GSpawnChildSetupFunc ()

void
(*GSpawnChildSetupFunc) (gpointer user_data);

Specifies the type of the setup function passed to g_spawn_async(), g_spawn_sync() and g_spawn_async_with_pipes(), which can, in very limited ways, be used to affect the child's execution.

On POSIX platforms, the function is called in the child after GLib has performed all the setup it plans to perform, but before calling exec(). Actions taken in this function will only affect the child, not the parent.

On Windows, the function is called in the parent. Its usefulness on Windows is thus questionable. In many cases executing the child setup function in the parent can have ill effects, and you should be very careful when porting software to Windows that uses child setup functions.

However, even on POSIX, you are extremely limited in what you can safely do from a GSpawnChildSetupFunc, because any mutexes that were held by other threads in the parent process at the time of the fork() will still be locked in the child process, and they will never be unlocked (since the threads that held them don't exist in the child). POSIX allows only async-signal-safe functions (see signal(7)) to be called in the child between fork() and exec(), which drastically limits the usefulness of child setup functions.

In particular, it is not safe to call any function which may call malloc(), which includes POSIX functions such as setenv(). If you need to set up the child environment differently from the parent, you should use g_get_environ(), g_environ_setenv(), and g_environ_unsetenv(), and then pass the complete environment list to the g_spawn... function.

g_spawn_async_with_pipes ()

gboolean
g_spawn_async_with_pipes (const gchar *working_directory,
                          gchar **argv,
                          gchar **envp,
                          GSpawnFlags flags,
                          GSpawnChildSetupFunc child_setup,
                          gpointer user_data,
                          GPid *child_pid,
                          gint *standard_input,
                          gint *standard_output,
                          gint *standard_error,
                          GError **error);

Executes a child program asynchronously (your program will not block waiting for the child to exit). The child program is specified by the only argument that must be provided, argv . argv should be a NULL-terminated array of strings, to be passed as the argument vector for the child. The first string in argv is of course the name of the program to execute. By default, the name of the program must be a full path. If flags contains the G_SPAWN_SEARCH_PATH flag, the PATH environment variable is used to search for the executable. If flags contains the G_SPAWN_SEARCH_PATH_FROM_ENVP flag, the PATH variable from envp is used to search for the executable. If both the G_SPAWN_SEARCH_PATH and G_SPAWN_SEARCH_PATH_FROM_ENVP flags are set, the PATH variable from envp takes precedence over the environment variable.

If the program name is not a full path and G_SPAWN_SEARCH_PATH flag is not used, then the program will be run from the current directory (or working_directory , if specified); this might be unexpected or even dangerous in some cases when the current directory is world-writable.

On Windows, note that all the string or string vector arguments to this function and the other g_spawn*() functions are in UTF-8, the GLib file name encoding. Unicode characters that are not part of the system codepage passed in these arguments will be correctly available in the spawned program only if it uses wide character API to retrieve its command line. For C programs built with Microsoft's tools it is enough to make the program have a wmain() instead of main(). wmain() has a wide character argument vector as parameter.

At least currently, mingw doesn't support wmain(), so if you use mingw to develop the spawned program, it will have to call the undocumented function __wgetmainargs() to get the wide character argument vector and environment. See gspawn-win32-helper.c in the GLib sources or init.c in the mingw runtime sources for a prototype for that function. Alternatively, you can retrieve the Win32 system level wide character command line passed to the spawned program using the GetCommandLineW() function.

On Windows the low-level child process creation API CreateProcess() doesn't use argument vectors, but a command line. The C runtime library's spawn*() family of functions (which g_spawn_async_with_pipes() eventually calls) paste the argument vector elements together into a command line, and the C runtime startup code does a corresponding reconstruction of an argument vector from the command line, to be passed to main(). Complications arise when you have argument vector elements that contain spaces of double quotes. The spawn*() functions don't do any quoting or escaping, but on the other hand the startup code does do unquoting and unescaping in order to enable receiving arguments with embedded spaces or double quotes. To work around this asymmetry, g_spawn_async_with_pipes() will do quoting and escaping on argument vector elements that need it before calling the C runtime spawn() function.

The returned child_pid on Windows is a handle to the child process, not its identifier. Process handles and process identifiers are different concepts on Windows.

envp is a NULL-terminated array of strings, where each string has the form KEY=VALUE. This will become the child's environment. If envp is NULL, the child inherits its parent's environment.

flags should be the bitwise OR of any flags you want to affect the function's behaviour. The G_SPAWN_DO_NOT_REAP_CHILD means that the child will not automatically be reaped; you must use a child watch to be notified about the death of the child process. Eventually you must call g_spawn_close_pid() on the child_pid , in order to free resources which may be associated with the child process. (On Unix, using a child watch is equivalent to calling waitpid() or handling the SIGCHLD signal manually. On Windows, calling g_spawn_close_pid() is equivalent to calling CloseHandle() on the process handle returned in child_pid ). See g_child_watch_add().

