Mercurial > hg > xemacs-beta
annotate src/process-unix.c @ 4974:fe0d3106cc36
fix compile problems in bytecode.c when no error-check-byte-code (issue 666)
-------------------- ChangeLog entries follow: --------------------
src/ChangeLog addition:
2010-02-05 Ben Wing <ben@xemacs.org>
* bytecode.c (bytecode_assert):
Use 0 instead of x in disabled_assert calls to avoid errors due
to using undefined vars/params (they are defined only when
ERROR_CHECK_BYTE_CODE, which also turns on/off the asserts).
| author | Ben Wing <ben@xemacs.org> |
|---|---|
| date | Fri, 05 Feb 2010 04:07:47 -0600 |
| parents | 304aebb79cd3 |
| children | 4aebb0131297 |
| rev | line source |
|---|---|
| 428 | 1 /* Asynchronous subprocess implementation for UNIX |
| 2 Copyright (C) 1985, 1986, 1987, 1988, 1992, 1993, 1994, 1995 | |
| 3 Free Software Foundation, Inc. | |
| 4 Copyright (C) 1995 Sun Microsystems, Inc. | |
| 1330 | 5 Copyright (C) 1995, 1996, 2001, 2002, 2003 Ben Wing. |
| 428 | 6 |
| 7 This file is part of XEmacs. | |
| 8 | |
| 9 XEmacs is free software; you can redistribute it and/or modify it | |
| 10 under the terms of the GNU General Public License as published by the | |
| 11 Free Software Foundation; either version 2, or (at your option) any | |
| 12 later version. | |
| 13 | |
| 14 XEmacs is distributed in the hope that it will be useful, but WITHOUT | |
| 15 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
| 16 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
| 17 for more details. | |
| 18 | |
| 19 You should have received a copy of the GNU General Public License | |
| 20 along with XEmacs; see the file COPYING. If not, write to | |
| 21 the Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
| 22 Boston, MA 02111-1307, USA. */ | |
| 23 | |
| 771 | 24 /* Mule-ized as of 6-14-00 */ |
| 428 | 25 |
| 26 /* This file has been split into process.c and process-unix.c by | |
| 27 Kirill M. Katsnelson <kkm@kis.ru>, so please bash him and not | |
| 28 the original author(s) */ | |
| 29 | |
| 440 | 30 /* The IPv6 support is derived from the code for GNU Emacs-20.3 |
| 31 written by Wolfgang S. Rupprecht */ | |
| 32 | |
| 428 | 33 #include <config.h> |
| 34 | |
| 35 #include "lisp.h" | |
| 36 | |
| 37 #include "buffer.h" | |
| 38 #include "events.h" | |
| 39 #include "frame.h" | |
| 40 #include "hash.h" | |
| 41 #include "lstream.h" | |
| 42 #include "opaque.h" | |
| 43 #include "process.h" | |
| 44 #include "procimpl.h" | |
| 45 #include "sysdep.h" | |
| 46 #include "window.h" | |
| 47 #include "file-coding.h" | |
| 48 | |
| 49 #include <setjmp.h> | |
| 853 | 50 #include "sysdir.h" |
| 428 | 51 #include "sysfile.h" |
| 52 #include "sysproc.h" | |
| 859 | 53 #include "syssignal.h" |
| 428 | 54 #include "systime.h" |
| 55 #include "systty.h" | |
| 56 #include "syswait.h" | |
| 57 | |
| 442 | 58 #ifdef HPUX |
| 59 #include <grp.h> /* See grantpt fixups for HPUX below. */ | |
| 60 #endif | |
| 428 | 61 |
| 502 | 62 #if defined (HAVE_GETADDRINFO) && defined (HAVE_GETNAMEINFO) |
| 63 #define USE_GETADDRINFO | |
| 64 #endif | |
| 65 | |
| 66 | |
| 428 | 67 /* |
| 68 * Implementation-specific data. Pointed to by Lisp_Process->process_data | |
| 69 */ | |
| 70 | |
| 71 struct unix_process_data | |
| 72 { | |
| 73 /* Non-0 if this is really a ToolTalk channel. */ | |
| 74 int connected_via_filedesc_p; | |
| 75 /* Descriptor by which we read from this process. -1 for dead process */ | |
| 76 int infd; | |
| 853 | 77 /* Descriptor by which we read stderr from this process. -1 for |
| 78 dead process */ | |
| 79 int errfd; | |
| 428 | 80 /* Descriptor for the tty which this process is using. |
| 81 -1 if we didn't record it (on some systems, there's no need). */ | |
| 82 int subtty; | |
| 83 /* Non-false if communicating through a pty. */ | |
| 84 char pty_flag; | |
| 85 }; | |
| 853 | 86 #define UNIX_DATA(p) ((struct unix_process_data*) ((p)->process_data)) |
| 428 | 87 |
| 88 | |
| 89 | |
| 90 /**********************************************************************/ | |
| 91 /* Static helper routines */ | |
| 92 /**********************************************************************/ | |
| 93 | |
| 94 static SIGTYPE | |
| 2286 | 95 close_safely_handler (int SIG_ARG_MAYBE_UNUSED (signo)) |
| 428 | 96 { |
| 97 EMACS_REESTABLISH_SIGNAL (signo, close_safely_handler); | |
| 98 SIGRETURN; | |
| 99 } | |
| 100 | |
| 101 static void | |
| 102 close_safely (int fd) | |
| 103 { | |
| 104 stop_interrupts (); | |
| 613 | 105 set_timeout_signal (SIGALRM, close_safely_handler); |
| 428 | 106 alarm (1); |
| 771 | 107 retry_close (fd); |
| 428 | 108 alarm (0); |
| 109 start_interrupts (); | |
| 110 } | |
| 111 | |
| 112 static void | |
| 113 close_descriptor_pair (int in, int out) | |
| 114 { | |
| 115 if (in >= 0) | |
| 771 | 116 retry_close (in); |
| 428 | 117 if (out != in && out >= 0) |
| 771 | 118 retry_close (out); |
| 428 | 119 } |
| 120 | |
| 121 /* Close all descriptors currently in use for communication | |
| 122 with subprocess. This is used in a newly-forked subprocess | |
| 123 to get rid of irrelevant descriptors. */ | |
| 124 | |
| 125 static int | |
| 2286 | 126 close_process_descs_mapfun (const void *UNUSED (key), void *contents, |
| 127 void *UNUSED (arg)) | |
| 428 | 128 { |
| 853 | 129 Lisp_Object proc = VOID_TO_LISP (contents); |
| 130 USID vaffan, culo; | |
| 131 | |
| 132 event_stream_delete_io_streams (XPROCESS (proc)->pipe_instream, | |
| 133 XPROCESS (proc)->pipe_outstream, | |
| 134 XPROCESS (proc)->pipe_errstream, | |
| 135 &vaffan, &culo); | |
| 428 | 136 return 0; |
| 137 } | |
| 138 | |
| 139 void | |
| 140 close_process_descs (void) | |
| 141 { | |
| 142 maphash (close_process_descs_mapfun, usid_to_process, 0); | |
| 143 } | |
| 144 | |
| 145 /* connect to an existing file descriptor. This is very similar to | |
| 146 open-network-stream except that it assumes that the connection has | |
| 147 already been initialized. It is currently used for ToolTalk | |
| 148 communication. */ | |
| 149 | |
| 150 /* This function used to be visible on the Lisp level, but there is no | |
| 151 real point in doing that. Here is the doc string: | |
| 152 | |
| 442 | 153 "Connect to an existing file descriptor. |
| 154 Return a subprocess-object to represent the connection. | |
| 155 Input and output work as for subprocesses; `delete-process' closes it. | |
| 156 Args are NAME BUFFER INFD OUTFD. | |
| 157 NAME is name for process. It is modified if necessary to make it unique. | |
| 158 BUFFER is the buffer (or buffer-name) to associate with the process. | |
| 159 Process output goes at end of that buffer, unless you specify | |
| 160 an output stream or filter function to handle the output. | |
| 161 BUFFER may also be nil, meaning that this process is not associated | |
| 162 with any buffer. | |
| 163 INFD and OUTFD specify the file descriptors to use for input and | |
| 428 | 164 output, respectively." |
| 165 */ | |
| 166 | |
| 167 Lisp_Object | |
| 168 connect_to_file_descriptor (Lisp_Object name, Lisp_Object buffer, | |
| 169 Lisp_Object infd, Lisp_Object outfd) | |
| 170 { | |
| 171 /* This function can GC */ | |
| 172 Lisp_Object proc; | |
| 4123 | 173 EMACS_INT inch; |
| 428 | 174 |
| 175 CHECK_STRING (name); | |
| 176 CHECK_INT (infd); | |
| 177 CHECK_INT (outfd); | |
| 178 | |
| 179 inch = XINT (infd); | |
| 442 | 180 if (get_process_from_usid (FD_TO_USID (inch))) |
| 181 invalid_operation ("There is already a process connected to fd", infd); | |
| 428 | 182 if (!NILP (buffer)) |
| 183 buffer = Fget_buffer_create (buffer); | |
| 184 proc = make_process_internal (name); | |
| 185 | |
| 186 XPROCESS (proc)->pid = Fcons (infd, name); | |
| 187 XPROCESS (proc)->buffer = buffer; | |
| 853 | 188 init_process_io_handles (XPROCESS (proc), (void *) inch, |
| 189 (void *) XINT (outfd), (void *) -1, 0); | |
| 428 | 190 UNIX_DATA (XPROCESS (proc))->connected_via_filedesc_p = 1; |
| 191 | |
| 853 | 192 event_stream_select_process (XPROCESS (proc), 1, 1); |
| 428 | 193 |
| 194 return proc; | |
| 195 } | |
| 196 | |
| 442 | 197 static int allocate_pty_the_old_fashioned_way (void); |
| 198 | |
| 199 /* The file name of the (slave) pty opened by allocate_pty(). */ | |
| 200 #ifndef MAX_PTYNAME_LEN | |
| 201 #define MAX_PTYNAME_LEN 64 | |
| 202 #endif | |
| 867 | 203 static Ibyte pty_name[MAX_PTYNAME_LEN]; |
| 428 | 204 |
| 205 /* Open an available pty, returning a file descriptor. | |
| 206 Return -1 on failure. | |
| 207 The file name of the terminal corresponding to the pty | |
| 442 | 208 is left in the variable `pty_name'. */ |
| 428 | 209 |
| 210 static int | |
| 211 allocate_pty (void) | |
| 212 { | |
| 442 | 213 /* Unix98 standardized grantpt, unlockpt, and ptsname, but not the |
| 214 functions required to open a master pty in the first place :-( | |
| 215 | |
| 216 Modern Unix systems all seems to have convenience methods to open | |
| 217 a master pty fd in one function call, but there is little | |
| 218 agreement on how to do it. | |
| 219 | |
| 220 allocate_pty() tries all the different known easy ways of opening | |
| 221 a pty. In case of failure, we resort to the old BSD-style pty | |
| 222 grovelling code in allocate_pty_the_old_fashioned_way(). */ | |
| 223 int master_fd = -1; | |
| 771 | 224 const Extbyte *slave_name = NULL; |
| 867 | 225 const CIbyte *clone = NULL; |
| 226 static const CIbyte * const clones[] = | |
| 771 | 227 /* Different pty master clone devices */ |
| 442 | 228 { |
| 229 "/dev/ptmx", /* Various systems */ | |
| 230 "/dev/ptm/clone", /* HPUX */ | |
| 231 "/dev/ptc", /* AIX */ | |
| 232 "/dev/ptmx_bsd" /* Tru64 */ | |
| 233 }; | |
| 234 | |
| 235 #ifdef HAVE_GETPT /* glibc */ | |
| 236 master_fd = getpt (); | |
| 237 if (master_fd >= 0) | |
| 238 goto have_master; | |
| 239 #endif /* HAVE_GETPT */ | |
| 240 | |
| 241 | |
| 242 #if defined(HAVE_OPENPTY) /* BSD, Tru64, glibc */ | |
| 243 { | |
| 244 int slave_fd = -1; | |
| 245 int rc; | |
| 246 EMACS_BLOCK_SIGNAL (SIGCHLD); | |
| 247 rc = openpty (&master_fd, &slave_fd, NULL, NULL, NULL); | |
| 248 EMACS_UNBLOCK_SIGNAL (SIGCHLD); | |
| 249 if (rc == 0) | |
| 250 { | |
| 251 slave_name = ttyname (slave_fd); | |
| 771 | 252 retry_close (slave_fd); |
| 442 | 253 goto have_slave_name; |
| 254 } | |
| 255 else | |
| 256 { | |
| 257 if (master_fd >= 0) | |
| 771 | 258 retry_close (master_fd); |
| 442 | 259 if (slave_fd >= 0) |
| 771 | 260 retry_close (slave_fd); |
| 442 | 261 } |
| 262 } | |
| 263 #endif /* HAVE_OPENPTY */ | |
| 264 | |
| 265 #if defined(HAVE__GETPTY) && defined (O_NDELAY) /* SGI */ | |
| 266 master_fd = -1; | |
| 267 EMACS_BLOCK_SIGNAL (SIGCHLD); | |
| 268 slave_name = _getpty (&master_fd, O_RDWR | O_NDELAY, 0600, 0); | |
| 269 EMACS_UNBLOCK_SIGNAL (SIGCHLD); | |
| 270 if (master_fd >= 0 && slave_name != NULL) | |
| 271 goto have_slave_name; | |
| 272 #endif /* HAVE__GETPTY */ | |
| 273 | |
| 274 /* Master clone devices are available on most systems */ | |
| 275 { | |
