1 /* vi:set ts=8 sts=4 sw=4 noet:
2 *
3 * VIM - Vi IMproved by Bram Moolenaar
4 *
5 * Do ":help uganda" in Vim to read copying and usage conditions.
6 * Do ":help credits" in Vim to see a list of people who contributed.
7 * See README.txt for an overview of the Vim source code.
8 */
9
10 /*
11 * fileio.c: read from and write to a file
12 */
13
14 #include "vim.h"
15
16 #if defined(__TANDEM)
17 # include <limits.h> // for SSIZE_MAX
18 #endif
19 #if (defined(UNIX) || defined(VMS)) && defined(FEAT_EVAL)
20 # include <pwd.h>
21 # include <grp.h>
22 #endif
23 #if defined(VMS) && defined(HAVE_XOS_R_H)
24 # include <x11/xos_r.h>
25 #endif
26
27 // Is there any system that doesn't have access()?
28 #define USE_MCH_ACCESS
29
30 #if defined(__hpux) && !defined(HAVE_DIRFD)
31 # define dirfd(x) ((x)->__dd_fd)
32 # define HAVE_DIRFD
33 #endif
34
35 static char_u *next_fenc(char_u **pp, int *alloced);
36 #ifdef FEAT_EVAL
37 static char_u *readfile_charconvert(char_u *fname, char_u *fenc, int *fdp);
38 #endif
39 #ifdef FEAT_CRYPT
40 static char_u *check_for_cryptkey(char_u *cryptkey, char_u *ptr, long *sizep, off_T *filesizep, int newfile, char_u *fname, int *did_ask);
41 #endif
42 static linenr_T readfile_linenr(linenr_T linecnt, char_u *p, char_u *endp);
43 static char_u *check_for_bom(char_u *p, long size, int *lenp, int flags);
44 static char *e_auchangedbuf = N_("E812: Autocommands changed buffer or buffer name");
45
46 #ifdef FEAT_EVAL
47 static int readdirex_sort;
48 #endif
49
50 void
filemess(buf_T * buf,char_u * name,char_u * s,int attr)51 filemess(
52 buf_T *buf,
53 char_u *name,
54 char_u *s,
55 int attr)
56 {
57 int msg_scroll_save;
58 int prev_msg_col = msg_col;
59
60 if (msg_silent != 0)
61 return;
62 msg_add_fname(buf, name); // put file name in IObuff with quotes
63
64 // If it's extremely long, truncate it.
65 if (STRLEN(IObuff) > IOSIZE - 100)
66 IObuff[IOSIZE - 100] = NUL;
67
68 // Avoid an over-long translation to cause trouble.
69 STRNCAT(IObuff, s, 99);
70
71 /*
72 * For the first message may have to start a new line.
73 * For further ones overwrite the previous one, reset msg_scroll before
74 * calling filemess().
75 */
76 msg_scroll_save = msg_scroll;
77 if (shortmess(SHM_OVERALL) && !exiting && p_verbose == 0)
78 msg_scroll = FALSE;
79 if (!msg_scroll) // wait a bit when overwriting an error msg
80 check_for_delay(FALSE);
81 msg_start();
82 if (prev_msg_col != 0 && msg_col == 0)
83 msg_putchar('\r'); // overwrite any previous message.
84 msg_scroll = msg_scroll_save;
85 msg_scrolled_ign = TRUE;
86 // may truncate the message to avoid a hit-return prompt
87 msg_outtrans_attr(msg_may_trunc(FALSE, IObuff), attr);
88 msg_clr_eos();
89 out_flush();
90 msg_scrolled_ign = FALSE;
91 }
92
93 /*
94 * Read lines from file "fname" into the buffer after line "from".
95 *
96 * 1. We allocate blocks with lalloc, as big as possible.
97 * 2. Each block is filled with characters from the file with a single read().
98 * 3. The lines are inserted in the buffer with ml_append().
99 *
100 * (caller must check that fname != NULL, unless READ_STDIN is used)
101 *
102 * "lines_to_skip" is the number of lines that must be skipped
103 * "lines_to_read" is the number of lines that are appended
104 * When not recovering lines_to_skip is 0 and lines_to_read MAXLNUM.
105 *
106 * flags:
107 * READ_NEW starting to edit a new buffer
108 * READ_FILTER reading filter output
109 * READ_STDIN read from stdin instead of a file
110 * READ_BUFFER read from curbuf instead of a file (converting after reading
111 * stdin)
112 * READ_DUMMY read into a dummy buffer (to check if file contents changed)
113 * READ_KEEP_UNDO don't clear undo info or read it from a file
114 * READ_FIFO read from fifo/socket instead of a file
115 *
116 * return FAIL for failure, NOTDONE for directory (failure), or OK
117 */
118 int
readfile(char_u * fname,char_u * sfname,linenr_T from,linenr_T lines_to_skip,linenr_T lines_to_read,exarg_T * eap,int flags)119 readfile(
120 char_u *fname,
121 char_u *sfname,
122 linenr_T from,
123 linenr_T lines_to_skip,
124 linenr_T lines_to_read,
125 exarg_T *eap, // can be NULL!
126 int flags)
127 {
128 int fd = 0;
129 int newfile = (flags & READ_NEW);
130 int check_readonly;
131 int filtering = (flags & READ_FILTER);
132 int read_stdin = (flags & READ_STDIN);
133 int read_buffer = (flags & READ_BUFFER);
134 int read_fifo = (flags & READ_FIFO);
135 int set_options = newfile || read_buffer
136 || (eap != NULL && eap->read_edit);
137 linenr_T read_buf_lnum = 1; // next line to read from curbuf
138 colnr_T read_buf_col = 0; // next char to read from this line
139 char_u c;
140 linenr_T lnum = from;
141 char_u *ptr = NULL; // pointer into read buffer
142 char_u *buffer = NULL; // read buffer
143 char_u *new_buffer = NULL; // init to shut up gcc
144 char_u *line_start = NULL; // init to shut up gcc
145 int wasempty; // buffer was empty before reading
146 colnr_T len;
147 long size = 0;
148 char_u *p;
149 off_T filesize = 0;
150 int skip_read = FALSE;
151 off_T filesize_disk = 0; // file size read from disk
152 off_T filesize_count = 0; // counter
153 #ifdef FEAT_CRYPT
154 char_u *cryptkey = NULL;
155 int did_ask_for_key = FALSE;
156 #endif
157 #ifdef FEAT_PERSISTENT_UNDO
158 context_sha256_T sha_ctx;
159 int read_undo_file = FALSE;
160 #endif
161 int split = 0; // number of split lines
162 #define UNKNOWN 0x0fffffff // file size is unknown
163 linenr_T linecnt;
164 int error = FALSE; // errors encountered
165 int ff_error = EOL_UNKNOWN; // file format with errors
166 long linerest = 0; // remaining chars in line
167 #ifdef UNIX
168 int perm = 0;
169 int swap_mode = -1; // protection bits for swap file
170 #else
171 int perm;
172 #endif
173 int fileformat = 0; // end-of-line format
174 int keep_fileformat = FALSE;
175 stat_T st;
176 int file_readonly;
177 linenr_T skip_count = 0;
178 linenr_T read_count = 0;
179 int msg_save = msg_scroll;
180 linenr_T read_no_eol_lnum = 0; // non-zero lnum when last line of
181 // last read was missing the eol
182 int try_mac;
183 int try_dos;
184 int try_unix;
185 int file_rewind = FALSE;
186 int can_retry;
187 linenr_T conv_error = 0; // line nr with conversion error
188 linenr_T illegal_byte = 0; // line nr with illegal byte
189 int keep_dest_enc = FALSE; // don't retry when char doesn't fit
190 // in destination encoding
191 int bad_char_behavior = BAD_REPLACE;
192 // BAD_KEEP, BAD_DROP or character to
193 // replace with
194 char_u *tmpname = NULL; // name of 'charconvert' output file
195 int fio_flags = 0;
196 char_u *fenc; // fileencoding to use
197 int fenc_alloced; // fenc_next is in allocated memory
198 char_u *fenc_next = NULL; // next item in 'fencs' or NULL
199 int advance_fenc = FALSE;
200 long real_size = 0;
201 #ifdef USE_ICONV
202 iconv_t iconv_fd = (iconv_t)-1; // descriptor for iconv() or -1
203 # ifdef FEAT_EVAL
204 int did_iconv = FALSE; // TRUE when iconv() failed and trying
205 // 'charconvert' next
206 # endif
207 #endif
208 int converted = FALSE; // TRUE if conversion done
209 int notconverted = FALSE; // TRUE if conversion wanted but it
210 // wasn't possible
211 char_u conv_rest[CONV_RESTLEN];
212 int conv_restlen = 0; // nr of bytes in conv_rest[]
213 pos_T orig_start;
214 buf_T *old_curbuf;
215 char_u *old_b_ffname;
216 char_u *old_b_fname;
217 int using_b_ffname;
218 int using_b_fname;
219 static char *msg_is_a_directory = N_("is a directory");
220 int eof;
221
222 au_did_filetype = FALSE; // reset before triggering any autocommands
223
224 curbuf->b_no_eol_lnum = 0; // in case it was set by the previous read
225
226 /*
227 * If there is no file name yet, use the one for the read file.
228 * BF_NOTEDITED is set to reflect this.
229 * Don't do this for a read from a filter.
230 * Only do this when 'cpoptions' contains the 'f' flag.
231 */
232 if (curbuf->b_ffname == NULL
233 && !filtering
234 && fname != NULL
235 && vim_strchr(p_cpo, CPO_FNAMER) != NULL
236 && !(flags & READ_DUMMY))
237 {
238 if (set_rw_fname(fname, sfname) == FAIL)
239 return FAIL;
240 }
241
242 // Remember the initial values of curbuf, curbuf->b_ffname and
243 // curbuf->b_fname to detect whether they are altered as a result of
244 // executing nasty autocommands. Also check if "fname" and "sfname"
245 // point to one of these values.
246 old_curbuf = curbuf;
247 old_b_ffname = curbuf->b_ffname;
248 old_b_fname = curbuf->b_fname;
249 using_b_ffname = (fname == curbuf->b_ffname)
250 || (sfname == curbuf->b_ffname);
251 using_b_fname = (fname == curbuf->b_fname) || (sfname == curbuf->b_fname);
252
253 // After reading a file the cursor line changes but we don't want to
254 // display the line.
255 ex_no_reprint = TRUE;
256
257 // don't display the file info for another buffer now
258 need_fileinfo = FALSE;
259
260 /*
261 * For Unix: Use the short file name whenever possible.
262 * Avoids problems with networks and when directory names are changed.
263 * Don't do this for MS-DOS, a "cd" in a sub-shell may have moved us to
264 * another directory, which we don't detect.
265 */
266 if (sfname == NULL)
267 sfname = fname;
268 #if defined(UNIX)
269 fname = sfname;
270 #endif
271
272 /*
273 * The BufReadCmd and FileReadCmd events intercept the reading process by
274 * executing the associated commands instead.
275 */
276 if (!filtering && !read_stdin && !read_buffer)
277 {
278 orig_start = curbuf->b_op_start;
279
280 // Set '[ mark to the line above where the lines go (line 1 if zero).
281 curbuf->b_op_start.lnum = ((from == 0) ? 1 : from);
282 curbuf->b_op_start.col = 0;
283
284 if (newfile)
285 {
286 if (apply_autocmds_exarg(EVENT_BUFREADCMD, NULL, sfname,
287 FALSE, curbuf, eap))
288 {
289 int status = OK;
290 #ifdef FEAT_EVAL
291 if (aborting())
292 status = FAIL;
293 #endif
294 // The BufReadCmd code usually uses ":read" to get the text and
295 // perhaps ":file" to change the buffer name. But we should
296 // consider this to work like ":edit", thus reset the
297 // BF_NOTEDITED flag. Then ":write" will work to overwrite the
298 // same file.
299 if (status == OK)
300 curbuf->b_flags &= ~BF_NOTEDITED;
301 return status;
302 }
303 }
304 else if (apply_autocmds_exarg(EVENT_FILEREADCMD, sfname, sfname,
305 FALSE, NULL, eap))
306 #ifdef FEAT_EVAL
307 return aborting() ? FAIL : OK;
308 #else
309 return OK;
310 #endif
311
312 curbuf->b_op_start = orig_start;
313 }
314
315 if ((shortmess(SHM_OVER) || curbuf->b_help) && p_verbose == 0)
316 msg_scroll = FALSE; // overwrite previous file message
317 else
318 msg_scroll = TRUE; // don't overwrite previous file message
319
320 if (fname != NULL && *fname != NUL)
321 {
322 size_t namelen = STRLEN(fname);
323
324 // If the name is too long we might crash further on, quit here.
325 if (namelen >= MAXPATHL)
326 {
327 filemess(curbuf, fname, (char_u *)_("Illegal file name"), 0);
328 msg_end();
329 msg_scroll = msg_save;
330 return FAIL;
331 }
332
333 // If the name ends in a path separator, we can't open it. Check here,
334 // because reading the file may actually work, but then creating the
335 // swap file may destroy it! Reported on MS-DOS and Win 95.
336 if (after_pathsep(fname, fname + namelen))
337 {
338 filemess(curbuf, fname, (char_u *)_(msg_is_a_directory), 0);
339 msg_end();
340 msg_scroll = msg_save;
341 return NOTDONE;
342 }
343 }
344
345 if (!read_stdin && !read_buffer && !read_fifo)
346 {
347 #if defined(UNIX) || defined(VMS)
348 /*
349 * On Unix it is possible to read a directory, so we have to
350 * check for it before the mch_open().
351 */
352 perm = mch_getperm(fname);
353 if (perm >= 0 && !S_ISREG(perm) // not a regular file ...
354 && !S_ISFIFO(perm) // ... or fifo
355 && !S_ISSOCK(perm) // ... or socket
356 # ifdef OPEN_CHR_FILES
357 && !(S_ISCHR(perm) && is_dev_fd_file(fname))
358 // ... or a character special file named /dev/fd/<n>
359 # endif
360 )
361 {
362 int retval = FAIL;
363
364 if (S_ISDIR(perm))
365 {
366 filemess(curbuf, fname, (char_u *)_(msg_is_a_directory), 0);
367 retval = NOTDONE;
368 }
369 else
370 filemess(curbuf, fname, (char_u *)_("is not a file"), 0);
371 msg_end();
372 msg_scroll = msg_save;
373 return retval;
374 }
375 #endif
376 #if defined(MSWIN)
377 /*
378 * MS-Windows allows opening a device, but we will probably get stuck
379 * trying to read it.
380 */
381 if (!p_odev && mch_nodetype(fname) == NODE_WRITABLE)
382 {
383 filemess(curbuf, fname, (char_u *)_("is a device (disabled with 'opendevice' option)"), 0);
384 msg_end();
385 msg_scroll = msg_save;
386 return FAIL;
387 }
388 #endif
389 }
390
391 // Set default or forced 'fileformat' and 'binary'.
392 set_file_options(set_options, eap);
393
394 /*
395 * When opening a new file we take the readonly flag from the file.
396 * Default is r/w, can be set to r/o below.
397 * Don't reset it when in readonly mode
398 * Only set/reset b_p_ro when BF_CHECK_RO is set.
399 */
400 check_readonly = (newfile && (curbuf->b_flags & BF_CHECK_RO));
401 if (check_readonly && !readonlymode)
402 curbuf->b_p_ro = FALSE;
403
404 if (newfile && !read_stdin && !read_buffer && !read_fifo)
405 {
406 // Remember time of file.
407 if (mch_stat((char *)fname, &st) >= 0)
408 {
409 buf_store_time(curbuf, &st, fname);
410 curbuf->b_mtime_read = curbuf->b_mtime;
411 curbuf->b_mtime_read_ns = curbuf->b_mtime_ns;
412 filesize_disk = st.st_size;
413 #ifdef UNIX
414 /*
415 * Use the protection bits of the original file for the swap file.
416 * This makes it possible for others to read the name of the
417 * edited file from the swapfile, but only if they can read the
418 * edited file.
419 * Remove the "write" and "execute" bits for group and others
420 * (they must not write the swapfile).
421 * Add the "read" and "write" bits for the user, otherwise we may
422 * not be able to write to the file ourselves.
423 * Setting the bits is done below, after creating the swap file.
424 */
425 swap_mode = (st.st_mode & 0644) | 0600;
426 #endif
427 #ifdef VMS
428 curbuf->b_fab_rfm = st.st_fab_rfm;
429 curbuf->b_fab_rat = st.st_fab_rat;
430 curbuf->b_fab_mrs = st.st_fab_mrs;
431 #endif
432 }
433 else
434 {
435 curbuf->b_mtime = 0;
436 curbuf->b_mtime_ns = 0;
437 curbuf->b_mtime_read = 0;
438 curbuf->b_mtime_read_ns = 0;
439 curbuf->b_orig_size = 0;
440 curbuf->b_orig_mode = 0;
441 }
442
443 // Reset the "new file" flag. It will be set again below when the
444 // file doesn't exist.
445 curbuf->b_flags &= ~(BF_NEW | BF_NEW_W);
446 }
447
448 /*
449 * for UNIX: check readonly with perm and mch_access()
450 * for Amiga: check readonly by trying to open the file for writing
451 */
452 file_readonly = FALSE;
453 if (read_stdin)
454 {
455 #if defined(MSWIN)
456 // Force binary I/O on stdin to avoid CR-LF -> LF conversion.
457 setmode(0, O_BINARY);
458 #endif
459 }
460 else if (!read_buffer)
461 {
462 #ifdef USE_MCH_ACCESS
463 if (
464 # ifdef UNIX
465 !(perm & 0222) ||
466 # endif
467 mch_access((char *)fname, W_OK))
468 file_readonly = TRUE;
469 fd = mch_open((char *)fname, O_RDONLY | O_EXTRA, 0);
470 #else
471 if (!newfile
472 || readonlymode
473 || (fd = mch_open((char *)fname, O_RDWR | O_EXTRA, 0)) < 0)
474 {
475 file_readonly = TRUE;
476 // try to open ro
477 fd = mch_open((char *)fname, O_RDONLY | O_EXTRA, 0);
478 }
479 #endif
480 }
481
482 if (fd < 0) // cannot open at all
483 {
484 #ifndef UNIX
485 int isdir_f;
486 #endif
487 msg_scroll = msg_save;
488 #ifndef UNIX
489 /*
490 * On Amiga we can't open a directory, check here.
491 */
492 isdir_f = (mch_isdir(fname));
493 perm = mch_getperm(fname); // check if the file exists
494 if (isdir_f)
495 {
496 filemess(curbuf, sfname, (char_u *)_(msg_is_a_directory), 0);
497 curbuf->b_p_ro = TRUE; // must use "w!" now
498 }
499 else
500 #endif
501 if (newfile)
502 {
503 if (perm < 0
504 #ifdef ENOENT
505 && errno == ENOENT
506 #endif
507 )
508 {
509 /*
510 * Set the 'new-file' flag, so that when the file has
511 * been created by someone else, a ":w" will complain.
512 */
513 curbuf->b_flags |= BF_NEW;
514
515 // Create a swap file now, so that other Vims are warned
516 // that we are editing this file. Don't do this for a
517 // "nofile" or "nowrite" buffer type.
518 #ifdef FEAT_QUICKFIX
519 if (!bt_dontwrite(curbuf))
520 #endif
521 {
522 check_need_swap(newfile);
523 // SwapExists autocommand may mess things up
524 if (curbuf != old_curbuf
525 || (using_b_ffname
526 && (old_b_ffname != curbuf->b_ffname))
527 || (using_b_fname
528 && (old_b_fname != curbuf->b_fname)))
529 {
530 emsg(_(e_auchangedbuf));
531 return FAIL;
532 }
533 }
534 if (dir_of_file_exists(fname))
535 filemess(curbuf, sfname,
536 (char_u *)new_file_message(), 0);
537 else
538 filemess(curbuf, sfname,
539 (char_u *)_("[New DIRECTORY]"), 0);
540 #ifdef FEAT_VIMINFO
541 // Even though this is a new file, it might have been
542 // edited before and deleted. Get the old marks.
543 check_marks_read();
544 #endif
545 // Set forced 'fileencoding'.
546 if (eap != NULL)
547 set_forced_fenc(eap);
548 apply_autocmds_exarg(EVENT_BUFNEWFILE, sfname, sfname,
549 FALSE, curbuf, eap);
550 // remember the current fileformat
551 save_file_ff(curbuf);
552
553 #if defined(FEAT_EVAL)
554 if (aborting()) // autocmds may abort script processing
555 return FAIL;
556 #endif
557 return OK; // a new file is not an error
558 }
559 else
560 {
561 filemess(curbuf, sfname, (char_u *)(
562 # ifdef EFBIG
563 (errno == EFBIG) ? _("[File too big]") :
564 # endif
565 # ifdef EOVERFLOW
566 (errno == EOVERFLOW) ? _("[File too big]") :
567 # endif
568 _("[Permission Denied]")), 0);
569 curbuf->b_p_ro = TRUE; // must use "w!" now
570 }
571 }
572
573 return FAIL;
574 }
575
576 /*
577 * Only set the 'ro' flag for readonly files the first time they are
578 * loaded. Help files always get readonly mode
579 */
580 if ((check_readonly && file_readonly) || curbuf->b_help)
581 curbuf->b_p_ro = TRUE;
582
583 if (set_options)
584 {
585 // Don't change 'eol' if reading from buffer as it will already be
586 // correctly set when reading stdin.
587 if (!read_buffer)
588 {
589 curbuf->b_p_eol = TRUE;
590 curbuf->b_start_eol = TRUE;
591 }
592 curbuf->b_p_bomb = FALSE;
593 curbuf->b_start_bomb = FALSE;
594 }
595
596 // Create a swap file now, so that other Vims are warned that we are
597 // editing this file.
598 // Don't do this for a "nofile" or "nowrite" buffer type.
599 #ifdef FEAT_QUICKFIX
600 if (!bt_dontwrite(curbuf))
601 #endif
602 {
603 check_need_swap(newfile);
604 if (!read_stdin && (curbuf != old_curbuf
605 || (using_b_ffname && (old_b_ffname != curbuf->b_ffname))
606 || (using_b_fname && (old_b_fname != curbuf->b_fname))))
607 {
608 emsg(_(e_auchangedbuf));
609 if (!read_buffer)
610 close(fd);
611 return FAIL;
612 }
613 #ifdef UNIX
614 // Set swap file protection bits after creating it.
615 if (swap_mode > 0 && curbuf->b_ml.ml_mfp != NULL
616 && curbuf->b_ml.ml_mfp->mf_fname != NULL)
617 {
618 char_u *swap_fname = curbuf->b_ml.ml_mfp->mf_fname;
619
620 /*
621 * If the group-read bit is set but not the world-read bit, then
622 * the group must be equal to the group of the original file. If
623 * we can't make that happen then reset the group-read bit. This
624 * avoids making the swap file readable to more users when the
625 * primary group of the user is too permissive.
626 */
627 if ((swap_mode & 044) == 040)
628 {
629 stat_T swap_st;
630
631 if (mch_stat((char *)swap_fname, &swap_st) >= 0
632 && st.st_gid != swap_st.st_gid
633 # ifdef HAVE_FCHOWN
634 && fchown(curbuf->b_ml.ml_mfp->mf_fd, -1, st.st_gid)
635 == -1
636 # endif
637 )
638 swap_mode &= 0600;
639 }
640
641 (void)mch_setperm(swap_fname, (long)swap_mode);
642 }
643 #endif
644 }
645
646 // If "Quit" selected at ATTENTION dialog, don't load the file
647 if (swap_exists_action == SEA_QUIT)
648 {
649 if (!read_buffer && !read_stdin)
650 close(fd);
651 return FAIL;
652 }
653
654 ++no_wait_return; // don't wait for return yet
655
656 /*
657 * Set '[ mark to the line above where the lines go (line 1 if zero).
658 */
659 orig_start = curbuf->b_op_start;
660 curbuf->b_op_start.lnum = ((from == 0) ? 1 : from);
661 curbuf->b_op_start.col = 0;
662
663 try_mac = (vim_strchr(p_ffs, 'm') != NULL);
664 try_dos = (vim_strchr(p_ffs, 'd') != NULL);
665 try_unix = (vim_strchr(p_ffs, 'x') != NULL);
666
667 if (!read_buffer)
668 {
669 int m = msg_scroll;
670 int n = msg_scrolled;
671
672 /*
673 * The file must be closed again, the autocommands may want to change
674 * the file before reading it.
675 */
676 if (!read_stdin)
677 close(fd); // ignore errors
678
679 /*
680 * The output from the autocommands should not overwrite anything and
681 * should not be overwritten: Set msg_scroll, restore its value if no
682 * output was done.
