indent.c revision 1.69 1 /* $NetBSD: indent.c,v 1.69 2021/09/25 10:41:03 rillig Exp $ */
2
3 /*-
4 * SPDX-License-Identifier: BSD-4-Clause
5 *
6 * Copyright (c) 1985 Sun Microsystems, Inc.
7 * Copyright (c) 1976 Board of Trustees of the University of Illinois.
8 * Copyright (c) 1980, 1993
9 * The Regents of the University of California. All rights reserved.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * SUCH DAMAGE.
38 */
39
40 #if 0
41 #ifndef lint
42 static char sccsid[] = "@(#)indent.c 5.17 (Berkeley) 6/7/93";
43 #endif /* not lint */
44 #endif
45
46 #include <sys/cdefs.h>
47 #ifndef lint
48 #if defined(__NetBSD__)
49 __RCSID("$NetBSD: indent.c,v 1.69 2021/09/25 10:41:03 rillig Exp $");
50 #elif defined(__FreeBSD__)
51 __FBSDID("$FreeBSD: head/usr.bin/indent/indent.c 340138 2018-11-04 19:24:49Z oshogbo $");
52 #endif
53 #endif
54
55 #include <sys/param.h>
56 #if HAVE_CAPSICUM
57 #include <sys/capsicum.h>
58 #include <capsicum_helpers.h>
59 #endif
60 #include <err.h>
61 #include <errno.h>
62 #include <fcntl.h>
63 #include <unistd.h>
64 #include <stdio.h>
65 #include <stdlib.h>
66 #include <string.h>
67 #include <ctype.h>
68
69 #include "indent.h"
70
71 struct options opt;
72 struct parser_state ps;
73
74 struct buffer lab;
75 struct buffer code;
76 struct buffer com;
77 struct buffer token;
78
79 char *in_buffer;
80 char *in_buffer_limit;
81 char *buf_ptr;
82 char *buf_end;
83
84 char sc_buf[sc_size];
85 char *save_com;
86 char *sc_end;
87
88 char *bp_save;
89 char *be_save;
90
91 int found_err;
92 int n_real_blanklines;
93 int prefix_blankline_requested;
94 int postfix_blankline_requested;
95 int break_comma;
96 float case_ind;
97 int had_eof;
98 int line_no;
99 int inhibit_formatting;
100 int suppress_blanklines;
101
102 int ifdef_level;
103 struct parser_state state_stack[5];
104 struct parser_state match_state[5];
105
106 FILE *input;
107 FILE *output;
108
109 static void bakcopy(void);
110 static void indent_declaration(int, int);
111
112 const char *in_name = "Standard Input"; /* will always point to name of input
113 * file */
114 const char *out_name = "Standard Output"; /* will always point to name
115 * of output file */
116 const char *simple_backup_suffix = ".BAK"; /* Suffix to use for backup
117 * files */
118 char bakfile[MAXPATHLEN] = "";
119
120 static void
121 check_size_code(size_t desired_size)
122 {
123 if (code.e + desired_size < code.l)
124 return;
125
126 size_t nsize = code.l - code.s + 400 + desired_size;
127 size_t code_len = code.e - code.s;
128 code.buf = xrealloc(code.buf, nsize);
129 code.e = code.buf + code_len + 1;
130 code.l = code.buf + nsize - 5;
131 code.s = code.buf + 1;
132 }
133
134 static void
135 check_size_label(size_t desired_size)
136 {
137 if (lab.e + (desired_size) < lab.l)
138 return;
139
140 size_t nsize = lab.l - lab.s + 400 + desired_size;
141 size_t label_len = lab.e - lab.s;
142 lab.buf = xrealloc(lab.buf, nsize);
143 lab.e = lab.buf + label_len + 1;
144 lab.l = lab.buf + nsize - 5;
