eSpeak NG is an open source speech synthesizer that supports more than hundred languages and accents.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

mbrowrap.c 13KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613
  1. /*
  2. * mbrowrap -- A wrapper library around the mbrola binary
  3. * providing a subset of the API from the Windows mbrola DLL.
  4. *
  5. * Copyright (C) 2010 by Nicolas Pitre <[email protected]>
  6. * Copyright (C) 2013 Reece H. Dunn
  7. *
  8. * This program is free software; you can redistribute it and/or modify
  9. * it under the terms of the GNU General Public License as published by
  10. * the Free Software Foundation; either version 3 of the License, or
  11. * (at your option) any later version.
  12. *
  13. * This program is distributed in the hope that it will be useful,
  14. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16. * GNU General Public License for more details.
  17. */
  18. #include "config.h"
  19. #include "speech.h"
  20. #include <stdarg.h>
  21. #include <stdio.h>
  22. #include <unistd.h>
  23. #include <string.h>
  24. #include <stdlib.h>
  25. #include <signal.h>
  26. #include <fcntl.h>
  27. #include <poll.h>
  28. #include <errno.h>
  29. #include <sys/types.h>
  30. #include <sys/wait.h>
  31. #include "mbrowrap.h"
  32. /*
  33. * mbrola instance parameters
  34. */
  35. enum mbr_state {
  36. MBR_INACTIVE = 0,
  37. MBR_IDLE,
  38. MBR_NEWDATA,
  39. MBR_AUDIO,
  40. MBR_WEDGED
  41. };
  42. static enum mbr_state mbr_state;
  43. static char *mbr_voice_path;
  44. static int mbr_cmd_fd, mbr_audio_fd, mbr_error_fd, mbr_proc_stat;
  45. static pid_t mbr_pid;
  46. static int mbr_samplerate;
  47. static float mbr_volume = 1.0;
  48. static char mbr_errorbuf[160];
  49. struct datablock {
  50. struct datablock *next;
  51. int done;
  52. int size;
  53. char buffer[1]; /* 1 or more, dynamically allocated */
  54. };
  55. static struct datablock *mbr_pending_data_head, *mbr_pending_data_tail;
  56. /*
  57. * Private support code.
  58. */
  59. static void log(const char *msg, ...)
  60. {
  61. va_list params;
  62. va_start(params, msg);
  63. vfprintf(stderr, msg, params);
  64. fputc('\n', stderr);
  65. va_end(params);
  66. }
  67. static void err(const char *errmsg, ...)
  68. {
  69. va_list params;
  70. va_start(params, errmsg);
  71. vsnprintf(mbr_errorbuf, sizeof(mbr_errorbuf), errmsg, params);
  72. va_end(params);
  73. log("mbrowrap error: %s", mbr_errorbuf);
  74. }
  75. static int create_pipes(int p1[2], int p2[2], int p3[2])
  76. {
  77. int error;
  78. if (pipe(p1) != -1) {
  79. if (pipe(p2) != -1) {
  80. if (pipe(p3) != -1) {
  81. return 0;
  82. } else
  83. error = errno;
  84. close(p2[0]);
  85. close(p2[1]);
  86. } else
  87. error = errno;
  88. close(p1[0]);
  89. close(p1[1]);
  90. } else
  91. error = errno;
  92. err("pipe(): %s", strerror(error));
  93. return -1;
  94. }
  95. static void close_pipes(int p1[2], int p2[2], int p3[2])
  96. {
  97. close(p1[0]);
  98. close(p1[1]);
  99. close(p2[0]);
  100. close(p2[1]);
  101. close(p3[0]);
  102. close(p3[1]);
  103. }
