Subprocess#

Added in version 2.40.

class Subprocess(**properties: Any)#

Superclasses: Object

Implemented Interfaces: Initable

GSubprocess allows the creation of and interaction with child processes.

Processes can be communicated with using standard GIO-style APIs (ie: InputStream, OutputStream). There are GIO-style APIs to wait for process termination (ie: cancellable and with an asynchronous variant).

There is an API to force a process to terminate, as well as a race-free API for sending UNIX signals to a subprocess.

One major advantage that GIO brings over the core GLib library is comprehensive API for asynchronous I/O, such splice_async. This makes GSubprocess significantly more powerful and flexible than equivalent APIs in some other languages such as the subprocess.py included with Python. For example, using GSubprocess one could create two child processes, reading standard output from the first, processing it, and writing to the input stream of the second, all without blocking the main loop.

A powerful communicate API is provided similar to the communicate() method of subprocess.py. This enables very easy interaction with a subprocess that has been opened with pipes.

GSubprocess defaults to tight control over the file descriptors open in the child process, avoiding dangling-FD issues that are caused by a simple fork()/exec(). The only open file descriptors in the spawned process are ones that were explicitly specified by the GSubprocess API (unless G_SUBPROCESS_FLAGS_INHERIT_FDS was specified).

GSubprocess will quickly reap all child processes as they exit, avoiding ‘zombie processes’ remaining around for long periods of time. wait can be used to wait for this to happen, but it will happen even without the call being explicitly made.

As a matter of principle, GSubprocess has no API that accepts shell-style space-separated strings. It will, however, match the typical shell behaviour of searching the PATH for executables that do not contain a directory separator in their name. By default, the PATH of the current process is used. You can specify G_SUBPROCESS_FLAGS_SEARCH_PATH_FROM_ENVP to use the PATH of the launcher environment instead.

GSubprocess attempts to have a very simple API for most uses (ie: spawning a subprocess with arguments and support for most typical kinds of input and output redirection). See new. The SubprocessLauncher API is provided for more complicated cases (advanced types of redirection, environment variable manipulation, change of working directory, child setup functions, etc).

A typical use of GSubprocess will involve calling new, followed by wait_async or wait. After the process exits, the status can be checked using functions such as get_if_exited (which are similar to the familiar WIFEXITED-style POSIX macros).

Note that as of GLib 2.82, creating a GSubprocess causes the signal SIGPIPE to be ignored for the remainder of the program. If you are writing a command-line utility that uses GSubprocess, you may need to take into account the fact that your program will not automatically be killed if it tries to write to stdout after it has been closed.

Constructors#

class Subprocess
classmethod new(argv: list[str], flags: SubprocessFlags) → Subprocess#

Create a new process with the given flags and varargs argument list. By default, matching the spawn_async() defaults, the child’s stdin will be set to the system null device, and stdout/stderr will be inherited from the parent. You can use flags to control this behavior.

The argument list must be terminated with None.

Added in version 2.40.

Parameters:

  • argv

  • flags – flags that define the behaviour of the subprocess

Methods#

class Subprocess
communicate(stdin_buf: Bytes | None = None, cancellable: Cancellable | None = None) → tuple[bool, Bytes | None, Bytes | None]#

Communicate with the subprocess until it terminates, and all input and output has been completed.

If stdin_buf is given, the subprocess must have been created with STDIN_PIPE. The given data is fed to the stdin of the subprocess and the pipe is closed (ie: EOF).

At the same time (as not to cause blocking when dealing with large amounts of data), if STDOUT_PIPE or STDERR_PIPE were used, reads from those streams. The data that was read is returned in stdout and/or the stderr.

If the subprocess was created with STDOUT_PIPE, stdout_buf will contain the data read from stdout. Otherwise, for subprocesses not created with STDOUT_PIPE, stdout_buf will be set to None. Similar provisions apply to stderr_buf and STDERR_PIPE.

As usual, any output variable may be given as None to ignore it.