| 276 int i; | |
| 277 for (i = 0; i < countof (clones); i++) | |
| 278 { | |
| 279 clone = clones[i]; | |
| 867 | 280 master_fd = qxe_open ((Ibyte *) clone, |
| 771 | 281 O_RDWR | O_NONBLOCK | OPEN_BINARY, 0); |
| 442 | 282 if (master_fd >= 0) |
| 283 goto have_master; | |
| 284 } | |
| 285 clone = NULL; | |
| 286 } | |
| 287 | |
| 288 goto lose; | |
| 289 | |
| 290 have_master: | |
| 291 | |
| 292 #if defined (HAVE_PTSNAME) | |
| 293 slave_name = ptsname (master_fd); | |
| 294 if (slave_name) | |
| 295 goto have_slave_name; | |
| 296 #endif | |
| 297 | |
| 298 /* AIX docs say to use ttyname, not ptsname, to get slave_name */ | |
| 299 if (clone | |
| 300 && !strcmp (clone, "/dev/ptc") | |
| 301 && (slave_name = ttyname (master_fd)) != NULL) | |
| 302 goto have_slave_name; | |
| 303 | |
| 304 goto lose; | |
| 305 | |
| 306 have_slave_name: | |
| 771 | 307 { |
| 867 | 308 Ibyte *slaveint; |
| 771 | 309 |
| 310 EXTERNAL_TO_C_STRING (slave_name, slaveint, Qfile_name); | |
| 311 qxestrncpy (pty_name, slaveint, sizeof (pty_name)); | |
| 312 } | |
| 313 | |
| 442 | 314 pty_name[sizeof (pty_name) - 1] = '\0'; |
| 315 setup_pty (master_fd); | |
| 316 | |
| 317 /* We jump through some hoops to frob the pty. | |
| 318 It's not obvious that checking the return code here is useful. */ | |
| 319 | |
| 320 /* "The grantpt() function will fail if it is unable to successfully | |
| 321 invoke the setuid root program. It may also fail if the | |
| 322 application has installed a signal handler to catch SIGCHLD | |
| 323 signals." */ | |
| 324 #if defined (HAVE_GRANTPT) || defined (HAVE_UNLOCKPT) | |
| 325 EMACS_BLOCK_SIGNAL (SIGCHLD); | |
| 326 | |
| 327 #if defined (HAVE_GRANTPT) | |
| 328 grantpt (master_fd); | |
| 329 #ifdef HPUX | |
| 330 /* grantpt() behavior on some versions of HP-UX differs from what's | |
| 331 specified in the man page: the group of the slave PTY is set to | |
| 332 the user's primary group, and we fix that. */ | |
| 333 { | |
| 334 struct group *tty_group = getgrnam ("tty"); | |
| 335 if (tty_group != NULL) | |
| 771 | 336 { |
| 337 Extbyte *ptyout; | |
| 338 | |
| 339 C_STRING_TO_EXTERNAL (pty_name, ptyout, Qfile_name); | |
| 340 chown (ptyout, (uid_t) -1, tty_group->gr_gid); | |
| 341 } | |
| 442 | 342 } |
| 343 #endif /* HPUX has broken grantpt() */ | |
| 344 #endif /* HAVE_GRANTPT */ | |
| 345 | |
| 346 #if defined (HAVE_UNLOCKPT) | |
| 347 unlockpt (master_fd); | |
| 348 #endif | |
| 349 | |
| 350 EMACS_UNBLOCK_SIGNAL (SIGCHLD); | |
| 351 #endif /* HAVE_GRANTPT || HAVE_UNLOCKPT */ | |
| 352 | |
| 353 return master_fd; | |
| 354 | |
| 355 lose: | |
| 356 if (master_fd >= 0) | |
| 771 | 357 retry_close (master_fd); |
| 442 | 358 return allocate_pty_the_old_fashioned_way (); |
| 359 } | |
| 360 | |
| 361 /* This function tries to allocate a pty by iterating through file | |
| 362 pairs with names like /dev/ptyp1 and /dev/ttyp1. */ | |
| 363 static int | |
| 364 allocate_pty_the_old_fashioned_way (void) | |
| 365 { | |
| 428 | 366 struct stat stb; |
| 367 | |
| 368 /* Some systems name their pseudoterminals so that there are gaps in | |
| 369 the usual sequence - for example, on HP9000/S700 systems, there | |
| 370 are no pseudoterminals with names ending in 'f'. So we wait for | |
| 371 three failures in a row before deciding that we've reached the | |
| 372 end of the ptys. */ | |
| 373 int failed_count = 0; | |
| 374 int fd; | |
| 375 int i; | |
| 376 int c; | |
| 377 | |
| 378 #ifdef PTY_ITERATION | |
| 379 PTY_ITERATION | |
| 380 #else | |
| 442 | 381 # ifndef FIRST_PTY_LETTER |
| 382 # define FIRST_PTY_LETTER 'p' | |
| 383 # endif | |
| 428 | 384 for (c = FIRST_PTY_LETTER; c <= 'z'; c++) |
| 385 for (i = 0; i < 16; i++) | |
| 442 | 386 #endif /* PTY_ITERATION */ |
| 387 | |
| 428 | 388 { |
| 389 #ifdef PTY_NAME_SPRINTF | |
| 390 PTY_NAME_SPRINTF | |
| 391 #else | |
| 771 | 392 qxesprintf (pty_name, "/dev/pty%c%x", c, i); |
| 428 | 393 #endif /* no PTY_NAME_SPRINTF */ |
| 394 | |
| 771 | 395 if (qxe_stat (pty_name, &stb) < 0) |
| 428 | 396 { |
| 442 | 397 if (++failed_count >= 3) |
| 428 | 398 return -1; |
| 399 } | |
| 400 else | |
| 401 failed_count = 0; | |
| 771 | 402 fd = qxe_open (pty_name, O_RDWR | O_NONBLOCK | OPEN_BINARY, 0); |
| 428 | 403 |
| 404 if (fd >= 0) | |
| 405 { | |
| 406 #ifdef PTY_TTY_NAME_SPRINTF | |
| 407 PTY_TTY_NAME_SPRINTF | |
| 408 #else | |
| 771 | 409 qxesprintf (pty_name, "/dev/tty%c%x", c, i); |
| 428 | 410 #endif /* no PTY_TTY_NAME_SPRINTF */ |
| 771 | 411 if (qxe_access (pty_name, R_OK | W_OK) == 0) |
| 428 | 412 { |
| 442 | 413 setup_pty (fd); |
| 414 return fd; | |
| 428 | 415 } |
| 771 | 416 retry_close (fd); |
| 428 | 417 } |
| 442 | 418 } /* iteration */ |
| 428 | 419 return -1; |
| 420 } | |
| 421 | |
| 422 static int | |
| 4123 | 423 create_bidirectional_pipe (EMACS_INT *inchannel, EMACS_INT *outchannel, |
| 424 volatile EMACS_INT *forkin, volatile EMACS_INT *forkout) | |
| 428 | 425 { |
| 426 int sv[2]; | |
| 427 | |
|
4759
aa5ed11f473b
Remove support for obsolete systems. See xemacs-patches message with ID
Jerry James <james@xemacs.org>
parents:
4123
diff
changeset
|
428 if (pipe (sv) < 0) return -1; |
| 428 | 429 *inchannel = sv[0]; |
| 430 *forkout = sv[1]; | |
|
4759
aa5ed11f473b
Remove support for obsolete systems. See xemacs-patches message with ID
Jerry James <james@xemacs.org>
parents:
4123
diff
changeset
|
431 if (pipe (sv) < 0) return -1; |
| 428 | 432 *outchannel = sv[1]; |
| 433 *forkin = sv[0]; | |
| 434 return 0; | |
| 435 } | |
| 436 | |
| 437 | |
| 438 #ifdef HAVE_SOCKETS | |
| 439 | |
| 502 | 440 #ifndef USE_GETADDRINFO |
| 428 | 441 static int |
| 442 get_internet_address (Lisp_Object host, struct sockaddr_in *address, | |
| 578 | 443 Error_Behavior errb) |
| 428 | 444 { |
| 445 struct hostent *host_info_ptr = NULL; | |
| 446 #ifdef TRY_AGAIN | |
| 447 int count = 0; | |
| 448 #endif | |
| 449 | |
| 450 xzero (*address); | |
| 451 | |
| 452 while (1) | |
| 453 { | |
| 771 | 454 Extbyte *hostext; |
| 455 | |
| 428 | 456 #ifdef TRY_AGAIN |
| 457 if (count++ > 10) break; | |
| 458 h_errno = 0; | |
| 459 #endif | |
| 771 | 460 |
| 1204 | 461 LISP_STRING_TO_EXTERNAL (host, hostext, Qunix_host_name_encoding); |
| 771 | 462 |
| 428 | 463 /* Some systems can't handle SIGIO/SIGALARM in gethostbyname. */ |
| 464 slow_down_interrupts (); | |
| 771 | 465 host_info_ptr = gethostbyname (hostext); |
| 428 | 466 speed_up_interrupts (); |
| 467 #ifdef TRY_AGAIN | |
| 468 if (! (host_info_ptr == 0 && h_errno == TRY_AGAIN)) | |
| 469 #endif | |
| 470 break; | |
| 471 Fsleep_for (make_int (1)); | |
| 472 } | |
| 473 if (host_info_ptr) | |
| 474 { | |
| 475 address->sin_family = host_info_ptr->h_addrtype; | |
| 502 | 476 memcpy (&address->sin_addr, host_info_ptr->h_addr, |
| 477 host_info_ptr->h_length); | |
| 428 | 478 } |
| 479 else | |
| 480 { | |
| 481 IN_ADDR numeric_addr; | |
| 1204 | 482 Extbyte *hostext; |
| 483 | |
| 428 | 484 /* Attempt to interpret host as numeric inet address */ |
| 1204 | 485 LISP_STRING_TO_EXTERNAL (host, hostext, Qunix_host_name_encoding); |
| 486 numeric_addr = inet_addr (hostext); | |
| 428 | 487 if (NUMERIC_ADDR_ERROR) |
| 488 { | |
| 563 | 489 maybe_signal_error (Qio_error, "Unknown host", host, |
| 1204 | 490 Qprocess, errb); |
| 428 | 491 return 0; |
| 492 } | |
| 493 | |
| 494 /* There was some broken code here that called strlen() here | |
| 495 on (char *) &numeric_addr and even sometimes accessed | |
| 496 uninitialized data. */ | |
| 497 address->sin_family = AF_INET; | |
| 498 * (IN_ADDR *) &address->sin_addr = numeric_addr; | |
| 499 } | |
| 500 | |
| 501 return 1; | |
| 502 } | |
| 502 | 503 #endif /* !USE_GETADDRINFO */ |
| 428 | 504 |
| 505 static void | |
| 2286 | 506 set_socket_nonblocking_maybe (int fd, |
| 507 #ifdef PROCESS_IO_BLOCKING | |
| 508 int port, const char *proto | |
| 509 #else | |
| 510 int UNUSED (port), const char *UNUSED (proto) | |
| 511 #endif | |
| 512 ) | |
| 428 | 513 { |
| 514 #ifdef PROCESS_IO_BLOCKING | |
| 515 Lisp_Object tail; | |
| 516 | |
| 517 for (tail = network_stream_blocking_port_list; CONSP (tail); tail = XCDR (tail)) | |
| 518 { | |
| 519 Lisp_Object tail_port = XCAR (tail); | |
| 520 | |
| 521 if (STRINGP (tail_port)) | |
| 522 { | |
| 523 struct servent *svc_info; | |
| 771 | 524 Extbyte *tailportext; |
| 525 | |
| 428 | 526 CHECK_STRING (tail_port); |
| 771 | 527 TO_EXTERNAL_FORMAT (LISP_STRING, tail_port, C_STRING_ALLOCA, |
|
4834
b3ea9c582280
Use new cygwin_conv_path API with Cygwin 1.7 for converting names between Win32 and POSIX, UTF-8-aware, with attendant changes elsewhere
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
528 tailportext, Qunix_service_name_encoding); |
| 771 | 529 |
| 530 svc_info = getservbyname (tailportext, proto); | |
| 428 | 531 if ((svc_info != 0) && (svc_info->s_port == port)) |
| 532 break; | |
| 533 else | |
| 534 continue; | |
| 535 } | |
| 536 else if (INTP (tail_port) && (htons ((unsigned short) XINT (tail_port)) == port)) | |
| 537 break; | |
| 538 } | |
| 539 | |
| 540 if (!CONSP (tail)) | |
| 541 { | |
| 542 set_descriptor_non_blocking (fd); | |
| 543 } | |
| 544 #else | |
| 545 set_descriptor_non_blocking (fd); | |
| 546 #endif /* PROCESS_IO_BLOCKING */ | |
| 547 } | |
| 548 | |
| 549 #endif /* HAVE_SOCKETS */ | |
| 550 | |
| 551 /* Compute the Lisp form of the process status from | |
| 552 the numeric status that was returned by `wait'. */ | |
| 553 | |
| 554 static void | |
| 440 | 555 update_status_from_wait_code (Lisp_Process *p, int *w_fmh) |
| 428 | 556 { |
| 557 /* C compiler lossage when attempting to pass w directly */ | |
| 558 int w = *w_fmh; | |
| 559 | |
| 560 if (WIFSTOPPED (w)) | |
| 561 { | |
| 562 p->status_symbol = Qstop; | |
| 563 p->exit_code = WSTOPSIG (w); | |
| 564 p->core_dumped = 0; | |
| 565 } | |
| 566 else if (WIFEXITED (w)) | |
| 567 { | |
| 568 p->status_symbol = Qexit; | |
| 569 p->exit_code = WEXITSTATUS (w); | |
| 570 p->core_dumped = 0; | |
| 571 } | |
| 572 else if (WIFSIGNALED (w)) | |
| 573 { | |
| 574 p->status_symbol = Qsignal; | |
| 575 p->exit_code = WTERMSIG (w); | |
| 576 p->core_dumped = WCOREDUMP (w); | |
| 577 } | |
| 578 else | |
| 579 { | |
| 580 p->status_symbol = Qrun; | |
| 581 p->exit_code = 0; | |
| 582 } | |
| 583 } | |
| 584 | |
| 585 #ifdef SIGCHLD | |
| 586 | |
| 587 #define MAX_EXITED_PROCESSES 1000 | |
| 588 static volatile pid_t exited_processes[MAX_EXITED_PROCESSES]; | |
| 589 static volatile int exited_processes_status[MAX_EXITED_PROCESSES]; | |
| 590 static volatile int exited_processes_index; | |
| 591 | |
| 592 static volatile int sigchld_happened; | |
| 593 | |
| 594 /* On receipt of a signal that a child status has changed, | |
| 595 loop asking about children with changed statuses until | |
| 596 the system says there are no more. All we do is record | |
| 597 the processes and wait status. | |
| 598 | |
| 599 This function could be called from within the SIGCHLD | |
| 600 handler, so it must be completely reentrant. When | |