145 lab.s = lab.buf + 1;
146 }
147
148 #if HAVE_CAPSICUM
149 static void
150 init_capsicum(void)
151 {
152 cap_rights_t rights;
153
154 /* Restrict input/output descriptors and enter Capsicum sandbox. */
155 cap_rights_init(&rights, CAP_FSTAT, CAP_WRITE);
156 if (caph_rights_limit(fileno(output), &rights) < 0)
157 err(EXIT_FAILURE, "unable to limit rights for %s", out_name);
158 cap_rights_init(&rights, CAP_FSTAT, CAP_READ);
159 if (caph_rights_limit(fileno(input), &rights) < 0)
160 err(EXIT_FAILURE, "unable to limit rights for %s", in_name);
161 if (caph_enter() < 0)
162 err(EXIT_FAILURE, "unable to enter capability mode");
163 }
164 #endif
165
166 static void
167 search_brace(token_type *inout_ttype, int *inout_force_nl,
168 int *inout_comment_buffered, int *inout_last_else)
169 {
170 while (ps.search_brace) {
171 switch (*inout_ttype) {
172 case newline:
173 if (sc_end == NULL) {
174 save_com = sc_buf;
175 save_com[0] = save_com[1] = ' ';
176 sc_end = &save_com[2];
177 }
178 *sc_end++ = '\n';
179 /*
180 * We may have inherited a force_nl == true from the previous
181 * token (like a semicolon). But once we know that a newline
182 * has been scanned in this loop, force_nl should be false.
183 *
184 * However, the force_nl == true must be preserved if newline
185 * is never scanned in this loop, so this assignment cannot be
186 * done earlier.
187 */
188 *inout_force_nl = false;
189 break;
190 case form_feed:
191 break;
192 case comment:
193 if (sc_end == NULL) {
194 /*
195 * Copy everything from the start of the line, because
196 * process_comment() will use that to calculate original
197 * indentation of a boxed comment.
198 */
199 memcpy(sc_buf, in_buffer, (size_t)(buf_ptr - in_buffer) - 4);
200 save_com = sc_buf + (buf_ptr - in_buffer - 4);
201 save_com[0] = save_com[1] = ' ';
202 sc_end = &save_com[2];
203 }
204 *inout_comment_buffered = true;
205 *sc_end++ = '/'; /* copy in start of comment */
206 *sc_end++ = '*';
207 for (;;) { /* loop until the end of the comment */
208 *sc_end = *buf_ptr++;
209 if (buf_ptr >= buf_end)
210 fill_buffer();
211 if (*sc_end++ == '*' && *buf_ptr == '/')
212 break; /* we are at end of comment */
213 if (sc_end >= &save_com[sc_size]) { /* check for temp buffer
214 * overflow */
215 diag(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
216 fflush(output);
217 exit(1);
218 }
219 }
220 *sc_end++ = '/'; /* add ending slash */
221 if (++buf_ptr >= buf_end) /* get past / in buffer */
222 fill_buffer();
223 break;
224 case lbrace:
225 /*
226 * Put KNF-style lbraces before the buffered up tokens and
227 * jump out of this loop in order to avoid copying the token
228 * again under the default case of the switch below.
229 */
230 if (sc_end != NULL && opt.btype_2) {
231 save_com[0] = '{';
232 /*
233 * Originally the lbrace may have been alone on its own
234 * line, but it will be moved into "the else's line", so
235 * if there was a newline resulting from the "{" before,
236 * it must be scanned now and ignored.