683 */
684 msg_scroll = TRUE;
685 if (filtering)
686 apply_autocmds_exarg(EVENT_FILTERREADPRE, NULL, sfname,
687 FALSE, curbuf, eap);
688 else if (read_stdin)
689 apply_autocmds_exarg(EVENT_STDINREADPRE, NULL, sfname,
690 FALSE, curbuf, eap);
691 else if (newfile)
692 apply_autocmds_exarg(EVENT_BUFREADPRE, NULL, sfname,
693 FALSE, curbuf, eap);
694 else
695 apply_autocmds_exarg(EVENT_FILEREADPRE, sfname, sfname,
696 FALSE, NULL, eap);
697 // autocommands may have changed it
698 try_mac = (vim_strchr(p_ffs, 'm') != NULL);
699 try_dos = (vim_strchr(p_ffs, 'd') != NULL);
700 try_unix = (vim_strchr(p_ffs, 'x') != NULL);
701 curbuf->b_op_start = orig_start;
702
703 if (msg_scrolled == n)
704 msg_scroll = m;
705
706 #ifdef FEAT_EVAL
707 if (aborting()) // autocmds may abort script processing
708 {
709 --no_wait_return;
710 msg_scroll = msg_save;
711 curbuf->b_p_ro = TRUE; // must use "w!" now
712 return FAIL;
713 }
714 #endif
715 /*
716 * Don't allow the autocommands to change the current buffer.
717 * Try to re-open the file.
718 *
719 * Don't allow the autocommands to change the buffer name either
720 * (cd for example) if it invalidates fname or sfname.
721 */
722 if (!read_stdin && (curbuf != old_curbuf
723 || (using_b_ffname && (old_b_ffname != curbuf->b_ffname))
724 || (using_b_fname && (old_b_fname != curbuf->b_fname))
725 || (fd = mch_open((char *)fname, O_RDONLY | O_EXTRA, 0)) < 0))
726 {
727 --no_wait_return;
728 msg_scroll = msg_save;
729 if (fd < 0)
730 emsg(_("E200: *ReadPre autocommands made the file unreadable"));
731 else
732 emsg(_("E201: *ReadPre autocommands must not change current buffer"));
733 curbuf->b_p_ro = TRUE; // must use "w!" now
734 return FAIL;
735 }
736 }
737
738 // Autocommands may add lines to the file, need to check if it is empty
739 wasempty = (curbuf->b_ml.ml_flags & ML_EMPTY);
740
741 if (!recoverymode && !filtering && !(flags & READ_DUMMY))
742 {
743 /*
744 * Show the user that we are busy reading the input. Sometimes this
745 * may take a while. When reading from stdin another program may
746 * still be running, don't move the cursor to the last line, unless
747 * always using the GUI.
748 */
749 if (read_stdin)
750 {
751 if (!is_not_a_term())
752 {
753 #ifndef ALWAYS_USE_GUI
754 # ifdef VIMDLL
755 if (!gui.in_use)
756 # endif
757 mch_msg(_("Vim: Reading from stdin...\n"));
758 #endif
759 #ifdef FEAT_GUI
760 // Also write a message in the GUI window, if there is one.
761 if (gui.in_use && !gui.dying && !gui.starting)
762 {
763 p = (char_u *)_("Reading from stdin...");
764 gui_write(p, (int)STRLEN(p));
765 }
766 #endif
767 }
768 }
769 else if (!read_buffer)
770 filemess(curbuf, sfname, (char_u *)"", 0);
771 }
772
773 msg_scroll = FALSE; // overwrite the file message
774
775 /*
776 * Set linecnt now, before the "retry" caused by a wrong guess for
777 * fileformat, and after the autocommands, which may change them.
778 */
779 linecnt = curbuf->b_ml.ml_line_count;
780
781 // "++bad=" argument.
782 if (eap != NULL && eap->bad_char != 0)
783 {
784 bad_char_behavior = eap->bad_char;
785 if (set_options)
786 curbuf->b_bad_char = eap->bad_char;
787 }
788 else
789 curbuf->b_bad_char = 0;
790
791 /*
792 * Decide which 'encoding' to use or use first.
793 */
794 if (eap != NULL && eap->force_enc != 0)
795 {
796 fenc = enc_canonize(eap->cmd + eap->force_enc);
797 fenc_alloced = TRUE;
798 keep_dest_enc = TRUE;
799 }
800 else if (curbuf->b_p_bin)
801 {
802 fenc = (char_u *)""; // binary: don't convert
803 fenc_alloced = FALSE;
804 }
805 else if (curbuf->b_help)
806 {
807 char_u firstline[80];
808 int fc;
809
810 // Help files are either utf-8 or latin1. Try utf-8 first, if this
811 // fails it must be latin1.
812 // Always do this when 'encoding' is "utf-8". Otherwise only do
813 // this when needed to avoid [converted] remarks all the time.
814 // It is needed when the first line contains non-ASCII characters.
815 // That is only in *.??x files.
816 fenc = (char_u *)"latin1";
817 c = enc_utf8;
818 if (!c && !read_stdin)
819 {
820 fc = fname[STRLEN(fname) - 1];
821 if (TOLOWER_ASC(fc) == 'x')
822 {
823 // Read the first line (and a bit more). Immediately rewind to
824 // the start of the file. If the read() fails "len" is -1.
825 len = read_eintr(fd, firstline, 80);
826 vim_lseek(fd, (off_T)0L, SEEK_SET);
827 for (p = firstline; p < firstline + len; ++p)
828 if (*p >= 0x80)
829 {
830 c = TRUE;
831 break;
832 }
833 }
834 }
835
836 if (c)
837 {
838 fenc_next = fenc;
839 fenc = (char_u *)"utf-8";
840
841 // When the file is utf-8 but a character doesn't fit in
842 // 'encoding' don't retry. In help text editing utf-8 bytes
843 // doesn't make sense.
844 if (!enc_utf8)
845 keep_dest_enc = TRUE;
846 }
847 fenc_alloced = FALSE;
848 }
849 else if (*p_fencs == NUL)
850 {
851 fenc = curbuf->b_p_fenc; // use format from buffer
852 fenc_alloced = FALSE;
853 }
854 else
855 {
856 fenc_next = p_fencs; // try items in 'fileencodings'
857 fenc = next_fenc(&fenc_next, &fenc_alloced);
858 }
859
860 /*
861 * Jump back here to retry reading the file in different ways.
862 * Reasons to retry:
863 * - encoding conversion failed: try another one from "fenc_next"
864 * - BOM detected and fenc was set, need to setup conversion
865 * - "fileformat" check failed: try another
866 *
867 * Variables set for special retry actions:
868 * "file_rewind" Rewind the file to start reading it again.
869 * "advance_fenc" Advance "fenc" using "fenc_next".
870 * "skip_read" Re-use already read bytes (BOM detected).
871 * "did_iconv" iconv() conversion failed, try 'charconvert'.
872 * "keep_fileformat" Don't reset "fileformat".
873 *
874 * Other status indicators:
875 * "tmpname" When != NULL did conversion with 'charconvert'.
876 * Output file has to be deleted afterwards.
877 * "iconv_fd" When != -1 did conversion with iconv().
878 */
879 retry:
880
881 if (file_rewind)
882 {
883 if (read_buffer)
884 {
885 read_buf_lnum = 1;
886 read_buf_col = 0;
887 }
888 else if (read_stdin || vim_lseek(fd, (off_T)0L, SEEK_SET) != 0)
889 {
890 // Can't rewind the file, give up.
891 error = TRUE;
892 goto failed;
893 }
894 // Delete the previously read lines.
895 while (lnum > from)
896 ml_delete(lnum--);
897 file_rewind = FALSE;
898 if (set_options)
899 {
900 curbuf->b_p_bomb = FALSE;
901 curbuf->b_start_bomb = FALSE;
902 }
903 conv_error = 0;
904 }
905
906 /*
907 * When retrying with another "fenc" and the first time "fileformat"
908 * will be reset.
909 */
910 if (keep_fileformat)
911 keep_fileformat = FALSE;
912 else
913 {
914 if (eap != NULL && eap->force_ff != 0)
915 {
916 fileformat = get_fileformat_force(curbuf, eap);
917 try_unix = try_dos = try_mac = FALSE;
918 }
919 else if (curbuf->b_p_bin)
920 fileformat = EOL_UNIX; // binary: use Unix format
921 else if (*p_ffs == NUL)
922 fileformat = get_fileformat(curbuf);// use format from buffer
923 else
924 fileformat = EOL_UNKNOWN; // detect from file
925 }
926
927 #ifdef USE_ICONV
928 if (iconv_fd != (iconv_t)-1)
929 {
930 // aborted conversion with iconv(), close the descriptor
931 iconv_close(iconv_fd);
932 iconv_fd = (iconv_t)-1;
933 }
934 #endif
935
936 if (advance_fenc)
937 {
938 /*
939 * Try the next entry in 'fileencodings'.
940 */
941 advance_fenc = FALSE;
942
943 if (eap != NULL && eap->force_enc != 0)
944 {
945 // Conversion given with "++cc=" wasn't possible, read
946 // without conversion.
947 notconverted = TRUE;
948 conv_error = 0;
949 if (fenc_alloced)
950 vim_free(fenc);
951 fenc = (char_u *)"";
952 fenc_alloced = FALSE;
953 }
954 else
955 {
956 if (fenc_alloced)
957 vim_free(fenc);
958 if (fenc_next != NULL)
959 {
960 fenc = next_fenc(&fenc_next, &fenc_alloced);
961 }
962 else
963 {
964 fenc = (char_u *)"";
965 fenc_alloced = FALSE;
966 }
967 }
968 if (tmpname != NULL)
969 {
970 mch_remove(tmpname); // delete converted file
971 VIM_CLEAR(tmpname);
972 }
973 }
974
975 /*
976 * Conversion may be required when the encoding of the file is different
977 * from 'encoding' or 'encoding' is UTF-16, UCS-2 or UCS-4.
978 */
979 fio_flags = 0;
980 converted = need_conversion(fenc);
981 if (converted)
982 {
983
984 // "ucs-bom" means we need to check the first bytes of the file
985 // for a BOM.
986 if (STRCMP(fenc, ENC_UCSBOM) == 0)
987 fio_flags = FIO_UCSBOM;
988
989 /*
990 * Check if UCS-2/4 or Latin1 to UTF-8 conversion needs to be
991 * done. This is handled below after read(). Prepare the
992 * fio_flags to avoid having to parse the string each time.
993 * Also check for Unicode to Latin1 conversion, because iconv()
994 * appears not to handle this correctly. This works just like
995 * conversion to UTF-8 except how the resulting character is put in
996 * the buffer.
997 */
998 else if (enc_utf8 || STRCMP(p_enc, "latin1") == 0)
999 fio_flags = get_fio_flags(fenc);
1000
1001 #ifdef MSWIN
1002 /*
1003 * Conversion from an MS-Windows codepage to UTF-8 or another codepage
1004 * is handled with MultiByteToWideChar().
1005 */
1006 if (fio_flags == 0)
1007 fio_flags = get_win_fio_flags(fenc);
1008 #endif
1009
1010 #ifdef MACOS_CONVERT
1011 // Conversion from Apple MacRoman to latin1 or UTF-8
1012 if (fio_flags == 0)
1013 fio_flags = get_mac_fio_flags(fenc);
1014 #endif
1015
1016 #ifdef USE_ICONV
1017 /*
1018 * Try using iconv() if we can't convert internally.
1019 */
1020 if (fio_flags == 0
1021 # ifdef FEAT_EVAL
1022 && !did_iconv
1023 # endif
1024 )
1025 iconv_fd = (iconv_t)my_iconv_open(
1026 enc_utf8 ? (char_u *)"utf-8" : p_enc, fenc);
1027 #endif
1028
1029 #ifdef FEAT_EVAL
1030 /*
1031 * Use the 'charconvert' expression when conversion is required
1032 * and we can't do it internally or with iconv().
1033 */
1034 if (fio_flags == 0 && !read_stdin && !read_buffer && *p_ccv != NUL
1035 && !read_fifo
1036 # ifdef USE_ICONV
1037 && iconv_fd == (iconv_t)-1
1038 # endif
1039 )
1040 {
1041 # ifdef USE_ICONV
1042 did_iconv = FALSE;
1043 # endif
1044 // Skip conversion when it's already done (retry for wrong
1045 // "fileformat").
1046 if (tmpname == NULL)
1047 {
1048 tmpname = readfile_charconvert(fname, fenc, &fd);
1049 if (tmpname == NULL)
1050 {
1051 // Conversion failed. Try another one.
1052 advance_fenc = TRUE;
1053 if (fd < 0)
1054 {
1055 // Re-opening the original file failed!
1056 emsg(_("E202: Conversion made file unreadable!"));
1057 error = TRUE;
1058 goto failed;
1059 }
1060 goto retry;
1061 }
1062 }
1063 }
1064 else
1065 #endif
1066 {
1067 if (fio_flags == 0
1068 #ifdef USE_ICONV
1069 && iconv_fd == (iconv_t)-1
1070 #endif
1071 )
1072 {
1073 // Conversion wanted but we can't.
1074 // Try the next conversion in 'fileencodings'
1075 advance_fenc = TRUE;
1076 goto retry;
1077 }
1078 }
1079 }
1080
1081 // Set "can_retry" when it's possible to rewind the file and try with
1082 // another "fenc" value. It's FALSE when no other "fenc" to try, reading
1083 // stdin or fixed at a specific encoding.
1084 can_retry = (*fenc != NUL && !read_stdin && !read_fifo && !keep_dest_enc);
1085
1086 if (!skip_read)
1087 {
1088 linerest = 0;
1089 filesize = 0;
1090 filesize_count = 0;
1091 skip_count = lines_to_skip;
1092 read_count = lines_to_read;
1093 conv_restlen = 0;
1094 #ifdef FEAT_PERSISTENT_UNDO
1095 read_undo_file = (newfile && (flags & READ_KEEP_UNDO) == 0
1096 && curbuf->b_ffname != NULL
1097 && curbuf->b_p_udf
1098 && !filtering
1099 && !read_fifo
1100 && !read_stdin
1101 && !read_buffer);
1102 if (read_undo_file)
1103 sha256_start(&sha_ctx);
1104 #endif
1105 #ifdef FEAT_CRYPT
1106 if (curbuf->b_cryptstate != NULL)
1107 {
1108 // Need to free the state, but keep the key, don't want to ask for
1109 // it again.
1110 crypt_free_state(curbuf->b_cryptstate);
1111 curbuf->b_cryptstate = NULL;
1112 }
1113 #endif
1114 }
1115
1116 while (!error && !got_int)
1117 {
1118 /*
1119 * We allocate as much space for the file as we can get, plus
1120 * space for the old line plus room for one terminating NUL.
1121 * The amount is limited by the fact that read() only can read
1122 * up to max_unsigned characters (and other things).
1123 */
1124 if (!skip_read)
1125 {
1126 #if defined(SSIZE_MAX) && (SSIZE_MAX < 0x10000L)
1127 size = SSIZE_MAX; // use max I/O size, 52K
1128 #else
1129 // Use buffer >= 64K. Add linerest to double the size if the
1130 // line gets very long, to avoid a lot of copying. But don't
1131 // read more than 1 Mbyte at a time, so we can be interrupted.
1132 size = 0x10000L + linerest;
1133 if (size > 0x100000L)
1134 size = 0x100000L;
1135 #endif
1136 }
1137
1138 // Protect against the argument of lalloc() going negative.
1139 if (size < 0 || size + linerest + 1 < 0 || linerest >= MAXCOL)
1140 {
1141 ++split;
1142 *ptr = NL; // split line by inserting a NL
1143 size = 1;
1144 }
1145 else
1146 {
1147 if (!skip_read)
1148 {
1149 for ( ; size >= 10; size = (long)((long_u)size >> 1))
1150 {
1151 if ((new_buffer = lalloc(size + linerest + 1,
1152 FALSE)) != NULL)
1153 break;
1154 }
1155 if (new_buffer == NULL)
1156 {
1157 do_outofmem_msg((long_u)(size * 2 + linerest + 1));
1158 error = TRUE;
1159 break;
1160 }
1161 if (linerest) // copy characters from the previous buffer
1162 mch_memmove(new_buffer, ptr - linerest, (size_t)linerest);
1163 vim_free(buffer);
1164 buffer = new_buffer;
1165 ptr = buffer + linerest;
1166 line_start = buffer;
1167
1168 // May need room to translate into.
1169 // For iconv() we don't really know the required space, use a
1170 // factor ICONV_MULT.
1171 // latin1 to utf-8: 1 byte becomes up to 2 bytes
1172 // utf-16 to utf-8: 2 bytes become up to 3 bytes, 4 bytes
1173 // become up to 4 bytes, size must be multiple of 2
1174 // ucs-2 to utf-8: 2 bytes become up to 3 bytes, size must be
1175 // multiple of 2
1176 // ucs-4 to utf-8: 4 bytes become up to 6 bytes, size must be
1177 // multiple of 4
1178 real_size = (int)size;
1179 #ifdef USE_ICONV
1180 if (iconv_fd != (iconv_t)-1)
1181 size = size / ICONV_MULT;
1182 else
1183 #endif
1184 if (fio_flags & FIO_LATIN1)
1185 size = size / 2;
1186 else if (fio_flags & (FIO_UCS2 | FIO_UTF16))
1187 size = (size * 2 / 3) & ~1;
1188 else if (fio_flags & FIO_UCS4)
1189 size = (size * 2 / 3) & ~3;
1190 else if (fio_flags == FIO_UCSBOM)
1191 size = size / ICONV_MULT; // worst case
1192 #ifdef MSWIN
1193 else if (fio_flags & FIO_CODEPAGE)
1194 size = size / ICONV_MULT; // also worst case
1195 #endif
1196 #ifdef MACOS_CONVERT
1197 else if (fio_flags & FIO_MACROMAN)
1198 size = size / ICONV_MULT; // also worst case
1199 #endif
1200
1201 if (conv_restlen > 0)
1202 {
1203 // Insert unconverted bytes from previous line.
1204 mch_memmove(ptr, conv_rest, conv_restlen);
1205 ptr += conv_restlen;
1206 size -= conv_restlen;
1207 }
1208
1209 if (read_buffer)
1210 {
1211 /*
1212 * Read bytes from curbuf. Used for converting text read
1213 * from stdin.
1214 */
1215 eof = FALSE;
1216 if (read_buf_lnum > from)
1217 size = 0;
1218 else
1219 {
1220 int n, ni;
1221 long tlen;
1222
1223 tlen = 0;
1224 for (;;)
1225 {
1226 p = ml_get(read_buf_lnum) + read_buf_col;
1227 n = (int)STRLEN(p);
1228 if ((int)tlen + n + 1 > size)
1229 {
1230 // Filled up to "size", append partial line.
1231 // Change NL to NUL to reverse the effect done
1232 // below.
1233 n = (int)(size - tlen);
1234 for (ni = 0; ni < n; ++ni)
1235 {
1236 if (p[ni] == NL)
1237 ptr[tlen++] = NUL;
1238 else
1239 ptr[tlen++] = p[ni];
1240 }
1241 read_buf_col += n;
1242 break;
1243 }
1244 else
1245 {
1246 // Append whole line and new-line. Change NL
1247 // to NUL to reverse the effect done below.
1248 for (ni = 0; ni < n; ++ni)
1249 {
1250 if (p[ni] == NL)
1251 ptr[tlen++] = NUL;
1252 else
1253 ptr[tlen++] = p[ni];
1254 }
1255 ptr[tlen++] = NL;
1256 read_buf_col = 0;
1257 if (++read_buf_lnum > from)
1258 {
1259 // When the last line didn't have an
1260 // end-of-line don't add it now either.
1261 if (!curbuf->b_p_eol)
1262 --tlen;
1263 size = tlen;
1264 eof = TRUE;
1265 break;
1266 }
1267 }
1268 }
1269 }
1270 }
1271 else
1272 {
1273 /*
1274 * Read bytes from the file.
1275 */
1276 # ifdef FEAT_SODIUM
1277 // Let the crypt layer work with a buffer size of 8192
1278 if (filesize == 0)
1279 // set size to 8K + Sodium Crypt Metadata
1280 size = WRITEBUFSIZE + crypt_get_max_header_len()
1281 + crypto_secretstream_xchacha20poly1305_HEADERBYTES
1282 + crypto_secretstream_xchacha20poly1305_ABYTES;
1283
1284 else if (filesize > 0 && (curbuf->b_cryptstate != NULL &&
1285 curbuf->b_cryptstate->method_nr == CRYPT_M_SOD))
1286 size = WRITEBUFSIZE + crypto_secretstream_xchacha20poly1305_ABYTES;
1287 # endif
1288 eof = size;
1289 size = read_eintr(fd, ptr, size);
1290 filesize_count += size;
1291 // hit end of file
1292 eof = (size < eof || filesize_count == filesize_disk);
1293 }
1294
1295 #ifdef FEAT_CRYPT
1296 /*
1297 * At start of file: Check for magic number of encryption.
1298 */
1299 if (filesize == 0 && size > 0)
1300 {
1301 cryptkey = check_for_cryptkey(cryptkey, ptr, &size,
1302 &filesize, newfile, sfname,
1303 &did_ask_for_key);
1304 # ifdef CRYPT_NOT_INPLACE
1305 if (curbuf->b_cryptstate != NULL
1306 && !crypt_works_inplace(curbuf->b_cryptstate))
1307 // reading undo file requires crypt_decode_inplace()
1308 read_undo_file = FALSE;
1309 # endif
1310 }
1311 /*
1312 * Decrypt the read bytes. This is done before checking for
1313 * EOF because the crypt layer may be buffering.
1314 */
1315 if (cryptkey != NULL && curbuf->b_cryptstate != NULL
1316 && size > 0)
1317 {
1318 # ifdef CRYPT_NOT_INPLACE
1319 if (crypt_works_inplace(curbuf->b_cryptstate))
1320 {
1321 # endif
1322 crypt_decode_inplace(curbuf->b_cryptstate, ptr,
1323 size, eof);
1324 # ifdef CRYPT_NOT_INPLACE
1325 }
1326 else
1327 {
1328 char_u *newptr = NULL;
1329 int decrypted_size;
1330
1331 decrypted_size = crypt_decode_alloc(
1332 curbuf->b_cryptstate, ptr, size,
1333 &newptr, eof);
1334
1335 if (decrypted_size < 0)
1336 {
1337 // error message already given
1338 error = TRUE;
1339 vim_free(newptr);
1340 break;
1341 }
1342 // If the crypt layer is buffering, not producing
1343 // anything yet, need to read more.
1344 if (decrypted_size == 0)
1345 continue;
1346
1347 if (linerest == 0)
1348 {
1349 // Simple case: reuse returned buffer (may be
1350 // NULL, checked later).
1351 new_buffer = newptr;
1352 }
1353 else
1354 {
1355 long_u new_size;
1356
1357 // Need new buffer to add bytes carried over.
1358 new_size = (long_u)(decrypted_size + linerest + 1);
1359 new_buffer = lalloc(new_size, FALSE);
1360 if (new_buffer == NULL)
1361 {
1362 do_outofmem_msg(new_size);
1363 error = TRUE;
1364 break;
1365 }
1366
1367 mch_memmove(new_buffer, buffer, linerest);
1368 if (newptr != NULL)
1369 mch_memmove(new_buffer + linerest, newptr,
1370 decrypted_size);
1371 vim_free(newptr);
1372 }
1373
1374 if (new_buffer != NULL)
1375 {
1376 vim_free(buffer);
1377 buffer = new_buffer;
1378 new_buffer = NULL;
1379 line_start = buffer;
1380 ptr = buffer + linerest;
1381 real_size = size;
1382 }
1383 size = decrypted_size;
1384 }
1385 # endif
1386 }
1387 #endif
1388
1389 if (size <= 0)
1390 {
1391 if (size < 0) // read error
1392 error = TRUE;
1393 else if (conv_restlen > 0)
1394 {
1395 /*
1396 * Reached end-of-file but some trailing bytes could
1397 * not be converted. Truncated file?
1398 */
1399
1400 // When we did a conversion report an error.
1401 if (fio_flags != 0
1402 #ifdef USE_ICONV
1403 || iconv_fd != (iconv_t)-1
1404 #endif
1405 )
1406 {
1407 if (can_retry)
1408 goto rewind_retry;
1409 if (conv_error == 0)
1410 conv_error = curbuf->b_ml.ml_line_count
1411 - linecnt + 1;
1412 }
1413 // Remember the first linenr with an illegal byte
1414 else if (illegal_byte == 0)
1415 illegal_byte = curbuf->b_ml.ml_line_count
1416 - linecnt + 1;
1417 if (bad_char_behavior == BAD_DROP)
1418 {
1419 *(ptr - conv_restlen) = NUL;
1420 conv_restlen = 0;
1421 }
1422 else
1423 {
1424 // Replace the trailing bytes with the replacement
1425 // character if we were converting; if we weren't,
1426 // leave the UTF8 checking code to do it, as it
1427 // works slightly differently.
