freshlybakedca.ke
Git fork
1#ifndef RUN_COMMAND_H
2#define RUN_COMMAND_H
3
4#include "thread-utils.h"
5
6#include "strvec.h"
7
8/**
9 * The run-command API offers a versatile tool to run sub-processes with
10 * redirected input and output as well as with a modified environment
11 * and an alternate current directory.
12 *
13 * A similar API offers the capability to run a function asynchronously,
14 * which is primarily used to capture the output that the function
15 * produces in the caller in order to process it.
16 */
17
18
19/**
20 * This describes the arguments, redirections, and environment of a
21 * command to run in a sub-process.
22 *
23 * The caller:
24 *
25 * 1. allocates and clears (using child_process_init() or
26 * CHILD_PROCESS_INIT) a struct child_process variable;
27 * 2. initializes the members;
28 * 3. calls start_command();
29 * 4. processes the data;
30 * 5. closes file descriptors (if necessary; see below);
31 * 6. calls finish_command().
32 *
33 * Special forms of redirection are available by setting these members
34 * to 1:
35 *
36 * .no_stdin, .no_stdout, .no_stderr: The respective channel is
37 * redirected to /dev/null.
38 *
39 * .stdout_to_stderr: stdout of the child is redirected to its
40 * stderr. This happens after stderr is itself redirected.
41 * So stdout will follow stderr to wherever it is
42 * redirected.
43 */
44struct child_process {
45
46 /**
47 * The .args is a `struct strvec', use that API to manipulate
48 * it, e.g. strvec_pushv() to add an existing "const char **"
49 * vector.
50 *
51 * If the command to run is a git command, set the first
52 * element in the strvec to the command name without the
53 * 'git-' prefix and set .git_cmd = 1.
54 *
55 * The memory in .args will be cleaned up automatically during
56 * `finish_command` (or during `start_command` when it is unsuccessful).
57 */
58 struct strvec args;
59
60 /**
61 * Like .args the .env is a `struct strvec'.
62 *
63 * To modify the environment of the sub-process, specify an array of
64 * environment settings. Each string in the array manipulates the
65 * environment.
66 *
67 * - If the string is of the form "VAR=value", i.e. it contains '='
68 * the variable is added to the child process's environment.
69 *
70 * - If the string does not contain '=', it names an environment
71 * variable that will be removed from the child process's environment.
72 *
73 * The memory in .env will be cleaned up automatically during
74 * `finish_command` (or during `start_command` when it is unsuccessful).
75 */
76 struct strvec env;
77 pid_t pid;
78
79 int trace2_child_id;
80 uint64_t trace2_child_us_start;
81 const char *trace2_child_class;
82 const char *trace2_hook_name;
83
84 /*
85 * Using .in, .out, .err:
86 * - Specify 0 for no redirections. No new file descriptor is allocated.
87 * (child inherits stdin, stdout, stderr from parent).
88 * - Specify -1 to have a pipe allocated as follows:
89 * .in: returns the writable pipe end; parent writes to it,
90 * the readable pipe end becomes child's stdin
91 * .out, .err: returns the readable pipe end; parent reads from
92 * it, the writable pipe end becomes child's stdout/stderr
93 * The caller of start_command() must close the returned FDs
94 * after it has completed reading from/writing to it!
95 * - Specify > 0 to set a channel to a particular FD as follows:
96 * .in: a readable FD, becomes child's stdin
97 * .out: a writable FD, becomes child's stdout/stderr
98 * .err: a writable FD, becomes child's stderr
99 * The specified FD is closed by start_command(), even in case
100 * of errors!
101 */
102 int in;
103 int out;
104 int err;
105
106 /**
107 * To specify a new initial working directory for the sub-process,
108 * specify it in the .dir member.