237 */
238 while (isspace((unsigned char)*buf_ptr)) {
239 if (++buf_ptr >= buf_end)
240 fill_buffer();
241 if (*buf_ptr == '\n')
242 break;
243 }
244 goto sw_buffer;
245 }
246 /* FALLTHROUGH */
247 default: /* it is the start of a normal statement */
248 {
249 int remove_newlines;
250
251 remove_newlines =
252 /* "} else" */
253 (*inout_ttype == keyword_do_else && *token.s == 'e' &&
254 code.e != code.s && code.e[-1] == '}')
255 /* "else if" */
256 || (*inout_ttype == keyword_for_if_while &&
257 *token.s == 'i' && *inout_last_else && opt.else_if);
258 if (remove_newlines)
259 *inout_force_nl = false;
260 if (sc_end == NULL) { /* ignore buffering if
261 * comment wasn't saved up */
262 ps.search_brace = false;
263 return;
264 }
265 while (sc_end > save_com && isblank((unsigned char)sc_end[-1])) {
266 sc_end--;
267 }
268 if (opt.swallow_optional_blanklines ||
269 (!*inout_comment_buffered && remove_newlines)) {
270 *inout_force_nl = !remove_newlines;
271 while (sc_end > save_com && sc_end[-1] == '\n') {
272 sc_end--;
273 }
274 }
275 if (*inout_force_nl) { /* if we should insert a nl here, put
276 * it into the buffer */
277 *inout_force_nl = false;
278 --line_no; /* this will be re-increased when the
279 * newline is read from the buffer */
280 *sc_end++ = '\n';
281 *sc_end++ = ' ';
282 if (opt.verbose) /* print error msg if the line was
283 * not already broken */
284 diag(0, "Line broken");
285 }
286 for (const char *t_ptr = token.s; *t_ptr; ++t_ptr)
287 *sc_end++ = *t_ptr;
288
289 sw_buffer:
290 ps.search_brace = false; /* stop looking for start of stmt */
291 bp_save = buf_ptr; /* save current input buffer */
292 be_save = buf_end;
293 buf_ptr = save_com; /* fix so that subsequent calls to
294 * lexi will take tokens out of save_com */
295 *sc_end++ = ' '; /* add trailing blank, just in case */
296 buf_end = sc_end;
297 sc_end = NULL;
298 debug_println("switched buf_ptr to save_com");
299 break;
300 }
301 } /* end of switch */
302 /*
303 * We must make this check, just in case there was an unexpected
304 * EOF.
305 */
306 if (*inout_ttype != end_of_file) {
307 /*
308 * The only intended purpose of calling lexi() below is to
309 * categorize the next token in order to decide whether to
310 * continue buffering forthcoming tokens. Once the buffering
311 * is over, lexi() will be called again elsewhere on all of
312 * the tokens - this time for normal processing.
313 *
314 * Calling it for this purpose is a bug, because lexi() also
315 * changes the parser state and discards leading whitespace,
316 * which is needed mostly for comment-related considerations.
317 *
318 * Work around the former problem by giving lexi() a copy of
319 * the current parser state and discard it if the call turned
320 * out to be just a look ahead.
321 *
322 * Work around the latter problem by copying all whitespace
323 * characters into the buffer so that the later lexi() call
324 * will read them.
325 */
326 if (sc_end != NULL) {
327 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
328 *sc_end++ = *buf_ptr++;
329 if (sc_end >= &save_com[sc_size]) {
330 errx(1, "input too long");
331 }
332 }
333 if (buf_ptr >= buf_end) {
334 fill_buffer();
335 }
336 }
337
338 struct parser_state transient_state;
339 transient_state = ps;
340 *inout_ttype = lexi(&transient_state); /* read another token */
341 if (*inout_ttype != newline && *inout_ttype != form_feed &&