| 601 not called from a SIGCHLD handler, BLOCK_SIGCHLD should | |
| 602 be non-zero so that SIGCHLD is blocked while this | |
| 603 function is running. (This is necessary so avoid | |
| 604 race conditions with the SIGCHLD_HAPPENED flag). */ | |
| 605 | |
| 606 static void | |
| 607 record_exited_processes (int block_sigchld) | |
| 608 { | |
| 609 if (!sigchld_happened) | |
| 610 { | |
| 611 return; | |
| 612 } | |
| 613 | |
| 614 #ifdef EMACS_BLOCK_SIGNAL | |
| 615 if (block_sigchld) | |
| 616 EMACS_BLOCK_SIGNAL (SIGCHLD); | |
| 617 #endif | |
| 618 | |
| 619 while (sigchld_happened) | |
| 620 { | |
| 621 int pid; | |
| 622 int w; | |
| 623 | |
| 624 /* Keep trying to get a status until we get a definitive result. */ | |
| 625 do | |
| 626 { | |
| 627 errno = 0; | |
| 628 #ifdef WNOHANG | |
| 629 # ifndef WUNTRACED | |
| 630 # define WUNTRACED 0 | |
| 631 # endif /* not WUNTRACED */ | |
| 632 # ifdef HAVE_WAITPID | |
| 633 pid = waitpid ((pid_t) -1, &w, WNOHANG | WUNTRACED); | |
| 634 # else | |
| 635 pid = wait3 (&w, WNOHANG | WUNTRACED, 0); | |
| 636 # endif | |
| 637 #else /* not WNOHANG */ | |
| 638 pid = wait (&w); | |
| 639 #endif /* not WNOHANG */ | |
| 640 } | |
| 641 while (pid <= 0 && errno == EINTR); | |
| 642 | |
| 643 if (pid <= 0) | |
| 644 break; | |
| 645 | |
| 646 if (exited_processes_index < MAX_EXITED_PROCESSES) | |
| 647 { | |
| 648 exited_processes[exited_processes_index] = pid; | |
| 649 exited_processes_status[exited_processes_index] = w; | |
| 650 exited_processes_index++; | |
| 651 } | |
| 652 | |
| 653 /* On systems with WNOHANG, we just ignore the number | |
| 654 of times that SIGCHLD was signalled, and keep looping | |
| 655 until there are no more processes to wait on. If we | |
| 656 don't have WNOHANG, we have to rely on the count in | |
| 657 SIGCHLD_HAPPENED. */ | |
| 658 #ifndef WNOHANG | |
| 659 sigchld_happened--; | |
| 660 #endif /* not WNOHANG */ | |
| 661 } | |
| 662 | |
| 663 sigchld_happened = 0; | |
| 664 | |
| 665 if (block_sigchld) | |
| 666 EMACS_UNBLOCK_SIGNAL (SIGCHLD); | |
| 667 } | |
| 668 | |
| 669 /* For any processes that have changed status and are recorded | |
| 440 | 670 and such, update the corresponding Lisp_Process. |
| 428 | 671 We separate this from record_exited_processes() so that |
| 672 we never have to call this function from within a signal | |
| 673 handler. We block SIGCHLD in case record_exited_processes() | |
| 674 is called from a signal handler. */ | |
| 675 | |
| 676 /** USG WARNING: Although it is not obvious from the documentation | |
| 677 in signal(2), on a USG system the SIGCLD handler MUST NOT call | |
| 678 signal() before executing at least one wait(), otherwise the handler | |
| 679 will be called again, resulting in an infinite loop. The relevant | |
| 680 portion of the documentation reads "SIGCLD signals will be queued | |
| 681 and the signal-catching function will be continually reentered until | |
| 682 the queue is empty". Invoking signal() causes the kernel to reexamine | |
| 683 the SIGCLD queue. Fred Fish, UniSoft Systems Inc. | |
| 684 | |
| 685 (Note that now this only applies in SYS V Release 2 and before. | |
| 686 On SYS V Release 3, we use sigset() to set the signal handler for | |
| 687 the first time, and so we don't have to reestablish the signal handler | |
| 688 in the handler below. On SYS V Release 4, we don't get this weirdo | |
| 689 behavior when we use sigaction(), which we do use.) */ | |
| 690 | |
| 691 static SIGTYPE | |
| 2286 | 692 sigchld_handler (int SIG_ARG_MAYBE_UNUSED (signo)) |
| 428 | 693 { |
| 694 #ifdef OBNOXIOUS_SYSV_SIGCLD_BEHAVIOR | |
| 695 int old_errno = errno; | |
| 696 | |
| 697 sigchld_happened++; | |
| 698 record_exited_processes (0); | |
| 699 errno = old_errno; | |
| 700 #else | |
| 701 sigchld_happened++; | |
| 702 #endif | |
| 703 #ifdef HAVE_UNIXOID_EVENT_LOOP | |
| 704 signal_fake_event (); | |
| 705 #endif | |
| 706 /* WARNING - must come after wait3() for USG systems */ | |
| 707 EMACS_REESTABLISH_SIGNAL (signo, sigchld_handler); | |
| 708 SIGRETURN; | |
| 709 } | |
| 710 | |
| 711 #endif /* SIGCHLD */ | |
| 712 | |
| 713 #ifdef SIGNALS_VIA_CHARACTERS | |
| 714 /* Get signal character to send to process if SIGNALS_VIA_CHARACTERS */ | |
| 715 | |
| 716 static int | |
| 717 process_signal_char (int tty_fd, int signo) | |
| 718 { | |
| 719 /* If it's not a tty, pray that these default values work */ | |
| 853 | 720 if (! isatty (tty_fd)) |
| 721 { | |
| 428 | 722 #define CNTL(ch) (037 & (ch)) |
| 853 | 723 switch (signo) |
| 724 { | |
| 725 case SIGINT: return CNTL ('C'); | |
| 726 case SIGQUIT: return CNTL ('\\'); | |
| 428 | 727 #ifdef SIGTSTP |
| 853 | 728 case SIGTSTP: return CNTL ('Z'); |
| 428 | 729 #endif |
| 853 | 730 } |
| 731 } | |
| 428 | 732 |
| 733 #ifdef HAVE_TERMIOS | |
| 734 /* TERMIOS is the latest and bestest, and seems most likely to work. | |
| 735 If the system has it, use it. */ | |
| 736 { | |
| 737 struct termios t; | |
| 738 tcgetattr (tty_fd, &t); | |
| 739 switch (signo) | |
| 740 { | |
| 741 case SIGINT: return t.c_cc[VINTR]; | |
| 742 case SIGQUIT: return t.c_cc[VQUIT]; | |
| 743 #if defined(SIGTSTP) && defined(VSUSP) | |
| 744 case SIGTSTP: return t.c_cc[VSUSP]; | |
| 745 #endif | |
| 746 } | |
| 747 } | |
| 748 | |
| 749 # elif defined (TIOCGLTC) && defined (TIOCGETC) /* not HAVE_TERMIOS */ | |
| 750 { | |
| 751 /* On Berkeley descendants, the following IOCTL's retrieve the | |
| 752 current control characters. */ | |
| 753 struct tchars c; | |
| 754 struct ltchars lc; | |
| 755 switch (signo) | |
| 756 { | |
| 757 case SIGINT: ioctl (tty_fd, TIOCGETC, &c); return c.t_intrc; | |
| 758 case SIGQUIT: ioctl (tty_fd, TIOCGETC, &c); return c.t_quitc; | |
| 759 # ifdef SIGTSTP | |
| 760 case SIGTSTP: ioctl (tty_fd, TIOCGLTC, &lc); return lc.t_suspc; | |
| 761 # endif /* SIGTSTP */ | |
| 762 } | |
| 763 } | |
| 764 | |
| 765 # elif defined (TCGETA) /* ! defined (TIOCGLTC) && defined (TIOCGETC) */ | |
| 766 { | |
| 767 /* On SYSV descendants, the TCGETA ioctl retrieves the current | |
| 768 control characters. */ | |
| 769 struct termio t; | |
| 770 ioctl (tty_fd, TCGETA, &t); | |
| 771 switch (signo) { | |
| 772 case SIGINT: return t.c_cc[VINTR]; | |
| 773 case SIGQUIT: return t.c_cc[VQUIT]; | |
| 774 # ifdef SIGTSTP | |
| 775 case SIGTSTP: return t.c_cc[VSWTCH]; | |
| 776 # endif /* SIGTSTP */ | |
| 777 } | |
| 778 } | |
| 779 # else /* ! defined (TCGETA) */ | |
| 780 #error ERROR! Using SIGNALS_VIA_CHARACTERS, but not HAVE_TERMIOS || (TIOCGLTC && TIOCGETC) || TCGETA | |
| 781 /* If your system configuration files define SIGNALS_VIA_CHARACTERS, | |
| 782 you'd better be using one of the alternatives above! */ | |
| 783 # endif /* ! defined (TCGETA) */ | |
| 784 return '\0'; | |
| 785 } | |
| 786 #endif /* SIGNALS_VIA_CHARACTERS */ | |
| 787 | |
| 788 | |
| 789 | |
| 790 | |
| 791 /**********************************************************************/ | |
| 792 /* Process implementation methods */ | |
| 793 /**********************************************************************/ | |
| 794 | |
| 795 /* | |
| 796 * Allocate and initialize Lisp_Process->process_data | |
| 797 */ | |
| 798 | |
| 799 static void | |
| 440 | 800 unix_alloc_process_data (Lisp_Process *p) |
| 428 | 801 { |
| 802 p->process_data = xnew (struct unix_process_data); | |
| 803 | |
| 1204 | 804 UNIX_DATA (p)->connected_via_filedesc_p = 0; |
| 805 UNIX_DATA (p)->infd = -1; | |
| 806 UNIX_DATA (p)->errfd = -1; | |
| 807 UNIX_DATA (p)->subtty = -1; | |
| 808 UNIX_DATA (p)->pty_flag = 0; | |
| 428 | 809 } |
| 810 | |
| 811 /* | |
| 812 * Initialize XEmacs process implementation once | |
| 813 */ | |
| 814 | |
| 815 #ifdef SIGCHLD | |
| 816 static void | |
| 817 unix_init_process (void) | |
| 818 { | |
| 819 if (! noninteractive || initialized) | |
| 613 | 820 EMACS_SIGNAL (SIGCHLD, sigchld_handler); |
| 428 | 821 } |
| 822 #endif /* SIGCHLD */ | |
| 823 | |
| 824 /* | |
| 825 * Initialize any process local data. This is called when newly | |
| 826 * created process is connected to real OS file handles. The | |
| 827 * handles are generally represented by void* type, but are | |
| 442 | 828 * of type int (file descriptors) for UNIX. |
| 428 | 829 */ |
| 830 | |
| 831 static void | |
| 2286 | 832 unix_init_process_io_handles (Lisp_Process *p, void *in, void *UNUSED (out), |
| 833 void *err, int UNUSED (flags)) | |
| 853 | 834 { |
| 4031 | 835 /* if sizeof(EMACS_INT) > sizeof(int) this truncates the value */ |
| 836 UNIX_DATA(p)->infd = (EMACS_INT) in; | |
| 837 UNIX_DATA(p)->errfd = (EMACS_INT) err; | |
| 853 | 838 } |
| 839 | |
| 840 /* Move the file descriptor FD so that its number is not less than MIN. * | |
| 841 The original file descriptor remains open. */ | |
| 842 static int | |
| 843 relocate_fd (int fd, int min) | |
| 844 { | |
| 845 if (fd >= min) | |
| 846 return fd; | |
| 847 else | |
| 848 { | |
| 849 int newfd = dup (fd); | |
| 850 if (newfd == -1) | |
| 851 { | |
| 867 | 852 Ibyte *errmess; |
| 853 | 853 GET_STRERROR (errmess, errno); |
| 854 stderr_out ("Error while setting up child: %s\n", errmess); | |
| 855 _exit (1); | |
| 856 } | |
| 857 return relocate_fd (newfd, min); | |
| 858 } | |
| 859 } | |
| 860 | |
| 861 /* This is the last thing run in a newly forked inferior process. | |
| 862 Copy descriptors IN, OUT and ERR | |
| 863 as descriptors STDIN_FILENO, STDOUT_FILENO, and STDERR_FILENO. | |
| 864 Initialize inferior's priority, pgrp, connected dir and environment. | |
| 865 then exec another program based on new_argv. | |
| 866 | |
| 867 XEmacs: We've removed the SET_PGRP argument because it's already | |
| 868 done by the callers of child_setup. | |
| 869 | |
| 870 CURRENT_DIR is an elisp string giving the path of the current | |
| 871 directory the subprocess should have. Since we can't really signal | |
| 872 a decent error from within the child (#### not quite correct in | |
| 873 XEmacs?), this should be verified as an executable directory by the | |
| 874 parent. */ | |
| 875 | |
| 2268 | 876 static DECLARE_DOESNT_RETURN (child_setup (int, int, int, Ibyte **, |
| 877 Lisp_Object)); | |
| 878 | |
| 879 static DOESNT_RETURN | |
| 867 | 880 child_setup (int in, int out, int err, Ibyte **new_argv, |
| 853 | 881 Lisp_Object current_dir) |
| 428 | 882 { |
| 867 | 883 Ibyte **env; |
| 884 Ibyte *pwd; | |
| 853 | 885 |
| 886 #ifdef SET_EMACS_PRIORITY | |
| 887 if (emacs_priority != 0) | |
| 888 nice (- emacs_priority); | |
| 889 #endif | |
| 890 | |
| 891 /* Close Emacs's descriptors that this process should not have. */ | |
| 892 close_process_descs (); | |
| 893 close_load_descs (); | |
| 894 | |
| 895 /* [[Note that use of alloca is always safe here. It's obvious for systems | |
| 896 that do not have true vfork or that have true (stack) alloca. | |
| 897 If using vfork and C_ALLOCA it is safe because that changes | |
| 898 the superior's static variables as if the superior had done alloca | |