G_SPAWN_LEAVE_DESCRIPTORS_OPEN means that the parent's open file descriptors will be inherited by the child; otherwise all descriptors except stdin/stdout/stderr will be closed before calling exec() in the child. G_SPAWN_SEARCH_PATH means that argv [0] need not be an absolute path, it will be looked for in the PATH environment variable. G_SPAWN_SEARCH_PATH_FROM_ENVP means need not be an absolute path, it will be looked for in the PATH variable from envp . If both G_SPAWN_SEARCH_PATH and G_SPAWN_SEARCH_PATH_FROM_ENVP are used, the value from envp takes precedence over the environment. G_SPAWN_STDOUT_TO_DEV_NULL means that the child's standard output will be discarded, instead of going to the same location as the parent's standard output. If you use this flag, standard_output must be NULL. G_SPAWN_STDERR_TO_DEV_NULL means that the child's standard error will be discarded, instead of going to the same location as the parent's standard error. If you use this flag, standard_error must be NULL. G_SPAWN_CHILD_INHERITS_STDIN means that the child will inherit the parent's standard input (by default, the child's standard input is attached to /dev/null). If you use this flag, standard_input must be NULL. G_SPAWN_FILE_AND_ARGV_ZERO means that the first element of argv is the file to execute, while the remaining elements are the actual argument vector to pass to the file. Normally g_spawn_async_with_pipes() uses argv [0] as the file to execute, and passes all of argv to the child.

child_setup and user_data are a function and user data. On POSIX platforms, the function is called in the child after GLib has performed all the setup it plans to perform (including creating pipes, closing file descriptors, etc.) but before calling exec(). That is, child_setup is called just before calling exec() in the child. Obviously actions taken in this function will only affect the child, not the parent.

On Windows, there is no separate fork() and exec() functionality. Child processes are created and run with a single API call, CreateProcess(). There is no sensible thing child_setup could be used for on Windows so it is ignored and not called.

If non-NULL, child_pid will on Unix be filled with the child's process ID. You can use the process ID to send signals to the child, or to use g_child_watch_add() (or waitpid()) if you specified the G_SPAWN_DO_NOT_REAP_CHILD flag. On Windows, child_pid will be filled with a handle to the child process only if you specified the G_SPAWN_DO_NOT_REAP_CHILD flag. You can then access the child process using the Win32 API, for example wait for its termination with the WaitFor*() functions, or examine its exit code with GetExitCodeProcess(). You should close the handle with CloseHandle() or g_spawn_close_pid() when you no longer need it.

If non-NULL, the standard_input , standard_output , standard_error locations will be filled with file descriptors for writing to the child's standard input or reading from its standard output or standard error. The caller of g_spawn_async_with_pipes() must close these file descriptors when they are no longer in use. If these parameters are NULL, the corresponding pipe won't be created.

If standard_input is NULL, the child's standard input is attached to /dev/null unless G_SPAWN_CHILD_INHERITS_STDIN is set.

If standard_error is NULL, the child's standard error goes to the same location as the parent's standard error unless G_SPAWN_STDERR_TO_DEV_NULL is set.

If standard_output is NULL, the child's standard output goes to the same location as the parent's standard output unless G_SPAWN_STDOUT_TO_DEV_NULL is set.

error can be NULL to ignore errors, or non-NULL to report errors. If an error is set, the function returns FALSE. Errors are reported even if they occur in the child (for example if the executable in argv [0] is not found). Typically the message field of returned errors should be displayed to users. Possible errors are those from the G_SPAWN_ERROR domain.

If an error occurs, child_pid , standard_input , standard_output , and standard_error will not be filled with valid values.

If child_pid is not NULL and an error does not occur then the returned process reference must be closed using g_spawn_close_pid().

If you are writing a GTK+ application, and the program you are spawning is a graphical application, too, then you may want to use gdk_spawn_on_screen_with_pipes() instead to ensure that the spawned program opens its windows on the right screen.

g_spawn_async ()

gboolean
g_spawn_async (const gchar *working_directory,
               gchar **argv,
               gchar **envp,
               GSpawnFlags flags,
               GSpawnChildSetupFunc child_setup,
               gpointer user_data,
               GPid *child_pid,
               GError **error);

See g_spawn_async_with_pipes() for a full description; this function simply calls the g_spawn_async_with_pipes() without any pipes.

You should call g_spawn_close_pid() on the returned child process reference when you don't need it any more.

If you are writing a GTK+ application, and the program you are spawning is a graphical application, too, then you may want to use gdk_spawn_on_screen() instead to ensure that the spawned program opens its windows on the right screen.

Note that the returned child_pid on Windows is a handle to the child process and not its identifier. Process handles and process identifiers are different concepts on Windows.

g_spawn_sync ()

gboolean
g_spawn_sync (const gchar *working_directory,
              gchar **argv,
              gchar **envp,
              GSpawnFlags flags,
              GSpawnChildSetupFunc child_setup,
              gpointer user_data,
              gchar **standard_output,
              gchar **standard_error,
              gint *exit_status,
              GError **error);

Executes a child synchronously (waits for the child to exit before returning). All output from the child is stored in standard_output and standard_error , if those parameters are non-NULL. Note that you must set the G_SPAWN_STDOUT_TO_DEV_NULL and G_SPAWN_STDERR_TO_DEV_NULL flags when passing NULL for standard_output and standard_error .