  104. static int start_mbrola(const char *voice_path)
  105. {
  106. int error, p_stdin[2], p_stdout[2], p_stderr[2];
  107. ssize_t written;
  108. char charbuf[20];
  109. if (mbr_state != MBR_INACTIVE) {
  110. err("mbrola init request when already initialized");
  111. return -1;
  112. }
  113. error = create_pipes(p_stdin, p_stdout, p_stderr);
  114. if (error)
  115. return -1;
  116. mbr_pid = fork();
  117. if (mbr_pid == -1) {
  118. error = errno;
  119. close_pipes(p_stdin, p_stdout, p_stderr);
  120. err("fork(): %s", strerror(error));
  121. return -1;
  122. }
  123. if (mbr_pid == 0) {
  124. int i;
  125. if (dup2(p_stdin[0], 0) == -1 ||
  126. dup2(p_stdout[1], 1) == -1 ||
  127. dup2(p_stderr[1], 2) == -1) {
  128. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
  129. "dup2(): %s\n", strerror(errno));
  130. written = write(p_stderr[1], mbr_errorbuf, strlen(mbr_errorbuf));
  131. (void)written; // suppress 'variable not used' warning
  132. _exit(1);
  133. }
  134. for (i = p_stderr[1]; i > 2; i--)
  135. close(i);
  136. signal(SIGHUP, SIG_IGN);
  137. signal(SIGINT, SIG_IGN);
  138. signal(SIGQUIT, SIG_IGN);
  139. signal(SIGTERM, SIG_IGN);
  140. snprintf(charbuf, sizeof(charbuf), "%g", mbr_volume);
  141. execlp("mbrola", "mbrola", "-e", "-v", charbuf,
  142. voice_path, "-", "-.wav", (char *)NULL);
  143. /* if execution reaches this point then the exec() failed */
  144. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
  145. "mbrola: %s\n", strerror(errno));
  146. written = write(2, mbr_errorbuf, strlen(mbr_errorbuf));
  147. (void)written; // suppress 'variable not used' warning
  148. _exit(1);
  149. }
  150. snprintf(charbuf, sizeof(charbuf), "/proc/%d/stat", mbr_pid);
  151. mbr_proc_stat = open(charbuf, O_RDONLY);
  152. if (mbr_proc_stat == -1) {
  153. error = errno;
  154. close_pipes(p_stdin, p_stdout, p_stderr);
  155. waitpid(mbr_pid, NULL, 0);
  156. mbr_pid = 0;
  157. err("/proc is unaccessible: %s", strerror(error));
  158. return -1;
  159. }
  160. signal(SIGPIPE, SIG_IGN);
  161. if (fcntl(p_stdin[1], F_SETFL, O_NONBLOCK) == -1 ||
  162. fcntl(p_stdout[0], F_SETFL, O_NONBLOCK) == -1 ||
  163. fcntl(p_stderr[0], F_SETFL, O_NONBLOCK) == -1) {
  164. error = errno;
  165. close_pipes(p_stdin, p_stdout, p_stderr);
  166. waitpid(mbr_pid, NULL, 0);
  167. mbr_pid = 0;
  168. err("fcntl(): %s", strerror(error));
  169. return -1;
  170. }
  171. mbr_cmd_fd = p_stdin[1];
  172. mbr_audio_fd = p_stdout[0];
  173. mbr_error_fd = p_stderr[0];
  174. close(p_stdin[0]);
  175. close(p_stdout[1]);
  176. close(p_stderr[1]);
  177. mbr_state = MBR_IDLE;
  178. return 0;
  179. }
  180. static void stop_mbrola(void)
  181. {
  182. if (mbr_state == MBR_INACTIVE)
  183. return;
  184. close(mbr_proc_stat);
  185. close(mbr_cmd_fd);
  186. close(mbr_audio_fd);
  187. close(mbr_error_fd);
  188. if (mbr_pid) {
  189. kill(mbr_pid, SIGTERM);
  190. waitpid(mbr_pid, NULL, 0);
  191. mbr_pid = 0;
  192. }
  193. mbr_state = MBR_INACTIVE;
  194. }
  195. static void free_pending_data(void)
  196. {
  197. struct datablock *p, *head = mbr_pending_data_head;
