Documentation/technical/api-run-command.txt

   1 run-command API
   2 ===============
   3
   4 The run-command API offers a versatile tool to run sub-processes with
   5 redirected input and output as well as with a modified environment
   6 and an alternate current directory.
   7
   8 A similar API offers the capability to run a function asynchronously,
   9 which is primarily used to capture the output that the function
  10 produces in the caller in order to process it.
  11
  12
  13 Functions
  14 ---------
  15
  16 `child_process_init`::
  17
  18         Initialize a struct child_process variable.
  19
  20 `start_command`::
  21
  22         Start a sub-process. Takes a pointer to a `struct child_process`
  23         that specifies the details and returns pipe FDs (if requested).
  24         See below for details.
  25
  26 `finish_command`::
  27
  28         Wait for the completion of a sub-process that was started with
  29         start_command().
  30
  31 `run_command`::
  32
  33         A convenience function that encapsulates a sequence of
  34         start_command() followed by finish_command(). Takes a pointer
  35         to a `struct child_process` that specifies the details.
  36
  37 `run_command_v_opt`, `run_command_v_opt_cd_env`::
  38
  39         Convenience functions that encapsulate a sequence of
  40         start_command() followed by finish_command(). The argument argv
  41         specifies the program and its arguments. The argument opt is zero
  42         or more of the flags `RUN_COMMAND_NO_STDIN`, `RUN_GIT_CMD`,
  43         `RUN_COMMAND_STDOUT_TO_STDERR`, or `RUN_SILENT_EXEC_FAILURE`
  44         that correspond to the members .no_stdin, .git_cmd,
  45         .stdout_to_stderr, .silent_exec_failure of `struct child_process`.
  46         The argument dir corresponds the member .dir. The argument env
  47         corresponds to the member .env.
  48
  49 `child_process_clear`::
  50
  51         Release the memory associated with the struct child_process.
  52         Most users of the run-command API don't need to call this
  53         function explicitly because `start_command` invokes it on
  54         failure and `finish_command` calls it automatically already.
  55
  56 The functions above do the following:
  57
  58 . If a system call failed, errno is set and -1 is returned. A diagnostic
  59   is printed.
  60
  61 . If the program was not found, then -1 is returned and errno is set to
  62   ENOENT; a diagnostic is printed only if .silent_exec_failure is 0.
  63
  64 . Otherwise, the program is run. If it terminates regularly, its exit
  65   code is returned. No diagnostic is printed, even if the exit code is
  66   non-zero.
  67
  68 . If the program terminated due to a signal, then the return value is the
  69   signal number + 128, ie. the same value that a POSIX shell's $? would
  70   report.  A diagnostic is printed.
  71
  72
  73 `start_async`::
  74
  75         Run a function asynchronously. Takes a pointer to a `struct
  76         async` that specifies the details and returns a set of pipe FDs
  77         for communication with the function. See below for details.
  78
  79 `finish_async`::
  80
  81         Wait for the completion of an asynchronous function that was
  82         started with start_async().
  83
  84 `run_hook`::
  85
  86         Run a hook.
  87         The first argument is a pathname to an index file, or NULL
  88         if the hook uses the default index file or no index is needed.
  89         The second argument is the name of the hook.
  90         The further arguments correspond to the hook arguments.
  91         The last argument has to be NULL to terminate the arguments list.
  92         If the hook does not exist or is not executable, the return
  93         value will be zero.
  94         If it is executable, the hook will be executed and the exit
  95         status of the hook is returned.
  96         On execution, .stdout_to_stderr and .no_stdin will be set.
  97         (See below.)
  98
  99
 100 Data structures
 101 ---------------
 102
 103 * `struct child_process`
 104
 105 This describes the arguments, redirections, and environment of a
 106 command to run in a sub-process.
 107
 108 The caller:
 109
 110 1. allocates and clears (using child_process_init() or
 111    CHILD_PROCESS_INIT) a struct child_process variable;
 112 2. initializes the members;
 113 3. calls start_command();
 114 4. processes the data;
 115 5. closes file descriptors (if necessary; see below);
 116 6. calls finish_command().
 117
 118 The .argv member is set up as an array of string pointers (NULL
 119 terminated), of which .argv[0] is the program name to run (usually
 120 without a path). If the command to run is a git command, set argv[0] to
 121 the command name without the 'git-' prefix and set .git_cmd = 1.
 122
 123 Note that the ownership of the memory pointed to by .argv stays with the
 124 caller, but it should survive until `finish_command` completes. If the
 125 .argv member is NULL, `start_command` will point it at the .args
 126 `argv_array` (so you may use one or the other, but you must use exactly
 127 one). The memory in .args will be cleaned up automatically during
 128 `finish_command` (or during `start_command` when it is unsuccessful).
 129
 130 The members .in, .out, .err are used to redirect stdin, stdout,
 131 stderr as follows:
 132
 133 . Specify 0 to request no special redirection. No new file descriptor
 134   is allocated. The child process simply inherits the channel from the
 135   parent.