| 899 and will be cleaned up in the usual way.]] -- irrelevant because | |
| 900 XEmacs does not use vfork. */ | |
| 901 { | |
| 902 REGISTER Bytecount i; | |
| 903 | |
| 904 i = XSTRING_LENGTH (current_dir); | |
| 2367 | 905 pwd = alloca_ibytes (i + 6); |
| 853 | 906 memcpy (pwd, "PWD=", 4); |
| 907 memcpy (pwd + 4, XSTRING_DATA (current_dir), i); | |
| 908 i += 4; | |
| 909 if (!IS_DIRECTORY_SEP (pwd[i - 1])) | |
| 910 pwd[i++] = DIRECTORY_SEP; | |
| 911 pwd[i] = 0; | |
| 912 | |
| 913 /* [[We can't signal an Elisp error here; we're in a vfork. Since | |
| 914 the callers check the current directory before forking, this | |
| 915 should only return an error if the directory's permissions | |
| 916 are changed between the check and this chdir, but we should | |
| 917 at least check.]] -- irrelevant because XEmacs does not use vfork. */ | |
| 918 if (qxe_chdir (pwd + 4) < 0) | |
| 919 { | |
| 920 /* Don't report the chdir error, or ange-ftp.el doesn't work. */ | |
| 921 /* (FSFmacs does _exit (errno) here.) */ | |
| 922 pwd = 0; | |
| 923 } | |
| 924 else | |
| 925 { | |
| 926 /* Strip trailing "/". Cretinous *[]&@$#^%@#$% Un*x */ | |
| 927 /* leave "//" (from FSF) */ | |
| 928 while (i > 6 && IS_DIRECTORY_SEP (pwd[i - 1])) | |
| 929 pwd[--i] = 0; | |
| 930 } | |
| 931 } | |
| 932 | |
| 933 /* Set `env' to a vector of the strings in Vprocess_environment. */ | |
| 934 /* + 2 to include PWD and terminating 0. */ | |
| 867 | 935 env = alloca_array (Ibyte *, XINT (Flength (Vprocess_environment)) + 2); |
| 853 | 936 { |
| 937 REGISTER Lisp_Object tail; | |
| 867 | 938 Ibyte **new_env = env; |
| 853 | 939 |
| 940 /* If we have a PWD envvar and we know the real current directory, | |
| 941 pass one down, but with corrected value. */ | |
| 942 if (pwd && egetenv ("PWD")) | |
| 943 *new_env++ = pwd; | |
| 944 | |
| 945 /* Copy the Vprocess_environment strings into new_env. */ | |
| 946 for (tail = Vprocess_environment; | |
| 947 CONSP (tail) && STRINGP (XCAR (tail)); | |
| 948 tail = XCDR (tail)) | |
| 949 { | |
| 867 | 950 Ibyte **ep = env; |
| 951 Ibyte *envvar = XSTRING_DATA (XCAR (tail)); | |
| 853 | 952 |
| 953 /* See if envvar duplicates any string already in the env. | |
| 954 If so, don't put it in. | |
| 955 When an env var has multiple definitions, | |
| 956 we keep the definition that comes first in process-environment. */ | |
| 957 for (; ep != new_env; ep++) | |
| 958 { | |
| 867 | 959 Ibyte *p = *ep, *q = envvar; |
| 853 | 960 while (1) |
| 961 { | |
| 962 if (*q == 0) | |
| 963 /* The string is malformed; might as well drop it. */ | |
| 964 goto duplicate; | |
| 965 if (*q != *p) | |
| 966 break; | |
| 967 if (*q == '=') | |
| 968 goto duplicate; | |
| 969 p++, q++; | |
| 970 } | |
| 971 } | |
| 867 | 972 if (pwd && !qxestrncmp ((Ibyte *) "PWD=", envvar, 4)) |
| 853 | 973 { |
| 974 *new_env++ = pwd; | |
| 975 pwd = 0; | |
| 976 } | |
| 977 else | |
| 978 *new_env++ = envvar; | |
| 979 | |
| 980 duplicate: ; | |
| 981 } | |
| 982 | |
| 983 *new_env = 0; | |
| 984 } | |
| 985 | |
| 986 /* Make sure that in, out, and err are not actually already in | |
| 987 descriptors zero, one, or two; this could happen if Emacs is | |
| 988 started with its standard in, out, or error closed, as might | |
| 989 happen under X. */ | |
| 990 in = relocate_fd (in, 3); | |
| 991 out = relocate_fd (out, 3); | |
| 992 err = relocate_fd (err, 3); | |
| 993 | |
| 994 /* Set the standard input/output channels of the new process. */ | |
| 995 retry_close (STDIN_FILENO); | |
| 996 retry_close (STDOUT_FILENO); | |
| 997 retry_close (STDERR_FILENO); | |
| 998 | |
| 999 dup2 (in, STDIN_FILENO); | |
| 1000 dup2 (out, STDOUT_FILENO); | |
| 1001 dup2 (err, STDERR_FILENO); | |
| 1002 | |
| 1003 retry_close (in); | |
| 1004 retry_close (out); | |
| 1005 retry_close (err); | |
| 1006 | |
| 1015 | 1007 /* Close non-process-related file descriptors. It would be cleaner to |
| 932 | 1008 close just the ones that need to be, but the following brute |
| 1015 | 1009 force approach is certainly effective, and not too slow. */ |
| 932 | 1010 |
| 1011 { | |
| 1012 int fd; | |
| 1015 | 1013 |
| 1014 for (fd = 3; fd < MAXDESC; fd++) | |
| 932 | 1015 retry_close (fd); |
| 1016 } | |
| 1017 | |
| 853 | 1018 /* we've wrapped execve; it translates its arguments */ |
| 1019 qxe_execve (new_argv[0], new_argv, env); | |
| 1020 | |
| 1021 stdout_out ("Can't exec program %s\n", new_argv[0]); | |
| 1022 _exit (1); | |
| 428 | 1023 } |
| 1024 | |
| 1025 /* | |
| 1026 * Fork off a subprocess. P is a pointer to a newly created subprocess | |
| 1027 * object. If this function signals, the caller is responsible for | |
| 1028 * deleting (and finalizing) the process object. | |
| 1029 * | |
| 1030 * The method must return PID of the new process, a (positive??? ####) number | |
| 1031 * which fits into Lisp_Int. No return value indicates an error, the method | |
| 1032 * must signal an error instead. | |
| 1033 */ | |
| 1034 | |
| 1035 static int | |
| 440 | 1036 unix_create_process (Lisp_Process *p, |
| 428 | 1037 Lisp_Object *argv, int nargv, |
| 853 | 1038 Lisp_Object program, Lisp_Object cur_dir, |
| 1039 int separate_err) | |
| 428 | 1040 { |
| 1041 int pid; | |
| 4123 | 1042 EMACS_INT inchannel = -1; |
| 1043 EMACS_INT outchannel = -1; | |
| 1044 EMACS_INT errchannel = -1; | |
| 428 | 1045 /* Use volatile to protect variables from being clobbered by longjmp. */ |
| 4123 | 1046 volatile EMACS_INT forkin = -1; |
| 1047 volatile EMACS_INT forkout = -1; | |
| 1048 volatile EMACS_INT forkerr = -1; | |
| 428 | 1049 volatile int pty_flag = 0; |
| 1050 | |
| 1051 if (!NILP (Vprocess_connection_type)) | |
| 1052 { | |
| 1053 /* find a new pty, open the master side, return the opened | |
| 1054 file handle, and store the name of the corresponding slave | |
| 1055 side in global variable pty_name. */ | |
| 1056 outchannel = inchannel = allocate_pty (); | |
| 1057 } | |
| 1058 | |
| 535 | 1059 if (inchannel >= 0) /* We successfully allocated a pty. */ |
| 428 | 1060 { |
| 1061 /* You're "supposed" to now open the slave in the child. | |
| 1062 On some systems, we can open it here; this allows for | |
| 1063 better error checking. */ | |
| 1064 #if !defined(USG) | |
| 1065 /* On USG systems it does not work to open the pty's tty here | |
| 1066 and then close and reopen it in the child. */ | |
| 853 | 1067 # ifdef O_NOCTTY |
| 428 | 1068 /* Don't let this terminal become our controlling terminal |
| 1069 (in case we don't have one). */ | |
| 771 | 1070 forkout = forkin = qxe_open (pty_name, |
| 1071 O_RDWR | O_NOCTTY | OPEN_BINARY, 0); | |
| 853 | 1072 # else |
| 771 | 1073 forkout = forkin = qxe_open (pty_name, O_RDWR | OPEN_BINARY, 0); |
| 853 | 1074 # endif |
| 428 | 1075 if (forkin < 0) |
| 1076 goto io_failure; | |
| 1077 #endif /* not USG */ | |
| 853 | 1078 UNIX_DATA (p)->pty_flag = pty_flag = 1; |
| 428 | 1079 } |
| 1080 else | |
| 1081 if (create_bidirectional_pipe (&inchannel, &outchannel, | |
| 1082 &forkin, &forkout) < 0) | |
| 1083 goto io_failure; | |
| 1084 | |
| 853 | 1085 if (separate_err) |
| 1086 { | |
| 1087 int sv[2]; | |
| 854 | 1088 |
| 853 | 1089 if (pipe (sv) < 0) |
| 1090 goto io_failure; | |
| 1091 forkerr = sv[1]; | |
| 1092 errchannel = sv[0]; | |
| 1093 } | |
| 854 | 1094 |
| 428 | 1095 #if 0 |
| 1096 /* Replaced by close_process_descs */ | |
| 1097 set_exclusive_use (inchannel); | |
| 1098 set_exclusive_use (outchannel); | |
| 1099 #endif | |
| 1100 | |
| 1101 set_descriptor_non_blocking (inchannel); | |
| 1192 | 1102 set_descriptor_non_blocking (outchannel); |
| 853 | 1103 if (errchannel >= 0) |
| 1104 set_descriptor_non_blocking (errchannel); | |
| 428 | 1105 |
| 1106 /* Record this as an active process, with its channels. | |
| 1107 As a result, child_setup will close Emacs's side of the pipes. */ | |
| 853 | 1108 init_process_io_handles (p, (void *) inchannel, (void *) outchannel, |
| 1109 (void *) errchannel, | |
| 428 | 1110 pty_flag ? STREAM_PTY_FLUSHING : 0); |
| 1111 /* Record the tty descriptor used in the subprocess. */ | |
| 853 | 1112 UNIX_DATA (p)->subtty = forkin; |
| 428 | 1113 |
| 1114 { | |
| 1115 pid = fork (); | |
| 1116 if (pid == 0) | |
| 1117 { | |
| 1118 /**** Now we're in the child process ****/ | |
| 1119 int xforkin = forkin; | |
| 1120 int xforkout = forkout; | |
| 853 | 1121 int xforkerr = forkerr; |
| 428 | 1122 |
| 1015 | 1123 /* Checking for quit in the child is bad because that will |
| 1124 cause I/O, and that, in turn, can confuse the X connection. */ | |
| 1125 begin_dont_check_for_quit(); | |
| 1126 | |
| 442 | 1127 /* Disconnect the current controlling terminal, pursuant to |
| 1128 making the pty be the controlling terminal of the process. | |
| 1129 Also put us in our own process group. */ | |
| 1130 | |
| 1131 disconnect_controlling_terminal (); | |
| 1132 | |
| 1133 if (pty_flag) | |
| 428 | 1134 { |
| 1135 /* Open the pty connection and make the pty's terminal | |
| 1136 our controlling terminal. | |
| 1137 | |
| 1138 On systems with TIOCSCTTY, we just use it to set | |
| 1139 the controlling terminal. On other systems, the | |
| 1140 first TTY we open becomes the controlling terminal. | |
| 1141 So, we end up with four possibilities: | |
| 1142 | |
| 1143 (1) on USG and TIOCSCTTY systems, we open the pty | |
| 1144 and use TIOCSCTTY. | |
| 1145 (2) on other USG systems, we just open the pty. | |
| 1146 (3) on non-USG systems with TIOCSCTTY, we | |
| 1147 just use TIOCSCTTY. (On non-USG systems, we | |
| 1148 already opened the pty in the parent process.) | |
| 1149 (4) on non-USG systems without TIOCSCTTY, we | |
| 1150 close the pty and reopen it. | |
| 1151 | |
| 1152 This would be cleaner if we didn't open the pty | |
| 1153 in the parent process, but doing it that way | |
| 1154 makes it possible to trap error conditions. | |
| 1155 It's harder to convey an error from the child | |
| 1156 process, and I don't feel like messing with | |
| 1157 this now. */ | |
| 1158 | |
| 1159 /* SunOS has TIOCSCTTY but the close/open method | |
| 1160 also works. */ | |
| 1161 | |
| 853 | 1162 #if defined (USG) || !defined (TIOCSCTTY) |
| 428 | 1163 /* Now close the pty (if we had it open) and reopen it. |
| 1164 This makes the pty the controlling terminal of the | |
| 1165 subprocess. */ | |
| 853 | 1166 /* I wonder if retry_close (qxe_open (pty_name, ...)) would |
| 1167 work? */ | |
| 428 | 1168 if (xforkin >= 0) |
| 771 | 1169 retry_close (xforkin); |
| 1170 xforkout = xforkin = qxe_open (pty_name, O_RDWR | OPEN_BINARY, 0); | |
| 428 | 1171 if (xforkin < 0) |
| 1172 { | |
| 771 | 1173 retry_write (1, "Couldn't open the pty terminal ", 31); |
| 1174 retry_write (1, pty_name, qxestrlen (pty_name)); | |
| 1175 retry_write (1, "\n", 1); | |
| 428 | 1176 _exit (1); |
| 1177 } | |
| 853 | 1178 #endif /* USG or not TIOCSCTTY */ |
| 428 | 1179 |
| 1180 /* Miscellaneous setup required for some systems. | |