1428 if (bad_char_behavior != BAD_KEEP && (fio_flags != 0
1429 #ifdef USE_ICONV
1430 || iconv_fd != (iconv_t)-1
1431 #endif
1432 ))
1433 {
1434 while (conv_restlen > 0)
1435 {
1436 *(--ptr) = bad_char_behavior;
1437 --conv_restlen;
1438 }
1439 }
1440 fio_flags = 0; // don't convert this
1441 #ifdef USE_ICONV
1442 if (iconv_fd != (iconv_t)-1)
1443 {
1444 iconv_close(iconv_fd);
1445 iconv_fd = (iconv_t)-1;
1446 }
1447 #endif
1448 }
1449 }
1450 }
1451 }
1452 skip_read = FALSE;
1453
1454 /*
1455 * At start of file (or after crypt magic number): Check for BOM.
1456 * Also check for a BOM for other Unicode encodings, but not after
1457 * converting with 'charconvert' or when a BOM has already been
1458 * found.
1459 */
1460 if ((filesize == 0
1461 #ifdef FEAT_CRYPT
1462 || (cryptkey != NULL
1463 && filesize == crypt_get_header_len(
1464 crypt_get_method_nr(curbuf)))
1465 #endif
1466 )
1467 && (fio_flags == FIO_UCSBOM
1468 || (!curbuf->b_p_bomb
1469 && tmpname == NULL
1470 && (*fenc == 'u' || (*fenc == NUL && enc_utf8)))))
1471 {
1472 char_u *ccname;
1473 int blen;
1474
1475 // no BOM detection in a short file or in binary mode
1476 if (size < 2 || curbuf->b_p_bin)
1477 ccname = NULL;
1478 else
1479 ccname = check_for_bom(ptr, size, &blen,
1480 fio_flags == FIO_UCSBOM ? FIO_ALL : get_fio_flags(fenc));
1481 if (ccname != NULL)
1482 {
1483 // Remove BOM from the text
1484 filesize += blen;
1485 size -= blen;
1486 mch_memmove(ptr, ptr + blen, (size_t)size);
1487 if (set_options)
1488 {
1489 curbuf->b_p_bomb = TRUE;
1490 curbuf->b_start_bomb = TRUE;
1491 }
1492 }
1493
1494 if (fio_flags == FIO_UCSBOM)
1495 {
1496 if (ccname == NULL)
1497 {
1498 // No BOM detected: retry with next encoding.
1499 advance_fenc = TRUE;
1500 }
1501 else
1502 {
1503 // BOM detected: set "fenc" and jump back
1504 if (fenc_alloced)
1505 vim_free(fenc);
1506 fenc = ccname;
1507 fenc_alloced = FALSE;
1508 }
1509 // retry reading without getting new bytes or rewinding
1510 skip_read = TRUE;
1511 goto retry;
1512 }
1513 }
1514
1515 // Include not converted bytes.
1516 ptr -= conv_restlen;
1517 size += conv_restlen;
1518 conv_restlen = 0;
1519 /*
1520 * Break here for a read error or end-of-file.
1521 */
1522 if (size <= 0)
1523 break;
1524
1525
1526 #ifdef USE_ICONV
1527 if (iconv_fd != (iconv_t)-1)
1528 {
1529 /*
1530 * Attempt conversion of the read bytes to 'encoding' using
1531 * iconv().
1532 */
1533 const char *fromp;
1534 char *top;
1535 size_t from_size;
1536 size_t to_size;
1537
1538 fromp = (char *)ptr;
1539 from_size = size;
1540 ptr += size;
1541 top = (char *)ptr;
1542 to_size = real_size - size;
1543
1544 /*
1545 * If there is conversion error or not enough room try using
1546 * another conversion. Except for when there is no
1547 * alternative (help files).
1548 */
1549 while ((iconv(iconv_fd, (void *)&fromp, &from_size,
1550 &top, &to_size)
1551 == (size_t)-1 && ICONV_ERRNO != ICONV_EINVAL)
1552 || from_size > CONV_RESTLEN)
1553 {
1554 if (can_retry)
1555 goto rewind_retry;
1556 if (conv_error == 0)
1557 conv_error = readfile_linenr(linecnt,
1558 ptr, (char_u *)top);
1559
1560 // Deal with a bad byte and continue with the next.
1561 ++fromp;
1562 --from_size;
1563 if (bad_char_behavior == BAD_KEEP)
1564 {
1565 *top++ = *(fromp - 1);
1566 --to_size;
1567 }
1568 else if (bad_char_behavior != BAD_DROP)
1569 {
1570 *top++ = bad_char_behavior;
1571 --to_size;
1572 }
1573 }
1574
1575 if (from_size > 0)
1576 {
1577 // Some remaining characters, keep them for the next
1578 // round.
1579 mch_memmove(conv_rest, (char_u *)fromp, from_size);
1580 conv_restlen = (int)from_size;
1581 }
1582
1583 // move the linerest to before the converted characters
1584 line_start = ptr - linerest;
1585 mch_memmove(line_start, buffer, (size_t)linerest);
1586 size = (long)((char_u *)top - ptr);
1587 }
1588 #endif
1589
1590 #ifdef MSWIN
1591 if (fio_flags & FIO_CODEPAGE)
1592 {
1593 char_u *src, *dst;
1594 WCHAR ucs2buf[3];
1595 int ucs2len;
1596 int codepage = FIO_GET_CP(fio_flags);
1597 int bytelen;
1598 int found_bad;
1599 char replstr[2];
1600
1601 /*
1602 * Conversion from an MS-Windows codepage or UTF-8 to UTF-8 or
1603 * a codepage, using standard MS-Windows functions. This
1604 * requires two steps:
1605 * 1. convert from 'fileencoding' to ucs-2
1606 * 2. convert from ucs-2 to 'encoding'
1607 *
1608 * Because there may be illegal bytes AND an incomplete byte
1609 * sequence at the end, we may have to do the conversion one
1610 * character at a time to get it right.
1611 */
1612
1613 // Replacement string for WideCharToMultiByte().
1614 if (bad_char_behavior > 0)
1615 replstr[0] = bad_char_behavior;
1616 else
1617 replstr[0] = '?';
1618 replstr[1] = NUL;
1619
1620 /*
1621 * Move the bytes to the end of the buffer, so that we have
1622 * room to put the result at the start.
1623 */
1624 src = ptr + real_size - size;
1625 mch_memmove(src, ptr, size);
1626
1627 /*
1628 * Do the conversion.
1629 */
1630 dst = ptr;
1631 size = size;
1632 while (size > 0)
1633 {
1634 found_bad = FALSE;
1635
1636 # ifdef CP_UTF8 // VC 4.1 doesn't define CP_UTF8
1637 if (codepage == CP_UTF8)
1638 {
1639 // Handle CP_UTF8 input ourselves to be able to handle
1640 // trailing bytes properly.
1641 // Get one UTF-8 character from src.
1642 bytelen = (int)utf_ptr2len_len(src, size);
1643 if (bytelen > size)
1644 {
1645 // Only got some bytes of a character. Normally
1646 // it's put in "conv_rest", but if it's too long
1647 // deal with it as if they were illegal bytes.
1648 if (bytelen <= CONV_RESTLEN)
1649 break;
1650
1651 // weird overlong byte sequence
1652 bytelen = size;
1653 found_bad = TRUE;
1654 }
1655 else
1656 {
1657 int u8c = utf_ptr2char(src);
1658
1659 if (u8c > 0xffff || (*src >= 0x80 && bytelen == 1))
1660 found_bad = TRUE;
1661 ucs2buf[0] = u8c;
1662 ucs2len = 1;
1663 }
1664 }
1665 else
1666 # endif
1667 {
1668 // We don't know how long the byte sequence is, try
1669 // from one to three bytes.
1670 for (bytelen = 1; bytelen <= size && bytelen <= 3;
1671 ++bytelen)
1672 {
1673 ucs2len = MultiByteToWideChar(codepage,
1674 MB_ERR_INVALID_CHARS,
1675 (LPCSTR)src, bytelen,
1676 ucs2buf, 3);
1677 if (ucs2len > 0)
1678 break;
1679 }
1680 if (ucs2len == 0)
1681 {
1682 // If we have only one byte then it's probably an
1683 // incomplete byte sequence. Otherwise discard
1684 // one byte as a bad character.
1685 if (size == 1)
1686 break;
1687 found_bad = TRUE;
1688 bytelen = 1;
1689 }
1690 }
1691
1692 if (!found_bad)
1693 {
1694 int i;
1695
1696 // Convert "ucs2buf[ucs2len]" to 'enc' in "dst".
1697 if (enc_utf8)
1698 {
1699 // From UCS-2 to UTF-8. Cannot fail.
1700 for (i = 0; i < ucs2len; ++i)
1701 dst += utf_char2bytes(ucs2buf[i], dst);
1702 }
1703 else
1704 {
1705 BOOL bad = FALSE;
1706 int dstlen;
1707
1708 // From UCS-2 to "enc_codepage". If the
1709 // conversion uses the default character "?",
1710 // the data doesn't fit in this encoding.
1711 dstlen = WideCharToMultiByte(enc_codepage, 0,
1712 (LPCWSTR)ucs2buf, ucs2len,
1713 (LPSTR)dst, (int)(src - dst),
1714 replstr, &bad);
1715 if (bad)
1716 found_bad = TRUE;
1717 else
1718 dst += dstlen;
1719 }
1720 }
1721
1722 if (found_bad)
1723 {
1724 // Deal with bytes we can't convert.
1725 if (can_retry)
1726 goto rewind_retry;
1727 if (conv_error == 0)
1728 conv_error = readfile_linenr(linecnt, ptr, dst);
1729 if (bad_char_behavior != BAD_DROP)
1730 {
1731 if (bad_char_behavior == BAD_KEEP)
1732 {
1733 mch_memmove(dst, src, bytelen);
1734 dst += bytelen;
1735 }
1736 else
1737 *dst++ = bad_char_behavior;
1738 }
1739 }
1740
1741 src += bytelen;
1742 size -= bytelen;
1743 }
1744
1745 if (size > 0)
1746 {
1747 // An incomplete byte sequence remaining.
1748 mch_memmove(conv_rest, src, size);
1749 conv_restlen = size;
1750 }
1751
1752 // The new size is equal to how much "dst" was advanced.
1753 size = (long)(dst - ptr);
1754 }
1755 else
1756 #endif
1757 #ifdef MACOS_CONVERT
1758 if (fio_flags & FIO_MACROMAN)
1759 {
1760 /*
1761 * Conversion from Apple MacRoman char encoding to UTF-8 or
1762 * latin1. This is in os_mac_conv.c.
1763 */
1764 if (macroman2enc(ptr, &size, real_size) == FAIL)
1765 goto rewind_retry;
1766 }
1767 else
1768 #endif
1769 if (fio_flags != 0)
1770 {
1771 int u8c;
1772 char_u *dest;
1773 char_u *tail = NULL;
1774
1775 /*
1776 * "enc_utf8" set: Convert Unicode or Latin1 to UTF-8.
1777 * "enc_utf8" not set: Convert Unicode to Latin1.
1778 * Go from end to start through the buffer, because the number
1779 * of bytes may increase.
1780 * "dest" points to after where the UTF-8 bytes go, "p" points
1781 * to after the next character to convert.
1782 */
1783 dest = ptr + real_size;
1784 if (fio_flags == FIO_LATIN1 || fio_flags == FIO_UTF8)
1785 {
1786 p = ptr + size;
1787 if (fio_flags == FIO_UTF8)
1788 {
1789 // Check for a trailing incomplete UTF-8 sequence
1790 tail = ptr + size - 1;
1791 while (tail > ptr && (*tail & 0xc0) == 0x80)
1792 --tail;
1793 if (tail + utf_byte2len(*tail) <= ptr + size)
1794 tail = NULL;
1795 else
1796 p = tail;
1797 }
1798 }
1799 else if (fio_flags & (FIO_UCS2 | FIO_UTF16))
1800 {
1801 // Check for a trailing byte
1802 p = ptr + (size & ~1);
1803 if (size & 1)
1804 tail = p;
1805 if ((fio_flags & FIO_UTF16) && p > ptr)
1806 {
1807 // Check for a trailing leading word
1808 if (fio_flags & FIO_ENDIAN_L)
1809 {
1810 u8c = (*--p << 8);
1811 u8c += *--p;
1812 }
1813 else
1814 {
1815 u8c = *--p;
1816 u8c += (*--p << 8);
1817 }
1818 if (u8c >= 0xd800 && u8c <= 0xdbff)
1819 tail = p;
1820 else
1821 p += 2;
1822 }
1823 }
1824 else // FIO_UCS4
1825 {
1826 // Check for trailing 1, 2 or 3 bytes
1827 p = ptr + (size & ~3);
1828 if (size & 3)
1829 tail = p;
1830 }
1831
1832 // If there is a trailing incomplete sequence move it to
1833 // conv_rest[].
1834 if (tail != NULL)
1835 {
1836 conv_restlen = (int)((ptr + size) - tail);
1837 mch_memmove(conv_rest, (char_u *)tail, conv_restlen);
1838 size -= conv_restlen;
1839 }
1840
1841
1842 while (p > ptr)
1843 {
1844 if (fio_flags & FIO_LATIN1)
1845 u8c = *--p;
1846 else if (fio_flags & (FIO_UCS2 | FIO_UTF16))
1847 {
1848 if (fio_flags & FIO_ENDIAN_L)
1849 {
1850 u8c = (*--p << 8);
1851 u8c += *--p;
1852 }
1853 else
1854 {
1855 u8c = *--p;
1856 u8c += (*--p << 8);
1857 }
1858 if ((fio_flags & FIO_UTF16)
1859 && u8c >= 0xdc00 && u8c <= 0xdfff)
1860 {
1861 int u16c;
1862
1863 if (p == ptr)
1864 {
1865 // Missing leading word.
1866 if (can_retry)
1867 goto rewind_retry;
1868 if (conv_error == 0)
1869 conv_error = readfile_linenr(linecnt,
1870 ptr, p);
1871 if (bad_char_behavior == BAD_DROP)
1872 continue;
1873 if (bad_char_behavior != BAD_KEEP)
1874 u8c = bad_char_behavior;
1875 }
1876
1877 // found second word of double-word, get the first
1878 // word and compute the resulting character
1879 if (fio_flags & FIO_ENDIAN_L)
1880 {
1881 u16c = (*--p << 8);
1882 u16c += *--p;
1883 }
1884 else
1885 {
1886 u16c = *--p;
1887 u16c += (*--p << 8);
1888 }
1889 u8c = 0x10000 + ((u16c & 0x3ff) << 10)
1890 + (u8c & 0x3ff);
1891
1892 // Check if the word is indeed a leading word.
1893 if (u16c < 0xd800 || u16c > 0xdbff)
1894 {
1895 if (can_retry)
1896 goto rewind_retry;
1897 if (conv_error == 0)
1898 conv_error = readfile_linenr(linecnt,
1899 ptr, p);
1900 if (bad_char_behavior == BAD_DROP)
1901 continue;
1902 if (bad_char_behavior != BAD_KEEP)
1903 u8c = bad_char_behavior;
1904 }
1905 }
1906 }
1907 else if (fio_flags & FIO_UCS4)
1908 {
1909 if (fio_flags & FIO_ENDIAN_L)
1910 {
1911 u8c = (unsigned)*--p << 24;
1912 u8c += (unsigned)*--p << 16;
1913 u8c += (unsigned)*--p << 8;
1914 u8c += *--p;
1915 }
1916 else // big endian
1917 {
1918 u8c = *--p;
1919 u8c += (unsigned)*--p << 8;
1920 u8c += (unsigned)*--p << 16;
1921 u8c += (unsigned)*--p << 24;
1922 }
1923 }
1924 else // UTF-8
1925 {
1926 if (*--p < 0x80)
1927 u8c = *p;
1928 else
1929 {
1930 len = utf_head_off(ptr, p);
1931 p -= len;
1932 u8c = utf_ptr2char(p);
1933 if (len == 0)
1934 {
1935 // Not a valid UTF-8 character, retry with
1936 // another fenc when possible, otherwise just
1937 // report the error.
1938 if (can_retry)
1939 goto rewind_retry;
1940 if (conv_error == 0)
1941 conv_error = readfile_linenr(linecnt,
1942 ptr, p);
1943 if (bad_char_behavior == BAD_DROP)
1944 continue;
1945 if (bad_char_behavior != BAD_KEEP)
1946 u8c = bad_char_behavior;
1947 }
1948 }
1949 }
1950 if (enc_utf8) // produce UTF-8
1951 {
1952 dest -= utf_char2len(u8c);
1953 (void)utf_char2bytes(u8c, dest);
1954 }
1955 else // produce Latin1
1956 {
1957 --dest;
1958 if (u8c >= 0x100)
1959 {
1960 // character doesn't fit in latin1, retry with
1961 // another fenc when possible, otherwise just
1962 // report the error.
1963 if (can_retry)
1964 goto rewind_retry;
1965 if (conv_error == 0)
1966 conv_error = readfile_linenr(linecnt, ptr, p);
1967 if (bad_char_behavior == BAD_DROP)
1968 ++dest;
1969 else if (bad_char_behavior == BAD_KEEP)
1970 *dest = u8c;
1971 else if (eap != NULL && eap->bad_char != 0)
1972 *dest = bad_char_behavior;
1973 else
1974 *dest = 0xBF;
1975 }
1976 else
1977 *dest = u8c;
1978 }
1979 }
1980
1981 // move the linerest to before the converted characters
1982 line_start = dest - linerest;
1983 mch_memmove(line_start, buffer, (size_t)linerest);
1984 size = (long)((ptr + real_size) - dest);
1985 ptr = dest;
1986 }
1987 else if (enc_utf8 && !curbuf->b_p_bin)
1988 {
1989 int incomplete_tail = FALSE;
1990
1991 // Reading UTF-8: Check if the bytes are valid UTF-8.
1992 for (p = ptr; ; ++p)
1993 {
1994 int todo = (int)((ptr + size) - p);
1995 int l;
1996
1997 if (todo <= 0)
1998 break;
1999 if (*p >= 0x80)
2000 {
2001 // A length of 1 means it's an illegal byte. Accept
2002 // an incomplete character at the end though, the next
2003 // read() will get the next bytes, we'll check it
2004 // then.
2005 l = utf_ptr2len_len(p, todo);
2006 if (l > todo && !incomplete_tail)
2007 {
2008 // Avoid retrying with a different encoding when
2009 // a truncated file is more likely, or attempting
2010 // to read the rest of an incomplete sequence when
2011 // we have already done so.
2012 if (p > ptr || filesize > 0)
2013 incomplete_tail = TRUE;
2014 // Incomplete byte sequence, move it to conv_rest[]
2015 // and try to read the rest of it, unless we've
2016 // already done so.
2017 if (p > ptr)
2018 {
2019 conv_restlen = todo;
2020 mch_memmove(conv_rest, p, conv_restlen);
2021 size -= conv_restlen;
2022 break;
2023 }
2024 }
2025 if (l == 1 || l > todo)
2026 {
2027 // Illegal byte. If we can try another encoding
2028 // do that, unless at EOF where a truncated
2029 // file is more likely than a conversion error.
2030 if (can_retry && !incomplete_tail)
2031 break;
2032 #ifdef USE_ICONV
2033 // When we did a conversion report an error.
2034 if (iconv_fd != (iconv_t)-1 && conv_error == 0)
2035 conv_error = readfile_linenr(linecnt, ptr, p);
2036 #endif
2037 // Remember the first linenr with an illegal byte
2038 if (conv_error == 0 && illegal_byte == 0)
2039 illegal_byte = readfile_linenr(linecnt, ptr, p);
2040
2041 // Drop, keep or replace the bad byte.
2042 if (bad_char_behavior == BAD_DROP)
2043 {
2044 mch_memmove(p, p + 1, todo - 1);
2045 --p;
2046 --size;
2047 }
2048 else if (bad_char_behavior != BAD_KEEP)
2049 *p = bad_char_behavior;
2050 }
2051 else
2052 p += l - 1;
2053 }
2054 }
2055 if (p < ptr + size && !incomplete_tail)
2056 {
2057 // Detected a UTF-8 error.
2058 rewind_retry:
2059 // Retry reading with another conversion.
2060 #if defined(FEAT_EVAL) && defined(USE_ICONV)
2061 if (*p_ccv != NUL && iconv_fd != (iconv_t)-1)
2062 // iconv() failed, try 'charconvert'
2063 did_iconv = TRUE;
2064 else
2065 #endif
2066 // use next item from 'fileencodings'
2067 advance_fenc = TRUE;
2068 file_rewind = TRUE;
2069 goto retry;
2070 }
2071 }
2072
2073 // count the number of characters (after conversion!)
2074 filesize += size;
2075
2076 /*
2077 * when reading the first part of a file: guess EOL type
2078 */
2079 if (fileformat == EOL_UNKNOWN)
2080 {
2081 // First try finding a NL, for Dos and Unix
2082 if (try_dos || try_unix)
2083 {
2084 // Reset the carriage return counter.
2085 if (try_mac)
2086 try_mac = 1;
2087
2088 for (p = ptr; p < ptr + size; ++p)
2089 {
2090 if (*p == NL)
2091 {
2092 if (!try_unix
2093 || (try_dos && p > ptr && p[-1] == CAR))
2094 fileformat = EOL_DOS;
2095 else
2096 fileformat = EOL_UNIX;
2097 break;
2098 }
2099 else if (*p == CAR && try_mac)
2100 try_mac++;
2101 }
2102
2103 // Don't give in to EOL_UNIX if EOL_MAC is more likely
2104 if (fileformat == EOL_UNIX && try_mac)
2105 {
2106 // Need to reset the counters when retrying fenc.
2107 try_mac = 1;
2108 try_unix = 1;
2109 for (; p >= ptr && *p != CAR; p--)
2110 ;
2111 if (p >= ptr)
2112 {
2113 for (p = ptr; p < ptr + size; ++p)
2114 {
2115 if (*p == NL)
2116 try_unix++;
2117 else if (*p == CAR)
2118 try_mac++;
2119 }
2120 if (try_mac > try_unix)
2121 fileformat = EOL_MAC;
2122 }
2123 }
2124 else if (fileformat == EOL_UNKNOWN && try_mac == 1)
2125 // Looking for CR but found no end-of-line markers at
2126 // all: use the default format.
2127 fileformat = default_fileformat();
2128 }
2129
2130 // No NL found: may use Mac format
2131 if (fileformat == EOL_UNKNOWN && try_mac)
2132 fileformat = EOL_MAC;
2133
2134 // Still nothing found? Use first format in 'ffs'
2135 if (fileformat == EOL_UNKNOWN)
2136 fileformat = default_fileformat();
2137
2138 // if editing a new file: may set p_tx and p_ff
2139 if (set_options)
2140 set_fileformat(fileformat, OPT_LOCAL);
2141 }
2142 }
2143
2144 /*
2145 * This loop is executed once for every character read.
2146 * Keep it fast!
2147 */
2148 if (fileformat == EOL_MAC)
2149 {
2150 --ptr;
2151 while (++ptr, --size >= 0)
2152 {
2153 // catch most common case first
2154 if ((c = *ptr) != NUL && c != CAR && c != NL)
2155 continue;
2156 if (c == NUL)
2157 *ptr = NL; // NULs are replaced by newlines!
2158 else if (c == NL)
2159 *ptr = CAR; // NLs are replaced by CRs!
2160 else
2161 {
2162 if (skip_count == 0)
2163 {
2164 *ptr = NUL; // end of line
2165 len = (colnr_T) (ptr - line_start + 1);
2166 if (ml_append(lnum, line_start, len, newfile) == FAIL)
2167 {
2168 error = TRUE;
2169 break;
2170 }
2171 #ifdef FEAT_PERSISTENT_UNDO
2172 if (read_undo_file)
2173 sha256_update(&sha_ctx, line_start, len);
2174 #endif
2175 ++lnum;
2176 if (--read_count == 0)
2177 {
2178 error = TRUE; // break loop
2179 line_start = ptr; // nothing left to write
2180 break;
2181 }
2182 }
2183 else
2184 --skip_count;
2185 line_start = ptr + 1;
2186 }
2187 }
2188 }
2189 else
2190 {
2191 --ptr;
2192 while (++ptr, --size >= 0)
2193 {
2194 if ((c = *ptr) != NUL && c != NL) // catch most common case
2195 continue;
2196 if (c == NUL)
2197 *ptr = NL; // NULs are replaced by newlines!