109 */
110 const char *dir;
111
112 unsigned no_stdin:1;
113 unsigned no_stdout:1;
114 unsigned no_stderr:1;
115 unsigned git_cmd:1; /* if this is to be git sub-command */
116
117 /**
118 * If the program cannot be found, the functions return -1 and set
119 * errno to ENOENT. Normally, an error message is printed, but if
120 * .silent_exec_failure is set to 1, no message is printed for this
121 * special error condition.
122 */
123 unsigned silent_exec_failure:1;
124
125 /**
126 * Run the command from argv[0] using a shell (but note that we may
127 * still optimize out the shell call if the command contains no
128 * metacharacters). Note that further arguments to the command in
129 * argv[1], etc, do not need to be shell-quoted.
130 */
131 unsigned use_shell:1;
132
133 /**
134 * Release any open file handles to the object store before running
135 * the command; This is necessary e.g. when the spawned process may
136 * want to repack because that would delete `.pack` files (and on
137 * Windows, you cannot delete files that are still in use).
138 */
139 unsigned close_object_store:1;
140
141 unsigned stdout_to_stderr:1;
142 unsigned clean_on_exit:1;
143 unsigned wait_after_clean:1;
144 void (*clean_on_exit_handler)(struct child_process *process);
145};
146
147#define CHILD_PROCESS_INIT { \
148 .args = STRVEC_INIT, \
149 .env = STRVEC_INIT, \
150}
151
152/**
153 * The functions: start_command, finish_command, run_command do the following:
154 *
155 * - If a system call failed, errno is set and -1 is returned. A diagnostic
156 * is printed.
157 *
158 * - If the program was not found, then -1 is returned and errno is set to
159 * ENOENT; a diagnostic is printed only if .silent_exec_failure is 0.
160 *
161 * - Otherwise, the program is run. If it terminates regularly, its exit
162 * code is returned. No diagnostic is printed, even if the exit code is
163 * non-zero.
164 *
165 * - If the program terminated due to a signal, then the return value is the
166 * signal number + 128, ie. the same value that a POSIX shell's $? would
167 * report. A diagnostic is printed.
168 *
169 */
170
171/**
172 * Initialize a struct child_process variable.
173 */
174void child_process_init(struct child_process *);
175
176/**
177 * Release the memory associated with the struct child_process.
178 * Most users of the run-command API don't need to call this
179 * function explicitly because `start_command` invokes it on
180 * failure and `finish_command` calls it automatically already.
181 */
182void child_process_clear(struct child_process *);
183
184int is_executable(const char *name);
185
186/**
187 * Check if the command exists on $PATH. This emulates the path search that
188 * execvp would perform, without actually executing the command so it
189 * can be used before fork() to prepare to run a command using
190 * execve() or after execvp() to diagnose why it failed.
191 *
192 * The caller should ensure that command contains no directory separators.
193 *
194 * Returns 1 if it is found in $PATH or 0 if the command could not be found.
195 */
196int exists_in_PATH(const char *command);
197
198/**
199 * Return the path that is used to execute Unix shell command-lines.
200 */
201char *git_shell_path(void);
202
203/**
204 * Start a sub-process. Takes a pointer to a `struct child_process`
205 * that specifies the details and returns pipe FDs (if requested).
206 * See below for details.
207 */
208int start_command(struct child_process *);
209
210/**
211 * Wait for the completion of a sub-process that was started with
212 * start_command().
213 */
214int finish_command(struct child_process *);
215
216int finish_command_in_signal(struct child_process *);
217
218/**
219 * A convenience function that encapsulates a sequence of
220 * start_command() followed by finish_command(). Takes a pointer
221 * to a `struct child_process` that specifies the details.
222 */
223int run_command(struct child_process *);
224
225/*
226 * Prepare a `struct child_process` to run auto-maintenance. Returns 1 if the
227 * process has been prepared and is ready to run, or 0 in case auto-maintenance
228 * should be skipped.
229 */
230int prepare_auto_maintenance(int quiet, struct child_process *maint);
231
232/*
233 * Trigger an auto-gc
234 */
235int run_auto_maintenance(int quiet);
236
237/**
238 * Execute the given command, sending "in" to its stdin, and capturing its
239 * stdout and stderr in the "out" and "err" strbufs. Any of the three may
240 * be NULL to skip processing.