  198. while (head) {
  199. p = head;
  200. head = head->next;
  201. free(p);
  202. }
  203. mbr_pending_data_head = NULL;
  204. mbr_pending_data_tail = NULL;
  205. }
  206. static int mbrola_died(void)
  207. {
  208. pid_t pid;
  209. int status, len;
  210. const char *msg;
  211. char msgbuf[80];
  212. pid = waitpid(mbr_pid, &status, WNOHANG);
  213. if (!pid) {
  214. msg = "mbrola closed stderr and did not exit";
  215. } else if (pid != mbr_pid) {
  216. msg = "waitpid() is confused";
  217. } else {
  218. mbr_pid = 0;
  219. if (WIFSIGNALED(status)) {
  220. int sig = WTERMSIG(status);
  221. snprintf(msgbuf, sizeof(msgbuf),
  222. "mbrola died by signal %d", sig);
  223. msg = msgbuf;
  224. } else if (WIFEXITED(status)) {
  225. int exst = WEXITSTATUS(status);
  226. snprintf(msgbuf, sizeof(msgbuf),
  227. "mbrola exited with status %d", exst);
  228. msg = msgbuf;
  229. } else {
  230. msg = "mbrola died and wait status is weird";
  231. }
  232. }
  233. log("mbrowrap error: %s", msg);
  234. len = strlen(mbr_errorbuf);
  235. if (!len)
  236. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf), "%s", msg);
  237. else
  238. snprintf(mbr_errorbuf + len, sizeof(mbr_errorbuf) - len,
  239. ", (%s)", msg);
  240. return -1;
  241. }
  242. static int mbrola_has_errors(void)
  243. {
  244. int result;
  245. char buffer[256];
  246. char *buf_ptr, *lf;
  247. buf_ptr = buffer;
  248. for (;;) {
  249. result = read(mbr_error_fd, buf_ptr,
  250. sizeof(buffer) - (buf_ptr - buffer) - 1);
  251. if (result == -1) {
  252. if (errno == EAGAIN)
  253. return 0;
  254. err("read(error): %s", strerror(errno));
  255. return -1;
  256. }
  257. if (result == 0) {
  258. /* EOF on stderr, assume mbrola died. */
  259. return mbrola_died();
  260. }
  261. buf_ptr[result] = 0;
  262. for (; (lf = strchr(buf_ptr, '\n')); buf_ptr = lf + 1) {
  263. /* inhibit the reset signal messages */
  264. if (strncmp(buf_ptr, "Got a reset signal", 18) == 0 ||
  265. strncmp(buf_ptr, "Input Flush Signal", 18) == 0)
  266. continue;
  267. *lf = 0;
  268. log("mbrola: %s", buf_ptr);
  269. /* is this the last line? */
  270. if (lf == &buf_ptr[result - 1]) {
  271. snprintf(mbr_errorbuf, sizeof(mbr_errorbuf),
  272. "%s", buf_ptr);
  273. /* don't consider this fatal at this point */
  274. return 0;
  275. }
  276. }
  277. memmove(buffer, buf_ptr, result);
  278. buf_ptr = buffer + result;
  279. }
  280. }
  281. static int send_to_mbrola(const char *cmd)
  282. {
  283. ssize_t result;
  284. int len;
  285. if (!mbr_pid)
  286. return -1;
  287. len = strlen(cmd);
  288. result = write(mbr_cmd_fd, cmd, len);
  289. if (result == -1) {
  290. int error = errno;
  291. if (error == EPIPE && mbrola_has_errors()) {
  292. return -1;
  293. } else if (error == EAGAIN) {
  294. result = 0;
  295. } else {
  296. err("write(): %s", strerror(error));
  297. return -1;
  298. }
  299. }
  300. if (result != len) {
  301. struct datablock *data;
  302. data = (struct datablock *)malloc(sizeof(*data) + len - result);
  303. if (data) {
  304. data->next = NULL;
  305. data->done = 0;
  306. data->size = len - result;