342 *inout_ttype != comment && !transient_state.search_brace) {
343 ps = transient_state;
344 }
345 }
346 }
347
348 *inout_last_else = 0;
349 }
350
351 static void
352 main_init_globals(void)
353 {
354 found_err = 0;
355
356 ps.p_stack[0] = stmt; /* this is the parser's stack */
357 ps.last_nl = true; /* this is true if the last thing scanned was
358 * a newline */
359 ps.last_token = semicolon;
360 com.buf = xmalloc(bufsize);
361 lab.buf = xmalloc(bufsize);
362 code.buf = xmalloc(bufsize);
363 token.buf = xmalloc(bufsize);
364 alloc_typenames();
365 init_constant_tt();
366 com.l = com.buf + bufsize - 5;
367 lab.l = lab.buf + bufsize - 5;
368 code.l = code.buf + bufsize - 5;
369 token.l = token.buf + bufsize - 5;
370 com.buf[0] = code.buf[0] = lab.buf[0] = ' '; /* set up code, label, and
371 * comment buffers */
372 com.buf[1] = code.buf[1] = lab.buf[1] = token.buf[1] = '\0';
373 opt.else_if = 1; /* Default else-if special processing to on */
374 lab.s = lab.e = lab.buf + 1;
375 code.s = code.e = code.buf + 1;
376 com.s = com.e = com.buf + 1;
377 token.s = token.e = token.buf + 1;
378
379 in_buffer = xmalloc(10);
380 in_buffer_limit = in_buffer + 8;
381 buf_ptr = buf_end = in_buffer;
382 line_no = 1;
383 had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
384 ps.in_or_st = false;
385 ps.bl_line = true;
386 ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
387
388 ps.pcase = false;
389 sc_end = NULL;
390 bp_save = NULL;
391 be_save = NULL;
392
393 output = NULL;
394
395 const char *suffix = getenv("SIMPLE_BACKUP_SUFFIX");
396 if (suffix != NULL)
397 simple_backup_suffix = suffix;
398 }
399
400 static void
401 main_parse_command_line(int argc, char **argv)
402 {
403 int i;
404 const char *profile_name = NULL;
405
406 #if 0
407 max_line_length = 78; /* -l78 */
408 lineup_to_parens = 1; /* -lp */
409 lineup_to_parens_always = 0; /* -nlpl */
410 ps.ljust_decl = 0; /* -ndj */
411 ps.com_ind = 33; /* -c33 */
412 star_comment_cont = 1; /* -sc */
413 ps.ind_size = 8; /* -i8 */
414 verbose = 0;
415 ps.decl_indent = 16; /* -di16 */
416 ps.local_decl_indent = -1; /* if this is not set to some nonnegative value
417 * by an arg, we will set this equal to
418 * ps.decl_ind */
419 ps.indent_parameters = 1; /* -ip */
420 ps.decl_com_ind = 0; /* if this is not set to some positive value
421 * by an arg, we will set this equal to
422 * ps.com_ind */
423 btype_2 = 1; /* -br */
424 cuddle_else = 1; /* -ce */
425 ps.unindent_displace = 0; /* -d0 */
426 ps.case_indent = 0; /* -cli0 */
427 format_block_comments = 1; /* -fcb */
428 format_col1_comments = 1; /* -fc1 */
429 procnames_start_line = 1; /* -psl */
430 proc_calls_space = 0; /* -npcs */
431 comment_delimiter_on_blankline = 1; /* -cdb */
432 ps.leave_comma = 1; /* -nbc */
433 #endif
434
435 for (i = 1; i < argc; ++i)
436 if (strcmp(argv[i], "-npro") == 0)
437 break;
438 else if (argv[i][0] == '-' && argv[i][1] == 'P' && argv[i][2] != '\0')
439 profile_name = argv[i]; /* non-empty -P (set profile) */
440 set_defaults();
441 if (i >= argc)
442 set_profile(profile_name);
443
444 for (i = 1; i < argc; ++i) {
445
446 /*
447 * look thru args (if any) for changes to defaults
448 */
449 if (argv[i][0] != '-') {/* no flag on parameter */
450 if (input == NULL) { /* we must have the input file */
451 in_name = argv[i]; /* remember name of input file */