If you desire the stdout and stderr data to be interleaved, create the subprocess with STDOUT_PIPE and STDERR_MERGE. The merged result will be returned in stdout_buf and stderr_buf will be set to None.

In case of any error (including cancellation), False will be returned with error set. Some or all of the stdin data may have been written. Any stdout or stderr data that has been read will be discarded. None of the out variables (aside from error) will have been set to anything in particular and should not be inspected.

In the case that True is returned, the subprocess has exited and the exit status inspection APIs (eg: get_if_exited(), get_exit_status()) may be used.

You should not attempt to use any of the subprocess pipes after starting this function, since they may be left in strange states, even if the operation was cancelled. You should especially not attempt to interact with the pipes while the operation is in progress (either from another thread or if using the asynchronous version).

Added in version 2.40.

Parameters:

  • stdin_buf – data to send to the stdin of the subprocess, or None

  • cancellable – a Cancellable

async communicate_async(self, stdin_buf: Bytes | None = None) → tuple[bool, Bytes | None, Bytes | None]#

This is the awaitable version of communicate_async().

Parameters:

stdin_buf – Input data, or None

communicate_async(stdin_buf: Bytes | None = None, cancellable: Cancellable | None = None, callback: Callable[[Object | None, AsyncResult, Any], None] | None = None, user_data: Any = None) → None#

Asynchronous version of communicate(). Complete invocation with communicate_finish().

Parameters:

  • stdin_buf – Input data, or None

  • cancellable – Cancellable

  • callback – Callback

  • user_data – User data

communicate_finish(result: AsyncResult) → tuple[bool, Bytes | None, Bytes | None]#

Complete an invocation of communicate_async().

Parameters:

result – Result

communicate_utf8(stdin_buf: str | None = None, cancellable: Cancellable | None = None) → tuple[bool, str | None, str | None]#

Like communicate(), but validates the output of the process as UTF-8, and returns it as a regular NUL terminated string.

On error, stdout_buf and stderr_buf will be set to undefined values and should not be used.

Parameters:

  • stdin_buf – data to send to the stdin of the subprocess, or None

  • cancellable – a Cancellable

async communicate_utf8_async(self, stdin_buf: str | None = None) → tuple[bool, str | None, str | None]#

This is the awaitable version of communicate_utf8_async().

Parameters:

stdin_buf – Input data, or None

communicate_utf8_async(stdin_buf: str | None = None, cancellable: Cancellable | None = None, callback: Callable[[Object | None, AsyncResult, Any], None] | None = None, user_data: Any = None) → None#

Asynchronous version of communicate_utf8(). Complete invocation with communicate_utf8_finish().

Parameters:

  • stdin_buf – Input data, or None

  • cancellable – Cancellable

  • callback – Callback

  • user_data – User data

communicate_utf8_finish(result: AsyncResult) → tuple[bool, str | None, str | None]#

Complete an invocation of communicate_utf8_async().

Parameters:

result – Result

force_exit() → None#

Use an operating-system specific method to attempt an immediate, forceful termination of the process. There is no mechanism to determine whether or not the request itself was successful; however, you can use wait() to monitor the status of the process after calling this function.

On Unix, this function sends %SIGKILL.

Added in version 2.40.

get_exit_status() → int#

Check the exit status of the subprocess, given that it exited normally. This is the value passed to the exit() system call or the return value from main.

This is equivalent to the system WEXITSTATUS macro.

It is an error to call this function before wait() and unless get_if_exited() returned True.

Added in version 2.40.

get_identifier() → str | None#

On UNIX, returns the process ID as a decimal string. On Windows, returns the result of GetProcessId() also as a string. If the subprocess has terminated, this will return None.

Added in version 2.40.

get_if_exited() → bool#

Check if the given subprocess exited normally (ie: by way of exit() or return from main()).

This is equivalent to the system WIFEXITED macro.

It is an error to call this function before wait() has returned.

Added in version 2.40.

get_if_signaled() → bool#

Check if the given subprocess terminated in response to a signal.

This is equivalent to the system WIFSIGNALED macro.

It is an error to call this function before wait() has returned.