241 *
242 * Returns -1 if starting the command fails or reading fails, and otherwise
243 * returns the exit code of the command. Any output collected in the
244 * buffers is kept even if the command returns a non-zero exit. The hint fields
245 * gives starting sizes for the strbuf allocations.
246 *
247 * The fields of "cmd" should be set up as they would for a normal run_command
248 * invocation. But note that there is no need to set the in, out, or err
249 * fields; pipe_command handles that automatically.
250 */
251int pipe_command(struct child_process *cmd,
252 const char *in, size_t in_len,
253 struct strbuf *out, size_t out_hint,
254 struct strbuf *err, size_t err_hint);
255
256/**
257 * Convenience wrapper around pipe_command for the common case
258 * of capturing only stdout.
259 */
260static inline int capture_command(struct child_process *cmd,
261 struct strbuf *out,
262 size_t hint)
263{
264 return pipe_command(cmd, NULL, 0, out, hint, NULL, 0);
265}
266
267/*
268 * The purpose of the following functions is to feed a pipe by running
269 * a function asynchronously and providing output that the caller reads.
270 *
271 * It is expected that no synchronization and mutual exclusion between
272 * the caller and the feed function is necessary so that the function
273 * can run in a thread without interfering with the caller.
274 *
275 * The caller:
276 *
277 * 1. allocates and clears (memset(&asy, 0, sizeof(asy));) a
278 * struct async variable;
279 * 2. initializes .proc and .data;
280 * 3. calls start_async();
281 * 4. processes communicates with proc through .in and .out;
282 * 5. closes .in and .out;
283 * 6. calls finish_async().
284 *
285 * There are serious restrictions on what the asynchronous function can do
286 * because this facility is implemented by a thread in the same address
287 * space on most platforms (when pthreads is available), but by a pipe to
288 * a forked process otherwise:
289 *
290 * - It cannot change the program's state (global variables, environment,
291 * etc.) in a way that the caller notices; in other words, .in and .out
292 * are the only communication channels to the caller.
293 *
294 * - It must not change the program's state that the caller of the
295 * facility also uses.
296 *
297 */
298struct async {
299
300 /**
301 * The function pointer in .proc has the following signature:
302 *
303 * int proc(int in, int out, void *data);
304 *
305 * - in, out specifies a set of file descriptors to which the function
306 * must read/write the data that it needs/produces. The function
307 * *must* close these descriptors before it returns. A descriptor
308 * may be -1 if the caller did not configure a descriptor for that
309 * direction.
310 *
311 * - data is the value that the caller has specified in the .data member
312 * of struct async.
313 *
314 * - The return value of the function is 0 on success and non-zero
315 * on failure. If the function indicates failure, finish_async() will
316 * report failure as well.
317 *
318 */
319 int (*proc)(int in, int out, void *data);
320
321 void *data;
322
323 /**
324 * The members .in, .out are used to provide a set of fd's for
325 * communication between the caller and the callee as follows:
326 *
327 * - Specify 0 to have no file descriptor passed. The callee will
328 * receive -1 in the corresponding argument.
329 *
330 * - Specify < 0 to have a pipe allocated; start_async() replaces
331 * with the pipe FD in the following way:
332 *
333 * .in: Returns the writable pipe end into which the caller
334 * writes; the readable end of the pipe becomes the function's
335 * in argument.
336 *
337 * .out: Returns the readable pipe end from which the caller
338 * reads; the writable end of the pipe becomes the function's
339 * out argument.
340 *
341 * The caller of start_async() must close the returned FDs after it
342 * has completed reading from/writing from them.
343 *
344 * - Specify a file descriptor > 0 to be used by the function:
345 *
346 * .in: The FD must be readable; it becomes the function's in.