| 1181 Must be done before using tc* functions on xforkin. | |
| 1182 This guarantees that isatty(xforkin) is true. */ | |
| 1183 | |
| 853 | 1184 #if defined (HAVE_ISASTREAM) && defined (I_PUSH) |
| 442 | 1185 if (isastream (xforkin)) |
| 1186 { | |
| 853 | 1187 # if defined (I_FIND) |
| 1188 # define stream_module_pushed(fd, module) (ioctl (fd, I_FIND, module) == 1) | |
| 1189 # else | |
| 1190 # define stream_module_pushed(fd, module) 0 | |
| 1191 # endif | |
| 442 | 1192 if (! stream_module_pushed (xforkin, "ptem")) |
| 1193 ioctl (xforkin, I_PUSH, "ptem"); | |
| 1194 if (! stream_module_pushed (xforkin, "ldterm")) | |
| 1195 ioctl (xforkin, I_PUSH, "ldterm"); | |
| 1196 if (! stream_module_pushed (xforkin, "ttcompat")) | |
| 1197 ioctl (xforkin, I_PUSH, "ttcompat"); | |
| 1198 } | |
| 853 | 1199 #endif /* defined (HAVE_ISASTREAM) && defined (I_PUSH) */ |
| 428 | 1200 |
| 853 | 1201 #ifdef TIOCSCTTY |
| 428 | 1202 /* We ignore the return value |
| 1203 because faith@cs.unc.edu says that is necessary on Linux. */ | |
| 1204 assert (isatty (xforkin)); | |
| 1205 ioctl (xforkin, TIOCSCTTY, 0); | |
| 853 | 1206 #endif /* TIOCSCTTY */ |
| 428 | 1207 |
| 1208 /* Change the line discipline. */ | |
| 1209 | |
| 853 | 1210 #if defined (HAVE_TERMIOS) && defined (LDISC1) |
| 428 | 1211 { |
| 1212 struct termios t; | |
| 1213 assert (isatty (xforkin)); | |
| 1214 tcgetattr (xforkin, &t); | |
| 1215 t.c_lflag = LDISC1; | |
| 1216 if (tcsetattr (xforkin, TCSANOW, &t) < 0) | |
| 1217 perror ("create_process/tcsetattr LDISC1 failed\n"); | |
| 1218 } | |
| 853 | 1219 #elif defined (NTTYDISC) && defined (TIOCSETD) |
| 428 | 1220 { |
| 1221 /* Use new line discipline. TIOCSETD is accepted and | |
| 1222 ignored on Sys5.4 systems with ttcompat. */ | |
| 1223 int ldisc = NTTYDISC; | |
| 1224 assert (isatty (xforkin)); | |
| 1225 ioctl (xforkin, TIOCSETD, &ldisc); | |
| 1226 } | |
| 853 | 1227 #endif /* TIOCSETD & NTTYDISC */ |
| 428 | 1228 |
| 1229 /* Make our process group be the foreground group | |
| 1230 of our new controlling terminal. */ | |
| 1231 | |
| 1232 { | |
| 442 | 1233 pid_t piddly = EMACS_GET_PROCESS_GROUP (); |
| 428 | 1234 EMACS_SET_TTY_PROCESS_GROUP (xforkin, &piddly); |
| 1235 } | |
| 1236 | |
| 1237 /* On AIX, we've disabled SIGHUP above once we start a | |
| 1238 child on a pty. Now reenable it in the child, so it | |
| 1239 will die when we want it to. | |
| 1240 JV: This needs to be done ALWAYS as we might have inherited | |
| 1241 a SIG_IGN handling from our parent (nohup) and we are in new | |
| 1242 process group. | |
| 1243 */ | |
| 613 | 1244 EMACS_SIGNAL (SIGHUP, SIG_DFL); |
| 428 | 1245 |
| 535 | 1246 /* Set up the terminal characteristics of the pty. */ |
| 1247 child_setup_tty (xforkout); | |
| 1248 } /* if (pty_flag) */ | |
| 428 | 1249 |
| 1250 | |
| 613 | 1251 EMACS_SIGNAL (SIGINT, SIG_DFL); |
| 1252 EMACS_SIGNAL (SIGQUIT, SIG_DFL); | |
| 428 | 1253 |
| 1254 { | |
| 867 | 1255 Ibyte **new_argv = alloca_array (Ibyte *, nargv + 2); |
| 428 | 1256 int i; |
| 1257 | |
| 1258 /* Nothing below here GCs so our string pointers shouldn't move. */ | |
| 771 | 1259 new_argv[0] = XSTRING_DATA (program); |
| 428 | 1260 for (i = 0; i < nargv; i++) |
| 1261 { | |
| 1262 CHECK_STRING (argv[i]); | |
| 771 | 1263 new_argv[i + 1] = XSTRING_DATA (argv[i]); |
| 428 | 1264 } |
| 1265 new_argv[i + 1] = 0; | |
| 1266 | |
| 853 | 1267 child_setup (xforkin, xforkout, separate_err ? xforkerr : xforkout, |
| 1268 new_argv, cur_dir); | |
| 428 | 1269 } |
| 1270 | |
| 1271 } /**** End of child code ****/ | |
| 1272 | |
| 1273 /**** Back in parent process ****/ | |
| 1274 } | |
| 1275 | |
| 1276 if (pid < 0) | |
| 1277 { | |
| 853 | 1278 /* Note: The caller set up an unwind-protect to automatically delete |
| 1279 the process if we fail. This will correctly deselect and close | |
| 1280 inchannel, outchannel, and errchannel. */ | |
| 442 | 1281 int save_errno = errno; |
| 428 | 1282 close_descriptor_pair (forkin, forkout); |
| 853 | 1283 if (separate_err) |
| 1284 retry_close (forkerr); | |
| 442 | 1285 errno = save_errno; |
| 563 | 1286 report_process_error ("Doing fork", Qunbound); |
| 428 | 1287 } |
| 1288 | |
| 1289 /* #### dmoore - why is this commented out, otherwise we leave | |
| 1290 subtty = forkin, but then we close forkin just below. */ | |
| 853 | 1291 /* UNIX_DATA (p)->subtty = -1; */ |
| 428 | 1292 |
| 1293 /* If the subfork execv fails, and it exits, | |
| 1294 this close hangs. I don't know why. | |
| 1295 So have an interrupt jar it loose. */ | |
| 1296 if (forkin >= 0) | |
| 1297 close_safely (forkin); | |
| 1298 if (forkin != forkout && forkout >= 0) | |
| 771 | 1299 retry_close (forkout); |
| 853 | 1300 if (separate_err) |
| 1301 retry_close (forkerr); | |
| 428 | 1302 |
|
4953
304aebb79cd3
function renamings to track names of char typedefs
Ben Wing <ben@xemacs.org>
parents:
4952
diff
changeset
|
1303 p->tty_name = pty_flag ? build_istring (pty_name) : Qnil; |
| 428 | 1304 |
| 1305 /* Notice that SIGCHLD was not blocked. (This is not possible on | |
| 1306 some systems.) No biggie if SIGCHLD occurs right around the | |
| 1307 time that this call happens, because SIGCHLD() does not actually | |
| 1308 deselect the process (that doesn't occur until the next time | |
| 1309 we're waiting for an event, when status_notify() is called). */ | |
| 1310 return pid; | |
| 1311 | |
| 853 | 1312 io_failure: |
| 428 | 1313 { |
| 1314 int save_errno = errno; | |
| 1315 close_descriptor_pair (forkin, forkout); | |
| 1316 close_descriptor_pair (inchannel, outchannel); | |
| 853 | 1317 close_descriptor_pair (forkerr, errchannel); |
| 428 | 1318 errno = save_errno; |
| 563 | 1319 report_process_error ("Opening pty or pipe", Qunbound); |
| 1204 | 1320 RETURN_NOT_REACHED (0); |
| 428 | 1321 } |
| 1322 } | |
| 1323 | |
| 1324 /* Return nonzero if this process is a ToolTalk connection. */ | |
| 1325 | |
| 1326 static int | |
| 440 | 1327 unix_tooltalk_connection_p (Lisp_Process *p) |
| 428 | 1328 { |
| 853 | 1329 return UNIX_DATA (p)->connected_via_filedesc_p; |
| 428 | 1330 } |
| 1331 | |
| 1332 /* This is called to set process' virtual terminal size */ | |
| 1333 | |
| 1334 static int | |
| 853 | 1335 unix_set_window_size (Lisp_Process *p, int cols, int rows) |
| 428 | 1336 { |
| 853 | 1337 return set_window_size (UNIX_DATA (p)->infd, cols, rows); |
| 428 | 1338 } |
| 1339 | |
| 1340 /* | |
| 1341 * This method is called to update status fields of the process | |
| 1342 * structure. If the process has not existed, this method is | |
| 1343 * expected to do nothing. | |
| 1344 * | |
| 1345 * The method is called only for real child processes. | |
| 1346 */ | |
| 1347 | |
| 1348 #ifdef HAVE_WAITPID | |
| 1349 static void | |
| 853 | 1350 unix_update_status_if_terminated (Lisp_Process *p) |
| 428 | 1351 { |
| 1352 int w; | |
| 1353 #ifdef SIGCHLD | |
| 1354 EMACS_BLOCK_SIGNAL (SIGCHLD); | |
| 1355 #endif | |
| 1356 if (waitpid (XINT (p->pid), &w, WNOHANG) == XINT (p->pid)) | |
| 1357 { | |
| 1358 p->tick++; | |
| 1359 update_status_from_wait_code (p, &w); | |
| 1360 } | |
| 1361 #ifdef SIGCHLD | |
| 1362 EMACS_UNBLOCK_SIGNAL (SIGCHLD); | |
| 1363 #endif | |
| 1364 } | |
| 1365 #endif | |
| 1366 | |
| 1367 /* | |
| 1368 * Update status of all exited processes. Called when SIGCLD has signaled. | |
| 1369 */ | |
| 1370 | |
| 1371 #ifdef SIGCHLD | |
| 1372 static void | |
| 1373 unix_reap_exited_processes (void) | |
| 1374 { | |
| 1375 int i; | |
| 440 | 1376 Lisp_Process *p; |
| 428 | 1377 |
| 1378 #ifndef OBNOXIOUS_SYSV_SIGCLD_BEHAVIOR | |
| 1379 record_exited_processes (1); | |
| 1380 #endif | |
| 1381 | |
| 1382 if (exited_processes_index <= 0) | |
| 1383 { | |
| 1384 return; | |
| 1385 } | |
| 1386 | |
| 853 | 1387 #ifdef EMACS_BLOCK_SIGNAL |
| 428 | 1388 EMACS_BLOCK_SIGNAL (SIGCHLD); |
| 1389 #endif | |
| 1390 for (i = 0; i < exited_processes_index; i++) | |
| 1391 { | |
| 1392 int pid = exited_processes[i]; | |
| 1393 int w = exited_processes_status[i]; | |
| 1394 | |
| 1395 /* Find the process that signaled us, and record its status. */ | |
| 1396 | |
| 1397 p = 0; | |
| 1398 { | |
| 1399 Lisp_Object tail; | |
| 1400 LIST_LOOP (tail, Vprocess_list) | |
| 1401 { | |
| 1402 Lisp_Object proc = XCAR (tail); | |
| 1403 p = XPROCESS (proc); | |
| 1404 if (INTP (p->pid) && XINT (p->pid) == pid) | |
| 1405 break; | |
| 1406 p = 0; | |
| 1407 } | |
| 1408 } | |
| 1409 | |
| 1410 if (p) | |
| 1411 { | |
| 1412 /* Change the status of the process that was found. */ | |
| 1413 p->tick++; | |
| 1414 process_tick++; | |
| 1415 update_status_from_wait_code (p, &w); | |
| 1416 | |
| 1417 /* If process has terminated, stop waiting for its output. */ | |
| 1418 if (WIFSIGNALED (w) || WIFEXITED (w)) | |
| 1419 { | |
| 853 | 1420 if (!NILP (p->pipe_instream)) |
| 428 | 1421 { |
| 1422 /* We can't just call event_stream->unselect_process_cb (p) | |
| 1423 here, because that calls XtRemoveInput, which is not | |
| 1424 necessarily reentrant, so we can't call this at interrupt | |
| 1425 level. | |
| 1426 */ | |
| 1427 } | |
| 1428 } | |
| 1429 } | |
| 853 | 1430 #ifdef NEED_SYNC_PROCESS_CODE |
| 428 | 1431 else |
| 1432 { | |
| 1433 /* There was no asynchronous process found for that id. Check | |
| 1434 if we have a synchronous process. Only set sync process status | |
| 1435 if there is one, so we work OK with the waitpid() call in | |
| 1436 wait_for_termination(). */ | |
| 1437 if (synch_process_alive != 0) | |
| 1438 { /* Set the global sync process status variables. */ | |
| 1439 synch_process_alive = 0; | |
| 1440 | |
| 1441 /* Report the status of the synchronous process. */ | |
| 1442 if (WIFEXITED (w)) | |
| 1443 synch_process_retcode = WEXITSTATUS (w); | |
| 1444 else if (WIFSIGNALED (w)) | |
| 1445 synch_process_death = signal_name (WTERMSIG (w)); | |
| 1446 } | |
| 1447 } | |
| 853 | 1448 #endif /* NEED_SYNC_PROCESS_CODE */ |
| 428 | 1449 } |
| 1450 | |
| 1451 exited_processes_index = 0; | |
| 1452 | |
| 1453 EMACS_UNBLOCK_SIGNAL (SIGCHLD); | |
| 1454 } | |
| 1455 #endif /* SIGCHLD */ | |
| 1456 | |
| 1457 /* | |
| 1458 * Stuff the entire contents of LSTREAM to the process output pipe | |
| 1459 */ | |
| 1460 | |
| 1461 static JMP_BUF send_process_frame; | |
| 1462 | |
| 1463 static SIGTYPE | |
| 1464 send_process_trap (int signum) | |
| 1465 { | |
| 1466 EMACS_REESTABLISH_SIGNAL (signum, send_process_trap); | |
| 1467 EMACS_UNBLOCK_SIGNAL (signum); | |
| 1468 LONGJMP (send_process_frame, 1); | |
| 1469 } | |
| 1470 | |
| 1471 static void | |
| 853 | 1472 unix_send_process (Lisp_Object proc, struct lstream *lstream) |
| 428 | 1473 { |
| 1111 | 1474 /* See comment lisp.h circa line 787 */ |
| 1475 SIGTYPE (*VOLATILE_IF_NOT_CPP old_sigpipe) (int) = 0; | |
| 1476 VOLATILE_IF_NOT_CPP Lisp_Object vol_proc = proc; | |
| 1477 Lisp_Process *VOLATILE_IF_NOT_CPP p = XPROCESS (proc); | |
| 428 | 1478 |
| 442 | 1479 /* #### JV: layering violation? |
| 1480 | |