Added in version 2.40.

get_status() → int#

Gets the raw status code of the process, as from waitpid().

This value has no particular meaning, but it can be used with the macros defined by the system headers such as WIFEXITED. It can also be used with spawn_check_wait_status().

It is more likely that you want to use get_if_exited() followed by get_exit_status().

It is an error to call this function before wait() has returned.

Added in version 2.40.

get_stderr_pipe() → InputStream | None#

Gets the InputStream from which to read the stderr output of subprocess.

The process must have been created with STDERR_PIPE, otherwise None will be returned.

Added in version 2.40.

get_stdin_pipe() → OutputStream | None#

Gets the OutputStream that you can write to in order to give data to the stdin of subprocess.

The process must have been created with STDIN_PIPE and not STDIN_INHERIT, otherwise None will be returned.

Added in version 2.40.

get_stdout_pipe() → InputStream | None#

Gets the InputStream from which to read the stdout output of subprocess.

The process must have been created with STDOUT_PIPE, otherwise None will be returned.

Added in version 2.40.

get_successful() → bool#

Checks if the process was “successful”. A process is considered successful if it exited cleanly with an exit status of 0, either by way of the exit() system call or return from main().

It is an error to call this function before wait() has returned.

Added in version 2.40.

get_term_sig() → int#

Get the signal number that caused the subprocess to terminate, given that it terminated due to a signal.

This is equivalent to the system WTERMSIG macro.

It is an error to call this function before wait() and unless get_if_signaled() returned True.

Added in version 2.40.

send_signal(signal_num: int) → None#

Sends the UNIX signal signal_num to the subprocess, if it is still running.

This API is race-free. If the subprocess has terminated, it will not be signalled.

This API is not available on Windows.

Added in version 2.40.

Parameters:

signal_num – the signal number to send

wait(cancellable: Cancellable | None = None) → bool#

Synchronously wait for the subprocess to terminate.

After the process terminates you can query its exit status with functions such as get_if_exited() and get_exit_status().

This function does not fail in the case of the subprocess having abnormal termination. See wait_check() for that.

Cancelling cancellable doesn’t kill the subprocess. Call force_exit() if it is desirable.

Added in version 2.40.

Parameters:

cancellable – a Cancellable

async wait_async(self) → bool#

This is the awaitable version of wait_async().

Added in version 2.40.

wait_async(cancellable: Cancellable | None = None, callback: Callable[[Object | None, AsyncResult, Any], None] | None = None, user_data: Any = None) → None#

Wait for the subprocess to terminate.

This is the asynchronous version of wait().

Added in version 2.40.

Parameters:

  • cancellable – a Cancellable, or None

  • callback – a AsyncReadyCallback to call when the operation is complete

  • user_data – user_data for callback

wait_check(cancellable: Cancellable | None = None) → bool#

Combines wait() with spawn_check_wait_status().

Added in version 2.40.

Parameters:

cancellable – a Cancellable

async wait_check_async(self) → bool#

This is the awaitable version of wait_check_async().

Added in version 2.40.

wait_check_async(cancellable: Cancellable | None = None, callback: Callable[[Object | None, AsyncResult, Any], None] | None = None, user_data: Any = None) → None#

Combines wait_async() with spawn_check_wait_status().

This is the asynchronous version of wait_check().

Added in version 2.40.

Parameters:

  • cancellable – a Cancellable, or None

  • callback – a AsyncReadyCallback to call when the operation is complete

  • user_data – user_data for callback

wait_check_finish(result: AsyncResult) → bool#

Collects the result of a previous call to wait_check_async().

Added in version 2.40.

Parameters:

result – the AsyncResult passed to your AsyncReadyCallback

wait_finish(result: AsyncResult) → bool#

Collects the result of a previous call to wait_async().

Added in version 2.40.

Parameters:

result – the AsyncResult passed to your AsyncReadyCallback

Properties#

class Subprocess
props.argv: list[str]#

The type of the None singleton.

Added in version 2.40.

props.flags: SubprocessFlags#

The type of the None singleton.

Added in version 2.40.