452 input = fopen(in_name, "r");
453 if (input == NULL) /* check for open error */
454 err(1, "%s", in_name);
455 continue;
456 } else if (output == NULL) { /* we have the output file */
457 out_name = argv[i]; /* remember name of output file */
458 if (strcmp(in_name, out_name) == 0) { /* attempt to overwrite
459 * the file */
460 errx(1, "input and output files must be different");
461 }
462 output = fopen(out_name, "w");
463 if (output == NULL) /* check for create error */
464 err(1, "%s", out_name);
465 continue;
466 }
467 errx(1, "unknown parameter: %s", argv[i]);
468 } else
469 set_option(argv[i]);
470 } /* end of for */
471 if (input == NULL)
472 input = stdin;
473 if (output == NULL) {
474 if (input == stdin)
475 output = stdout;
476 else {
477 out_name = in_name;
478 bakcopy();
479 }
480 }
481
482 if (opt.comment_column <= 1)
483 opt.comment_column = 2; /* don't put normal comments before column 2 */
484 if (opt.block_comment_max_line_length <= 0)
485 opt.block_comment_max_line_length = opt.max_line_length;
486 if (opt.local_decl_indent < 0) /* if not specified by user, set this */
487 opt.local_decl_indent = opt.decl_indent;
488 if (opt.decl_comment_column <= 0) /* if not specified by user, set this */
489 opt.decl_comment_column = opt.ljust_decl
490 ? (opt.comment_column <= 10 ? 2 : opt.comment_column - 8)
491 : opt.comment_column;
492 if (opt.continuation_indent == 0)
493 opt.continuation_indent = opt.indent_size;
494 }
495
496 static void
497 main_prepare_parsing(void)
498 {
499 fill_buffer(); /* get first batch of stuff into input buffer */
500
501 parse(semicolon);
502
503 char *p = buf_ptr;
504 int col = 1;
505
506 for (;;) {
507 if (*p == ' ')
508 col++;
509 else if (*p == '\t')
510 col = opt.tabsize * (1 + (col - 1) / opt.tabsize) + 1;
511 else
512 break;
513 p++;
514 }
515 if (col > opt.indent_size)
516 ps.ind_level = ps.i_l_follow = col / opt.indent_size;
517 }
518
519 static void __attribute__((__noreturn__))
520 process_end_of_file(void)
521 {
522 if (lab.s != lab.e || code.s != code.e || com.s != com.e)
523 dump_line();
524
525 if (ps.tos > 1) /* check for balanced braces */
526 diag(1, "Stuff missing from end of file");
527
528 if (opt.verbose) {
529 printf("There were %d output lines and %d comments\n",
530 ps.stats.lines, ps.stats.comments);
531 printf("(Lines with comments)/(Lines with code): %6.3f\n",
532 (1.0 * ps.stats.comment_lines) / ps.stats.code_lines);
533 }
534
535 fflush(output);
536 exit(found_err);
537 }
538
539 static void
540 process_comment_in_code(token_type ttype, int *inout_force_nl)
541 {
542 if (*inout_force_nl &&
543 ttype != semicolon &&
544 (ttype != lbrace || !opt.btype_2)) {
545
546 /* we should force a broken line here */
547 if (opt.verbose)
548 diag(0, "Line broken");
549 dump_line();
550 ps.want_blank = false; /* dont insert blank at line start */
551 *inout_force_nl = false;
552 }
553
554 ps.in_stmt = true; /* turn on flag which causes an extra level of
555 * indentation. this is turned off by a ; or
556 * '}' */
557 if (com.s != com.e) { /* the turkey has embedded a comment
558 * in a line. fix it */
559 size_t len = com.e - com.s;
560
561 check_size_code(len + 3);
562 *code.e++ = ' ';
563 memcpy(code.e, com.s, len);
564 code.e += len;
565 *code.e++ = ' ';
566 *code.e = '\0';
567 ps.want_blank = false;
568 com.e = com.s;
569 }
570 }
571
572 static void