  307. memcpy(data->buffer, cmd + result, len - result);
  308. result = len;
  309. if (!mbr_pending_data_head)
  310. mbr_pending_data_head = data;
  311. else
  312. mbr_pending_data_tail->next = data;
  313. mbr_pending_data_tail = data;
  314. }
  315. }
  316. return result;
  317. }
  318. static int mbrola_is_idle(void)
  319. {
  320. char *p;
  321. char buffer[20]; /* looking for "12345 (mbrola) S" so 20 is plenty*/
  322. /* look in /proc to determine if mbrola is still running or sleeping */
  323. if (lseek(mbr_proc_stat, 0, SEEK_SET) != 0)
  324. return 0;
  325. if (read(mbr_proc_stat, buffer, sizeof(buffer)) != sizeof(buffer))
  326. return 0;
  327. p = (char *)memchr(buffer, ')', sizeof(buffer));
  328. if (!p || (unsigned)(p - buffer) >= sizeof(buffer) - 2)
  329. return 0;
  330. return p[1] == ' ' && p[2] == 'S';
  331. }
  332. static ssize_t receive_from_mbrola(void *buffer, size_t bufsize)
  333. {
  334. int result, wait = 1;
  335. size_t cursize = 0;
  336. if (!mbr_pid)
  337. return -1;
  338. do {
  339. struct pollfd pollfd[3];
  340. nfds_t nfds = 0;
  341. int idle;
  342. pollfd[0].fd = mbr_audio_fd;
  343. pollfd[0].events = POLLIN;
  344. nfds++;
  345. pollfd[1].fd = mbr_error_fd;
  346. pollfd[1].events = POLLIN;
  347. nfds++;
  348. if (mbr_pending_data_head) {
  349. pollfd[2].fd = mbr_cmd_fd;
  350. pollfd[2].events = POLLOUT;
  351. nfds++;
  352. }
  353. idle = mbrola_is_idle();
  354. result = poll(pollfd, nfds, idle ? 0 : wait);
  355. if (result == -1) {
  356. err("poll(): %s", strerror(errno));
  357. return -1;
  358. }
  359. if (result == 0) {
  360. if (idle) {
  361. mbr_state = MBR_IDLE;
  362. break;
  363. } else {
  364. if (wait >= 5000 * (4-1)/4) {
  365. mbr_state = MBR_WEDGED;
  366. err("mbrola process is stalled");
  367. break;
  368. } else {
  369. wait *= 4;
  370. continue;
  371. }
  372. }
  373. }
  374. wait = 1;
  375. if (pollfd[1].revents && mbrola_has_errors())
  376. return -1;
  377. if (mbr_pending_data_head && pollfd[2].revents) {
  378. struct datablock *head = mbr_pending_data_head;
  379. char *data = head->buffer + head->done;
  380. int left = head->size - head->done;
  381. result = write(mbr_cmd_fd, data, left);
  382. if (result == -1) {
  383. int error = errno;
  384. if (error == EPIPE && mbrola_has_errors())
  385. return -1;
  386. err("write(): %s", strerror(error));
  387. return -1;
  388. }
  389. if (result != left) {
  390. head->done += result;
  391. } else {
  392. mbr_pending_data_head = head->next;
  393. free(head);
  394. if (!mbr_pending_data_head)
  395. mbr_pending_data_tail = NULL;
  396. else
  397. continue;
  398. }
  399. }
  400. if (pollfd[0].revents) {
  401. char *curpos = (char *)buffer + cursize;
  402. size_t space = bufsize - cursize;
  403. ssize_t obtained = read(mbr_audio_fd, curpos, space);
  404. if (obtained == -1) {
  405. err("read(): %s", strerror(errno));
  406. return -1;
  407. }
  408. cursize += obtained;
  409. mbr_state = MBR_AUDIO;
  410. }
  411. } while (cursize < bufsize);
  412. return cursize;
  413. }
  414. /*
  415. * API functions.