347 * .out: The FD must be writable; it becomes the function's out.
348 *
349 * The specified FD is closed by start_async(), even if it fails to
350 * run the function.
351 */
352 int in; /* caller writes here and closes it */
353 int out; /* caller reads from here and closes it */
354#ifdef NO_PTHREADS
355 pid_t pid;
356#else
357 pthread_t tid;
358 int proc_in;
359 int proc_out;
360#endif
361 int isolate_sigpipe;
362};
363
364/**
365 * Run a function asynchronously. Takes a pointer to a `struct
366 * async` that specifies the details and returns a set of pipe FDs
367 * for communication with the function. See below for details.
368 */
369int start_async(struct async *async);
370
371/**
372 * Wait for the completion of an asynchronous function that was
373 * started with start_async().
374 */
375int finish_async(struct async *async);
376
377int in_async(void);
378int async_with_fork(void);
379void check_pipe(int err);
380
381/**
382 * This callback should initialize the child process and preload the
383 * error channel if desired. The preloading of is useful if you want to
384 * have a message printed directly before the output of the child process.
385 * pp_cb is the callback cookie as passed to run_processes_parallel.
386 * You can store a child process specific callback cookie in pp_task_cb.
387 *
388 * See run_processes_parallel() below for a discussion of the "struct
389 * strbuf *out" parameter.
390 *
391 * Even after returning 0 to indicate that there are no more processes,
392 * this function will be called again until there are no more running
393 * child processes.
394 *
395 * Return 1 if the next child is ready to run.
396 * Return 0 if there are currently no more tasks to be processed.
397 * To send a signal to other child processes for abortion,
398 * return the negative signal number.
399 */
400typedef int (*get_next_task_fn)(struct child_process *cp,
401 struct strbuf *out,
402 void *pp_cb,
403 void **pp_task_cb);
404
405/**
406 * This callback is called whenever there are problems starting
407 * a new process.
408 *
409 * See run_processes_parallel() below for a discussion of the "struct
410 * strbuf *out" parameter.
411 *
412 * pp_cb is the callback cookie as passed into run_processes_parallel,
413 * pp_task_cb is the callback cookie as passed into get_next_task_fn.
414 *
415 * Return 0 to continue the parallel processing. To abort return non zero.
416 * To send a signal to other child processes for abortion, return
417 * the negative signal number.
418 */
419typedef int (*start_failure_fn)(struct strbuf *out,
420 void *pp_cb,
421 void *pp_task_cb);
422
423/**
424 * This callback is called on every child process that finished processing.
425 *
426 * See run_processes_parallel() below for a discussion of the "struct
427 * strbuf *out" parameter.
428 *
429 * pp_cb is the callback cookie as passed into run_processes_parallel,
430 * pp_task_cb is the callback cookie as passed into get_next_task_fn.
431 *
432 * Return 0 to continue the parallel processing. To abort return non zero.
433 * To send a signal to other child processes for abortion, return
434 * the negative signal number.
435 */
436typedef int (*task_finished_fn)(int result,
437 struct strbuf *out,
438 void *pp_cb,
439 void *pp_task_cb);
440
441/**
442 * Option used by run_processes_parallel(), { 0 }-initialized means no
443 * options.
444 */
445struct run_process_parallel_opts
446{
447 /**
448 * tr2_category & tr2_label: sets the trace2 category and label for
449 * logging. These must either be unset, or both of them must be set.
450 */
451 const char *tr2_category;
452 const char *tr2_label;
453
454 /**
455 * processes: see 'processes' in run_processes_parallel() below.
456 */
457 size_t processes;
458
459 /**
460 * ungroup: see 'ungroup' in run_processes_parallel() below.
461 */
462 unsigned int ungroup:1;
463
464 /**
465 * get_next_task: See get_next_task_fn() above. This must be
466 * specified.
467 */
468 get_next_task_fn get_next_task;
469
470 /**
471 * start_failure: See start_failure_fn() above. This can be
472 * NULL to omit any special handling.