573 process_form_feed(void)
574 {
575 ps.use_ff = true; /* a form feed is treated much like a newline */
576 dump_line();
577 ps.want_blank = false;
578 }
579
580 static void
581 process_newline(void)
582 {
583 if (ps.last_token != comma || ps.p_l_follow > 0
584 || !opt.leave_comma || ps.block_init || !break_comma || com.s != com.e) {
585 dump_line();
586 ps.want_blank = false;
587 }
588 ++line_no; /* keep track of input line number */
589 }
590
591 static void
592 process_lparen_or_lbracket(int dec_ind, int tabs_to_var, int sp_sw)
593 {
594 /* count parens to make Healy happy */
595 if (++ps.p_l_follow == nitems(ps.paren_indents)) {
596 diag(0, "Reached internal limit of %zu unclosed parens",
597 nitems(ps.paren_indents));
598 ps.p_l_follow--;
599 }
600 if (*token.s == '[')
601 /* not a function pointer declaration or a function call */;
602 else if (ps.in_decl && !ps.block_init && !ps.dumped_decl_indent &&
603 ps.procname[0] == '\0' && ps.paren_level == 0) {
604 /* function pointer declarations */
605 indent_declaration(dec_ind, tabs_to_var);
606 ps.dumped_decl_indent = true;
607 } else if (ps.want_blank &&
608 ((ps.last_token != ident && ps.last_token != funcname) ||
609 opt.proc_calls_space ||
610 (ps.keyword == rw_sizeof ? opt.Bill_Shannon :
611 ps.keyword != rw_0 && ps.keyword != rw_offsetof)))
612 *code.e++ = ' ';
613 ps.want_blank = false;
614 *code.e++ = token.s[0];
615
616 ps.paren_indents[ps.p_l_follow - 1] =
617 indentation_after_range(0, code.s, code.e);
618 debug_println("paren_indent[%d] is now %d",
619 ps.p_l_follow - 1, ps.paren_indents[ps.p_l_follow - 1]);
620
621 if (sp_sw && ps.p_l_follow == 1 && opt.extra_expression_indent
622 && ps.paren_indents[0] < 2 * opt.indent_size) {
623 ps.paren_indents[0] = 2 * opt.indent_size;
624 debug_println("paren_indent[0] is now %d", ps.paren_indents[0]);
625 }
626 if (ps.in_or_st && *token.s == '(' && ps.tos <= 2) {
627 /*
628 * this is a kluge to make sure that declarations will be
629 * aligned right if proc decl has an explicit type on it, i.e.
630 * "int a(x) {..."
631 */
632 parse(semicolon); /* I said this was a kluge... */
633 ps.in_or_st = false; /* turn off flag for structure decl or
634 * initialization */
635 }
636 /* parenthesized type following sizeof or offsetof is not a cast */
637 if (ps.keyword == rw_offsetof || ps.keyword == rw_sizeof)
638 ps.not_cast_mask |= 1 << ps.p_l_follow;
639 }
640
641 static void
642 process_rparen_or_rbracket(int *inout_sp_sw, int *inout_force_nl,
643 token_type hd_type)
644 {
645 if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.not_cast_mask) {
646 ps.last_u_d = true;
647 ps.cast_mask &= (1 << ps.p_l_follow) - 1;
648 ps.want_blank = opt.space_after_cast;
649 } else
650 ps.want_blank = true;
651 ps.not_cast_mask &= (1 << ps.p_l_follow) - 1;
652
653 if (--ps.p_l_follow < 0) {
654 ps.p_l_follow = 0;
655 diag(0, "Extra %c", *token.s);
656 }
657
658 if (code.e == code.s) /* if the paren starts the line */
659 ps.paren_level = ps.p_l_follow; /* then indent it */
660
661 *code.e++ = token.s[0];
662
663 if (*inout_sp_sw && (ps.p_l_follow == 0)) { /* check for end of if
664 * (...), or some such */
665 *inout_sp_sw = false;
666 *inout_force_nl = true; /* must force newline after if */
667 ps.last_u_d = true; /* inform lexi that a following
668 * operator is unary */
669 ps.in_stmt = false; /* dont use stmt continuation indentation */