2198 else
2199 {
2200 if (skip_count == 0)
2201 {
2202 *ptr = NUL; // end of line
2203 len = (colnr_T)(ptr - line_start + 1);
2204 if (fileformat == EOL_DOS)
2205 {
2206 if (ptr > line_start && ptr[-1] == CAR)
2207 {
2208 // remove CR before NL
2209 ptr[-1] = NUL;
2210 --len;
2211 }
2212 /*
2213 * Reading in Dos format, but no CR-LF found!
2214 * When 'fileformats' includes "unix", delete all
2215 * the lines read so far and start all over again.
2216 * Otherwise give an error message later.
2217 */
2218 else if (ff_error != EOL_DOS)
2219 {
2220 if ( try_unix
2221 && !read_stdin
2222 && (read_buffer
2223 || vim_lseek(fd, (off_T)0L, SEEK_SET)
2224 == 0))
2225 {
2226 fileformat = EOL_UNIX;
2227 if (set_options)
2228 set_fileformat(EOL_UNIX, OPT_LOCAL);
2229 file_rewind = TRUE;
2230 keep_fileformat = TRUE;
2231 goto retry;
2232 }
2233 ff_error = EOL_DOS;
2234 }
2235 }
2236 if (ml_append(lnum, line_start, len, newfile) == FAIL)
2237 {
2238 error = TRUE;
2239 break;
2240 }
2241 #ifdef FEAT_PERSISTENT_UNDO
2242 if (read_undo_file)
2243 sha256_update(&sha_ctx, line_start, len);
2244 #endif
2245 ++lnum;
2246 if (--read_count == 0)
2247 {
2248 error = TRUE; // break loop
2249 line_start = ptr; // nothing left to write
2250 break;
2251 }
2252 }
2253 else
2254 --skip_count;
2255 line_start = ptr + 1;
2256 }
2257 }
2258 }
2259 linerest = (long)(ptr - line_start);
2260 ui_breakcheck();
2261 }
2262
2263 failed:
2264 // not an error, max. number of lines reached
2265 if (error && read_count == 0)
2266 error = FALSE;
2267
2268 /*
2269 * If we get EOF in the middle of a line, note the fact and
2270 * complete the line ourselves.
2271 * In Dos format ignore a trailing CTRL-Z, unless 'binary' set.
2272 */
2273 if (!error
2274 && !got_int
2275 && linerest != 0
2276 && !(!curbuf->b_p_bin
2277 && fileformat == EOL_DOS
2278 && *line_start == Ctrl_Z
2279 && ptr == line_start + 1))
2280 {
2281 // remember for when writing
2282 if (set_options)
2283 curbuf->b_p_eol = FALSE;
2284 *ptr = NUL;
2285 len = (colnr_T)(ptr - line_start + 1);
2286 if (ml_append(lnum, line_start, len, newfile) == FAIL)
2287 error = TRUE;
2288 else
2289 {
2290 #ifdef FEAT_PERSISTENT_UNDO
2291 if (read_undo_file)
2292 sha256_update(&sha_ctx, line_start, len);
2293 #endif
2294 read_no_eol_lnum = ++lnum;
2295 }
2296 }
2297
2298 if (set_options)
2299 save_file_ff(curbuf); // remember the current file format
2300
2301 #ifdef FEAT_CRYPT
2302 if (curbuf->b_cryptstate != NULL)
2303 {
2304 crypt_free_state(curbuf->b_cryptstate);
2305 curbuf->b_cryptstate = NULL;
2306 }
2307 if (cryptkey != NULL && cryptkey != curbuf->b_p_key)
2308 crypt_free_key(cryptkey);
2309 // Don't set cryptkey to NULL, it's used below as a flag that
2310 // encryption was used.
2311 #endif
2312
2313 // If editing a new file: set 'fenc' for the current buffer.
2314 // Also for ":read ++edit file".
2315 if (set_options)
2316 set_string_option_direct((char_u *)"fenc", -1, fenc,
2317 OPT_FREE|OPT_LOCAL, 0);
2318 if (fenc_alloced)
2319 vim_free(fenc);
2320 #ifdef USE_ICONV
2321 if (iconv_fd != (iconv_t)-1)
2322 iconv_close(iconv_fd);
2323 #endif
2324
2325 if (!read_buffer && !read_stdin)
2326 close(fd); // errors are ignored
2327 #ifdef HAVE_FD_CLOEXEC
2328 else
2329 {
2330 int fdflags = fcntl(fd, F_GETFD);
2331
2332 if (fdflags >= 0 && (fdflags & FD_CLOEXEC) == 0)
2333 (void)fcntl(fd, F_SETFD, fdflags | FD_CLOEXEC);
2334 }
2335 #endif
2336 vim_free(buffer);
2337
2338 #ifdef HAVE_DUP
2339 if (read_stdin)
2340 {
2341 // Use stderr for stdin, makes shell commands work.
2342 close(0);
2343 vim_ignored = dup(2);
2344 }
2345 #endif
2346
2347 if (tmpname != NULL)
2348 {
2349 mch_remove(tmpname); // delete converted file
2350 vim_free(tmpname);
2351 }
2352 --no_wait_return; // may wait for return now
2353
2354 /*
2355 * In recovery mode everything but autocommands is skipped.
2356 */
2357 if (!recoverymode)
2358 {
2359 // need to delete the last line, which comes from the empty buffer
2360 if (newfile && wasempty && !(curbuf->b_ml.ml_flags & ML_EMPTY))
2361 {
2362 #ifdef FEAT_NETBEANS_INTG
2363 netbeansFireChanges = 0;
2364 #endif
2365 ml_delete(curbuf->b_ml.ml_line_count);
2366 #ifdef FEAT_NETBEANS_INTG
2367 netbeansFireChanges = 1;
2368 #endif
2369 --linecnt;
2370 }
2371 linecnt = curbuf->b_ml.ml_line_count - linecnt;
2372 if (filesize == 0)
2373 linecnt = 0;
2374 if (newfile || read_buffer)
2375 {
2376 redraw_curbuf_later(NOT_VALID);
2377 #ifdef FEAT_DIFF
2378 // After reading the text into the buffer the diff info needs to
2379 // be updated.
2380 diff_invalidate(curbuf);
2381 #endif
2382 #ifdef FEAT_FOLDING
2383 // All folds in the window are invalid now. Mark them for update
2384 // before triggering autocommands.
2385 foldUpdateAll(curwin);
2386 #endif
2387 }
2388 else if (linecnt) // appended at least one line
2389 appended_lines_mark(from, linecnt);
2390
2391 #ifndef ALWAYS_USE_GUI
2392 /*
2393 * If we were reading from the same terminal as where messages go,
2394 * the screen will have been messed up.
2395 * Switch on raw mode now and clear the screen.
2396 */
2397 if (read_stdin)
2398 {
2399 settmode(TMODE_RAW); // set to raw mode
2400 starttermcap();
2401 screenclear();
2402 }
2403 #endif
2404
2405 if (got_int)
2406 {
2407 if (!(flags & READ_DUMMY))
2408 {
2409 filemess(curbuf, sfname, (char_u *)_(e_interr), 0);
2410 if (newfile)
2411 curbuf->b_p_ro = TRUE; // must use "w!" now
2412 }
2413 msg_scroll = msg_save;
2414 #ifdef FEAT_VIMINFO
2415 check_marks_read();
2416 #endif
2417 return OK; // an interrupt isn't really an error
2418 }
2419
2420 if (!filtering && !(flags & READ_DUMMY))
2421 {
2422 msg_add_fname(curbuf, sfname); // fname in IObuff with quotes
2423 c = FALSE;
2424
2425 #ifdef UNIX
2426 if (S_ISFIFO(perm)) // fifo
2427 {
2428 STRCAT(IObuff, _("[fifo]"));
2429 c = TRUE;
2430 }
2431 if (S_ISSOCK(perm)) // or socket
2432 {
2433 STRCAT(IObuff, _("[socket]"));
2434 c = TRUE;
2435 }
2436 # ifdef OPEN_CHR_FILES
2437 if (S_ISCHR(perm)) // or character special
2438 {
2439 STRCAT(IObuff, _("[character special]"));
2440 c = TRUE;
2441 }
2442 # endif
2443 #endif
2444 if (curbuf->b_p_ro)
2445 {
2446 STRCAT(IObuff, shortmess(SHM_RO) ? _("[RO]") : _("[readonly]"));
2447 c = TRUE;
2448 }
2449 if (read_no_eol_lnum)
2450 {
2451 msg_add_eol();
2452 c = TRUE;
2453 }
2454 if (ff_error == EOL_DOS)
2455 {
2456 STRCAT(IObuff, _("[CR missing]"));
2457 c = TRUE;
2458 }
2459 if (split)
2460 {
2461 STRCAT(IObuff, _("[long lines split]"));
2462 c = TRUE;
2463 }
2464 if (notconverted)
2465 {
2466 STRCAT(IObuff, _("[NOT converted]"));
2467 c = TRUE;
2468 }
2469 else if (converted)
2470 {
2471 STRCAT(IObuff, _("[converted]"));
2472 c = TRUE;
2473 }
2474 #ifdef FEAT_CRYPT
2475 if (cryptkey != NULL)
2476 {
2477 crypt_append_msg(curbuf);
2478 c = TRUE;
2479 }
2480 #endif
2481 if (conv_error != 0)
2482 {
2483 sprintf((char *)IObuff + STRLEN(IObuff),
2484 _("[CONVERSION ERROR in line %ld]"), (long)conv_error);
2485 c = TRUE;
2486 }
2487 else if (illegal_byte > 0)
2488 {
2489 sprintf((char *)IObuff + STRLEN(IObuff),
2490 _("[ILLEGAL BYTE in line %ld]"), (long)illegal_byte);
2491 c = TRUE;
2492 }
2493 else if (error)
2494 {
2495 STRCAT(IObuff, _("[READ ERRORS]"));
2496 c = TRUE;
2497 }
2498 if (msg_add_fileformat(fileformat))
2499 c = TRUE;
2500 #ifdef FEAT_CRYPT
2501 if (cryptkey != NULL)
2502 msg_add_lines(c, (long)linecnt, filesize
2503 - crypt_get_header_len(crypt_get_method_nr(curbuf)));
2504 else
2505 #endif
2506 msg_add_lines(c, (long)linecnt, filesize);
2507
2508 VIM_CLEAR(keep_msg);
2509 msg_scrolled_ign = TRUE;
2510 #ifdef ALWAYS_USE_GUI
2511 // Don't show the message when reading stdin, it would end up in a
2512 // message box (which might be shown when exiting!)
2513 if (read_stdin || read_buffer)
2514 p = msg_may_trunc(FALSE, IObuff);
2515 else
2516 #endif
2517 {
2518 if (msg_col > 0)
2519 msg_putchar('\r'); // overwrite previous message
2520 p = (char_u *)msg_trunc_attr((char *)IObuff, FALSE, 0);
2521 }
2522 if (read_stdin || read_buffer || restart_edit != 0
2523 || (msg_scrolled != 0 && !need_wait_return))
2524 // Need to repeat the message after redrawing when:
2525 // - When reading from stdin (the screen will be cleared next).
2526 // - When restart_edit is set (otherwise there will be a delay
2527 // before redrawing).
2528 // - When the screen was scrolled but there is no wait-return
2529 // prompt.
2530 set_keep_msg(p, 0);
2531 msg_scrolled_ign = FALSE;
2532 }
2533
2534 // with errors writing the file requires ":w!"
2535 if (newfile && (error
2536 || conv_error != 0
2537 || (illegal_byte > 0 && bad_char_behavior != BAD_KEEP)))
2538 curbuf->b_p_ro = TRUE;
2539
2540 u_clearline(); // cannot use "U" command after adding lines
2541
2542 /*
2543 * In Ex mode: cursor at last new line.
2544 * Otherwise: cursor at first new line.
2545 */
2546 if (exmode_active)
2547 curwin->w_cursor.lnum = from + linecnt;
2548 else
2549 curwin->w_cursor.lnum = from + 1;
2550 check_cursor_lnum();
2551 beginline(BL_WHITE | BL_FIX); // on first non-blank
2552
2553 if ((cmdmod.cmod_flags & CMOD_LOCKMARKS) == 0)
2554 {
2555 // Set '[ and '] marks to the newly read lines.
2556 curbuf->b_op_start.lnum = from + 1;
2557 curbuf->b_op_start.col = 0;
2558 curbuf->b_op_end.lnum = from + linecnt;
2559 curbuf->b_op_end.col = 0;
2560 }
2561
2562 #ifdef MSWIN
2563 /*
2564 * Work around a weird problem: When a file has two links (only
2565 * possible on NTFS) and we write through one link, then stat() it
2566 * through the other link, the timestamp information may be wrong.
2567 * It's correct again after reading the file, thus reset the timestamp
2568 * here.
2569 */
2570 if (newfile && !read_stdin && !read_buffer
2571 && mch_stat((char *)fname, &st) >= 0)
2572 {
2573 buf_store_time(curbuf, &st, fname);
2574 curbuf->b_mtime_read = curbuf->b_mtime;
2575 curbuf->b_mtime_read_ns = curbuf->b_mtime_ns;
2576 }
2577 #endif
2578 }
2579 msg_scroll = msg_save;
2580
2581 #ifdef FEAT_VIMINFO
2582 /*
2583 * Get the marks before executing autocommands, so they can be used there.
2584 */
2585 check_marks_read();
2586 #endif
2587
2588 /*
2589 * We remember if the last line of the read didn't have
2590 * an eol even when 'binary' is off, to support turning 'fixeol' off,
2591 * or writing the read again with 'binary' on. The latter is required
2592 * for ":autocmd FileReadPost *.gz set bin|'[,']!gunzip" to work.
2593 */
2594 curbuf->b_no_eol_lnum = read_no_eol_lnum;
2595
2596 // When reloading a buffer put the cursor at the first line that is
2597 // different.
2598 if (flags & READ_KEEP_UNDO)
2599 u_find_first_changed();
2600
2601 #ifdef FEAT_PERSISTENT_UNDO
2602 /*
2603 * When opening a new file locate undo info and read it.
2604 */
2605 if (read_undo_file)
2606 {
2607 char_u hash[UNDO_HASH_SIZE];
2608
2609 sha256_finish(&sha_ctx, hash);
2610 u_read_undo(NULL, hash, fname);
2611 }
2612 #endif
2613
2614 if (!read_stdin && !read_fifo && (!read_buffer || sfname != NULL))
2615 {
2616 int m = msg_scroll;
2617 int n = msg_scrolled;
2618
2619 // Save the fileformat now, otherwise the buffer will be considered
2620 // modified if the format/encoding was automatically detected.
2621 if (set_options)
2622 save_file_ff(curbuf);
2623
2624 /*
2625 * The output from the autocommands should not overwrite anything and
2626 * should not be overwritten: Set msg_scroll, restore its value if no
2627 * output was done.
2628 */
2629 msg_scroll = TRUE;
2630 if (filtering)
2631 apply_autocmds_exarg(EVENT_FILTERREADPOST, NULL, sfname,
2632 FALSE, curbuf, eap);
2633 else if (newfile || (read_buffer && sfname != NULL))
2634 {
2635 apply_autocmds_exarg(EVENT_BUFREADPOST, NULL, sfname,
2636 FALSE, curbuf, eap);
2637 if (!au_did_filetype && *curbuf->b_p_ft != NUL)
2638 /*
2639 * EVENT_FILETYPE was not triggered but the buffer already has a
2640 * filetype. Trigger EVENT_FILETYPE using the existing filetype.
2641 */
2642 apply_autocmds(EVENT_FILETYPE, curbuf->b_p_ft, curbuf->b_fname,
2643 TRUE, curbuf);
2644 }
2645 else
2646 apply_autocmds_exarg(EVENT_FILEREADPOST, sfname, sfname,
2647 FALSE, NULL, eap);
2648 if (msg_scrolled == n)
2649 msg_scroll = m;
2650 # ifdef FEAT_EVAL
2651 if (aborting()) // autocmds may abort script processing
2652 return FAIL;
2653 # endif
2654 }
2655
2656 if (recoverymode && error)
2657 return FAIL;
2658 return OK;
2659 }
2660
2661 #if defined(OPEN_CHR_FILES) || defined(PROTO)
2662 /*
2663 * Returns TRUE if the file name argument is of the form "/dev/fd/\d\+",
2664 * which is the name of files used for process substitution output by
2665 * some shells on some operating systems, e.g., bash on SunOS.
2666 * Do not accept "/dev/fd/[012]", opening these may hang Vim.
2667 */
2668 int
is_dev_fd_file(char_u * fname)2669 is_dev_fd_file(char_u *fname)
2670 {
2671 return (STRNCMP(fname, "/dev/fd/", 8) == 0
2672 && VIM_ISDIGIT(fname[8])
2673 && *skipdigits(fname + 9) == NUL
2674 && (fname[9] != NUL
2675 || (fname[8] != '0' && fname[8] != '1' && fname[8] != '2')));
2676 }
2677 #endif
2678
2679 /*
2680 * From the current line count and characters read after that, estimate the
2681 * line number where we are now.
2682 * Used for error messages that include a line number.
2683 */
2684 static linenr_T
readfile_linenr(linenr_T linecnt,char_u * p,char_u * endp)2685 readfile_linenr(
2686 linenr_T linecnt, // line count before reading more bytes
2687 char_u *p, // start of more bytes read
2688 char_u *endp) // end of more bytes read
2689 {
2690 char_u *s;
2691 linenr_T lnum;
2692
2693 lnum = curbuf->b_ml.ml_line_count - linecnt + 1;
2694 for (s = p; s < endp; ++s)
2695 if (*s == '\n')
2696 ++lnum;
2697 return lnum;
2698 }
2699
2700 /*
2701 * Fill "*eap" to force the 'fileencoding', 'fileformat' and 'binary to be
2702 * equal to the buffer "buf". Used for calling readfile().
2703 * Returns OK or FAIL.
2704 */
2705 int
prep_exarg(exarg_T * eap,buf_T * buf)2706 prep_exarg(exarg_T *eap, buf_T *buf)
2707 {
2708 eap->cmd = alloc(15 + (unsigned)STRLEN(buf->b_p_fenc));
2709 if (eap->cmd == NULL)
2710 return FAIL;
2711
2712 sprintf((char *)eap->cmd, "e ++enc=%s", buf->b_p_fenc);
2713 eap->force_enc = 8;
2714 eap->bad_char = buf->b_bad_char;
2715 eap->force_ff = *buf->b_p_ff;
2716
2717 eap->force_bin = buf->b_p_bin ? FORCE_BIN : FORCE_NOBIN;
2718 eap->read_edit = FALSE;
2719 eap->forceit = FALSE;
2720 return OK;
2721 }
2722
2723 /*
2724 * Set default or forced 'fileformat' and 'binary'.
2725 */
2726 void
set_file_options(int set_options,exarg_T * eap)2727 set_file_options(int set_options, exarg_T *eap)
2728 {
2729 // set default 'fileformat'
2730 if (set_options)
2731 {
2732 if (eap != NULL && eap->force_ff != 0)
2733 set_fileformat(get_fileformat_force(curbuf, eap), OPT_LOCAL);
2734 else if (*p_ffs != NUL)
2735 set_fileformat(default_fileformat(), OPT_LOCAL);
2736 }
2737
2738 // set or reset 'binary'
2739 if (eap != NULL && eap->force_bin != 0)
2740 {
2741 int oldval = curbuf->b_p_bin;
2742
2743 curbuf->b_p_bin = (eap->force_bin == FORCE_BIN);
2744 set_options_bin(oldval, curbuf->b_p_bin, OPT_LOCAL);
2745 }
2746 }
2747
2748 /*
2749 * Set forced 'fileencoding'.
2750 */
2751 void
set_forced_fenc(exarg_T * eap)2752 set_forced_fenc(exarg_T *eap)
2753 {
2754 if (eap->force_enc != 0)
2755 {
2756 char_u *fenc = enc_canonize(eap->cmd + eap->force_enc);
2757
2758 if (fenc != NULL)
2759 set_string_option_direct((char_u *)"fenc", -1,
2760 fenc, OPT_FREE|OPT_LOCAL, 0);
2761 vim_free(fenc);
2762 }
2763 }
2764
2765 /*
2766 * Find next fileencoding to use from 'fileencodings'.
2767 * "pp" points to fenc_next. It's advanced to the next item.
2768 * When there are no more items, an empty string is returned and *pp is set to
2769 * NULL.
2770 * When *pp is not set to NULL, the result is in allocated memory and "alloced"
2771 * is set to TRUE.
2772 */
2773 static char_u *
next_fenc(char_u ** pp,int * alloced)2774 next_fenc(char_u **pp, int *alloced)
2775 {
2776 char_u *p;
2777 char_u *r;
2778
2779 *alloced = FALSE;
2780 if (**pp == NUL)
2781 {
2782 *pp = NULL;
2783 return (char_u *)"";
2784 }
2785 p = vim_strchr(*pp, ',');
2786 if (p == NULL)
2787 {
2788 r = enc_canonize(*pp);
2789 *pp += STRLEN(*pp);
2790 }
2791 else
2792 {
2793 r = vim_strnsave(*pp, p - *pp);
2794 *pp = p + 1;
2795 if (r != NULL)
2796 {
2797 p = enc_canonize(r);
2798 vim_free(r);
2799 r = p;
2800 }
2801 }
2802 if (r != NULL)
2803 *alloced = TRUE;
2804 else
2805 {
2806 // out of memory
2807 r = (char_u *)"";
2808 *pp = NULL;
2809 }
2810 return r;
2811 }
2812
2813 #ifdef FEAT_EVAL
2814 /*
2815 * Convert a file with the 'charconvert' expression.
2816 * This closes the file which is to be read, converts it and opens the
2817 * resulting file for reading.
2818 * Returns name of the resulting converted file (the caller should delete it
2819 * after reading it).
2820 * Returns NULL if the conversion failed ("*fdp" is not set) .
2821 */
2822 static char_u *
readfile_charconvert(char_u * fname,char_u * fenc,int * fdp)2823 readfile_charconvert(
2824 char_u *fname, // name of input file
2825 char_u *fenc, // converted from
2826 int *fdp) // in/out: file descriptor of file
2827 {
2828 char_u *tmpname;
2829 char *errmsg = NULL;
2830
2831 tmpname = vim_tempname('r', FALSE);
2832 if (tmpname == NULL)
2833 errmsg = _("Can't find temp file for conversion");
2834 else
2835 {
2836 close(*fdp); // close the input file, ignore errors
2837 *fdp = -1;
2838 if (eval_charconvert(fenc, enc_utf8 ? (char_u *)"utf-8" : p_enc,
2839 fname, tmpname) == FAIL)
2840 errmsg = _("Conversion with 'charconvert' failed");
2841 if (errmsg == NULL && (*fdp = mch_open((char *)tmpname,
2842 O_RDONLY | O_EXTRA, 0)) < 0)
2843 errmsg = _("can't read output of 'charconvert'");
2844 }
2845
2846 if (errmsg != NULL)
2847 {
2848 // Don't use emsg(), it breaks mappings, the retry with
2849 // another type of conversion might still work.
2850 msg(errmsg);
2851 if (tmpname != NULL)
2852 {
2853 mch_remove(tmpname); // delete converted file
2854 VIM_CLEAR(tmpname);
2855 }
2856 }
2857
2858 // If the input file is closed, open it (caller should check for error).
2859 if (*fdp < 0)
2860 *fdp = mch_open((char *)fname, O_RDONLY | O_EXTRA, 0);
2861
2862 return tmpname;
2863 }
2864 #endif
2865
2866 #if defined(FEAT_CRYPT) || defined(PROTO)
2867 /*
2868 * Check for magic number used for encryption. Applies to the current buffer.
2869 * If found, the magic number is removed from ptr[*sizep] and *sizep and
2870 * *filesizep are updated.
2871 * Return the (new) encryption key, NULL for no encryption.
2872 */
2873 static char_u *
check_for_cryptkey(char_u * cryptkey,char_u * ptr,long * sizep,off_T * filesizep,int newfile,char_u * fname,int * did_ask)2874 check_for_cryptkey(
2875 char_u *cryptkey, // previous encryption key or NULL
2876 char_u *ptr, // pointer to read bytes
2877 long *sizep, // length of read bytes
2878 off_T *filesizep, // nr of bytes used from file
2879 int newfile, // editing a new buffer
2880 char_u *fname, // file name to display
2881 int *did_ask) // flag: whether already asked for key
2882 {
2883 int method = crypt_method_nr_from_magic((char *)ptr, *sizep);
2884 int b_p_ro = curbuf->b_p_ro;
2885
2886 if (method >= 0)
2887 {
2888 // Mark the buffer as read-only until the decryption has taken place.