| 1481 This function knows too much about the relation between the encoding | |
| 1482 stream (DATA_OUTSTREAM) and the actual output stream p->output_stream. | |
| 1483 | |
| 1484 If encoding streams properly forwarded all calls, we could simply | |
| 1485 use DATA_OUTSTREAM everywhere. */ | |
| 1486 | |
| 428 | 1487 if (!SETJMP (send_process_frame)) |
| 1488 { | |
| 1489 /* use a reasonable-sized buffer (somewhere around the size of the | |
| 1490 stream buffer) so as to avoid inundating the stream with blocked | |
| 1491 data. */ | |
| 867 | 1492 Ibyte chunkbuf[512]; |
| 428 | 1493 Bytecount chunklen; |
| 1494 | |
| 2566 | 1495 do |
| 428 | 1496 { |
| 771 | 1497 int writeret; |
| 428 | 1498 |
| 1499 chunklen = Lstream_read (lstream, chunkbuf, 512); | |
| 1500 old_sigpipe = | |
| 613 | 1501 (SIGTYPE (*) (int)) EMACS_SIGNAL (SIGPIPE, send_process_trap); |
| 2566 | 1502 if (chunklen > 0) |
| 1503 { | |
| 1504 int save_errno; | |
| 1505 | |
| 1506 /* Lstream_write() will never successfully write less than | |
| 1507 the amount sent in. In the worst case, it just buffers | |
| 1508 the unwritten data. */ | |
| 1509 writeret = Lstream_write (XLSTREAM (DATA_OUTSTREAM(p)), chunkbuf, | |
| 1510 chunklen); | |
| 1511 save_errno = errno; | |
| 1512 EMACS_SIGNAL (SIGPIPE, old_sigpipe); | |
| 1513 errno = save_errno; | |
| 1514 if (writeret < 0) | |
| 1515 /* This is a real error. Blocking errors are handled | |
| 1516 specially inside of the filedesc stream. */ | |
| 1517 report_file_error ("writing to process", list1 (proc)); | |
| 1518 } | |
| 1519 else | |
| 1520 { | |
| 1521 /* Need to make sure that everything up to and including the | |
| 1522 last chunk is flushed, even when the pipe is currently | |
| 1523 blocked. */ | |
| 1524 Lstream_flush (XLSTREAM (DATA_OUTSTREAM(p))); | |
| 1525 EMACS_SIGNAL (SIGPIPE, old_sigpipe); | |
| 1526 } | |
| 428 | 1527 while (Lstream_was_blocked_p (XLSTREAM (p->pipe_outstream))) |
| 1528 { | |
| 3325 | 1529 /* Buffer is full. Wait 10ms, accepting input; that may |
| 1530 allow the program to finish doing output and read more. | |
| 1531 Used to be 1s, but that's excruciating. nt_send_process | |
| 1532 uses geometrically increasing timeouts (up to 1s). This | |
| 1533 might be a good idea here. | |
| 1534 N.B. timeout_secs = Qnil is faster than Qzero. */ | |
| 1535 Faccept_process_output (Qnil, Qnil, make_int (10)); | |
| 442 | 1536 /* It could have *really* finished, deleting the process */ |
| 1537 if (NILP(p->pipe_outstream)) | |
| 1538 return; | |
| 428 | 1539 old_sigpipe = |
| 613 | 1540 (SIGTYPE (*) (int)) EMACS_SIGNAL (SIGPIPE, send_process_trap); |
| 428 | 1541 Lstream_flush (XLSTREAM (p->pipe_outstream)); |
| 613 | 1542 EMACS_SIGNAL (SIGPIPE, old_sigpipe); |
| 428 | 1543 } |
| 2566 | 1544 /* Perhaps should ABORT() if < 0? This should never happen. */ |
| 428 | 1545 } |
| 2566 | 1546 while (chunklen > 0); |
| 428 | 1547 } |
| 1548 else | |
| 1549 { /* We got here from a longjmp() from the SIGPIPE handler */ | |
| 613 | 1550 EMACS_SIGNAL (SIGPIPE, old_sigpipe); |
| 428 | 1551 /* Close the file lstream so we don't attempt to write to it further */ |
| 1552 /* #### There is controversy over whether this might cause fd leakage */ | |
| 1553 /* my tests say no. -slb */ | |
| 1554 XLSTREAM (p->pipe_outstream)->flags &= ~LSTREAM_FL_IS_OPEN; | |
| 898 | 1555 XLSTREAM (p->coding_outstream)->flags &= ~LSTREAM_FL_IS_OPEN; |
| 428 | 1556 p->status_symbol = Qexit; |
| 1557 p->exit_code = 256; /* #### SIGPIPE ??? */ | |
| 1558 p->core_dumped = 0; | |
| 1559 p->tick++; | |
| 1560 process_tick++; | |
| 898 | 1561 deactivate_process (vol_proc); |
| 442 | 1562 invalid_operation ("SIGPIPE raised on process; closed it", p->name); |
| 428 | 1563 } |
| 1564 | |
| 613 | 1565 old_sigpipe = (SIGTYPE (*) (int)) EMACS_SIGNAL (SIGPIPE, send_process_trap); |
| 800 | 1566 Lstream_flush (XLSTREAM (DATA_OUTSTREAM (p))); |
| 613 | 1567 EMACS_SIGNAL (SIGPIPE, old_sigpipe); |
| 428 | 1568 } |
| 1569 | |
| 1570 /* | |
| 1571 * Send EOF to the process. The default implementation simply | |
| 1572 * closes the output stream. The method must return 0 to call | |
| 1573 * the default implementation, or 1 if it has taken all care about | |
| 1574 * sending EOF to the process. | |
| 1575 */ | |
| 1576 | |
| 1577 static int | |
| 1578 unix_process_send_eof (Lisp_Object proc) | |
| 1579 { | |
| 1580 if (!UNIX_DATA (XPROCESS (proc))->pty_flag) | |
| 1581 return 0; | |
| 1582 | |
| 1583 /* #### get_eof_char simply doesn't return the correct character | |
| 1584 here. Maybe it is needed to determine the right eof | |
| 1585 character in init_process_io_handles but here it simply screws | |
| 1586 things up. */ | |
| 1587 #if 0 | |
| 867 | 1588 Ibyte eof_char = get_eof_char (XPROCESS (proc)); |
| 428 | 1589 send_process (proc, Qnil, &eof_char, 0, 1); |
| 1590 #else | |
| 867 | 1591 send_process (proc, Qnil, (const Ibyte *) "\004", 0, 1); |
| 428 | 1592 #endif |
| 1593 return 1; | |
| 1594 } | |
| 1595 | |
| 1596 /* | |
| 1597 * Called before the process is deactivated. The process object | |
| 1598 * is not immediately finalized, just undergoes a transition to | |
| 1599 * inactive state. | |
| 1600 * | |
| 1601 * The return value is a unique stream ID, as returned by | |
| 853 | 1602 * event_stream_delete_io_streams |
| 428 | 1603 * |
| 853 | 1604 * In the lack of this method, only event_stream_delete_io_streams |
| 428 | 1605 * is called on both I/O streams of the process. |
| 1606 * | |
| 1607 * The UNIX version guards this by ignoring possible SIGPIPE. | |
| 1608 */ | |
| 1609 | |
| 853 | 1610 static void |
| 1611 unix_deactivate_process (Lisp_Process *p, | |
| 1612 USID *in_usid, | |
| 1613 USID *err_usid) | |
| 428 | 1614 { |
| 1615 SIGTYPE (*old_sigpipe) (int) = 0; | |
| 1616 | |
| 2367 | 1617 if (UNIX_DATA (p)->infd >= 0) |
| 1618 flush_pending_output (UNIX_DATA (p)->infd); | |
| 1619 if (UNIX_DATA (p)->errfd >= 0) | |
| 1620 flush_pending_output (UNIX_DATA (p)->errfd); | |
| 428 | 1621 |
| 1622 /* closing the outstream could result in SIGPIPE, so ignore it. */ | |
| 613 | 1623 old_sigpipe = (SIGTYPE (*) (int)) EMACS_SIGNAL (SIGPIPE, SIG_IGN); |
| 853 | 1624 event_stream_delete_io_streams (p->pipe_instream, p->pipe_outstream, |
| 1625 p->pipe_errstream, in_usid, err_usid); | |
| 613 | 1626 EMACS_SIGNAL (SIGPIPE, old_sigpipe); |
| 428 | 1627 |
| 2367 | 1628 UNIX_DATA (p)->infd = -1; |
| 1629 UNIX_DATA (p)->errfd = -1; | |
| 428 | 1630 } |
| 1631 | |
| 442 | 1632 /* If the subtty field of the process data is not filled in, do so now. */ |
| 1633 static void | |
| 1204 | 1634 try_to_initialize_subtty (Lisp_Process *p) |
| 442 | 1635 { |
| 1204 | 1636 struct unix_process_data *upd = UNIX_DATA (p); |
| 442 | 1637 if (upd->pty_flag |
| 444 | 1638 && (upd->subtty == -1 || ! isatty (upd->subtty)) |
| 1204 | 1639 && STRINGP (p->tty_name)) |
| 1640 upd->subtty = qxe_open (XSTRING_DATA (p->tty_name), O_RDWR, 0); | |
| 442 | 1641 } |
| 1642 | |
| 1643 /* Send signal number SIGNO to PROCESS. | |
| 428 | 1644 CURRENT_GROUP means send to the process group that currently owns |
| 1645 the terminal being used to communicate with PROCESS. | |
| 1646 This is used for various commands in shell mode. | |
| 1647 If NOMSG is zero, insert signal-announcements into process's buffers | |
| 1648 right away. | |
| 1649 | |
| 1650 If we can, we try to signal PROCESS by sending control characters | |
| 1651 down the pty. This allows us to signal inferiors who have changed | |
| 442 | 1652 their uid, for which killpg would return an EPERM error, |
| 1653 or processes running on other machines via remote login. | |
| 428 | 1654 |
| 442 | 1655 The method signals an error if the given SIGNO is not valid. */ |
| 428 | 1656 |
| 1657 static void | |
| 1658 unix_kill_child_process (Lisp_Object proc, int signo, | |
| 1659 int current_group, int nomsg) | |
| 1660 { | |
| 442 | 1661 pid_t pgid = -1; |
| 440 | 1662 Lisp_Process *p = XPROCESS (proc); |
| 442 | 1663 struct unix_process_data *d = UNIX_DATA (p); |
| 428 | 1664 |
| 1665 switch (signo) | |
| 1666 { | |
| 1667 #ifdef SIGCONT | |
| 1668 case SIGCONT: | |
| 1669 p->status_symbol = Qrun; | |
| 1670 p->exit_code = 0; | |
| 1671 p->tick++; | |
| 1672 process_tick++; | |
| 1673 if (!nomsg) | |
| 1674 status_notify (); | |
| 1675 break; | |
| 1676 #endif /* ! defined (SIGCONT) */ | |
| 1677 case SIGINT: | |
| 1678 case SIGQUIT: | |
| 1679 case SIGKILL: | |
| 442 | 1680 flush_pending_output (d->infd); |
| 853 | 1681 flush_pending_output (d->errfd); |
| 428 | 1682 break; |
| 1683 } | |
| 1684 | |
| 442 | 1685 if (! d->pty_flag) |
| 1686 current_group = 0; | |
| 1687 | |
| 1688 /* If current_group is true, we want to send a signal to the | |
| 1689 foreground process group of the terminal our child process is | |
| 1690 running on. You would think that would be easy. | |
| 1691 | |
| 1692 The BSD people invented the TIOCPGRP ioctl to get the foreground | |
| 1693 process group of a tty. That, combined with killpg, gives us | |
| 1694 what we want. | |
| 1695 | |
| 1696 However, the POSIX standards people, in their infinite wisdom, | |
| 1697 have seen fit to only allow this for processes which have the | |
| 1698 terminal as controlling terminal, which doesn't apply to us. | |
| 1699 | |
| 1700 Sooo..., we have to do something non-standard. The ioctls | |
| 1701 TIOCSIGNAL, TIOCSIG, and TIOCSIGSEND send the signal directly on | |
| 1702 many systems. POSIX tcgetpgrp(), since it is *documented* as not | |
| 1703 doing what we want, is actually less likely to work than the BSD | |
| 1704 ioctl TIOCGPGRP it is supposed to obsolete. Sometimes we have to | |
| 1705 use TIOCGPGRP on the master end, sometimes the slave end | |
| 1706 (probably an AIX bug). So we better get a fd for the slave if we | |
| 444 | 1707 haven't got it yet. |
| 1708 | |
| 1709 Anal operating systems like SGI Irix and Compaq Tru64 adhere | |
| 1710 strictly to the letter of the law, so our hack doesn't work. | |
| 1711 The following fragment from an Irix header file is suggestive: | |
| 1712 | |
| 1713 #ifdef __notdef__ | |
| 1714 // this is not currently supported | |
| 1715 #define TIOCSIGNAL (tIOC|31) // pty: send signal to slave | |
| 1716 #endif | |
| 1717 | |
| 1718 On those systems where none of our tricks work, we just fall back | |
| 1719 to the non-current_group behavior and kill the process group of | |
| 1720 the child. | |
| 1721 */ | |
| 442 | 1722 if (current_group) |
| 428 | 1723 { |
| 1204 | 1724 try_to_initialize_subtty (p); |
| 442 | 1725 |
| 1726 #ifdef SIGNALS_VIA_CHARACTERS | |
| 1727 /* If possible, send signals to the entire pgrp | |
| 1728 by sending an input character to it. */ | |
| 1729 { | |
| 867 | 1730 Ibyte sigchar = process_signal_char (d->subtty, signo); |
| 442 | 1731 if (sigchar) |
| 1732 { | |
| 853 | 1733 send_process (proc, Qnil, &sigchar, 0, 1); |
| 442 | 1734 return; |
| 1735 } | |
| 1736 } | |
| 1737 #endif /* SIGNALS_VIA_CHARACTERS */ | |
| 1738 | |