670
671 parse(hd_type); /* let parser worry about if, or whatever */
672 }
673 ps.search_brace = opt.btype_2; /* this should ensure that constructs such
674 * as main(){...} and int[]{...} have their
675 * braces put in the right place */
676 }
677
678 static void
679 process_unary_op(int dec_ind, int tabs_to_var)
680 {
681 if (!ps.dumped_decl_indent && ps.in_decl && !ps.block_init &&
682 ps.procname[0] == '\0' && ps.paren_level == 0) {
683 /* pointer declarations */
684
685 /*
686 * if this is a unary op in a declaration, we should indent
687 * this token
688 */
689 int i;
690 for (i = 0; token.s[i]; ++i)
691 /* find length of token */;
692 indent_declaration(dec_ind - i, tabs_to_var);
693 ps.dumped_decl_indent = true;
694 } else if (ps.want_blank)
695 *code.e++ = ' ';
696
697 {
698 size_t len = token.e - token.s;
699
700 check_size_code(len);
701 memcpy(code.e, token.s, len);
702 code.e += len;
703 }
704 ps.want_blank = false;
705 }
706
707 static void
708 process_binary_op(void)
709 {
710 size_t len = token.e - token.s;
711
712 check_size_code(len + 1);
713 if (ps.want_blank)
714 *code.e++ = ' ';
715 memcpy(code.e, token.s, len);
716 code.e += len;
717
718 ps.want_blank = true;
719 }
720
721 static void
722 process_postfix_op(void)
723 {
724 *code.e++ = token.s[0];
725 *code.e++ = token.s[1];
726 ps.want_blank = true;
727 }
728
729 static void
730 process_question(int *inout_squest)
731 {
732 (*inout_squest)++; /* this will be used when a later colon
733 * appears so we can distinguish the
734 * <c>?<n>:<n> construct */
735 if (ps.want_blank)
736 *code.e++ = ' ';
737 *code.e++ = '?';
738 ps.want_blank = true;
739 }
740
741 static void
742 process_colon(int *inout_squest, int *inout_force_nl, int *inout_scase)
743 {
744 if (*inout_squest > 0) { /* it is part of the <c>?<n>: <n> construct */
745 --*inout_squest;
746 if (ps.want_blank)
747 *code.e++ = ' ';
748 *code.e++ = ':';
749 ps.want_blank = true;
750 return;
751 }
752 if (ps.in_or_st) {
753 *code.e++ = ':';
754 ps.want_blank = false;
755 return;
756 }
757 ps.in_stmt = false; /* seeing a label does not imply we are in a
758 * stmt */
759 /*
760 * turn everything so far into a label
761 */
762 {
763 size_t len = code.e - code.s;
764
765 check_size_label(len + 3);
766 memcpy(lab.e, code.s, len);
767 lab.e += len;
768 *lab.e++ = ':';
769 *lab.e = '\0';
770 code.e = code.s;
771 }
772 *inout_force_nl = ps.pcase = *inout_scase; /* ps.pcase will be used by
773 * dump_line to decide how to
774 * indent the label. force_nl
775 * will force a case n: to be
776 * on a line by itself */
777 *inout_scase = false;
778 ps.want_blank = false;
779 }
780
781 static void
782 process_semicolon(int *inout_scase, int *inout_squest, int const dec_ind,
783 int const tabs_to_var, int *inout_sp_sw,
784 token_type const hd_type,
785 int *inout_force_nl)
786 {
787 if (ps.dec_nest == 0)
788 ps.in_or_st = false; /* we are not in an initialization or
789 * structure declaration */
790 *inout_scase = false; /* these will only need resetting in an error */
791 *inout_squest = 0;
792 if (ps.last_token == rparen)
793 ps.in_parameter_declaration = 0;
794 ps.cast_mask = 0;
795 ps.not_cast_mask = 0;
796 ps.block_init = 0;
797 ps.block_init_level = 0;
798 ps.just_saw_decl--;
799
800 if (ps.in_decl && code.s == code.e && !ps.block_init &&
801 !ps.dumped_decl_indent && ps.paren_level == 0) {