2889 // Avoids accidentally overwriting the file with garbage.
2890 curbuf->b_p_ro = TRUE;
2891
2892 // Set the cryptmethod local to the buffer.
2893 crypt_set_cm_option(curbuf, method);
2894 if (cryptkey == NULL && !*did_ask)
2895 {
2896 if (*curbuf->b_p_key)
2897 cryptkey = curbuf->b_p_key;
2898 else
2899 {
2900 // When newfile is TRUE, store the typed key in the 'key'
2901 // option and don't free it. bf needs hash of the key saved.
2902 // Don't ask for the key again when first time Enter was hit.
2903 // Happens when retrying to detect encoding.
2904 smsg(_(need_key_msg), fname);
2905 msg_scroll = TRUE;
2906 crypt_check_method(method);
2907 cryptkey = crypt_get_key(newfile, FALSE);
2908 *did_ask = TRUE;
2909
2910 // check if empty key entered
2911 if (cryptkey != NULL && *cryptkey == NUL)
2912 {
2913 if (cryptkey != curbuf->b_p_key)
2914 vim_free(cryptkey);
2915 cryptkey = NULL;
2916 }
2917 }
2918 }
2919
2920 if (cryptkey != NULL)
2921 {
2922 int header_len;
2923
2924 header_len = crypt_get_header_len(method);
2925 if (*sizep <= header_len)
2926 // invalid header, buffer can't be encrypted
2927 return NULL;
2928
2929 curbuf->b_cryptstate = crypt_create_from_header(
2930 method, cryptkey, ptr);
2931 crypt_set_cm_option(curbuf, method);
2932
2933 // Remove cryptmethod specific header from the text.
2934 *filesizep += header_len;
2935 *sizep -= header_len;
2936 mch_memmove(ptr, ptr + header_len, (size_t)*sizep);
2937
2938 // Restore the read-only flag.
2939 curbuf->b_p_ro = b_p_ro;
2940 }
2941 }
2942 // When starting to edit a new file which does not have encryption, clear
2943 // the 'key' option, except when starting up (called with -x argument)
2944 else if (newfile && *curbuf->b_p_key != NUL && !starting)
2945 set_option_value((char_u *)"key", 0L, (char_u *)"", OPT_LOCAL);
2946
2947 return cryptkey;
2948 }
2949 #endif // FEAT_CRYPT
2950
2951 /*
2952 * Return TRUE if a file appears to be read-only from the file permissions.
2953 */
2954 int
check_file_readonly(char_u * fname,int perm UNUSED)2955 check_file_readonly(
2956 char_u *fname, // full path to file
2957 int perm UNUSED) // known permissions on file
2958 {
2959 #ifndef USE_MCH_ACCESS
2960 int fd = 0;
2961 #endif
2962
2963 return (
2964 #ifdef USE_MCH_ACCESS
2965 # ifdef UNIX
2966 (perm & 0222) == 0 ||
2967 # endif
2968 mch_access((char *)fname, W_OK)
2969 #else
2970 (fd = mch_open((char *)fname, O_RDWR | O_EXTRA, 0)) < 0
2971 ? TRUE : (close(fd), FALSE)
2972 #endif
2973 );
2974 }
2975
2976 #if defined(HAVE_FSYNC) || defined(PROTO)
2977 /*
2978 * Call fsync() with Mac-specific exception.
2979 * Return fsync() result: zero for success.
2980 */
2981 int
vim_fsync(int fd)2982 vim_fsync(int fd)
2983 {
2984 int r;
2985
2986 # ifdef MACOS_X
2987 r = fcntl(fd, F_FULLFSYNC);
2988 if (r != 0) // F_FULLFSYNC not working or not supported
2989 # endif
2990 r = fsync(fd);
2991 return r;
2992 }
2993 #endif
2994
2995 /*
2996 * Set the name of the current buffer. Use when the buffer doesn't have a
2997 * name and a ":r" or ":w" command with a file name is used.
2998 */
2999 int
set_rw_fname(char_u * fname,char_u * sfname)3000 set_rw_fname(char_u *fname, char_u *sfname)
3001 {
3002 buf_T *buf = curbuf;
3003
3004 // It's like the unnamed buffer is deleted....
3005 if (curbuf->b_p_bl)
3006 apply_autocmds(EVENT_BUFDELETE, NULL, NULL, FALSE, curbuf);
3007 apply_autocmds(EVENT_BUFWIPEOUT, NULL, NULL, FALSE, curbuf);
3008 #ifdef FEAT_EVAL
3009 if (aborting()) // autocmds may abort script processing
3010 return FAIL;
3011 #endif
3012 if (curbuf != buf)
3013 {
3014 // We are in another buffer now, don't do the renaming.
3015 emsg(_(e_auchangedbuf));
3016 return FAIL;
3017 }
3018
3019 if (setfname(curbuf, fname, sfname, FALSE) == OK)
3020 curbuf->b_flags |= BF_NOTEDITED;
3021
3022 // ....and a new named one is created
3023 apply_autocmds(EVENT_BUFNEW, NULL, NULL, FALSE, curbuf);
3024 if (curbuf->b_p_bl)
3025 apply_autocmds(EVENT_BUFADD, NULL, NULL, FALSE, curbuf);
3026 #ifdef FEAT_EVAL
3027 if (aborting()) // autocmds may abort script processing
3028 return FAIL;
3029 #endif
3030
3031 // Do filetype detection now if 'filetype' is empty.
3032 if (*curbuf->b_p_ft == NUL)
3033 {
3034 if (au_has_group((char_u *)"filetypedetect"))
3035 (void)do_doautocmd((char_u *)"filetypedetect BufRead", FALSE, NULL);
3036 do_modelines(0);
3037 }
3038
3039 return OK;
3040 }
3041
3042 /*
3043 * Put file name into IObuff with quotes.
3044 */
3045 void
msg_add_fname(buf_T * buf,char_u * fname)3046 msg_add_fname(buf_T *buf, char_u *fname)
3047 {
3048 if (fname == NULL)
3049 fname = (char_u *)"-stdin-";
3050 home_replace(buf, fname, IObuff + 1, IOSIZE - 4, TRUE);
3051 IObuff[0] = '"';
3052 STRCAT(IObuff, "\" ");
3053 }
3054
3055 /*
3056 * Append message for text mode to IObuff.
3057 * Return TRUE if something appended.
3058 */
3059 int
msg_add_fileformat(int eol_type)3060 msg_add_fileformat(int eol_type)
3061 {
3062 #ifndef USE_CRNL
3063 if (eol_type == EOL_DOS)
3064 {
3065 STRCAT(IObuff, shortmess(SHM_TEXT) ? _("[dos]") : _("[dos format]"));
3066 return TRUE;
3067 }
3068 #endif
3069 if (eol_type == EOL_MAC)
3070 {
3071 STRCAT(IObuff, shortmess(SHM_TEXT) ? _("[mac]") : _("[mac format]"));
3072 return TRUE;
3073 }
3074 #ifdef USE_CRNL
3075 if (eol_type == EOL_UNIX)
3076 {
3077 STRCAT(IObuff, shortmess(SHM_TEXT) ? _("[unix]") : _("[unix format]"));
3078 return TRUE;
3079 }
3080 #endif
3081 return FALSE;
3082 }
3083
3084 /*
3085 * Append line and character count to IObuff.
3086 */
3087 void
msg_add_lines(int insert_space,long lnum,off_T nchars)3088 msg_add_lines(
3089 int insert_space,
3090 long lnum,
3091 off_T nchars)
3092 {
3093 char_u *p;
3094
3095 p = IObuff + STRLEN(IObuff);
3096
3097 if (insert_space)
3098 *p++ = ' ';
3099 if (shortmess(SHM_LINES))
3100 vim_snprintf((char *)p, IOSIZE - (p - IObuff),
3101 "%ldL, %lldB", lnum, (varnumber_T)nchars);
3102 else
3103 {
3104 sprintf((char *)p, NGETTEXT("%ld line, ", "%ld lines, ", lnum), lnum);
3105 p += STRLEN(p);
3106 vim_snprintf((char *)p, IOSIZE - (p - IObuff),
3107 NGETTEXT("%lld byte", "%lld bytes", nchars),
3108 (varnumber_T)nchars);
3109 }
3110 }
3111
3112 /*
3113 * Append message for missing line separator to IObuff.
3114 */
3115 void
msg_add_eol(void)3116 msg_add_eol(void)
3117 {
3118 STRCAT(IObuff, shortmess(SHM_LAST) ? _("[noeol]") : _("[Incomplete last line]"));
3119 }
3120
3121 int
time_differs(stat_T * st,long mtime,long mtime_ns UNUSED)3122 time_differs(stat_T *st, long mtime, long mtime_ns UNUSED)
3123 {
3124 return
3125 #ifdef ST_MTIM_NSEC
3126 (long)st->ST_MTIM_NSEC != mtime_ns ||
3127 #endif
3128 #if defined(__linux__) || defined(MSWIN)
3129 // On a FAT filesystem, esp. under Linux, there are only 5 bits to store
3130 // the seconds. Since the roundoff is done when flushing the inode, the
3131 // time may change unexpectedly by one second!!!
3132 (long)st->st_mtime - mtime > 1 || mtime - (long)st->st_mtime > 1
3133 #else
3134 (long)st->st_mtime != mtime
3135 #endif
3136 ;
3137 }
3138
3139 /*
3140 * Return TRUE if file encoding "fenc" requires conversion from or to
3141 * 'encoding'.
3142 */
3143 int
need_conversion(char_u * fenc)3144 need_conversion(char_u *fenc)
3145 {
3146 int same_encoding;
3147 int enc_flags;
3148 int fenc_flags;
3149
3150 if (*fenc == NUL || STRCMP(p_enc, fenc) == 0)
3151 {
3152 same_encoding = TRUE;
3153 fenc_flags = 0;
3154 }
3155 else
3156 {
3157 // Ignore difference between "ansi" and "latin1", "ucs-4" and
3158 // "ucs-4be", etc.
3159 enc_flags = get_fio_flags(p_enc);
3160 fenc_flags = get_fio_flags(fenc);
3161 same_encoding = (enc_flags != 0 && fenc_flags == enc_flags);
3162 }
3163 if (same_encoding)
3164 {
3165 // Specified encoding matches with 'encoding'. This requires
3166 // conversion when 'encoding' is Unicode but not UTF-8.
3167 return enc_unicode != 0;
3168 }
3169
3170 // Encodings differ. However, conversion is not needed when 'enc' is any
3171 // Unicode encoding and the file is UTF-8.
3172 return !(enc_utf8 && fenc_flags == FIO_UTF8);
3173 }
3174
3175 /*
3176 * Check "ptr" for a unicode encoding and return the FIO_ flags needed for the
3177 * internal conversion.
3178 * if "ptr" is an empty string, use 'encoding'.
3179 */
3180 int
get_fio_flags(char_u * ptr)3181 get_fio_flags(char_u *ptr)
3182 {
3183 int prop;
3184
3185 if (*ptr == NUL)
3186 ptr = p_enc;
3187
3188 prop = enc_canon_props(ptr);
3189 if (prop & ENC_UNICODE)
3190 {
3191 if (prop & ENC_2BYTE)
3192 {
3193 if (prop & ENC_ENDIAN_L)
3194 return FIO_UCS2 | FIO_ENDIAN_L;
3195 return FIO_UCS2;
3196 }
3197 if (prop & ENC_4BYTE)
3198 {
3199 if (prop & ENC_ENDIAN_L)
3200 return FIO_UCS4 | FIO_ENDIAN_L;
3201 return FIO_UCS4;
3202 }
3203 if (prop & ENC_2WORD)
3204 {
3205 if (prop & ENC_ENDIAN_L)
3206 return FIO_UTF16 | FIO_ENDIAN_L;
3207 return FIO_UTF16;
3208 }
3209 return FIO_UTF8;
3210 }
3211 if (prop & ENC_LATIN1)
3212 return FIO_LATIN1;
3213 // must be ENC_DBCS, requires iconv()
3214 return 0;
3215 }
3216
3217 #if defined(MSWIN) || defined(PROTO)
3218 /*
3219 * Check "ptr" for a MS-Windows codepage name and return the FIO_ flags needed
3220 * for the conversion MS-Windows can do for us. Also accept "utf-8".
3221 * Used for conversion between 'encoding' and 'fileencoding'.
3222 */
3223 int
get_win_fio_flags(char_u * ptr)3224 get_win_fio_flags(char_u *ptr)
3225 {
3226 int cp;
3227
3228 // Cannot do this when 'encoding' is not utf-8 and not a codepage.
3229 if (!enc_utf8 && enc_codepage <= 0)
3230 return 0;
3231
3232 cp = encname2codepage(ptr);
3233 if (cp == 0)
3234 {
3235 # ifdef CP_UTF8 // VC 4.1 doesn't define CP_UTF8
3236 if (STRCMP(ptr, "utf-8") == 0)
3237 cp = CP_UTF8;
3238 else
3239 # endif
3240 return 0;
3241 }
3242 return FIO_PUT_CP(cp) | FIO_CODEPAGE;
3243 }
3244 #endif
3245
3246 #if defined(MACOS_CONVERT) || defined(PROTO)
3247 /*
3248 * Check "ptr" for a Carbon supported encoding and return the FIO_ flags
3249 * needed for the internal conversion to/from utf-8 or latin1.
3250 */
3251 int
get_mac_fio_flags(char_u * ptr)3252 get_mac_fio_flags(char_u *ptr)
3253 {
3254 if ((enc_utf8 || STRCMP(p_enc, "latin1") == 0)
3255 && (enc_canon_props(ptr) & ENC_MACROMAN))
3256 return FIO_MACROMAN;
3257 return 0;
3258 }
3259 #endif
3260
3261 /*
3262 * Check for a Unicode BOM (Byte Order Mark) at the start of p[size].
3263 * "size" must be at least 2.
3264 * Return the name of the encoding and set "*lenp" to the length.
3265 * Returns NULL when no BOM found.
3266 */
3267 static char_u *
check_for_bom(char_u * p,long size,int * lenp,int flags)3268 check_for_bom(
3269 char_u *p,
3270 long size,
3271 int *lenp,
3272 int flags)
3273 {
3274 char *name = NULL;
3275 int len = 2;
3276
3277 if (p[0] == 0xef && p[1] == 0xbb && size >= 3 && p[2] == 0xbf
3278 && (flags == FIO_ALL || flags == FIO_UTF8 || flags == 0))
3279 {
3280 name = "utf-8"; // EF BB BF
3281 len = 3;
3282 }
3283 else if (p[0] == 0xff && p[1] == 0xfe)
3284 {
3285 if (size >= 4 && p[2] == 0 && p[3] == 0
3286 && (flags == FIO_ALL || flags == (FIO_UCS4 | FIO_ENDIAN_L)))
3287 {
3288 name = "ucs-4le"; // FF FE 00 00
3289 len = 4;
3290 }
3291 else if (flags == (FIO_UCS2 | FIO_ENDIAN_L))
3292 name = "ucs-2le"; // FF FE
3293 else if (flags == FIO_ALL || flags == (FIO_UTF16 | FIO_ENDIAN_L))
3294 // utf-16le is preferred, it also works for ucs-2le text
3295 name = "utf-16le"; // FF FE
3296 }
3297 else if (p[0] == 0xfe && p[1] == 0xff
3298 && (flags == FIO_ALL || flags == FIO_UCS2 || flags == FIO_UTF16))
3299 {
3300 // Default to utf-16, it works also for ucs-2 text.
3301 if (flags == FIO_UCS2)
3302 name = "ucs-2"; // FE FF
3303 else
3304 name = "utf-16"; // FE FF
3305 }
3306 else if (size >= 4 && p[0] == 0 && p[1] == 0 && p[2] == 0xfe
3307 && p[3] == 0xff && (flags == FIO_ALL || flags == FIO_UCS4))
3308 {
3309 name = "ucs-4"; // 00 00 FE FF
3310 len = 4;
3311 }
3312
3313 *lenp = len;
3314 return (char_u *)name;
3315 }
3316
3317 /*
3318 * Try to find a shortname by comparing the fullname with the current
3319 * directory.
3320 * Returns "full_path" or pointer into "full_path" if shortened.
3321 */
3322 char_u *
shorten_fname1(char_u * full_path)3323 shorten_fname1(char_u *full_path)
3324 {
3325 char_u *dirname;
3326 char_u *p = full_path;
3327
3328 dirname = alloc(MAXPATHL);
3329 if (dirname == NULL)
3330 return full_path;
3331 if (mch_dirname(dirname, MAXPATHL) == OK)
3332 {
3333 p = shorten_fname(full_path, dirname);
3334 if (p == NULL || *p == NUL)
3335 p = full_path;
3336 }
3337 vim_free(dirname);
3338 return p;
3339 }
3340
3341 /*
3342 * Try to find a shortname by comparing the fullname with the current
3343 * directory.
3344 * Returns NULL if not shorter name possible, pointer into "full_path"
3345 * otherwise.
3346 */
3347 char_u *
shorten_fname(char_u * full_path,char_u * dir_name)3348 shorten_fname(char_u *full_path, char_u *dir_name)
3349 {
3350 int len;
3351 char_u *p;
3352
3353 if (full_path == NULL)
3354 return NULL;
3355 len = (int)STRLEN(dir_name);
3356 if (fnamencmp(dir_name, full_path, len) == 0)
3357 {
3358 p = full_path + len;
3359 #if defined(MSWIN)
3360 /*
3361 * MS-Windows: when a file is in the root directory, dir_name will end
3362 * in a slash, since C: by itself does not define a specific dir. In
3363 * this case p may already be correct. <negri>
3364 */
3365 if (!((len > 2) && (*(p - 2) == ':')))
3366 #endif
3367 {
3368 if (vim_ispathsep(*p))
3369 ++p;
3370 #ifndef VMS // the path separator is always part of the path
3371 else
3372 p = NULL;
3373 #endif
3374 }
3375 }
3376 #if defined(MSWIN)
3377 /*
3378 * When using a file in the current drive, remove the drive name:
3379 * "A:\dir\file" -> "\dir\file". This helps when moving a session file on
3380 * a floppy from "A:\dir" to "B:\dir".
3381 */
3382 else if (len > 3
3383 && TOUPPER_LOC(full_path[0]) == TOUPPER_LOC(dir_name[0])
3384 && full_path[1] == ':'
3385 && vim_ispathsep(full_path[2]))
3386 p = full_path + 2;
3387 #endif
3388 else
3389 p = NULL;
3390 return p;
3391 }
3392
3393 /*
3394 * Shorten filename of a buffer.
3395 * When "force" is TRUE: Use full path from now on for files currently being
3396 * edited, both for file name and swap file name. Try to shorten the file
3397 * names a bit, if safe to do so.
3398 * When "force" is FALSE: Only try to shorten absolute file names.
3399 * For buffers that have buftype "nofile" or "scratch": never change the file
3400 * name.
3401 */
3402 void
shorten_buf_fname(buf_T * buf,char_u * dirname,int force)3403 shorten_buf_fname(buf_T *buf, char_u *dirname, int force)
3404 {
3405 char_u *p;
3406
3407 if (buf->b_fname != NULL
3408 #ifdef FEAT_QUICKFIX
3409 && !bt_nofilename(buf)
3410 #endif
3411 && !path_with_url(buf->b_fname)
3412 && (force
3413 || buf->b_sfname == NULL
3414 || mch_isFullName(buf->b_sfname)))
3415 {
3416 if (buf->b_sfname != buf->b_ffname)
3417 VIM_CLEAR(buf->b_sfname);
3418 p = shorten_fname(buf->b_ffname, dirname);
3419 if (p != NULL)
3420 {
3421 buf->b_sfname = vim_strsave(p);
3422 buf->b_fname = buf->b_sfname;
3423 }
3424 if (p == NULL || buf->b_fname == NULL)
3425 buf->b_fname = buf->b_ffname;
3426 }
3427 }
3428
3429 /*
3430 * Shorten filenames for all buffers.
3431 */
3432 void
shorten_fnames(int force)3433 shorten_fnames(int force)
3434 {
3435 char_u dirname[MAXPATHL];
3436 buf_T *buf;
3437
3438 mch_dirname(dirname, MAXPATHL);
3439 FOR_ALL_BUFFERS(buf)
3440 {
3441 shorten_buf_fname(buf, dirname, force);
3442
3443 // Always make the swap file name a full path, a "nofile" buffer may
3444 // also have a swap file.
3445 mf_fullname(buf->b_ml.ml_mfp);
3446 }
3447 status_redraw_all();
3448 redraw_tabline = TRUE;
3449 #if defined(FEAT_PROP_POPUP) && defined(FEAT_QUICKFIX)
3450 popup_update_preview_title();
3451 #endif
3452 }
3453
3454 #if (defined(FEAT_DND) && defined(FEAT_GUI_GTK)) \
3455 || defined(FEAT_GUI_MSWIN) \
3456 || defined(FEAT_GUI_HAIKU) \
3457 || defined(PROTO)
3458 /*
3459 * Shorten all filenames in "fnames[count]" by current directory.
3460 */
3461 void
shorten_filenames(char_u ** fnames,int count)3462 shorten_filenames(char_u **fnames, int count)
3463 {
3464 int i;
3465 char_u dirname[MAXPATHL];
3466 char_u *p;
3467
3468 if (fnames == NULL || count < 1)
3469 return;
3470 mch_dirname(dirname, sizeof(dirname));
3471 for (i = 0; i < count; ++i)
3472 {
3473 if ((p = shorten_fname(fnames[i], dirname)) != NULL)
3474 {
3475 // shorten_fname() returns pointer in given "fnames[i]". If free
3476 // "fnames[i]" first, "p" becomes invalid. So we need to copy
3477 // "p" first then free fnames[i].
3478 p = vim_strsave(p);
3479 vim_free(fnames[i]);
3480 fnames[i] = p;
3481 }
3482 }
3483 }
3484 #endif
3485
3486 /*
3487 * Add extension to file name - change path/fo.o.h to path/fo.o.h.ext or
3488 * fo_o_h.ext for MSDOS or when shortname option set.
3489 *
3490 * Assumed that fname is a valid name found in the filesystem we assure that
3491 * the return value is a different name and ends in 'ext'.
3492 * "ext" MUST be at most 4 characters long if it starts with a dot, 3
3493 * characters otherwise.
3494 * Space for the returned name is allocated, must be freed later.
3495 * Returns NULL when out of memory.
3496 */
3497 char_u *
modname(char_u * fname,char_u * ext,int prepend_dot)3498 modname(
3499 char_u *fname,
3500 char_u *ext,
3501 int prepend_dot) // may prepend a '.' to file name
3502 {
3503 return buf_modname((curbuf->b_p_sn || curbuf->b_shortname),
3504 fname, ext, prepend_dot);
3505 }
3506
3507 char_u *
buf_modname(int shortname,char_u * fname,char_u * ext,int prepend_dot)3508 buf_modname(
3509 int shortname, // use 8.3 file name
3510 char_u *fname,
3511 char_u *ext,
3512 int prepend_dot) // may prepend a '.' to file name
3513 {
3514 char_u *retval;
3515 char_u *s;
3516 char_u *e;
3517 char_u *ptr;
3518 int fnamelen, extlen;
3519
3520 extlen = (int)STRLEN(ext);
3521
3522 /*
3523 * If there is no file name we must get the name of the current directory
3524 * (we need the full path in case :cd is used).
3525 */
3526 if (fname == NULL || *fname == NUL)
3527 {
3528 retval = alloc(MAXPATHL + extlen + 3);
3529 if (retval == NULL)
3530 return NULL;
3531 if (mch_dirname(retval, MAXPATHL) == FAIL ||
3532 (fnamelen = (int)STRLEN(retval)) == 0)
3533 {
3534 vim_free(retval);
3535 return NULL;
3536 }
3537 if (!after_pathsep(retval, retval + fnamelen))
3538 {
3539 retval[fnamelen++] = PATHSEP;
3540 retval[fnamelen] = NUL;
3541 }
3542 prepend_dot = FALSE; // nothing to prepend a dot to
3543 }
3544 else
3545 {
3546 fnamelen = (int)STRLEN(fname);
3547 retval = alloc(fnamelen + extlen + 3);
3548 if (retval == NULL)
3549 return NULL;
3550 STRCPY(retval, fname);
3551 #ifdef VMS
3552 vms_remove_version(retval); // we do not need versions here
3553 #endif
3554 }
3555
3556 /*
3557 * search backwards until we hit a '/', '\' or ':' replacing all '.'
