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