  416. */
  417. int init_MBR(const char *voice_path)
  418. {
  419. int error, result;
  420. unsigned char wavhdr[45];
  421. error = start_mbrola(voice_path);
  422. if (error)
  423. return -1;
  424. result = send_to_mbrola("#\n");
  425. if (result != 2) {
  426. stop_mbrola();
  427. return -1;
  428. }
  429. /* we should actually be getting only 44 bytes */
  430. result = receive_from_mbrola(wavhdr, 45);
  431. if (result != 44) {
  432. if (result >= 0)
  433. err("unable to get .wav header from mbrola");
  434. stop_mbrola();
  435. return -1;
  436. }
  437. /* parse wavhdr to get mbrola voice samplerate */
  438. if (memcmp(wavhdr, "RIFF", 4) != 0 ||
  439. memcmp(wavhdr+8, "WAVEfmt ", 8) != 0) {
  440. err("mbrola did not return a .wav header");
  441. stop_mbrola();
  442. return -1;
  443. }
  444. mbr_samplerate = wavhdr[24] + (wavhdr[25]<<8) +
  445. (wavhdr[26]<<16) + (wavhdr[27]<<24);
  446. // log("mbrowrap: voice samplerate = %d", mbr_samplerate);
  447. /* remember the voice path for setVolumeRatio_MBR() */
  448. if (mbr_voice_path != voice_path) {
  449. free(mbr_voice_path);
  450. mbr_voice_path = strdup(voice_path);
  451. }
  452. return 0;
  453. }
  454. void close_MBR(void)
  455. {
  456. stop_mbrola();
  457. free_pending_data();
  458. free(mbr_voice_path);
  459. mbr_voice_path = NULL;
  460. mbr_volume = 1.0;
  461. }
  462. int reset_MBR()
  463. {
  464. int result, success = 1;
  465. char dummybuf[4096];
  466. if (mbr_state == MBR_IDLE)
  467. return 1;
  468. if (!mbr_pid)
  469. return 0;
  470. if (kill(mbr_pid, SIGUSR1) == -1)
  471. success = 0;
  472. free_pending_data();
  473. result = write(mbr_cmd_fd, "\n#\n", 3);
  474. if (result != 3)
  475. success = 0;
  476. do {
  477. result = read(mbr_audio_fd, dummybuf, sizeof(dummybuf));
  478. } while (result > 0);
  479. if (result != -1 || errno != EAGAIN)
  480. success = 0;
  481. if (!mbrola_has_errors() && success)
  482. mbr_state = MBR_IDLE;
  483. return success;
  484. }
  485. int read_MBR(void *buffer, int nb_samples)
  486. {
  487. int result = receive_from_mbrola(buffer, nb_samples * 2);
  488. if (result > 0)
  489. result /= 2;
  490. return result;
  491. }
  492. int write_MBR(const char *data)
  493. {
  494. mbr_state = MBR_NEWDATA;
  495. return send_to_mbrola(data);
  496. }
  497. int flush_MBR(void)
  498. {
  499. return send_to_mbrola("\n#\n") == 3;
  500. }
  501. int getFreq_MBR(void)
  502. {
  503. return mbr_samplerate;
  504. }
  505. void setVolumeRatio_MBR(float value)
  506. {
  507. if (value == mbr_volume)
  508. return;
  509. mbr_volume = value;
  510. if (mbr_state != MBR_IDLE)
  511. return;
  512. /*
  513. * We have no choice but to kill and restart mbrola with
  514. * the new argument here.
  515. */
  516. stop_mbrola();
  517. init_MBR(mbr_voice_path);
  518. }
  519. int lastErrorStr_MBR(char *buffer, int bufsize)
  520. {
  521. int result;
  522. if (mbr_pid)
  523. mbrola_has_errors();
  524. result = snprintf(buffer, bufsize, "%s", mbr_errorbuf);
  525. return result >= bufsize ? (bufsize - 1) : result;
  526. }
  527. void resetError_MBR(void)
  528. {
  529. mbr_errorbuf[0] = 0;
  530. }