3558 * by '_' for MSDOS or when shortname option set and ext starts with a dot.
3559 * Then truncate what is after the '/', '\' or ':' to 8 characters for
3560 * MSDOS and 26 characters for AMIGA, a lot more for UNIX.
3561 */
3562 for (ptr = retval + fnamelen; ptr > retval; MB_PTR_BACK(retval, ptr))
3563 {
3564 if (*ext == '.' && shortname)
3565 if (*ptr == '.') // replace '.' by '_'
3566 *ptr = '_';
3567 if (vim_ispathsep(*ptr))
3568 {
3569 ++ptr;
3570 break;
3571 }
3572 }
3573
3574 // the file name has at most BASENAMELEN characters.
3575 if (STRLEN(ptr) > (unsigned)BASENAMELEN)
3576 ptr[BASENAMELEN] = '\0';
3577
3578 s = ptr + STRLEN(ptr);
3579
3580 /*
3581 * For 8.3 file names we may have to reduce the length.
3582 */
3583 if (shortname)
3584 {
3585 /*
3586 * If there is no file name, or the file name ends in '/', and the
3587 * extension starts with '.', put a '_' before the dot, because just
3588 * ".ext" is invalid.
3589 */
3590 if (fname == NULL || *fname == NUL
3591 || vim_ispathsep(fname[STRLEN(fname) - 1]))
3592 {
3593 if (*ext == '.')
3594 *s++ = '_';
3595 }
3596 /*
3597 * If the extension starts with '.', truncate the base name at 8
3598 * characters
3599 */
3600 else if (*ext == '.')
3601 {
3602 if ((size_t)(s - ptr) > (size_t)8)
3603 {
3604 s = ptr + 8;
3605 *s = '\0';
3606 }
3607 }
3608 /*
3609 * If the extension doesn't start with '.', and the file name
3610 * doesn't have an extension yet, append a '.'
3611 */
3612 else if ((e = vim_strchr(ptr, '.')) == NULL)
3613 *s++ = '.';
3614 /*
3615 * If the extension doesn't start with '.', and there already is an
3616 * extension, it may need to be truncated
3617 */
3618 else if ((int)STRLEN(e) + extlen > 4)
3619 s = e + 4 - extlen;
3620 }
3621 #ifdef MSWIN
3622 /*
3623 * If there is no file name, and the extension starts with '.', put a
3624 * '_' before the dot, because just ".ext" may be invalid if it's on a
3625 * FAT partition, and on HPFS it doesn't matter.
3626 */
3627 else if ((fname == NULL || *fname == NUL) && *ext == '.')
3628 *s++ = '_';
3629 #endif
3630
3631 /*
3632 * Append the extension.
3633 * ext can start with '.' and cannot exceed 3 more characters.
3634 */
3635 STRCPY(s, ext);
3636
3637 /*
3638 * Prepend the dot.
3639 */
3640 if (prepend_dot && !shortname && *(e = gettail(retval)) != '.')
3641 {
3642 STRMOVE(e + 1, e);
3643 *e = '.';
3644 }
3645
3646 /*
3647 * Check that, after appending the extension, the file name is really
3648 * different.
3649 */
3650 if (fname != NULL && STRCMP(fname, retval) == 0)
3651 {
3652 // we search for a character that can be replaced by '_'
3653 while (--s >= ptr)
3654 {
3655 if (*s != '_')
3656 {
3657 *s = '_';
3658 break;
3659 }
3660 }
3661 if (s < ptr) // fname was "________.<ext>", how tricky!
3662 *ptr = 'v';
3663 }
3664 return retval;
3665 }
3666
3667 /*
3668 * Like fgets(), but if the file line is too long, it is truncated and the
3669 * rest of the line is thrown away. Returns TRUE for end-of-file.
3670 * If the line is truncated then buf[size - 2] will not be NUL.
3671 */
3672 int
vim_fgets(char_u * buf,int size,FILE * fp)3673 vim_fgets(char_u *buf, int size, FILE *fp)
3674 {
3675 char *eof;
3676 #define FGETS_SIZE 200
3677 char tbuf[FGETS_SIZE];
3678
3679 buf[size - 2] = NUL;
3680 eof = fgets((char *)buf, size, fp);
3681 if (buf[size - 2] != NUL && buf[size - 2] != '\n')
3682 {
3683 buf[size - 1] = NUL; // Truncate the line
3684
3685 // Now throw away the rest of the line:
3686 do
3687 {
3688 tbuf[FGETS_SIZE - 2] = NUL;
3689 vim_ignoredp = fgets((char *)tbuf, FGETS_SIZE, fp);
3690 } while (tbuf[FGETS_SIZE - 2] != NUL && tbuf[FGETS_SIZE - 2] != '\n');
3691 }
3692 return (eof == NULL);
3693 }
3694
3695 /*
3696 * rename() only works if both files are on the same file system, this
3697 * function will (attempts to?) copy the file across if rename fails -- webb
3698 * Return -1 for failure, 0 for success.
3699 */
3700 int
vim_rename(char_u * from,char_u * to)3701 vim_rename(char_u *from, char_u *to)
3702 {
3703 int fd_in;
3704 int fd_out;
3705 int n;
3706 char *errmsg = NULL;
3707 char *buffer;
3708 #ifdef AMIGA
3709 BPTR flock;
3710 #endif
3711 stat_T st;
3712 long perm;
3713 #ifdef HAVE_ACL
3714 vim_acl_T acl; // ACL from original file
3715 #endif
3716 int use_tmp_file = FALSE;
3717
3718 /*
3719 * When the names are identical, there is nothing to do. When they refer
3720 * to the same file (ignoring case and slash/backslash differences) but
3721 * the file name differs we need to go through a temp file.
3722 */
3723 if (fnamecmp(from, to) == 0)
3724 {
3725 if (p_fic && STRCMP(gettail(from), gettail(to)) != 0)
3726 use_tmp_file = TRUE;
3727 else
3728 return 0;
3729 }
3730
3731 /*
3732 * Fail if the "from" file doesn't exist. Avoids that "to" is deleted.
3733 */
3734 if (mch_stat((char *)from, &st) < 0)
3735 return -1;
3736
3737 #ifdef UNIX
3738 {
3739 stat_T st_to;
3740
3741 // It's possible for the source and destination to be the same file.
3742 // This happens when "from" and "to" differ in case and are on a FAT32
3743 // filesystem. In that case go through a temp file name.
3744 if (mch_stat((char *)to, &st_to) >= 0
3745 && st.st_dev == st_to.st_dev
3746 && st.st_ino == st_to.st_ino)
3747 use_tmp_file = TRUE;
3748 }
3749 #endif
3750 #ifdef MSWIN
3751 {
3752 BY_HANDLE_FILE_INFORMATION info1, info2;
3753
3754 // It's possible for the source and destination to be the same file.
3755 // In that case go through a temp file name. This makes rename("foo",
3756 // "./foo") a no-op (in a complicated way).
3757 if (win32_fileinfo(from, &info1) == FILEINFO_OK
3758 && win32_fileinfo(to, &info2) == FILEINFO_OK
3759 && info1.dwVolumeSerialNumber == info2.dwVolumeSerialNumber
3760 && info1.nFileIndexHigh == info2.nFileIndexHigh
3761 && info1.nFileIndexLow == info2.nFileIndexLow)
3762 use_tmp_file = TRUE;
3763 }
3764 #endif
3765
3766 if (use_tmp_file)
3767 {
3768 char tempname[MAXPATHL + 1];
3769
3770 /*
3771 * Find a name that doesn't exist and is in the same directory.
3772 * Rename "from" to "tempname" and then rename "tempname" to "to".
3773 */
3774 if (STRLEN(from) >= MAXPATHL - 5)
3775 return -1;
3776 STRCPY(tempname, from);
3777 for (n = 123; n < 99999; ++n)
3778 {
3779 sprintf((char *)gettail((char_u *)tempname), "%d", n);
3780 if (mch_stat(tempname, &st) < 0)
3781 {
3782 if (mch_rename((char *)from, tempname) == 0)
3783 {
3784 if (mch_rename(tempname, (char *)to) == 0)
3785 return 0;
3786 // Strange, the second step failed. Try moving the
3787 // file back and return failure.
3788 (void)mch_rename(tempname, (char *)from);
3789 return -1;
3790 }
3791 // If it fails for one temp name it will most likely fail
3792 // for any temp name, give up.
3793 return -1;
3794 }
3795 }
3796 return -1;
3797 }
3798
3799 /*
3800 * Delete the "to" file, this is required on some systems to make the
3801 * mch_rename() work, on other systems it makes sure that we don't have
3802 * two files when the mch_rename() fails.
3803 */
3804
3805 #ifdef AMIGA
3806 /*
3807 * With MSDOS-compatible filesystems (crossdos, messydos) it is possible
3808 * that the name of the "to" file is the same as the "from" file, even
3809 * though the names are different. To avoid the chance of accidentally
3810 * deleting the "from" file (horror!) we lock it during the remove.
3811 *
3812 * When used for making a backup before writing the file: This should not
3813 * happen with ":w", because startscript() should detect this problem and
3814 * set buf->b_shortname, causing modname() to return a correct ".bak" file
3815 * name. This problem does exist with ":w filename", but then the
3816 * original file will be somewhere else so the backup isn't really
3817 * important. If autoscripting is off the rename may fail.
3818 */
3819 flock = Lock((UBYTE *)from, (long)ACCESS_READ);
3820 #endif
3821 mch_remove(to);
3822 #ifdef AMIGA
3823 if (flock)
3824 UnLock(flock);
3825 #endif
3826
3827 /*
3828 * First try a normal rename, return if it works.
3829 */
3830 if (mch_rename((char *)from, (char *)to) == 0)
3831 return 0;
3832
3833 /*
3834 * Rename() failed, try copying the file.
3835 */
3836 perm = mch_getperm(from);
3837 #ifdef HAVE_ACL
3838 // For systems that support ACL: get the ACL from the original file.
3839 acl = mch_get_acl(from);
3840 #endif
3841 fd_in = mch_open((char *)from, O_RDONLY|O_EXTRA, 0);
3842 if (fd_in == -1)
3843 {
3844 #ifdef HAVE_ACL
3845 mch_free_acl(acl);
3846 #endif
3847 return -1;
3848 }
3849
3850 // Create the new file with same permissions as the original.
3851 fd_out = mch_open((char *)to,
3852 O_CREAT|O_EXCL|O_WRONLY|O_EXTRA|O_NOFOLLOW, (int)perm);
3853 if (fd_out == -1)
3854 {
3855 close(fd_in);
3856 #ifdef HAVE_ACL
3857 mch_free_acl(acl);
3858 #endif
3859 return -1;
3860 }
3861
3862 buffer = alloc(WRITEBUFSIZE);
3863 if (buffer == NULL)
3864 {
3865 close(fd_out);
3866 close(fd_in);
3867 #ifdef HAVE_ACL
3868 mch_free_acl(acl);
3869 #endif
3870 return -1;
3871 }
3872
3873 while ((n = read_eintr(fd_in, buffer, WRITEBUFSIZE)) > 0)
3874 if (write_eintr(fd_out, buffer, n) != n)
3875 {
3876 errmsg = _("E208: Error writing to \"%s\"");
3877 break;
3878 }
3879
3880 vim_free(buffer);
3881 close(fd_in);
3882 if (close(fd_out) < 0)
3883 errmsg = _("E209: Error closing \"%s\"");
3884 if (n < 0)
3885 {
3886 errmsg = _("E210: Error reading \"%s\"");
3887 to = from;
3888 }
3889 #ifndef UNIX // for Unix mch_open() already set the permission
3890 mch_setperm(to, perm);
3891 #endif
3892 #ifdef HAVE_ACL
3893 mch_set_acl(to, acl);
3894 mch_free_acl(acl);
3895 #endif
3896 #if defined(HAVE_SELINUX) || defined(HAVE_SMACK)
3897 mch_copy_sec(from, to);
3898 #endif
3899 if (errmsg != NULL)
3900 {
3901 semsg(errmsg, to);
3902 return -1;
3903 }
3904 mch_remove(from);
3905 return 0;
3906 }
3907
3908 static int already_warned = FALSE;
3909
3910 /*
3911 * Check if any not hidden buffer has been changed.
3912 * Postpone the check if there are characters in the stuff buffer, a global
3913 * command is being executed, a mapping is being executed or an autocommand is
3914 * busy.
3915 * Returns TRUE if some message was written (screen should be redrawn and
3916 * cursor positioned).
3917 */
3918 int
check_timestamps(int focus)3919 check_timestamps(
3920 int focus) // called for GUI focus event
3921 {
3922 buf_T *buf;
3923 int didit = 0;
3924 int n;
3925
3926 // Don't check timestamps while system() or another low-level function may
3927 // cause us to lose and gain focus.
3928 if (no_check_timestamps > 0)
3929 return FALSE;
3930
3931 // Avoid doing a check twice. The OK/Reload dialog can cause a focus
3932 // event and we would keep on checking if the file is steadily growing.
3933 // Do check again after typing something.
3934 if (focus && did_check_timestamps)
3935 {
3936 need_check_timestamps = TRUE;
3937 return FALSE;
3938 }
3939
3940 if (!stuff_empty() || global_busy || !typebuf_typed()
3941 || autocmd_busy || curbuf_lock > 0 || allbuf_lock > 0)
3942 need_check_timestamps = TRUE; // check later
3943 else
3944 {
3945 ++no_wait_return;
3946 did_check_timestamps = TRUE;
3947 already_warned = FALSE;
3948 FOR_ALL_BUFFERS(buf)
3949 {
3950 // Only check buffers in a window.
3951 if (buf->b_nwindows > 0)
3952 {
3953 bufref_T bufref;
3954
3955 set_bufref(&bufref, buf);
3956 n = buf_check_timestamp(buf, focus);
3957 if (didit < n)
3958 didit = n;
3959 if (n > 0 && !bufref_valid(&bufref))
3960 {
3961 // Autocommands have removed the buffer, start at the
3962 // first one again.
3963 buf = firstbuf;
3964 continue;
3965 }
3966 }
3967 }
3968 --no_wait_return;
3969 need_check_timestamps = FALSE;
3970 if (need_wait_return && didit == 2)
3971 {
3972 // make sure msg isn't overwritten
3973 msg_puts("\n");
3974 out_flush();
3975 }
3976 }
3977 return didit;
3978 }
3979
3980 /*
3981 * Move all the lines from buffer "frombuf" to buffer "tobuf".
3982 * Return OK or FAIL. When FAIL "tobuf" is incomplete and/or "frombuf" is not
3983 * empty.
3984 */
3985 static int
move_lines(buf_T * frombuf,buf_T * tobuf)3986 move_lines(buf_T *frombuf, buf_T *tobuf)
3987 {
3988 buf_T *tbuf = curbuf;
3989 int retval = OK;
3990 linenr_T lnum;
3991 char_u *p;
3992
3993 // Copy the lines in "frombuf" to "tobuf".
3994 curbuf = tobuf;
3995 for (lnum = 1; lnum <= frombuf->b_ml.ml_line_count; ++lnum)
3996 {
3997 p = vim_strsave(ml_get_buf(frombuf, lnum, FALSE));
3998 if (p == NULL || ml_append(lnum - 1, p, 0, FALSE) == FAIL)
3999 {
4000 vim_free(p);
4001 retval = FAIL;
4002 break;
4003 }
4004 vim_free(p);
4005 }
4006
4007 // Delete all the lines in "frombuf".
4008 if (retval != FAIL)
4009 {
4010 curbuf = frombuf;
4011 for (lnum = curbuf->b_ml.ml_line_count; lnum > 0; --lnum)
4012 if (ml_delete(lnum) == FAIL)
4013 {
4014 // Oops! We could try putting back the saved lines, but that
4015 // might fail again...
4016 retval = FAIL;
4017 break;
4018 }
4019 }
4020
4021 curbuf = tbuf;
4022 return retval;
4023 }
4024
4025 /*
4026 * Check if buffer "buf" has been changed.
4027 * Also check if the file for a new buffer unexpectedly appeared.
4028 * return 1 if a changed buffer was found.
4029 * return 2 if a message has been displayed.
4030 * return 0 otherwise.
4031 */
4032 int
buf_check_timestamp(buf_T * buf,int focus UNUSED)4033 buf_check_timestamp(
4034 buf_T *buf,
4035 int focus UNUSED) // called for GUI focus event
4036 {
4037 stat_T st;
4038 int stat_res;
4039 int retval = 0;
4040 char_u *path;
4041 char *tbuf;
4042 char *mesg = NULL;
4043 char *mesg2 = "";
4044 int helpmesg = FALSE;
4045 int reload = FALSE;
4046 char *reason;
4047 #if defined(FEAT_CON_DIALOG) || defined(FEAT_GUI_DIALOG)
4048 int can_reload = FALSE;
4049 #endif
4050 off_T orig_size = buf->b_orig_size;
4051 int orig_mode = buf->b_orig_mode;
4052 #ifdef FEAT_GUI
4053 int save_mouse_correct = need_mouse_correct;
4054 #endif
4055 static int busy = FALSE;
4056 int n;
4057 #ifdef FEAT_EVAL
4058 char_u *s;
4059 #endif
4060 bufref_T bufref;
4061
4062 set_bufref(&bufref, buf);
4063
4064 // If there is no file name, the buffer is not loaded, 'buftype' is
4065 // set, we are in the middle of a save or being called recursively: ignore
4066 // this buffer.
4067 if (buf->b_ffname == NULL
4068 || buf->b_ml.ml_mfp == NULL
4069 || !bt_normal(buf)
4070 || buf->b_saving
4071 || busy
4072 #ifdef FEAT_NETBEANS_INTG
4073 || isNetbeansBuffer(buf)
4074 #endif
4075 #ifdef FEAT_TERMINAL
4076 || buf->b_term != NULL
4077 #endif
4078 )
4079 return 0;
4080
4081 if ( !(buf->b_flags & BF_NOTEDITED)
4082 && buf->b_mtime != 0
4083 && ((stat_res = mch_stat((char *)buf->b_ffname, &st)) < 0
4084 || time_differs(&st, buf->b_mtime, buf->b_mtime_ns)
4085 || st.st_size != buf->b_orig_size
4086 #ifdef HAVE_ST_MODE
4087 || (int)st.st_mode != buf->b_orig_mode
4088 #else
4089 || mch_getperm(buf->b_ffname) != buf->b_orig_mode
4090 #endif
4091 ))
4092 {
4093 long prev_b_mtime = buf->b_mtime;
4094
4095 retval = 1;
4096
4097 // set b_mtime to stop further warnings (e.g., when executing
4098 // FileChangedShell autocmd)
4099 if (stat_res < 0)
4100 {
4101 // Check the file again later to see if it re-appears.
4102 buf->b_mtime = -1;
4103 buf->b_orig_size = 0;
4104 buf->b_orig_mode = 0;
4105 }
4106 else
4107 buf_store_time(buf, &st, buf->b_ffname);
4108
4109 // Don't do anything for a directory. Might contain the file
4110 // explorer.
4111 if (mch_isdir(buf->b_fname))
4112 ;
4113
4114 /*
4115 * If 'autoread' is set, the buffer has no changes and the file still
4116 * exists, reload the buffer. Use the buffer-local option value if it
4117 * was set, the global option value otherwise.
4118 */
4119 else if ((buf->b_p_ar >= 0 ? buf->b_p_ar : p_ar)
4120 && !bufIsChanged(buf) && stat_res >= 0)
4121 reload = TRUE;
4122 else
4123 {
4124 if (stat_res < 0)
4125 reason = "deleted";
4126 else if (bufIsChanged(buf))
4127 reason = "conflict";
4128 /*
4129 * Check if the file contents really changed to avoid giving a
4130 * warning when only the timestamp was set (e.g., checked out of
4131 * CVS). Always warn when the buffer was changed.
4132 */
4133 else if (orig_size != buf->b_orig_size || buf_contents_changed(buf))
4134 reason = "changed";
4135 else if (orig_mode != buf->b_orig_mode)
4136 reason = "mode";
4137 else
4138 reason = "time";
4139
4140 /*
4141 * Only give the warning if there are no FileChangedShell
4142 * autocommands.
4143 * Avoid being called recursively by setting "busy".
4144 */
4145 busy = TRUE;
4146 #ifdef FEAT_EVAL
4147 set_vim_var_string(VV_FCS_REASON, (char_u *)reason, -1);
4148 set_vim_var_string(VV_FCS_CHOICE, (char_u *)"", -1);
4149 #endif
4150 ++allbuf_lock;
4151 n = apply_autocmds(EVENT_FILECHANGEDSHELL,
4152 buf->b_fname, buf->b_fname, FALSE, buf);
4153 --allbuf_lock;
4154 busy = FALSE;
4155 if (n)
4156 {
4157 if (!bufref_valid(&bufref))
4158 emsg(_("E246: FileChangedShell autocommand deleted buffer"));
4159 #ifdef FEAT_EVAL
4160 s = get_vim_var_str(VV_FCS_CHOICE);
4161 if (STRCMP(s, "reload") == 0 && *reason != 'd')
4162 reload = TRUE;
4163 else if (STRCMP(s, "ask") == 0)
4164 n = FALSE;
4165 else
4166 #endif
4167 return 2;
4168 }
4169 if (!n)
4170 {
4171 if (*reason == 'd')
4172 {
4173 // Only give the message once.
4174 if (prev_b_mtime != -1)
4175 mesg = _("E211: File \"%s\" no longer available");
4176 }
4177 else
4178 {
4179 helpmesg = TRUE;
4180 #if defined(FEAT_CON_DIALOG) || defined(FEAT_GUI_DIALOG)
4181 can_reload = TRUE;
4182 #endif
4183 if (reason[2] == 'n')
4184 {
4185 mesg = _("W12: Warning: File \"%s\" has changed and the buffer was changed in Vim as well");
4186 mesg2 = _("See \":help W12\" for more info.");
4187 }
4188 else if (reason[1] == 'h')
4189 {
4190 mesg = _("W11: Warning: File \"%s\" has changed since editing started");
4191 mesg2 = _("See \":help W11\" for more info.");
4192 }
4193 else if (*reason == 'm')
4194 {
4195 mesg = _("W16: Warning: Mode of file \"%s\" has changed since editing started");
4196 mesg2 = _("See \":help W16\" for more info.");
4197 }
4198 else
4199 {
4200 // Only timestamp changed, store it to avoid a warning
4201 // in check_mtime() later.
4202 buf->b_mtime_read = buf->b_mtime;
4203 buf->b_mtime_read_ns = buf->b_mtime_ns;
4204 }
4205 }
4206 }
4207 }
4208
4209 }
4210 else if ((buf->b_flags & BF_NEW) && !(buf->b_flags & BF_NEW_W)
4211 && vim_fexists(buf->b_ffname))
4212 {
4213 retval = 1;
4214 mesg = _("W13: Warning: File \"%s\" has been created after editing started");
4215 buf->b_flags |= BF_NEW_W;
4216 #if defined(FEAT_CON_DIALOG) || defined(FEAT_GUI_DIALOG)
4217 can_reload = TRUE;
4218 #endif
4219 }
4220
4221 if (mesg != NULL)
4222 {
4223 path = home_replace_save(buf, buf->b_fname);
4224 if (path != NULL)
4225 {
4226 if (!helpmesg)
4227 mesg2 = "";
4228 tbuf = alloc(STRLEN(path) + STRLEN(mesg) + STRLEN(mesg2) + 2);
4229 sprintf(tbuf, mesg, path);
4230 #ifdef FEAT_EVAL
4231 // Set warningmsg here, before the unimportant and output-specific
4232 // mesg2 has been appended.
