated. This applies also if the process could not be started. (On operating systems that have zombie processes, Qt will also wait() on the old process.) \sa launch() closeStdin() */ /*! \fn void Q3Process::tryTerminate() const Asks the process to terminate. Processes can ignore this if they wish. If you want to be certain that the process really terminates, you can use kill() instead. The slot returns immediately: it does not wait until the process has finished. When the process terminates, the processExited() signal is emitted. \sa kill() processExited() */ /*! \fn void Q3Process::kill() const Terminates the process. This is not a safe way to end a process since the process will not be able to do any cleanup. tryTerminate() is safer, but processes can ignore a tryTerminate(). The nice way to end a process and to be sure that it is finished, is to do something like this: \snippet doc/src/snippets/code/src_qt3support_other_q3process_unix.cpp 0 This tries to terminate the process the nice way. If the process is still running after 5 seconds, it terminates the process the hard way. The timeout should be chosen depending on the time the process needs to do all its cleanup: use a higher value if the process is likely to do a lot of computation or I/O on cleanup. The slot returns immediately: it does not wait until the process has finished. When the process terminates, the processExited() signal is emitted. \sa tryTerminate() processExited() */ /*! \fn bool Q3Process::isRunning() const Returns true if the process is running; otherwise returns false. \sa normalExit() exitStatus() processExited() */ /*! \fn bool Q3Process::canReadLineStdout() const Returns true if it's possible to read an entire line of text from standard output at this time; otherwise returns false. \sa readLineStdout() canReadLineStderr() */ /*! \fn bool Q3Process::canReadLineStderr() const Returns true if it's possible to read an entire line of text from standard error at this time; otherwise returns false. \sa readLineStderr() canReadLineStdout() */ /*! \fn void Q3Process::writeToStdin( const QByteArray& buf ) Writes the data \a buf to the process's standard input. The process may or may not read this data. This function returns immediately; the Q3Process class might write the data at a later point (you must enter the event loop for this to occur). When all the data is written to the process, the signal wroteToStdin() is emitted. This does not mean that the process actually read the data, since this class only detects when it was able to write the data to the operating system. \sa wroteToStdin() closeStdin() readStdout() readStderr() */ /*! \fn void Q3Process::closeStdin() Closes the process's standard input. This function also deletes any pending data that has not been written to standard input. \sa wroteToStdin() */ /*! \fn Q3Process::PID Q3Process::processIdentifier() Returns platform dependent information about the process. This can be used together with platform specific system calls. Under Unix the return value is the PID of the process, or -1 if no process belongs to this object. Under Windows it is a pointer to the \c PROCESS_INFORMATION struct, or 0 if no process is belongs to this object. Use of this function's return value is likely to be non-portable. */ /*! \fn void Q3Process::launchFinished() This signal is emitted when the process was started with launch(). If the start was successful, this signal is emitted after all the data has been written to standard input. If the start failed, then this signal is emitted immediately. This signal is especially useful if you want to know when you can safely delete the Q3Process object when you are not interested in reading from standard output or standard error. \sa launch() QObject::deleteLater() */ /*! Runs the process and writes the data \a buf to the process's standard input. If all the data is written to standard input, standard input is closed. The command is searched for in the path for executable programs; you can also use an absolute path in the command itself. If \a env is null, then the process is started with the same environment as the starting process. If \a env is non-null, then the values in the string list are interpreted as environment setttings of the form \c {key=value} and the process is started with these environment settings. For convenience, there is a small exception to this rule under Unix: if \a env does not contain any settings for the environment variable \c LD_LIBRARY_PATH, then this variable is inherited from the starting process. Returns true if the process could be started; otherwise returns false. Note that you should not use the slots writeToStdin() and closeStdin() on processes started with launch(), since the result is not well-defined. If you need these slots, use start() instead. The process may or may not read the \a buf data sent to its standard input. You can call this function even when a process that was started with this instance is still running. Be aware that if you do this the standard input of the process that was launched first will be closed, with any pending data being deleted, and the process will be left to run out of your control. Similarly, if the process could not be started the standard input will