If exit_status is non-NULL, the platform-specific exit status of the child is stored there; see the documentation of g_spawn_check_exit_status() for how to use and interpret this. Note that it is invalid to pass G_SPAWN_DO_NOT_REAP_CHILD in flags .

If an error occurs, no data is returned in standard_output , standard_error , or exit_status .

This function calls g_spawn_async_with_pipes() internally; see that function for full details on the other parameters and details on how these functions work on Windows.

g_spawn_check_exit_status ()

gboolean
g_spawn_check_exit_status (gint exit_status,
                           GError **error);

Set error if exit_status indicates the child exited abnormally (e.g. with a nonzero exit code, or via a fatal signal).

The g_spawn_sync() and g_child_watch_add() family of APIs return an exit status for subprocesses encoded in a platform-specific way. On Unix, this is guaranteed to be in the same format waitpid() returns, and on Windows it is guaranteed to be the result of GetExitCodeProcess().

Prior to the introduction of this function in GLib 2.34, interpreting exit_status required use of platform-specific APIs, which is problematic for software using GLib as a cross-platform layer.

Additionally, many programs simply want to determine whether or not the child exited successfully, and either propagate a GError or print a message to standard error. In that common case, this function can be used. Note that the error message in error will contain human-readable information about the exit status.

The domain and code of error have special semantics in the case where the process has an "exit code", as opposed to being killed by a signal. On Unix, this happens if WIFEXITED() would be true of exit_status . On Windows, it is always the case.

The special semantics are that the actual exit code will be the code set in error , and the domain will be G_SPAWN_EXIT_ERROR. This allows you to differentiate between different exit codes.

If the process was terminated by some means other than an exit status, the domain will be G_SPAWN_ERROR, and the code will be G_SPAWN_ERROR_FAILED.

This function just offers convenience; you can of course also check the available platform via a macro such as G_OS_UNIX, and use WIFEXITED() and WEXITSTATUS() on exit_status directly. Do not attempt to scan or parse the error message string; it may be translated and/or change in future versions of GLib.

Since 2.34

g_spawn_command_line_async ()

gboolean
g_spawn_command_line_async (const gchar *command_line,
                            GError **error);

A simple version of g_spawn_async() that parses a command line with g_shell_parse_argv() and passes it to g_spawn_async(). Runs a command line in the background. Unlike g_spawn_async(), the G_SPAWN_SEARCH_PATH flag is enabled, other flags are not. Note that G_SPAWN_SEARCH_PATH can have security implications, so consider using g_spawn_async() directly if appropriate. Possible errors are those from g_shell_parse_argv() and g_spawn_async().

The same concerns on Windows apply as for g_spawn_command_line_sync().

g_spawn_command_line_sync ()

gboolean
g_spawn_command_line_sync (const gchar *command_line,
                           gchar **standard_output,
                           gchar **standard_error,
                           gint *exit_status,
                           GError **error);

A simple version of g_spawn_sync() with little-used parameters removed, taking a command line instead of an argument vector. See g_spawn_sync() for full details. command_line will be parsed by g_shell_parse_argv(). Unlike g_spawn_sync(), the G_SPAWN_SEARCH_PATH flag is enabled. Note that G_SPAWN_SEARCH_PATH can have security implications, so consider using g_spawn_sync() directly if appropriate. Possible errors are those from g_spawn_sync() and those from g_shell_parse_argv().

If exit_status is non-NULL, the platform-specific exit status of the child is stored there; see the documentation of g_spawn_check_exit_status() for how to use and interpret this.

On Windows, please note the implications of g_shell_parse_argv() parsing command_line . Parsing is done according to Unix shell rules, not Windows command interpreter rules. Space is a separator, and backslashes are special. Thus you cannot simply pass a command_line containing canonical Windows paths, like "c:\program files\app\app.exe", as the backslashes will be eaten, and the space will act as a separator. You need to enclose such paths with single quotes, like "'c:\program files\app\app.exe' 'e:\folder\argument.txt'".

g_spawn_close_pid ()

void
g_spawn_close_pid (GPid pid);

On some platforms, notably Windows, the GPid type represents a resource which must be closed to prevent resource leaking. g_spawn_close_pid() is provided for this purpose. It should be used on all platforms, even though it doesn't do anything under UNIX.

enum GSpawnError

Error codes returned by spawning processes.

G_SPAWN_ERROR

#define G_SPAWN_ERROR g_spawn_error_quark ()

Error domain for spawning processes. Errors in this domain will be from the GSpawnError enumeration. See GError for information on error domains.

enum GSpawnFlags

Flags passed to g_spawn_sync(), g_spawn_async() and g_spawn_async_with_pipes().

G_SPAWN_EXIT_ERROR

#define G_SPAWN_EXIT_ERROR g_spawn_exit_error_quark ()

Error domain used by g_spawn_check_exit_status(). The code will be the program exit code.