4233 set_vim_var_string(VV_WARNINGMSG, (char_u *)tbuf, -1);
4234 #endif
4235 #if defined(FEAT_CON_DIALOG) || defined(FEAT_GUI_DIALOG)
4236 if (can_reload)
4237 {
4238 if (*mesg2 != NUL)
4239 {
4240 STRCAT(tbuf, "\n");
4241 STRCAT(tbuf, mesg2);
4242 }
4243 if (do_dialog(VIM_WARNING, (char_u *)_("Warning"),
4244 (char_u *)tbuf,
4245 (char_u *)_("&OK\n&Load File"), 1, NULL, TRUE) == 2)
4246 reload = TRUE;
4247 }
4248 else
4249 #endif
4250 if (State > NORMAL_BUSY || (State & CMDLINE) || already_warned)
4251 {
4252 if (*mesg2 != NUL)
4253 {
4254 STRCAT(tbuf, "; ");
4255 STRCAT(tbuf, mesg2);
4256 }
4257 emsg(tbuf);
4258 retval = 2;
4259 }
4260 else
4261 {
4262 if (!autocmd_busy)
4263 {
4264 msg_start();
4265 msg_puts_attr(tbuf, HL_ATTR(HLF_E) + MSG_HIST);
4266 if (*mesg2 != NUL)
4267 msg_puts_attr(mesg2, HL_ATTR(HLF_W) + MSG_HIST);
4268 msg_clr_eos();
4269 (void)msg_end();
4270 if (emsg_silent == 0 && !in_assert_fails)
4271 {
4272 out_flush();
4273 #ifdef FEAT_GUI
4274 if (!focus)
4275 #endif
4276 // give the user some time to think about it
4277 ui_delay(1004L, TRUE);
4278
4279 // don't redraw and erase the message
4280 redraw_cmdline = FALSE;
4281 }
4282 }
4283 already_warned = TRUE;
4284 }
4285
4286 vim_free(path);
4287 vim_free(tbuf);
4288 }
4289 }
4290
4291 if (reload)
4292 {
4293 // Reload the buffer.
4294 buf_reload(buf, orig_mode);
4295 #ifdef FEAT_PERSISTENT_UNDO
4296 if (buf->b_p_udf && buf->b_ffname != NULL)
4297 {
4298 char_u hash[UNDO_HASH_SIZE];
4299 buf_T *save_curbuf = curbuf;
4300
4301 // Any existing undo file is unusable, write it now.
4302 curbuf = buf;
4303 u_compute_hash(hash);
4304 u_write_undo(NULL, FALSE, buf, hash);
4305 curbuf = save_curbuf;
4306 }
4307 #endif
4308 }
4309
4310 // Trigger FileChangedShell when the file was changed in any way.
4311 if (bufref_valid(&bufref) && retval != 0)
4312 (void)apply_autocmds(EVENT_FILECHANGEDSHELLPOST,
4313 buf->b_fname, buf->b_fname, FALSE, buf);
4314 #ifdef FEAT_GUI
4315 // restore this in case an autocommand has set it; it would break
4316 // 'mousefocus'
4317 need_mouse_correct = save_mouse_correct;
4318 #endif
4319
4320 return retval;
4321 }
4322
4323 /*
4324 * Reload a buffer that is already loaded.
4325 * Used when the file was changed outside of Vim.
4326 * "orig_mode" is buf->b_orig_mode before the need for reloading was detected.
4327 * buf->b_orig_mode may have been reset already.
4328 */
4329 void
buf_reload(buf_T * buf,int orig_mode)4330 buf_reload(buf_T *buf, int orig_mode)
4331 {
4332 exarg_T ea;
4333 pos_T old_cursor;
4334 linenr_T old_topline;
4335 int old_ro = buf->b_p_ro;
4336 buf_T *savebuf;
4337 bufref_T bufref;
4338 int saved = OK;
4339 aco_save_T aco;
4340 int flags = READ_NEW;
4341
4342 // set curwin/curbuf for "buf" and save some things
4343 aucmd_prepbuf(&aco, buf);
4344
4345 // We only want to read the text from the file, not reset the syntax
4346 // highlighting, clear marks, diff status, etc. Force the fileformat
4347 // and encoding to be the same.
4348 if (prep_exarg(&ea, buf) == OK)
4349 {
4350 old_cursor = curwin->w_cursor;
4351 old_topline = curwin->w_topline;
4352
4353 if (p_ur < 0 || curbuf->b_ml.ml_line_count <= p_ur)
4354 {
4355 // Save all the text, so that the reload can be undone.
4356 // Sync first so that this is a separate undo-able action.
4357 u_sync(FALSE);
4358 saved = u_savecommon(0, curbuf->b_ml.ml_line_count + 1, 0, TRUE);
4359 flags |= READ_KEEP_UNDO;
4360 }
4361
4362 /*
4363 * To behave like when a new file is edited (matters for
4364 * BufReadPost autocommands) we first need to delete the current
4365 * buffer contents. But if reading the file fails we should keep
4366 * the old contents. Can't use memory only, the file might be
4367 * too big. Use a hidden buffer to move the buffer contents to.
4368 */
4369 if (BUFEMPTY() || saved == FAIL)
4370 savebuf = NULL;
4371 else
4372 {
4373 // Allocate a buffer without putting it in the buffer list.
4374 savebuf = buflist_new(NULL, NULL, (linenr_T)1, BLN_DUMMY);
4375 set_bufref(&bufref, savebuf);
4376 if (savebuf != NULL && buf == curbuf)
4377 {
4378 // Open the memline.
4379 curbuf = savebuf;
4380 curwin->w_buffer = savebuf;
4381 saved = ml_open(curbuf);
4382 curbuf = buf;
4383 curwin->w_buffer = buf;
4384 }
4385 if (savebuf == NULL || saved == FAIL || buf != curbuf
4386 || move_lines(buf, savebuf) == FAIL)
4387 {
4388 semsg(_("E462: Could not prepare for reloading \"%s\""),
4389 buf->b_fname);
4390 saved = FAIL;
4391 }
4392 }
4393
4394 if (saved == OK)
4395 {
4396 curbuf->b_flags |= BF_CHECK_RO; // check for RO again
4397 keep_filetype = TRUE; // don't detect 'filetype'
4398 if (readfile(buf->b_ffname, buf->b_fname, (linenr_T)0,
4399 (linenr_T)0,
4400 (linenr_T)MAXLNUM, &ea, flags) != OK)
4401 {
4402 #if defined(FEAT_EVAL)
4403 if (!aborting())
4404 #endif
4405 semsg(_("E321: Could not reload \"%s\""), buf->b_fname);
4406 if (savebuf != NULL && bufref_valid(&bufref) && buf == curbuf)
4407 {
4408 // Put the text back from the save buffer. First
4409 // delete any lines that readfile() added.
4410 while (!BUFEMPTY())
4411 if (ml_delete(buf->b_ml.ml_line_count) == FAIL)
4412 break;
4413 (void)move_lines(savebuf, buf);
4414 }
4415 }
4416 else if (buf == curbuf) // "buf" still valid
4417 {
4418 // Mark the buffer as unmodified and free undo info.
4419 unchanged(buf, TRUE, TRUE);
4420 if ((flags & READ_KEEP_UNDO) == 0)
4421 {
4422 u_blockfree(buf);
4423 u_clearall(buf);
4424 }
4425 else
4426 {
4427 // Mark all undo states as changed.
4428 u_unchanged(curbuf);
4429 }
4430 }
4431 }
4432 vim_free(ea.cmd);
4433
4434 if (savebuf != NULL && bufref_valid(&bufref))
4435 wipe_buffer(savebuf, FALSE);
4436
4437 #ifdef FEAT_DIFF
4438 // Invalidate diff info if necessary.
4439 diff_invalidate(curbuf);
4440 #endif
4441
4442 // Restore the topline and cursor position and check it (lines may
4443 // have been removed).
4444 if (old_topline > curbuf->b_ml.ml_line_count)
4445 curwin->w_topline = curbuf->b_ml.ml_line_count;
4446 else
4447 curwin->w_topline = old_topline;
4448 curwin->w_cursor = old_cursor;
4449 check_cursor();
4450 update_topline();
4451 keep_filetype = FALSE;
4452 #ifdef FEAT_FOLDING
4453 {
4454 win_T *wp;
4455 tabpage_T *tp;
4456
4457 // Update folds unless they are defined manually.
4458 FOR_ALL_TAB_WINDOWS(tp, wp)
4459 if (wp->w_buffer == curwin->w_buffer
4460 && !foldmethodIsManual(wp))
4461 foldUpdateAll(wp);
4462 }
4463 #endif
4464 // If the mode didn't change and 'readonly' was set, keep the old
4465 // value; the user probably used the ":view" command. But don't
4466 // reset it, might have had a read error.
4467 if (orig_mode == curbuf->b_orig_mode)
4468 curbuf->b_p_ro |= old_ro;
4469
4470 // Modelines must override settings done by autocommands.
4471 do_modelines(0);
4472 }
4473
4474 // restore curwin/curbuf and a few other things
4475 aucmd_restbuf(&aco);
4476 // Careful: autocommands may have made "buf" invalid!
4477 }
4478
4479 void
buf_store_time(buf_T * buf,stat_T * st,char_u * fname UNUSED)4480 buf_store_time(buf_T *buf, stat_T *st, char_u *fname UNUSED)
4481 {
4482 buf->b_mtime = (long)st->st_mtime;
4483 #ifdef ST_MTIM_NSEC
4484 buf->b_mtime_ns = (long)st->ST_MTIM_NSEC;
4485 #else
4486 buf->b_mtime_ns = 0;
4487 #endif
4488 buf->b_orig_size = st->st_size;
4489 #ifdef HAVE_ST_MODE
4490 buf->b_orig_mode = (int)st->st_mode;
4491 #else
4492 buf->b_orig_mode = mch_getperm(fname);
4493 #endif
4494 }
4495
4496 /*
4497 * Adjust the line with missing eol, used for the next write.
4498 * Used for do_filter(), when the input lines for the filter are deleted.
4499 */
4500 void
write_lnum_adjust(linenr_T offset)4501 write_lnum_adjust(linenr_T offset)
4502 {
4503 if (curbuf->b_no_eol_lnum != 0) // only if there is a missing eol
4504 curbuf->b_no_eol_lnum += offset;
4505 }
4506
4507 // Subfuncions for readdirex()
4508 #ifdef FEAT_EVAL
4509 # ifdef MSWIN
4510 static char_u *
getfpermwfd(WIN32_FIND_DATAW * wfd,char_u * perm)4511 getfpermwfd(WIN32_FIND_DATAW *wfd, char_u *perm)
4512 {
4513 stat_T st;
4514 unsigned short st_mode;
4515 DWORD flag = wfd->dwFileAttributes;
4516 WCHAR *wp;
4517
4518 st_mode = (flag & FILE_ATTRIBUTE_DIRECTORY)
4519 ? (_S_IFDIR | _S_IEXEC) : _S_IFREG;
4520 st_mode |= (flag & FILE_ATTRIBUTE_READONLY)
4521 ? _S_IREAD : (_S_IREAD | _S_IWRITE);
4522
4523 wp = wcsrchr(wfd->cFileName, L'.');
4524 if (wp != NULL)
4525 {
4526 if (_wcsicmp(wp, L".exe") == 0 ||
4527 _wcsicmp(wp, L".com") == 0 ||
4528 _wcsicmp(wp, L".cmd") == 0 ||
4529 _wcsicmp(wp, L".bat") == 0)
4530 st_mode |= _S_IEXEC;
4531 }
4532
4533 // Copy user bits to group/other.
4534 st_mode |= (st_mode & 0700) >> 3;
4535 st_mode |= (st_mode & 0700) >> 6;
4536
4537 st.st_mode = st_mode;
4538 return getfpermst(&st, perm);
4539 }
4540
4541 static char_u *
getftypewfd(WIN32_FIND_DATAW * wfd)4542 getftypewfd(WIN32_FIND_DATAW *wfd)
4543 {
4544 DWORD flag = wfd->dwFileAttributes;
4545 DWORD tag = wfd->dwReserved0;
4546
4547 if (flag & FILE_ATTRIBUTE_REPARSE_POINT)
4548 {
4549 if (tag == IO_REPARSE_TAG_MOUNT_POINT)
4550 return (char_u*)"junction";
4551 else if (tag == IO_REPARSE_TAG_SYMLINK)
4552 {
4553 if (flag & FILE_ATTRIBUTE_DIRECTORY)
4554 return (char_u*)"linkd";
4555 else
4556 return (char_u*)"link";
4557 }
4558 return (char_u*)"reparse"; // unknown reparse point type
4559 }
4560 if (flag & FILE_ATTRIBUTE_DIRECTORY)
4561 return (char_u*)"dir";
4562 else
4563 return (char_u*)"file";
4564 }
4565
4566 static dict_T *
create_readdirex_item(WIN32_FIND_DATAW * wfd)4567 create_readdirex_item(WIN32_FIND_DATAW *wfd)
4568 {
4569 dict_T *item;
4570 char_u *p;
4571 varnumber_T size, time;
4572 char_u permbuf[] = "---------";
4573
4574 item = dict_alloc();
4575 if (item == NULL)
4576 return NULL;
4577 item->dv_refcount++;
4578
4579 p = utf16_to_enc(wfd->cFileName, NULL);
4580 if (p == NULL)
4581 goto theend;
4582 if (dict_add_string(item, "name", p) == FAIL)
4583 {
4584 vim_free(p);
4585 goto theend;
4586 }
4587 vim_free(p);
4588
4589 size = (((varnumber_T)wfd->nFileSizeHigh) << 32) | wfd->nFileSizeLow;
4590 if (dict_add_number(item, "size", size) == FAIL)
4591 goto theend;
4592
4593 // Convert FILETIME to unix time.
4594 time = (((((varnumber_T)wfd->ftLastWriteTime.dwHighDateTime) << 32) |
4595 wfd->ftLastWriteTime.dwLowDateTime)
4596 - 116444736000000000) / 10000000;
4597 if (dict_add_number(item, "time", time) == FAIL)
4598 goto theend;
4599
4600 if (dict_add_string(item, "type", getftypewfd(wfd)) == FAIL)
4601 goto theend;
4602 if (dict_add_string(item, "perm", getfpermwfd(wfd, permbuf)) == FAIL)
4603 goto theend;
4604
4605 if (dict_add_string(item, "user", (char_u*)"") == FAIL)
4606 goto theend;
4607 if (dict_add_string(item, "group", (char_u*)"") == FAIL)
4608 goto theend;
4609
4610 return item;
4611
4612 theend:
4613 dict_unref(item);
4614 return NULL;
4615 }
4616 # else
4617 static dict_T *
create_readdirex_item(char_u * path,char_u * name)4618 create_readdirex_item(char_u *path, char_u *name)
4619 {
4620 dict_T *item;
4621 char *p;
4622 size_t len;
4623 stat_T st;
4624 int ret, link = FALSE;
4625 varnumber_T size;
4626 char_u permbuf[] = "---------";
4627 char_u *q = NULL;
4628 struct passwd *pw;
4629 struct group *gr;
4630
4631 item = dict_alloc();
4632 if (item == NULL)
4633 return NULL;
4634 item->dv_refcount++;
4635
4636 len = STRLEN(path) + 1 + STRLEN(name) + 1;
4637 p = alloc(len);
4638 if (p == NULL)
4639 goto theend;
4640 vim_snprintf(p, len, "%s/%s", path, name);
4641 ret = mch_lstat(p, &st);
4642 if (ret >= 0 && S_ISLNK(st.st_mode))
4643 {
4644 link = TRUE;
4645 ret = mch_stat(p, &st);
4646 if (ret < 0)
4647 q = (char_u*)"link";
4648
4649 }
4650 vim_free(p);
4651
4652 if (dict_add_string(item, "name", name) == FAIL)
4653 goto theend;
4654
4655 if (ret >= 0)
4656 {
4657 size = (varnumber_T)st.st_size;
4658 if (S_ISDIR(st.st_mode))
4659 size = 0;
4660 // non-perfect check for overflow
4661 else if ((off_T)size != (off_T)st.st_size)
4662 size = -2;
4663 if (dict_add_number(item, "size", size) == FAIL)
4664 goto theend;
4665 if (dict_add_number(item, "time", (varnumber_T)st.st_mtime) == FAIL)
4666 goto theend;
4667
4668 if (link)
4669 {
4670 if (S_ISDIR(st.st_mode))
4671 q = (char_u*)"linkd";
4672 else
4673 q = (char_u*)"link";
4674 }
4675 else
4676 q = getftypest(&st);
4677 if (dict_add_string(item, "type", q) == FAIL)
4678 goto theend;
4679 if (dict_add_string(item, "perm", getfpermst(&st, permbuf)) == FAIL)
4680 goto theend;
4681
4682 pw = getpwuid(st.st_uid);
4683 if (pw == NULL)
4684 q = (char_u*)"";
4685 else
4686 q = (char_u*)pw->pw_name;
4687 if (dict_add_string(item, "user", q) == FAIL)
4688 goto theend;
4689 # if !defined(VMS) || (defined(VMS) && defined(HAVE_XOS_R_H))
4690 gr = getgrgid(st.st_gid);
4691 if (gr == NULL)
4692 q = (char_u*)"";
4693 else
4694 q = (char_u*)gr->gr_name;
4695 # endif
4696 if (dict_add_string(item, "group", q) == FAIL)
4697 goto theend;
4698 }
4699 else
4700 {
4701 if (dict_add_number(item, "size", -1) == FAIL)
4702 goto theend;
4703 if (dict_add_number(item, "time", -1) == FAIL)
4704 goto theend;
4705 if (dict_add_string(item, "type", q == NULL ? (char_u*)"" : q) == FAIL)
4706 goto theend;
4707 if (dict_add_string(item, "perm", (char_u*)"") == FAIL)
4708 goto theend;
4709 if (dict_add_string(item, "user", (char_u*)"") == FAIL)
4710 goto theend;
4711 if (dict_add_string(item, "group", (char_u*)"") == FAIL)
4712 goto theend;
4713 }
4714 return item;
4715
4716 theend:
4717 dict_unref(item);
4718 return NULL;
4719 }
4720 # endif
4721
4722 static int
compare_readdirex_item(const void * p1,const void * p2)4723 compare_readdirex_item(const void *p1, const void *p2)
4724 {
4725 char_u *name1, *name2;
4726
4727 name1 = dict_get_string(*(dict_T**)p1, (char_u*)"name", FALSE);
4728 name2 = dict_get_string(*(dict_T**)p2, (char_u*)"name", FALSE);
4729 if (readdirex_sort == READDIR_SORT_BYTE)
4730 return STRCMP(name1, name2);
4731 else if (readdirex_sort == READDIR_SORT_IC)
4732 return STRICMP(name1, name2);
4733 else
4734 return STRCOLL(name1, name2);
4735 }
4736
4737 static int
compare_readdir_item(const void * s1,const void * s2)4738 compare_readdir_item(const void *s1, const void *s2)
4739 {
4740 if (readdirex_sort == READDIR_SORT_BYTE)
4741 return STRCMP(*(char **)s1, *(char **)s2);
4742 else if (readdirex_sort == READDIR_SORT_IC)
4743 return STRICMP(*(char **)s1, *(char **)s2);
4744 else
4745 return STRCOLL(*(char **)s1, *(char **)s2);
4746 }
4747 #endif
4748
4749 #if defined(TEMPDIRNAMES) || defined(FEAT_EVAL) || defined(PROTO)
4750 /*
4751 * Core part of "readdir()" and "readdirex()" function.
4752 * Retrieve the list of files/directories of "path" into "gap".
4753 * If "withattr" is TRUE, retrieve the names and their attributes.
4754 * If "withattr" is FALSE, retrieve the names only.
4755 * Return OK for success, FAIL for failure.
4756 */
4757 int
readdir_core(garray_T * gap,char_u * path,int withattr UNUSED,void * context,int (* checkitem)(void * context,void * item),int sort)4758 readdir_core(
4759 garray_T *gap,
4760 char_u *path,
4761 int withattr UNUSED,
4762 void *context,
4763 int (*checkitem)(void *context, void *item),
4764 int sort)
4765 {
4766 int failed = FALSE;
4767 char_u *p;
4768 # ifdef MSWIN
4769 char_u *buf;
4770 int ok;
4771 HANDLE hFind = INVALID_HANDLE_VALUE;
4772 WIN32_FIND_DATAW wfd;
4773 WCHAR *wn = NULL; // UTF-16 name, NULL when not used.
4774 # else
4775 DIR *dirp;
4776 struct dirent *dp;
4777 # endif
4778
4779 ga_init2(gap, (int)sizeof(void *), 20);
4780
4781 # ifdef FEAT_EVAL
4782 # define FREE_ITEM(item) do { \
4783 if (withattr) \
4784 dict_unref((dict_T*)item); \
4785 else \
4786 vim_free(item); \
4787 } while (0)
4788
4789 readdirex_sort = READDIR_SORT_BYTE;
4790 # else
4791 # define FREE_ITEM(item) vim_free(item)
4792 # endif
4793
4794 # ifdef MSWIN
4795 buf = alloc(MAXPATHL);
4796 if (buf == NULL)
4797 return FAIL;
4798 STRNCPY(buf, path, MAXPATHL-5);
4799 p = buf + STRLEN(buf);
4800 MB_PTR_BACK(buf, p);
4801 if (*p == '\\' || *p == '/')
4802 *p = NUL;
4803 STRCAT(p, "\\*");
4804
4805 wn = enc_to_utf16(buf, NULL);
4806 if (wn != NULL)
4807 hFind = FindFirstFileW(wn, &wfd);
4808 ok = (hFind != INVALID_HANDLE_VALUE);
4809 if (!ok)
4810 {
4811 failed = TRUE;
4812 semsg(_(e_notopen), path);
4813 }
4814 else
4815 {
4816 while (ok)
4817 {
4818 int ignore;
4819 void *item;
4820 WCHAR *wp;
4821
4822 wp = wfd.cFileName;
4823 ignore = wp[0] == L'.' &&
4824 (wp[1] == NUL ||
4825 (wp[1] == L'.' && wp[2] == NUL));
4826 if (ignore)
4827 {
4828 ok = FindNextFileW(hFind, &wfd);
4829 continue;
4830 }
4831 # ifdef FEAT_EVAL
4832 if (withattr)
4833 item = (void*)create_readdirex_item(&wfd);
4834 else
4835 # endif
4836 item = (void*)utf16_to_enc(wfd.cFileName, NULL);
4837 if (item == NULL)
4838 {
4839 failed = TRUE;
4840 break;
4841 }
4842
4843 if (!ignore && checkitem != NULL)
4844 {
4845 int r = checkitem(context, item);
4846
4847 if (r < 0)
4848 {
4849 FREE_ITEM(item);
4850 break;
4851 }
4852 if (r == 0)
4853 ignore = TRUE;
4854 }
4855
4856 if (!ignore)
4857 {
4858 if (ga_grow(gap, 1) == OK)
4859 ((void**)gap->ga_data)[gap->ga_len++] = item;
4860 else
4861 {
4862 failed = TRUE;
4863 FREE_ITEM(item);
4864 break;
4865 }
4866 }
4867 else
4868 FREE_ITEM(item);
4869
4870 ok = FindNextFileW(hFind, &wfd);
4871 }
4872 FindClose(hFind);
4873 }
4874
4875 vim_free(buf);
4876 vim_free(wn);
4877 # else // MSWIN
4878 dirp = opendir((char *)path);
4879 if (dirp == NULL)
4880 {
4881 failed = TRUE;
4882 semsg(_(e_notopen), path);
4883 }
4884 else
4885 {
4886 for (;;)
4887 {
4888 int ignore;
4889 void *item;
4890
4891 dp = readdir(dirp);
4892 if (dp == NULL)
4893 break;
4894 p = (char_u *)dp->d_name;
4895
4896 ignore = p[0] == '.' &&
4897 (p[1] == NUL ||
4898 (p[1] == '.' && p[2] == NUL));
4899 if (ignore)
4900 continue;
4901 # ifdef FEAT_EVAL
4902 if (withattr)
4903 item = (void*)create_readdirex_item(path, p);
4904 else
4905 # endif
4906 item = (void*)vim_strsave(p);
4907 if (item == NULL)
4908 {
4909 failed = TRUE;
4910 break;
4911 }
4912
4913 if (!ignore && checkitem != NULL)
4914 {
4915 int r = checkitem(context, item);
4916
4917 if (r < 0)
4918 {
4919 FREE_ITEM(item);
4920 break;
4921 }
4922 if (r == 0)
4923 ignore = TRUE;
4924 }
4925
4926 if (!ignore)
4927 {
4928 if (ga_grow(gap, 1) == OK)
4929 ((void**)gap->ga_data)[gap->ga_len++] = item;
4930 else
4931 {
4932 failed = TRUE;
4933 FREE_ITEM(item);
4934 break;
4935 }
4936 }
4937 else
4938 FREE_ITEM(item);
4939 }
4940
4941 closedir(dirp);
4942 }
4943 # endif // MSWIN
4944
4945 # undef FREE_ITEM
4946
4947 if (!failed && gap->ga_len > 0 && sort > READDIR_SORT_NONE)
4948 {
4949 # ifdef FEAT_EVAL
4950 readdirex_sort = sort;
4951 if (withattr)
4952 qsort((void*)gap->ga_data, (size_t)gap->ga_len, sizeof(dict_T*),
4953 compare_readdirex_item);
4954 else
4955 qsort((void*)gap->ga_data, (size_t)gap->ga_len, sizeof(char_u *),
4956 compare_readdir_item);
4957 # else
4958 sort_strings((char_u **)gap->ga_data, gap->ga_len);
4959 # endif
4960 }
4961
4962 return failed ? FAIL : OK;
4963 }
4964
4965 /*
4966 * Delete "name" and everything in it, recursively.