 136
 137 . Specify -1 to have a pipe allocated; start_command() replaces -1
 138   by the pipe FD in the following way:
 139
 140         .in: Returns the writable pipe end into which the caller writes;
 141                 the readable end of the pipe becomes the child's stdin.
 142
 143         .out, .err: Returns the readable pipe end from which the caller
 144                 reads; the writable end of the pipe end becomes child's
 145                 stdout/stderr.
 146
 147   The caller of start_command() must close the so returned FDs
 148   after it has completed reading from/writing to it!
 149
 150 . Specify a file descriptor > 0 to be used by the child:
 151
 152         .in: The FD must be readable; it becomes child's stdin.
 153         .out: The FD must be writable; it becomes child's stdout.
 154         .err: The FD must be writable; it becomes child's stderr.
 155
 156   The specified FD is closed by start_command(), even if it fails to
 157   run the sub-process!
 158
 159 . Special forms of redirection are available by setting these members
 160   to 1:
 161
 162         .no_stdin, .no_stdout, .no_stderr: The respective channel is
 163                 redirected to /dev/null.
 164
 165         .stdout_to_stderr: stdout of the child is redirected to its
 166                 stderr. This happens after stderr is itself redirected.
 167                 So stdout will follow stderr to wherever it is
 168                 redirected.
 169
 170 To modify the environment of the sub-process, specify an array of
 171 string pointers (NULL terminated) in .env:
 172
 173 . If the string is of the form "VAR=value", i.e. it contains '='
 174   the variable is added to the child process's environment.
 175
 176 . If the string does not contain '=', it names an environment
 177   variable that will be removed from the child process's environment.
 178
 179 If the .env member is NULL, `start_command` will point it at the
 180 .env_array `argv_array` (so you may use one or the other, but not both).
 181 The memory in .env_array will be cleaned up automatically during
 182 `finish_command` (or during `start_command` when it is unsuccessful).
 183
 184 To specify a new initial working directory for the sub-process,
 185 specify it in the .dir member.
 186
 187 If the program cannot be found, the functions return -1 and set
 188 errno to ENOENT. Normally, an error message is printed, but if
 189 .silent_exec_failure is set to 1, no message is printed for this
 190 special error condition.
 191
 192
 193 * `struct async`
 194
 195 This describes a function to run asynchronously, whose purpose is
 196 to produce output that the caller reads.
 197
 198 The caller:
 199
 200 1. allocates and clears (memset(&asy, 0, sizeof(asy));) a
 201    struct async variable;
 202 2. initializes .proc and .data;
 203 3. calls start_async();
 204 4. processes communicates with proc through .in and .out;
 205 5. closes .in and .out;
 206 6. calls finish_async().
 207
 208 The members .in, .out are used to provide a set of fd's for
 209 communication between the caller and the callee as follows:
 210
 211 . Specify 0 to have no file descriptor passed.  The callee will
 212   receive -1 in the corresponding argument.
 213
 214 . Specify < 0 to have a pipe allocated; start_async() replaces
 215   with the pipe FD in the following way:
 216
 217         .in: Returns the writable pipe end into which the caller
 218         writes; the readable end of the pipe becomes the function's
 219         in argument.
 220
 221         .out: Returns the readable pipe end from which the caller
 222         reads; the writable end of the pipe becomes the function's
 223         out argument.
 224
 225   The caller of start_async() must close the returned FDs after it
 226   has completed reading from/writing from them.
 227
 228 . Specify a file descriptor > 0 to be used by the function:
 229
 230         .in: The FD must be readable; it becomes the function's in.
 231         .out: The FD must be writable; it becomes the function's out.
 232
 233   The specified FD is closed by start_async(), even if it fails to
 234   run the function.
 235
 236 The function pointer in .proc has the following signature:
 237
 238         int proc(int in, int out, void *data);
 239
 240 . in, out specifies a set of file descriptors to which the function
 241   must read/write the data that it needs/produces.  The function
 242   *must* close these descriptors before it returns.  A descriptor
 243   may be -1 if the caller did not configure a descriptor for that
 244   direction.
 245
 246 . data is the value that the caller has specified in the .data member
 247   of struct async.
 248
 249 . The return value of the function is 0 on success and non-zero
 250   on failure. If the function indicates failure, finish_async() will
 251   report failure as well.
 252
 253
 254 There are serious restrictions on what the asynchronous function can do
 255 because this facility is implemented by a thread in the same address
 256 space on most platforms (when pthreads is available), but by a pipe to
 257 a forked process otherwise:
 258
 259 . It cannot change the program's state (global variables, environment,
 260   etc.) in a way that the caller notices; in other words, .in and .out
 261   are the only communication channels to the caller.
 262
 263 . It must not change the program's state that the caller of the
 264   facility also uses.