| 1739 #ifdef TIOCGPGRP | |
| 1740 if (pgid == -1) | |
| 1741 ioctl (d->infd, TIOCGPGRP, &pgid); /* BSD */ | |
| 1742 if (pgid == -1 && d->subtty != -1) | |
| 1743 ioctl (d->subtty, TIOCGPGRP, &pgid); /* Only this works on AIX! */ | |
| 1744 #endif /* TIOCGPGRP */ | |
| 1745 | |
| 1746 if (pgid == -1) | |
| 428 | 1747 { |
| 442 | 1748 /* Many systems provide an ioctl to send a signal directly */ |
| 1749 #ifdef TIOCSIGNAL /* Solaris, HP-UX */ | |
| 1750 if (ioctl (d->infd, TIOCSIGNAL, signo) != -1) | |
| 1751 return; | |
| 1752 #endif /* TIOCSIGNAL */ | |
| 1753 | |
| 1754 #ifdef TIOCSIG /* BSD */ | |
| 1755 if (ioctl (d->infd, TIOCSIG, signo) != -1) | |
| 1756 return; | |
| 1757 #endif /* TIOCSIG */ | |
| 428 | 1758 } |
| 442 | 1759 } /* current_group */ |
| 428 | 1760 |
| 442 | 1761 if (pgid == -1) |
| 1762 /* Either current_group is 0, or we failed to get the foreground | |
| 1763 process group using the trickery above. So we fall back to | |
| 1764 sending the signal to the process group of our child process. | |
| 1765 Since this is often a shell that ignores signals like SIGINT, | |
| 1766 the shell's subprocess is killed, which is the desired effect. | |
| 1767 The process group of p->pid is always p->pid, since it was | |
| 1768 created as a process group leader. */ | |
| 1769 pgid = XINT (p->pid); | |
| 1770 | |
| 1771 /* Finally send the signal. */ | |
| 1772 if (EMACS_KILLPG (pgid, signo) == -1) | |
| 458 | 1773 { |
| 1774 /* It's not an error if our victim is already dead. | |
| 462 | 1775 And we can't rely on the result of killing a zombie, since |
| 1776 XPG 4.2 requires that killing a zombie fail with ESRCH, | |
| 1777 while FIPS 151-2 requires that it succeeds! */ | |
| 458 | 1778 #ifdef ESRCH |
| 1779 if (errno != ESRCH) | |
| 1780 #endif | |
| 563 | 1781 signal_ferror_with_frob (Qio_error, lisp_strerror (errno), |
| 1782 "kill (pgid=%ld, signo=%ld) failed", | |
| 1783 (long) pgid, (long) signo); | |
| 458 | 1784 } |
| 428 | 1785 } |
| 1786 | |
| 442 | 1787 /* Send signal SIGCODE to any process in the system given its PID. |
| 1788 Return zero if successful, a negative number upon failure. */ | |
| 428 | 1789 |
| 1790 static int | |
| 1791 unix_kill_process_by_pid (int pid, int sigcode) | |
| 1792 { | |
| 1793 return kill (pid, sigcode); | |
| 1794 } | |
| 1795 | |
| 442 | 1796 /* Canonicalize host name HOST, and return its canonical form. |
| 1797 The default implementation just takes HOST for a canonical name. */ | |
| 428 | 1798 |
| 1799 #ifdef HAVE_SOCKETS | |
| 1800 static Lisp_Object | |
| 1801 unix_canonicalize_host_name (Lisp_Object host) | |
| 1802 { | |
| 502 | 1803 #ifdef USE_GETADDRINFO |
| 440 | 1804 struct addrinfo hints, *res; |
| 1805 static char addrbuf[NI_MAXHOST]; | |
| 1806 Lisp_Object canonname; | |
| 1807 int retval; | |
| 1808 char *ext_host; | |
| 1809 | |
| 1810 xzero (hints); | |
| 1811 hints.ai_flags = AI_CANONNAME; | |
| 724 | 1812 #ifdef IPV6_CANONICALIZE |
| 440 | 1813 hints.ai_family = AF_UNSPEC; |
| 724 | 1814 #else |
| 1815 hints.ai_family = PF_INET; | |
| 1816 #endif | |
| 440 | 1817 hints.ai_socktype = SOCK_STREAM; |
| 1818 hints.ai_protocol = 0; | |
|
4834
b3ea9c582280
Use new cygwin_conv_path API with Cygwin 1.7 for converting names between Win32 and POSIX, UTF-8-aware, with attendant changes elsewhere
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
1819 LISP_STRING_TO_EXTERNAL (host, ext_host, Qunix_host_name_encoding); |
| 440 | 1820 retval = getaddrinfo (ext_host, NULL, &hints, &res); |
| 1821 if (retval != 0) | |
| 1822 { | |
| 867 | 1823 CIbyte *gai_err; |
| 440 | 1824 |
| 855 | 1825 EXTERNAL_TO_C_STRING (gai_strerror (retval), gai_err, |
| 771 | 1826 Qstrerror_encoding); |
| 855 | 1827 maybe_signal_error (Qio_error, gai_err, host, |
| 793 | 1828 Qprocess, ERROR_ME_DEBUG_WARN); |
| 440 | 1829 canonname = host; |
| 1830 } | |
| 1831 else | |
| 1832 { | |
| 1833 int gni = getnameinfo (res->ai_addr, res->ai_addrlen, | |
| 1834 addrbuf, sizeof(addrbuf), | |
| 1835 NULL, 0, NI_NUMERICHOST); | |
|
4953
304aebb79cd3
function renamings to track names of char typedefs
Ben Wing <ben@xemacs.org>
parents:
4952
diff
changeset
|
1836 canonname = gni ? host : build_extstring (addrbuf, |
|
4834
b3ea9c582280
Use new cygwin_conv_path API with Cygwin 1.7 for converting names between Win32 and POSIX, UTF-8-aware, with attendant changes elsewhere
Ben Wing <ben@xemacs.org>
parents:
4759
diff
changeset
|
1837 Qunix_host_name_encoding); |
| 440 | 1838 |
| 1839 freeaddrinfo (res); | |
| 1840 } | |
| 1841 | |
| 1842 return canonname; | |
| 502 | 1843 #else /* ! USE_GETADDRINFO */ |
| 428 | 1844 struct sockaddr_in address; |
| 1845 | |
| 1846 if (!get_internet_address (host, &address, ERROR_ME_NOT)) | |
| 1847 return host; | |
| 1848 | |
| 1849 if (address.sin_family == AF_INET) | |
|
4953
304aebb79cd3
function renamings to track names of char typedefs
Ben Wing <ben@xemacs.org>
parents:
4952
diff
changeset
|
1850 return build_extstring (inet_ntoa (address.sin_addr), |
|
4952
19a72041c5ed
Mule-izing, various fixes related to char * arguments
Ben Wing <ben@xemacs.org>
parents:
4834
diff
changeset
|
1851 Qunix_host_name_encoding); |
| 428 | 1852 else |
| 1853 /* #### any clue what to do here? */ | |
| 1854 return host; | |
| 502 | 1855 #endif /* ! USE_GETADDRINFO */ |
| 428 | 1856 } |
| 1857 | |
| 442 | 1858 /* Open a TCP network connection to a given HOST/SERVICE. |
| 1859 Treated exactly like a normal process when reading and writing. | |
| 1860 Only differences are in status display and process deletion. | |
| 1861 A network connection has no PID; you cannot signal it. All you can | |
| 1862 do is deactivate and close it via delete-process. */ | |
| 428 | 1863 |
| 1864 static void | |
| 502 | 1865 unix_open_network_stream (Lisp_Object name, Lisp_Object host, |
| 1866 Lisp_Object service, Lisp_Object protocol, | |
| 1867 void **vinfd, void **voutfd) | |
| 428 | 1868 { |
| 4123 | 1869 EMACS_INT inch; |
| 1870 EMACS_INT outch; | |
| 502 | 1871 volatile int s = -1; |
| 428 | 1872 volatile int port; |
| 1873 volatile int retry = 0; | |
| 502 | 1874 volatile int xerrno = 0; |
| 1875 volatile int failed_connect = 0; | |
| 428 | 1876 int retval; |
| 1877 | |
| 1878 CHECK_STRING (host); | |
| 1879 | |
| 1880 if (!EQ (protocol, Qtcp) && !EQ (protocol, Qudp)) | |
| 563 | 1881 invalid_constant ("Unsupported protocol", protocol); |
| 428 | 1882 |
| 440 | 1883 { |
| 502 | 1884 #ifdef USE_GETADDRINFO |
| 1885 | |
| 440 | 1886 struct addrinfo hints, *res; |
| 1887 struct addrinfo * volatile lres; | |
| 771 | 1888 Extbyte *portstring; |
| 1889 Extbyte *ext_host; | |
| 1890 Extbyte portbuf[128]; | |
| 440 | 1891 /* |
| 1892 * Caution: service can either be a string or int. | |
| 1893 * Convert to a C string for later use by getaddrinfo. | |
| 1894 */ | |
| 1895 if (INTP (service)) | |
| 1896 { | |
| 1897 snprintf (portbuf, sizeof (portbuf), "%ld", (long) XINT (service)); | |
| 1898 portstring = portbuf; | |
| 1899 port = htons ((unsigned short) XINT (service)); | |
| 1900 } | |
| 1901 else | |
| 1902 { | |
| 1903 CHECK_STRING (service); | |
| 771 | 1904 LISP_STRING_TO_EXTERNAL (service, portstring, |
| 1905 Qunix_service_name_encoding); | |
| 440 | 1906 port = 0; |
| 1907 } | |
| 1908 | |
| 1909 xzero (hints); | |
| 1910 hints.ai_flags = 0; | |
| 1911 hints.ai_family = AF_UNSPEC; | |
| 1912 if (EQ (protocol, Qtcp)) | |
| 1913 hints.ai_socktype = SOCK_STREAM; | |
| 1914 else /* EQ (protocol, Qudp) */ | |
| 1915 hints.ai_socktype = SOCK_DGRAM; | |
| 1916 hints.ai_protocol = 0; | |
| 771 | 1917 LISP_STRING_TO_EXTERNAL (host, ext_host, Qunix_host_name_encoding); |
| 440 | 1918 retval = getaddrinfo (ext_host, portstring, &hints, &res); |
| 1919 if (retval != 0) | |
| 1920 { | |
| 867 | 1921 CIbyte *gai_err; |
| 440 | 1922 |
| 855 | 1923 EXTERNAL_TO_C_STRING (gai_strerror (retval), gai_err, |
| 771 | 1924 Qstrerror_encoding); |
| 855 | 1925 signal_error (Qio_error, gai_err, list2 (host, service)); |
| 440 | 1926 } |
| 1927 | |
| 1928 /* address loop */ | |
| 1929 for (lres = res; lres ; lres = lres->ai_next) | |
| 1930 | |
| 502 | 1931 #else /* !USE_GETADDRINFO */ |
| 440 | 1932 |
| 1933 struct sockaddr_in address; | |
| 502 | 1934 volatile int i; |
| 440 | 1935 |
| 1936 if (INTP (service)) | |
| 1937 port = htons ((unsigned short) XINT (service)); | |
| 1938 else | |
| 1939 { | |
| 1940 struct servent *svc_info; | |
| 771 | 1941 Extbyte *servext; |
| 1942 | |
| 440 | 1943 CHECK_STRING (service); |
| 771 | 1944 LISP_STRING_TO_EXTERNAL (service, servext, |
| 1945 Qunix_service_name_encoding); | |
| 440 | 1946 |
| 1947 if (EQ (protocol, Qtcp)) | |
| 771 | 1948 svc_info = getservbyname (servext, "tcp"); |
| 440 | 1949 else /* EQ (protocol, Qudp) */ |
| 771 | 1950 svc_info = getservbyname (servext, "udp"); |
| 428 | 1951 |
| 440 | 1952 if (svc_info == 0) |
| 442 | 1953 invalid_argument ("Unknown service", service); |
| 440 | 1954 port = svc_info->s_port; |
| 1955 } | |
| 428 | 1956 |
| 440 | 1957 get_internet_address (host, &address, ERROR_ME); |
| 1958 address.sin_port = port; | |
| 428 | 1959 |
| 502 | 1960 /* use a trivial address loop */ |
| 1961 for (i = 0; i < 1; i++) | |
| 1962 | |
| 1963 #endif /* !USE_GETADDRINFO */ | |
| 1964 { | |
| 1965 #ifdef USE_GETADDRINFO | |
| 1966 int family = lres->ai_family; | |
| 1967 #else | |
| 1968 int family = address.sin_family; | |
| 1969 #endif | |
| 1970 | |
| 1971 if (EQ (protocol, Qtcp)) | |
| 1972 s = socket (family, SOCK_STREAM, 0); | |
| 1973 else /* EQ (protocol, Qudp) */ | |
| 1974 s = socket (family, SOCK_DGRAM, 0); | |
| 1975 | |
| 1976 if (s < 0) | |
| 1977 { | |
| 1978 xerrno = errno; | |
| 1979 failed_connect = 0; | |
| 1980 continue; | |
| 1981 } | |
| 1982 | |
| 1983 loop: | |
| 1984 | |
| 1985 /* A system call interrupted with a SIGALRM or SIGIO comes back | |
| 1986 here, with can_break_system_calls reset to 0. */ | |
| 1987 SETJMP (break_system_call_jump); | |
| 1988 if (QUITP) | |
| 1989 { | |
| 853 | 1990 QUIT; |
| 502 | 1991 /* In case something really weird happens ... */ |
| 1992 } | |
| 1993 | |
| 1994 /* Break out of connect with a signal (it isn't otherwise possible). | |
| 1995 Thus you don't get screwed with a hung network. */ | |
| 1996 can_break_system_calls = 1; | |
| 1997 | |
| 1998 #ifdef USE_GETADDRINFO | |
| 1999 retval = connect (s, lres->ai_addr, lres->ai_addrlen); | |
| 2000 #else | |
| 2001 retval = connect (s, (struct sockaddr *) &address, sizeof (address)); | |
| 2002 #endif | |
| 2003 can_break_system_calls = 0; | |
| 2004 if (retval == -1 && errno != EISCONN) | |
| 2005 { | |
| 2006 xerrno = errno; | |
| 859 | 2007 |
| 2008 if (errno == EINTR || errno == EINPROGRESS || errno == EALREADY) | |
| 502 | 2009 goto loop; |
| 2010 if (errno == EADDRINUSE && retry < 20) | |
| 2011 { | |
| 2012 #ifdef __FreeBSD__ | |
| 2013 /* A delay here is needed on some FreeBSD systems, | |
| 2014 and it is harmless, since this retrying takes | |
| 2015 time anyway and should be infrequent. | |