4967 * return 0 for success, -1 if some file was not deleted.
4968 */
4969 int
delete_recursive(char_u * name)4970 delete_recursive(char_u *name)
4971 {
4972 int result = 0;
4973 int i;
4974 char_u *exp;
4975 garray_T ga;
4976
4977 // A symbolic link to a directory itself is deleted, not the directory it
4978 // points to.
4979 if (
4980 # if defined(UNIX) || defined(MSWIN)
4981 mch_isrealdir(name)
4982 # else
4983 mch_isdir(name)
4984 # endif
4985 )
4986 {
4987 exp = vim_strsave(name);
4988 if (exp == NULL)
4989 return -1;
4990 if (readdir_core(&ga, exp, FALSE, NULL, NULL, READDIR_SORT_NONE) == OK)
4991 {
4992 for (i = 0; i < ga.ga_len; ++i)
4993 {
4994 vim_snprintf((char *)NameBuff, MAXPATHL, "%s/%s", exp,
4995 ((char_u **)ga.ga_data)[i]);
4996 if (delete_recursive(NameBuff) != 0)
4997 result = -1;
4998 }
4999 ga_clear_strings(&ga);
5000 }
5001 else
5002 result = -1;
5003 (void)mch_rmdir(exp);
5004 vim_free(exp);
5005 }
5006 else
5007 result = mch_remove(name) == 0 ? 0 : -1;
5008
5009 return result;
5010 }
5011 #endif
5012
5013 #if defined(TEMPDIRNAMES) || defined(PROTO)
5014 static long temp_count = 0; // Temp filename counter.
5015
5016 # if defined(UNIX) && defined(HAVE_FLOCK) && defined(HAVE_DIRFD)
5017 /*
5018 * Open temporary directory and take file lock to prevent
5019 * to be auto-cleaned.
5020 */
5021 static void
vim_opentempdir(void)5022 vim_opentempdir(void)
5023 {
5024 DIR *dp = NULL;
5025
5026 if (vim_tempdir_dp != NULL)
5027 return;
5028
5029 dp = opendir((const char*)vim_tempdir);
5030
5031 if (dp != NULL)
5032 {
5033 vim_tempdir_dp = dp;
5034 flock(dirfd(vim_tempdir_dp), LOCK_SH);
5035 }
5036 }
5037
5038 /*
5039 * Close temporary directory - it automatically release file lock.
5040 */
5041 static void
vim_closetempdir(void)5042 vim_closetempdir(void)
5043 {
5044 if (vim_tempdir_dp != NULL)
5045 {
5046 closedir(vim_tempdir_dp);
5047 vim_tempdir_dp = NULL;
5048 }
5049 }
5050 # endif
5051
5052 /*
5053 * Delete the temp directory and all files it contains.
5054 */
5055 void
vim_deltempdir(void)5056 vim_deltempdir(void)
5057 {
5058 if (vim_tempdir != NULL)
5059 {
5060 # if defined(UNIX) && defined(HAVE_FLOCK) && defined(HAVE_DIRFD)
5061 vim_closetempdir();
5062 # endif
5063 // remove the trailing path separator
5064 gettail(vim_tempdir)[-1] = NUL;
5065 delete_recursive(vim_tempdir);
5066 VIM_CLEAR(vim_tempdir);
5067 }
5068 }
5069
5070 /*
5071 * Directory "tempdir" was created. Expand this name to a full path and put
5072 * it in "vim_tempdir". This avoids that using ":cd" would confuse us.
5073 * "tempdir" must be no longer than MAXPATHL.
5074 */
5075 static void
vim_settempdir(char_u * tempdir)5076 vim_settempdir(char_u *tempdir)
5077 {
5078 char_u *buf;
5079
5080 buf = alloc(MAXPATHL + 2);
5081 if (buf != NULL)
5082 {
5083 if (vim_FullName(tempdir, buf, MAXPATHL, FALSE) == FAIL)
5084 STRCPY(buf, tempdir);
5085 add_pathsep(buf);
5086 vim_tempdir = vim_strsave(buf);
5087 # if defined(UNIX) && defined(HAVE_FLOCK) && defined(HAVE_DIRFD)
5088 vim_opentempdir();
5089 # endif
5090 vim_free(buf);
5091 }
5092 }
5093 #endif
5094
5095 /*
5096 * vim_tempname(): Return a unique name that can be used for a temp file.
5097 *
5098 * The temp file is NOT guaranteed to be created. If "keep" is FALSE it is
5099 * guaranteed to NOT be created.
5100 *
5101 * The returned pointer is to allocated memory.
5102 * The returned pointer is NULL if no valid name was found.
5103 */
5104 char_u *
vim_tempname(int extra_char UNUSED,int keep UNUSED)5105 vim_tempname(
5106 int extra_char UNUSED, // char to use in the name instead of '?'
5107 int keep UNUSED)
5108 {
5109 #ifdef USE_TMPNAM
5110 char_u itmp[L_tmpnam]; // use tmpnam()
5111 #elif defined(MSWIN)
5112 WCHAR itmp[TEMPNAMELEN];
5113 #else
5114 char_u itmp[TEMPNAMELEN];
5115 #endif
5116
5117 #ifdef TEMPDIRNAMES
5118 static char *(tempdirs[]) = {TEMPDIRNAMES};
5119 int i;
5120 # ifndef EEXIST
5121 stat_T st;
5122 # endif
5123
5124 /*
5125 * This will create a directory for private use by this instance of Vim.
5126 * This is done once, and the same directory is used for all temp files.
5127 * This method avoids security problems because of symlink attacks et al.
5128 * It's also a bit faster, because we only need to check for an existing
5129 * file when creating the directory and not for each temp file.
5130 */
5131 if (vim_tempdir == NULL)
5132 {
5133 /*
5134 * Try the entries in TEMPDIRNAMES to create the temp directory.
5135 */
5136 for (i = 0; i < (int)ARRAY_LENGTH(tempdirs); ++i)
5137 {
5138 # ifndef HAVE_MKDTEMP
5139 size_t itmplen;
5140 long nr;
5141 long off;
5142 # endif
5143
5144 // Expand $TMP, leave room for "/v1100000/999999999".
5145 // Skip the directory check if the expansion fails.
5146 expand_env((char_u *)tempdirs[i], itmp, TEMPNAMELEN - 20);
5147 if (itmp[0] != '$' && mch_isdir(itmp))
5148 {
5149 // directory exists
5150 add_pathsep(itmp);
5151
5152 # ifdef HAVE_MKDTEMP
5153 {
5154 # if defined(UNIX) || defined(VMS)
5155 // Make sure the umask doesn't remove the executable bit.
5156 // "repl" has been reported to use "177".
5157 mode_t umask_save = umask(077);
5158 # endif
5159 // Leave room for filename
5160 STRCAT(itmp, "vXXXXXX");
5161 if (mkdtemp((char *)itmp) != NULL)
5162 vim_settempdir(itmp);
5163 # if defined(UNIX) || defined(VMS)
5164 (void)umask(umask_save);
5165 # endif
5166 }
5167 # else
5168 // Get an arbitrary number of up to 6 digits. When it's
5169 // unlikely that it already exists it will be faster,
5170 // otherwise it doesn't matter. The use of mkdir() avoids any
5171 // security problems because of the predictable number.
5172 nr = (mch_get_pid() + (long)time(NULL)) % 1000000L;
5173 itmplen = STRLEN(itmp);
5174
5175 // Try up to 10000 different values until we find a name that
5176 // doesn't exist.
5177 for (off = 0; off < 10000L; ++off)
5178 {
5179 int r;
5180 # if defined(UNIX) || defined(VMS)
5181 mode_t umask_save;
5182 # endif
5183
5184 sprintf((char *)itmp + itmplen, "v%ld", nr + off);
5185 # ifndef EEXIST
5186 // If mkdir() does not set errno to EEXIST, check for
5187 // existing file here. There is a race condition then,
5188 // although it's fail-safe.
5189 if (mch_stat((char *)itmp, &st) >= 0)
5190 continue;
5191 # endif
5192 # if defined(UNIX) || defined(VMS)
5193 // Make sure the umask doesn't remove the executable bit.
5194 // "repl" has been reported to use "177".
5195 umask_save = umask(077);
5196 # endif
5197 r = vim_mkdir(itmp, 0700);
5198 # if defined(UNIX) || defined(VMS)
5199 (void)umask(umask_save);
5200 # endif
5201 if (r == 0)
5202 {
5203 vim_settempdir(itmp);
5204 break;
5205 }
5206 # ifdef EEXIST
5207 // If the mkdir() didn't fail because the file/dir exists,
5208 // we probably can't create any dir here, try another
5209 // place.
5210 if (errno != EEXIST)
5211 # endif
5212 break;
5213 }
5214 # endif // HAVE_MKDTEMP
5215 if (vim_tempdir != NULL)
5216 break;
5217 }
5218 }
5219 }
5220
5221 if (vim_tempdir != NULL)
5222 {
5223 // There is no need to check if the file exists, because we own the
5224 // directory and nobody else creates a file in it.
5225 sprintf((char *)itmp, "%s%ld", vim_tempdir, temp_count++);
5226 return vim_strsave(itmp);
5227 }
5228
5229 return NULL;
5230
5231 #else // TEMPDIRNAMES
5232
5233 # ifdef MSWIN
5234 WCHAR wszTempFile[_MAX_PATH + 1];
5235 WCHAR buf4[4];
5236 WCHAR *chartab = L"0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
5237 char_u *retval;
5238 char_u *p;
5239 char_u *shname;
5240 long i;
5241
5242 wcscpy(itmp, L"");
5243 if (GetTempPathW(_MAX_PATH, wszTempFile) == 0)
5244 {
5245 wszTempFile[0] = L'.'; // GetTempPathW() failed, use current dir
5246 wszTempFile[1] = L'\\';
5247 wszTempFile[2] = NUL;
5248 }
5249 wcscpy(buf4, L"VIM");
5250
5251 // randomize the name to avoid collisions
5252 i = mch_get_pid() + extra_char;
5253 buf4[1] = chartab[i % 36];
5254 buf4[2] = chartab[101 * i % 36];
5255 if (GetTempFileNameW(wszTempFile, buf4, 0, itmp) == 0)
5256 return NULL;
5257 if (!keep)
5258 // GetTempFileName() will create the file, we don't want that
5259 (void)DeleteFileW(itmp);
5260
5261 // Backslashes in a temp file name cause problems when filtering with
5262 // "sh". NOTE: This also checks 'shellcmdflag' to help those people who
5263 // didn't set 'shellslash' but only if not using PowerShell.
5264 retval = utf16_to_enc(itmp, NULL);
5265 shname = gettail(p_sh);
5266 if ((*p_shcf == '-' && !(strstr((char *)shname, "powershell") != NULL
5267 || strstr((char *)shname, "pwsh") != NULL ))
5268 || p_ssl)
5269 for (p = retval; *p; ++p)
5270 if (*p == '\\')
5271 *p = '/';
5272 return retval;
5273
5274 # else // MSWIN
5275
5276 # ifdef USE_TMPNAM
5277 char_u *p;
5278
5279 // tmpnam() will make its own name
5280 p = tmpnam((char *)itmp);
5281 if (p == NULL || *p == NUL)
5282 return NULL;
5283 # else
5284 char_u *p;
5285
5286 # ifdef VMS_TEMPNAM
5287 // mktemp() is not working on VMS. It seems to be
5288 // a do-nothing function. Therefore we use tempnam().
5289 sprintf((char *)itmp, "VIM%c", extra_char);
5290 p = (char_u *)tempnam("tmp:", (char *)itmp);
5291 if (p != NULL)
5292 {
5293 // VMS will use '.LIS' if we don't explicitly specify an extension,
5294 // and VIM will then be unable to find the file later
5295 STRCPY(itmp, p);
5296 STRCAT(itmp, ".txt");
5297 free(p);
5298 }
5299 else
5300 return NULL;
5301 # else
5302 STRCPY(itmp, TEMPNAME);
5303 if ((p = vim_strchr(itmp, '?')) != NULL)
5304 *p = extra_char;
5305 if (mktemp((char *)itmp) == NULL)
5306 return NULL;
5307 # endif
5308 # endif
5309
5310 return vim_strsave(itmp);
5311 # endif // MSWIN
5312 #endif // TEMPDIRNAMES
5313 }
5314
5315 #if defined(BACKSLASH_IN_FILENAME) || defined(PROTO)
5316 /*
5317 * Convert all backslashes in fname to forward slashes in-place, unless when
5318 * it looks like a URL.
5319 */
5320 void
forward_slash(char_u * fname)5321 forward_slash(char_u *fname)
5322 {
5323 char_u *p;
5324
5325 if (path_with_url(fname))
5326 return;
5327 for (p = fname; *p != NUL; ++p)
5328 // The Big5 encoding can have '\' in the trail byte.
5329 if (enc_dbcs != 0 && (*mb_ptr2len)(p) > 1)
5330 ++p;
5331 else if (*p == '\\')
5332 *p = '/';
5333 }
5334 #endif
5335
5336 /*
5337 * Try matching a filename with a "pattern" ("prog" is NULL), or use the
5338 * precompiled regprog "prog" ("pattern" is NULL). That avoids calling
5339 * vim_regcomp() often.
5340 * Used for autocommands and 'wildignore'.
5341 * Returns TRUE if there is a match, FALSE otherwise.
5342 */
5343 int
match_file_pat(char_u * pattern,regprog_T ** prog,char_u * fname,char_u * sfname,char_u * tail,int allow_dirs)5344 match_file_pat(
5345 char_u *pattern, // pattern to match with
5346 regprog_T **prog, // pre-compiled regprog or NULL
5347 char_u *fname, // full path of file name
5348 char_u *sfname, // short file name or NULL
5349 char_u *tail, // tail of path
5350 int allow_dirs) // allow matching with dir
5351 {
5352 regmatch_T regmatch;
5353 int result = FALSE;
5354
5355 regmatch.rm_ic = p_fic; // ignore case if 'fileignorecase' is set
5356 if (prog != NULL)
5357 regmatch.regprog = *prog;
5358 else
5359 regmatch.regprog = vim_regcomp(pattern, RE_MAGIC);
5360
5361 /*
5362 * Try for a match with the pattern with:
5363 * 1. the full file name, when the pattern has a '/'.
5364 * 2. the short file name, when the pattern has a '/'.
5365 * 3. the tail of the file name, when the pattern has no '/'.
5366 */
5367 if (regmatch.regprog != NULL
5368 && ((allow_dirs
5369 && (vim_regexec(®match, fname, (colnr_T)0)
5370 || (sfname != NULL
5371 && vim_regexec(®match, sfname, (colnr_T)0))))
5372 || (!allow_dirs && vim_regexec(®match, tail, (colnr_T)0))))
5373 result = TRUE;
5374
5375 if (prog != NULL)
5376 *prog = regmatch.regprog;
5377 else
5378 vim_regfree(regmatch.regprog);
5379 return result;
5380 }
5381
5382 #if defined(FEAT_WILDIGN) || defined(PROTO)
5383 /*
5384 * Return TRUE if a file matches with a pattern in "list".
5385 * "list" is a comma-separated list of patterns, like 'wildignore'.
5386 * "sfname" is the short file name or NULL, "ffname" the long file name.
5387 */
5388 int
match_file_list(char_u * list,char_u * sfname,char_u * ffname)5389 match_file_list(char_u *list, char_u *sfname, char_u *ffname)
5390 {
5391 char_u buf[100];
5392 char_u *tail;
5393 char_u *regpat;
5394 char allow_dirs;
5395 int match;
5396 char_u *p;
5397
5398 tail = gettail(sfname);
5399
5400 // try all patterns in 'wildignore'
5401 p = list;
5402 while (*p)
5403 {
5404 copy_option_part(&p, buf, 100, ",");
5405 regpat = file_pat_to_reg_pat(buf, NULL, &allow_dirs, FALSE);
5406 if (regpat == NULL)
5407 break;
5408 match = match_file_pat(regpat, NULL, ffname, sfname,
5409 tail, (int)allow_dirs);
5410 vim_free(regpat);
5411 if (match)
5412 return TRUE;
5413 }
5414 return FALSE;
5415 }
5416 #endif
5417
5418 /*
5419 * Convert the given pattern "pat" which has shell style wildcards in it, into
5420 * a regular expression, and return the result in allocated memory. If there
5421 * is a directory path separator to be matched, then TRUE is put in
5422 * allow_dirs, otherwise FALSE is put there -- webb.
5423 * Handle backslashes before special characters, like "\*" and "\ ".
5424 *
5425 * Returns NULL when out of memory.
5426 */
5427 char_u *
file_pat_to_reg_pat(char_u * pat,char_u * pat_end,char * allow_dirs,int no_bslash UNUSED)5428 file_pat_to_reg_pat(
5429 char_u *pat,
5430 char_u *pat_end, // first char after pattern or NULL
5431 char *allow_dirs, // Result passed back out in here
5432 int no_bslash UNUSED) // Don't use a backward slash as pathsep
5433 {
5434 int size = 2; // '^' at start, '$' at end
5435 char_u *endp;
5436 char_u *reg_pat;
5437 char_u *p;
5438 int i;
5439 int nested = 0;
5440 int add_dollar = TRUE;
5441
5442 if (allow_dirs != NULL)
5443 *allow_dirs = FALSE;
5444 if (pat_end == NULL)
5445 pat_end = pat + STRLEN(pat);
5446
5447 for (p = pat; p < pat_end; p++)
5448 {
5449 switch (*p)
5450 {
5451 case '*':
5452 case '.':
5453 case ',':
5454 case '{':
5455 case '}':
5456 case '~':
5457 size += 2; // extra backslash
5458 break;
5459 #ifdef BACKSLASH_IN_FILENAME
5460 case '\\':
5461 case '/':
5462 size += 4; // could become "[\/]"
5463 break;
5464 #endif
5465 default:
5466 size++;
5467 if (enc_dbcs != 0 && (*mb_ptr2len)(p) > 1)
5468 {
5469 ++p;
5470 ++size;
5471 }
5472 break;
5473 }
5474 }
5475 reg_pat = alloc(size + 1);
5476 if (reg_pat == NULL)
5477 return NULL;
5478
5479 i = 0;
5480
5481 if (pat[0] == '*')
5482 while (pat[0] == '*' && pat < pat_end - 1)
5483 pat++;
5484 else
5485 reg_pat[i++] = '^';
5486 endp = pat_end - 1;
5487 if (endp >= pat && *endp == '*')
5488 {
5489 while (endp - pat > 0 && *endp == '*')
5490 endp--;
5491 add_dollar = FALSE;
5492 }
5493 for (p = pat; *p && nested >= 0 && p <= endp; p++)
5494 {
5495 switch (*p)
5496 {
5497 case '*':
5498 reg_pat[i++] = '.';
5499 reg_pat[i++] = '*';
5500 while (p[1] == '*') // "**" matches like "*"
5501 ++p;
5502 break;
5503 case '.':
5504 case '~':
5505 reg_pat[i++] = '\\';
5506 reg_pat[i++] = *p;
5507 break;
5508 case '?':
5509 reg_pat[i++] = '.';
5510 break;
5511 case '\\':
5512 if (p[1] == NUL)
5513 break;
5514 #ifdef BACKSLASH_IN_FILENAME
5515 if (!no_bslash)
5516 {
5517 // translate:
5518 // "\x" to "\\x" e.g., "dir\file"
5519 // "\*" to "\\.*" e.g., "dir\*.c"
5520 // "\?" to "\\." e.g., "dir\??.c"
5521 // "\+" to "\+" e.g., "fileX\+.c"
5522 if ((vim_isfilec(p[1]) || p[1] == '*' || p[1] == '?')
5523 && p[1] != '+')
5524 {
5525 reg_pat[i++] = '[';
5526 reg_pat[i++] = '\\';
5527 reg_pat[i++] = '/';
5528 reg_pat[i++] = ']';
5529 if (allow_dirs != NULL)
5530 *allow_dirs = TRUE;
5531 break;
5532 }
5533 }
5534 #endif
5535 // Undo escaping from ExpandEscape():
5536 // foo\?bar -> foo?bar
5537 // foo\%bar -> foo%bar
5538 // foo\,bar -> foo,bar
5539 // foo\ bar -> foo bar
5540 // Don't unescape \, * and others that are also special in a
5541 // regexp.
5542 // An escaped { must be unescaped since we use magic not
5543 // verymagic. Use "\\\{n,m\}"" to get "\{n,m}".
5544 if (*++p == '?'
5545 #ifdef BACKSLASH_IN_FILENAME
5546 && no_bslash
5547 #endif
5548 )
5549 reg_pat[i++] = '?';
5550 else
5551 if (*p == ',' || *p == '%' || *p == '#'
5552 || vim_isspace(*p) || *p == '{' || *p == '}')
5553 reg_pat[i++] = *p;
5554 else if (*p == '\\' && p[1] == '\\' && p[2] == '{')
5555 {
5556 reg_pat[i++] = '\\';
5557 reg_pat[i++] = '{';
5558 p += 2;
5559 }
5560 else
5561 {
5562 if (allow_dirs != NULL && vim_ispathsep(*p)
5563 #ifdef BACKSLASH_IN_FILENAME
5564 && (!no_bslash || *p != '\\')
5565 #endif
5566 )
5567 *allow_dirs = TRUE;
5568 reg_pat[i++] = '\\';
5569 reg_pat[i++] = *p;
5570 }
5571 break;
5572 #ifdef BACKSLASH_IN_FILENAME
5573 case '/':
5574 reg_pat[i++] = '[';
5575 reg_pat[i++] = '\\';
5576 reg_pat[i++] = '/';
5577 reg_pat[i++] = ']';
5578 if (allow_dirs != NULL)
5579 *allow_dirs = TRUE;
5580 break;
5581 #endif
5582 case '{':
5583 reg_pat[i++] = '\\';
5584 reg_pat[i++] = '(';
5585 nested++;
5586 break;
5587 case '}':
5588 reg_pat[i++] = '\\';
5589 reg_pat[i++] = ')';
5590 --nested;
5591 break;
5592 case ',':
5593 if (nested)
5594 {
5595 reg_pat[i++] = '\\';
5596 reg_pat[i++] = '|';
5597 }
5598 else
5599 reg_pat[i++] = ',';
5600 break;
5601 default:
5602 if (enc_dbcs != 0 && (*mb_ptr2len)(p) > 1)
5603 reg_pat[i++] = *p++;
5604 else if (allow_dirs != NULL && vim_ispathsep(*p))
5605 *allow_dirs = TRUE;
5606 reg_pat[i++] = *p;
5607 break;
5608 }
5609 }
5610 if (add_dollar)
5611 reg_pat[i++] = '$';
5612 reg_pat[i] = NUL;
5613 if (nested != 0)
5614 {
5615 if (nested < 0)
5616 emsg(_("E219: Missing {."));
5617 else
5618 emsg(_("E220: Missing }."));
5619 VIM_CLEAR(reg_pat);
5620 }
5621 return reg_pat;
5622 }
5623
5624 #if defined(EINTR) || defined(PROTO)
5625 /*
5626 * Version of read() that retries when interrupted by EINTR (possibly
5627 * by a SIGWINCH).
5628 */
5629 long
read_eintr(int fd,void * buf,size_t bufsize)5630 read_eintr(int fd, void *buf, size_t bufsize)
5631 {
5632 long ret;
5633
5634 for (;;)
5635 {
5636 ret = vim_read(fd, buf, bufsize);
5637 if (ret >= 0 || errno != EINTR)
5638 break;
5639 }
5640 return ret;
5641 }
5642
5643 /*
5644 * Version of write() that retries when interrupted by EINTR (possibly
5645 * by a SIGWINCH).
5646 */
5647 long
write_eintr(int fd,void * buf,size_t bufsize)5648 write_eintr(int fd, void *buf, size_t bufsize)
5649 {
5650 long ret = 0;
5651 long wlen;
5652
5653 // Repeat the write() so long it didn't fail, other than being interrupted
5654 // by a signal.
5655 while (ret < (long)bufsize)
5656 {
5657 wlen = vim_write(fd, (char *)buf + ret, bufsize - ret);
5658 if (wlen < 0)
5659 {
5660 if (errno != EINTR)
5661 break;
5662 }
5663 else
5664 ret += wlen;
5665 }
5666 return ret;
5667 }
5668 #endif
5669