be closed and the pending data deleted. (On operating systems that have zombie processes, Qt will also wait() on the old process.) The object emits the signal launchFinished() when this function call is finished. If the start was successful, this signal is emitted after all the data has been written to standard input. If the start failed, then this signal is emitted immediately. \sa start() launchFinished() */ bool Q3Process::launch( const QByteArray& buf, QStringList *env ) { if ( start( env ) ) { if ( !buf.isEmpty() ) { connect( this, SIGNAL(wroteToStdin()), this, SLOT(closeStdinLaunch()) ); writeToStdin( buf ); } else { closeStdin(); emit launchFinished(); } return true; } else { emit launchFinished(); return false; } } /*! \overload The data \a buf is written to standard input with writeToStdin() using the QString::local8Bit() representation of the strings. */ bool Q3Process::launch( const QString& buf, QStringList *env ) { if ( start( env ) ) { if ( !buf.isEmpty() ) { connect( this, SIGNAL(wroteToStdin()), this, SLOT(closeStdinLaunch()) ); writeToStdin( buf ); } else { closeStdin(); emit launchFinished(); } return true; } else { emit launchFinished(); return false; } } /* This private slot is used by the launch() functions to close standard input. */ void Q3Process::closeStdinLaunch() { disconnect( this, SIGNAL(wroteToStdin()), this, SLOT(closeStdinLaunch()) ); closeStdin(); emit launchFinished(); } /*! \fn void Q3Process::readyReadStdout() This signal is emitted when the process has written data to standard output. You can read the data with readStdout(). Note that this signal is only emitted when there is new data and not when there is old, but unread data. In the slot connected to this signal, you should always read everything that is available at that moment to make sure that you don't lose any data. \sa readStdout() readLineStdout() readyReadStderr() */ /*! \fn void Q3Process::readyReadStderr() This signal is emitted when the process has written data to standard error. You can read the data with readStderr(). Note that this signal is only emitted when there is new data and not when there is old, but unread data. In the slot connected to this signal, you should always read everything that is available at that moment to make sure that you don't lose any data. \sa readStderr() readLineStderr() readyReadStdout() */ /*! \fn void Q3Process::processExited() This signal is emitted when the process has exited. \sa isRunning() normalExit() exitStatus() start() launch() */ /*! \fn void Q3Process::wroteToStdin() This signal is emitted if the data sent to standard input (via writeToStdin()) was actually written to the process. This does not imply that the process really read the data, since this class only detects when it was able to write the data to the operating system. But it is now safe to close standard input without losing pending data. \sa writeToStdin() closeStdin() */ /*! \overload The string \a buf is handled as text using the QString::local8Bit() representation. */ void Q3Process::writeToStdin( const QString& buf ) { QByteArray tmp = buf.local8Bit(); tmp.resize( buf.length() ); writeToStdin( tmp ); } /* * Under Windows the implementation is not so nice: it is not that easy to * detect when one of the signals should be emitted; therefore there are some * timers that query the information. * To keep it a little efficient, use the timers only when they are needed. * They are needed, if you are interested in the signals. So use * connectNotify() and disconnectNotify() to keep track of your interest. */ /*! \reimp */ void Q3Process::connectNotify( const char * signal ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::connectNotify(): signal %s has been connected", signal ); #endif if ( !ioRedirection ) if ( qstrcmp( signal, SIGNAL(readyReadStdout()) )==0 || qstrcmp( signal, SIGNAL(readyReadStderr()) )==0 ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::connectNotify(): set ioRedirection to true" ); #endif setIoRedirection( true ); return; } if ( !notifyOnExit && qstrcmp( signal, SIGNAL(processExited()) )==0 ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::connectNotify(): set notifyOnExit to true" ); #endif setNotifyOnExit( true ); return; } if ( !wroteToStdinConnected && qstrcmp( signal, SIGNAL(wroteToStdin()) )==0 ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::connectNotify(): set wroteToStdinConnected to true" ); #endif setWroteStdinConnected( true ); return; } } /*! \reimp */ void Q3Process::disconnectNotify( const char * ) { if ( ioRedirection && receivers( SIGNAL(readyReadStdout()) ) ==0 && receivers( SIGNAL(readyReadStderr()) ) ==0 ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::disconnectNotify(): set ioRedirection to false" ); #endif setIoRedirection( false ); } if ( notifyOnExit && receivers( SIGNAL(processExited()) ) == 0 ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::disconnectNotify(): set notifyOnExit to false" ); #endif setNotifyOnExit( false ); } if ( wroteToStdinConnected && receivers( SIGNAL(wroteToStdin()) ) == 0 ) { #if defined(QT_Q3PROCESS_DEBUG) qDebug( "Q3Process::disconnectNotify(): set wroteToStdinConnected to false" ); #endif setWroteStdinConnected( false ); } } QT_END_NAMESPACE #endif // QT_NO_PROCESS