473 */
474 start_failure_fn start_failure;
475
476 /**
477 * task_finished: See task_finished_fn() above. This can be
478 * NULL to omit any special handling.
479 */
480 task_finished_fn task_finished;
481
482 /**
483 * data: user data, will be passed as "pp_cb" to the callback
484 * parameters.
485 */
486 void *data;
487};
488
489/**
490 * Options are passed via the "struct run_process_parallel_opts" above.
491 *
492 * Runs N 'processes' at the same time. Whenever a process can be
493 * started, the callback opts.get_next_task is called to obtain the data
494 * required to start another child process.
495 *
496 * The children started via this function run in parallel. Their output
497 * (both stdout and stderr) is routed to stderr in a manner that output
498 * from different tasks does not interleave (but see "ungroup" below).
499 *
500 * If the "ungroup" option isn't specified, the API will set the
501 * "stdout_to_stderr" parameter in "struct child_process" and provide
502 * the callbacks with a "struct strbuf *out" parameter to write output
503 * to. In this case the callbacks must not write to stdout or
504 * stderr as such output will mess up the output of the other parallel
505 * processes. If "ungroup" option is specified callbacks will get a
506 * NULL "struct strbuf *out" parameter, and are responsible for
507 * emitting their own output, including dealing with any race
508 * conditions due to writing in parallel to stdout and stderr.
509 */
510void run_processes_parallel(const struct run_process_parallel_opts *opts);
511
512/**
513 * Convenience function which prepares env for a command to be run in a
514 * new repo. This adds all GIT_* environment variables to env with the
515 * exception of GIT_CONFIG_PARAMETERS and GIT_CONFIG_COUNT (which cause the
516 * corresponding environment variables to be unset in the subprocess) and adds
517 * an environment variable pointing to new_git_dir. See local_repo_env in
518 * environment.h for more information.
519 */
520void prepare_other_repo_env(struct strvec *env, const char *new_git_dir);
521
522/**
523 * Possible return values for start_bg_command().
524 */
525enum start_bg_result {
526 /* child process is "ready" */
527 SBGR_READY = 0,
528
529 /* child process could not be started */
530 SBGR_ERROR,
531
532 /* callback error when testing for "ready" */
533 SBGR_CB_ERROR,
534
535 /* timeout expired waiting for child to become "ready" */
536 SBGR_TIMEOUT,
537
538 /* child process exited or was signalled before becoming "ready" */
539 SBGR_DIED,
540};
541
542/**
543 * Callback used by start_bg_command() to ask whether the
544 * child process is ready or needs more time to become "ready".
545 *
546 * The callback will receive the cmd and cb_data arguments given to
547 * start_bg_command().
548 *
549 * Returns 1 is child needs more time (subject to the requested timeout).
550 * Returns 0 if child is "ready".
551 * Returns -1 on any error and cause start_bg_command() to also error out.
552 */
553typedef int(start_bg_wait_cb)(const struct child_process *cmd, void *cb_data);
554
555/**
556 * Start a command in the background. Wait long enough for the child
557 * to become "ready" (as defined by the provided callback). Capture
558 * immediate errors (like failure to start) and any immediate exit
559 * status (such as a shutdown/signal before the child became "ready")
560 * and return this like start_command().
561 *
562 * We run a custom wait loop using the provided callback to wait for
563 * the child to start and become "ready". This is limited by the given
564 * timeout value.
565 *
566 * If the child does successfully start and become "ready", we orphan
567 * it into the background.
568 *
569 * The caller must not call finish_command().
570 *
571 * The opaque cb_data argument will be forwarded to the callback for
572 * any instance data that it might require. This may be NULL.
573 */
574enum start_bg_result start_bg_command(struct child_process *cmd,
575 start_bg_wait_cb *wait_cb,
576 void *cb_data,
577 unsigned int timeout_sec);
578
579int sane_execvp(const char *file, char *const argv[]);
580
581#endif