| 2016 `sleep-for' allowed for quitting this loop with | |
| 2017 interrupts slowed down so it can't be used | |
| 2018 here. Async timers should already be disabled | |
| 2019 at this point so we can use `sleep'. | |
| 2020 | |
| 2021 (Again, this was not conditionalized on FreeBSD. | |
| 854 | 2022 Let's not mess up systems without the problem. --ben) |
| 502 | 2023 */ |
| 2024 sleep (1); | |
| 2025 #endif | |
| 2026 retry++; | |
| 2027 goto loop; | |
| 2028 } | |
| 2029 | |
| 2030 failed_connect = 1; | |
| 771 | 2031 retry_close (s); |
| 502 | 2032 s = -1; |
| 2033 continue; | |
| 2034 } | |
| 2035 | |
| 2036 #ifdef USE_GETADDRINFO | |
| 2037 if (port == 0) | |
| 2038 { | |
| 2039 int gni; | |
| 2040 char servbuf[NI_MAXSERV]; | |
| 2041 | |
| 2042 if (EQ (protocol, Qtcp)) | |
| 2043 gni = getnameinfo (lres->ai_addr, lres->ai_addrlen, | |
| 2044 NULL, 0, servbuf, sizeof(servbuf), | |
| 2045 NI_NUMERICSERV); | |
| 2046 else /* EQ (protocol, Qudp) */ | |
| 2047 gni = getnameinfo (lres->ai_addr, lres->ai_addrlen, | |
| 2048 NULL, 0, servbuf, sizeof(servbuf), | |
| 2049 NI_NUMERICSERV | NI_DGRAM); | |
| 2050 | |
| 2051 if (gni == 0) | |
| 2052 port = strtol (servbuf, NULL, 10); | |
| 2053 } | |
| 2054 | |
| 2055 break; | |
| 2056 #endif /* USE_GETADDRINFO */ | |
| 2057 } /* address loop */ | |
| 2058 | |
| 2059 #ifdef USE_GETADDRINFO | |
| 2060 freeaddrinfo (res); | |
| 2061 #endif | |
| 428 | 2062 |
| 440 | 2063 if (s < 0) |
| 502 | 2064 { |
| 2065 errno = xerrno; | |
| 428 | 2066 |
| 502 | 2067 if (failed_connect) |
| 563 | 2068 report_network_error ("connection failed", list3 (Qunbound, host, |
| 2069 name)); | |
| 502 | 2070 else |
| 563 | 2071 report_network_error ("error creating socket", name); |
| 440 | 2072 } |
| 2073 } | |
| 428 | 2074 |
| 2075 inch = s; | |
| 2076 outch = dup (s); | |
| 2077 if (outch < 0) | |
| 2078 { | |
| 563 | 2079 int save_errno = errno; |
| 771 | 2080 retry_close (s); /* this used to be leaked; from Kyle Jones */ |
| 563 | 2081 errno = save_errno; |
| 2082 report_network_error ("error duplicating socket", name); | |
| 428 | 2083 } |
| 2084 | |
| 2085 set_socket_nonblocking_maybe (inch, port, "tcp"); | |
| 2086 | |
| 502 | 2087 *vinfd = (void *) inch; |
| 2088 *voutfd = (void *) outch; | |
| 428 | 2089 } |
| 2090 | |
| 2091 | |
| 2092 #ifdef HAVE_MULTICAST | |
| 2093 | |
| 442 | 2094 /* Didier Verna <didier@xemacs.org> Nov. 28 1997. |
| 428 | 2095 |
| 2096 This function is similar to open-network-stream-internal, but provides a | |
| 2097 mean to open an UDP multicast connection instead of a TCP one. Like in the | |
| 2098 TCP case, the multicast connection will be seen as a sub-process, | |
| 2099 | |
| 2100 Some notes: | |
| 2101 - Normally, we should use sendto and recvfrom with non connected | |
| 2102 sockets. The current code doesn't allow us to do this. In the future, it | |
| 2103 would be a good idea to extend the process data structure in order to deal | |
| 2104 properly with the different types network connections. | |
| 2105 - For the same reason, when leaving a multicast group, it is better to make | |
| 2106 a setsockopt - IP_DROP_MEMBERSHIP before closing the descriptors. | |
| 2107 Unfortunately, this can't be done here because delete_process doesn't know | |
| 2108 about the kind of connection we have. However, this is not such an | |
| 2109 important issue. | |
| 2110 */ | |
| 2111 | |
| 2112 static void | |
| 442 | 2113 unix_open_multicast_group (Lisp_Object name, Lisp_Object dest, |
| 853 | 2114 Lisp_Object port, Lisp_Object ttl, void **vinfd, |
| 2115 void **voutfd) | |
| 428 | 2116 { |
| 2117 struct ip_mreq imr; | |
| 2118 struct sockaddr_in sa; | |
| 2119 struct protoent *udp; | |
| 4123 | 2120 EMACS_INT ws, rs; |
| 428 | 2121 int theport; |
| 2122 unsigned char thettl; | |
| 2123 int one = 1; /* For REUSEADDR */ | |
| 2124 int ret; | |
| 2125 volatile int retry = 0; | |
| 2126 | |
| 2127 CHECK_STRING (dest); | |
| 2128 | |
| 2129 CHECK_NATNUM (port); | |
| 2130 theport = htons ((unsigned short) XINT (port)); | |
| 2131 | |
| 2132 CHECK_NATNUM (ttl); | |
| 2133 thettl = (unsigned char) XINT (ttl); | |
| 2134 | |
| 2135 if ((udp = getprotobyname ("udp")) == NULL) | |
| 563 | 2136 invalid_operation ("No info available for UDP protocol", Qunbound); |
| 428 | 2137 |
| 2138 /* Init the sockets. Yes, I need 2 sockets. I couldn't duplicate one. */ | |
| 2139 if ((rs = socket (PF_INET, SOCK_DGRAM, udp->p_proto)) < 0) | |
| 563 | 2140 report_network_error ("error creating socket", name); |
| 428 | 2141 if ((ws = socket (PF_INET, SOCK_DGRAM, udp->p_proto)) < 0) |
| 2142 { | |
| 563 | 2143 int save_errno = errno; |
| 771 | 2144 retry_close (rs); |
| 563 | 2145 errno = save_errno; |
| 2146 report_network_error ("error creating socket", name); | |
| 428 | 2147 } |
| 2148 | |
| 2149 /* This will be used for both sockets */ | |
| 2150 memset (&sa, 0, sizeof(sa)); | |
| 2151 sa.sin_family = AF_INET; | |
| 2152 sa.sin_port = theport; | |
| 671 | 2153 sa.sin_addr.s_addr = inet_addr ((char *) XSTRING_DATA (dest)); |
| 428 | 2154 |
| 2155 /* Socket configuration for reading ------------------------ */ | |
| 2156 | |
| 2157 /* Multiple connections from the same machine. This must be done before | |
| 2158 bind. If it fails, it shouldn't be fatal. The only consequence is that | |
| 2159 people won't be able to connect twice from the same machine. */ | |
| 2160 if (setsockopt (rs, SOL_SOCKET, SO_REUSEADDR, (char *) &one, sizeof (one)) | |
| 2161 < 0) | |
| 2162 warn_when_safe (Qmulticast, Qwarning, "Cannot reuse socket address"); | |
| 2163 | |
| 2164 /* bind socket name */ | |
| 2165 if (bind (rs, (struct sockaddr *)&sa, sizeof(sa))) | |
| 2166 { | |
| 563 | 2167 int save_errno = errno; |
| 771 | 2168 retry_close (rs); |
| 2169 retry_close (ws); | |
| 563 | 2170 errno = save_errno; |
| 2171 report_network_error ("error binding socket", list3 (Qunbound, name, | |
| 2172 port)); | |
| 428 | 2173 } |
| 2174 | |
| 2175 /* join multicast group */ | |
| 671 | 2176 imr.imr_multiaddr.s_addr = inet_addr ((char *) XSTRING_DATA (dest)); |
| 428 | 2177 imr.imr_interface.s_addr = htonl (INADDR_ANY); |
| 2178 if (setsockopt (rs, IPPROTO_IP, IP_ADD_MEMBERSHIP, | |
| 442 | 2179 &imr, sizeof (struct ip_mreq)) < 0) |
| 428 | 2180 { |
| 563 | 2181 int save_errno = errno; |
| 771 | 2182 retry_close (ws); |
| 2183 retry_close (rs); | |
| 563 | 2184 errno = save_errno; |
| 2185 report_network_error ("error adding membership", list3 (Qunbound, name, | |
| 2186 dest)); | |
| 428 | 2187 } |
| 2188 | |
| 2189 /* Socket configuration for writing ----------------------- */ | |
| 2190 | |
| 2191 /* Normally, there's no 'connect' in multicast, since we prefer to use | |
| 2192 'sendto' and 'recvfrom'. However, in order to handle this connection in | |
| 2193 the process-like way it is done for TCP, we must be able to use 'write' | |
| 2194 instead of 'sendto'. Consequently, we 'connect' this socket. */ | |
| 2195 | |
| 2196 /* See open-network-stream-internal for comments on this part of the code */ | |
| 2197 loop: | |
| 2198 | |
| 2199 /* A system call interrupted with a SIGALRM or SIGIO comes back | |
| 2200 here, with can_break_system_calls reset to 0. */ | |
| 2201 SETJMP (break_system_call_jump); | |
| 2202 if (QUITP) | |
| 2203 { | |
| 853 | 2204 QUIT; |
| 428 | 2205 /* In case something really weird happens ... */ |
| 2206 } | |
| 2207 | |
| 2208 /* Break out of connect with a signal (it isn't otherwise possible). | |
| 2209 Thus you don't get screwed with a hung network. */ | |
| 2210 can_break_system_calls = 1; | |
| 2211 ret = connect (ws, (struct sockaddr *) &sa, sizeof (sa)); | |
| 2212 can_break_system_calls = 0; | |
| 2213 if (ret == -1 && errno != EISCONN) | |
| 2214 { | |
| 2215 int xerrno = errno; | |
| 2216 | |
| 859 | 2217 if (errno == EINTR || errno == EINPROGRESS || errno == EALREADY) |
| 428 | 2218 goto loop; |
| 2219 if (errno == EADDRINUSE && retry < 20) | |
| 2220 { | |
| 859 | 2221 #ifdef __FreeBSD__ |
| 428 | 2222 /* A delay here is needed on some FreeBSD systems, |
| 2223 and it is harmless, since this retrying takes time anyway | |
| 2224 and should be infrequent. | |
| 2225 `sleep-for' allowed for quitting this loop with interrupts | |
| 2226 slowed down so it can't be used here. Async timers should | |
| 2227 already be disabled at this point so we can use `sleep'. */ | |
| 2228 sleep (1); | |
| 859 | 2229 #endif |
| 428 | 2230 retry++; |
| 2231 goto loop; | |
| 2232 } | |
| 2233 | |
| 771 | 2234 retry_close (rs); |
| 2235 retry_close (ws); | |
| 428 | 2236 |
| 2237 errno = xerrno; | |
| 563 | 2238 report_network_error ("error connecting socket", list3 (Qunbound, name, |
| 2239 port)); | |
| 428 | 2240 } |
| 2241 | |
| 2242 /* scope */ | |
| 2243 if (setsockopt (ws, IPPROTO_IP, IP_MULTICAST_TTL, | |
| 442 | 2244 &thettl, sizeof (thettl)) < 0) |
| 428 | 2245 { |
| 563 | 2246 int save_errno = errno; |
| 771 | 2247 retry_close (rs); |
| 2248 retry_close (ws); | |
| 563 | 2249 errno = save_errno; |
| 2250 report_network_error ("error setting ttl", list3 (Qunbound, name, ttl)); | |
| 428 | 2251 } |
| 2252 | |
| 2253 set_socket_nonblocking_maybe (rs, theport, "udp"); | |
| 2254 | |
| 2255 *vinfd = (void*)rs; | |
| 2256 *voutfd = (void*)ws; | |
| 2257 } | |
| 2258 | |
| 2259 #endif /* HAVE_MULTICAST */ | |
| 2260 | |
| 2261 #endif /* HAVE_SOCKETS */ | |
| 2262 | |
| 2263 | |
| 2264 /**********************************************************************/ | |
| 2265 /* Initialization */ | |
| 2266 /**********************************************************************/ | |
| 2267 | |
| 2268 void | |
| 2269 process_type_create_unix (void) | |
| 2270 { | |
| 2271 PROCESS_HAS_METHOD (unix, alloc_process_data); | |
| 2272 #ifdef SIGCHLD | |
| 2273 PROCESS_HAS_METHOD (unix, init_process); | |
| 2274 PROCESS_HAS_METHOD (unix, reap_exited_processes); | |
| 2275 #endif | |
| 2276 PROCESS_HAS_METHOD (unix, init_process_io_handles); | |
| 2277 PROCESS_HAS_METHOD (unix, create_process); | |
| 2278 PROCESS_HAS_METHOD (unix, tooltalk_connection_p); | |
| 2279 PROCESS_HAS_METHOD (unix, set_window_size); | |
| 2280 #ifdef HAVE_WAITPID | |
| 2281 PROCESS_HAS_METHOD (unix, update_status_if_terminated); | |
| 2282 #endif | |
| 2283 PROCESS_HAS_METHOD (unix, send_process); | |
| 2284 PROCESS_HAS_METHOD (unix, process_send_eof); | |
| 2285 PROCESS_HAS_METHOD (unix, deactivate_process); | |
| 2286 PROCESS_HAS_METHOD (unix, kill_child_process); | |
| 2287 PROCESS_HAS_METHOD (unix, kill_process_by_pid); | |
| 2288 #ifdef HAVE_SOCKETS | |
| 2289 PROCESS_HAS_METHOD (unix, canonicalize_host_name); | |
| 2290 PROCESS_HAS_METHOD (unix, open_network_stream); | |
| 2291 #ifdef HAVE_MULTICAST | |
| 2292 PROCESS_HAS_METHOD (unix, open_multicast_group); | |
| 2293 #endif | |
| 2294 #endif | |
| 2295 } | |
| 2296 | |
| 2297 void | |
| 2298 vars_of_process_unix (void) | |
| 2299 { | |
| 2300 Fprovide (intern ("unix-processes")); | |
| 2301 } |