802 /* indent stray semicolons in declarations */
803 indent_declaration(dec_ind - 1, tabs_to_var);
804 ps.dumped_decl_indent = true;
805 }
806
807 ps.in_decl = (ps.dec_nest > 0); /* if we were in a first level
808 * structure declaration, we
809 * arent any more */
810
811 if ((!*inout_sp_sw || hd_type != for_exprs) && ps.p_l_follow > 0) {
812
813 /*
814 * This should be true iff there were unbalanced parens in the
815 * stmt. It is a bit complicated, because the semicolon might
816 * be in a for stmt
817 */
818 diag(1, "Unbalanced parens");
819 ps.p_l_follow = 0;
820 if (*inout_sp_sw) { /* this is a check for an if, while, etc. with
821 * unbalanced parens */
822 *inout_sp_sw = false;
823 parse(hd_type); /* dont lose the if, or whatever */
824 }
825 }
826 *code.e++ = ';';
827 ps.want_blank = true;
828 ps.in_stmt = (ps.p_l_follow > 0); /* we are no longer in the
829 * middle of a stmt */
830
831 if (!*inout_sp_sw) { /* if not if for (;;) */
832 parse(semicolon); /* let parser know about end of stmt */
833 *inout_force_nl = true;/* force newline after an end of stmt */
834 }
835 }
836
837 static void
838 process_lbrace(int *inout_force_nl, int *inout_sp_sw, token_type hd_type,
839 int *di_stack, int di_stack_cap, int *inout_dec_ind)
840 {
841 ps.in_stmt = false; /* dont indent the {} */
842 if (!ps.block_init)
843 *inout_force_nl = true; /* force other stuff on same line as '{' onto
844 * new line */
845 else if (ps.block_init_level <= 0)
846 ps.block_init_level = 1;
847 else
848 ps.block_init_level++;
849
850 if (code.s != code.e && !ps.block_init) {
851 if (!opt.btype_2) {
852 dump_line();
853 ps.want_blank = false;
854 } else if (ps.in_parameter_declaration && !ps.in_or_st) {
855 ps.i_l_follow = 0;
856 if (opt.function_brace_split) { /* dump the line prior
857 * to the brace ... */
858 dump_line();
859 ps.want_blank = false;
860 } else /* add a space between the decl and brace */
861 ps.want_blank = true;
862 }
863 }
864 if (ps.in_parameter_declaration)
865 prefix_blankline_requested = 0;
866
867 if (ps.p_l_follow > 0) { /* check for preceding unbalanced
868 * parens */
869 diag(1, "Unbalanced parens");
870 ps.p_l_follow = 0;
871 if (*inout_sp_sw) { /* check for unclosed if, for, etc. */
872 *inout_sp_sw = false;
873 parse(hd_type);
874 ps.ind_level = ps.i_l_follow;
875 }
876 }
877 if (code.s == code.e)
878 ps.ind_stmt = false; /* dont put extra indentation on line
879 * with '{' */
880 if (ps.in_decl && ps.in_or_st) { /* this is either a structure
881 * declaration or an init */
882 di_stack[ps.dec_nest] = *inout_dec_ind;
883 if (++ps.dec_nest == di_stack_cap) {
884 diag(0, "Reached internal limit of %d struct levels",
885 di_stack_cap);
886 ps.dec_nest--;
887 }
888 /* ? dec_ind = 0; */
889 } else {
890 ps.decl_on_line = false; /* we can't be in the middle of
891 * a declaration, so don't do
892 * special indentation of
893 * comments */
894 if (opt.blanklines_after_declarations_at_proctop
895 && ps.in_parameter_declaration)
896 postfix_blankline_requested = 1;
897 ps.in_parameter_declaration = 0;
898 ps.in_decl = false;
899 }
900 *inout_dec_ind = 0;
901 parse(lbrace); /* let parser know about this */
902 if (ps.want_blank) /* put a blank before '{' if '{' is not at
903 * start of line */
904 *code.e++ = ' ';
905 ps.want_blank = false;
906 *code.e++ = '{';
907 ps.just_saw_decl = 0;
908 }
909
910 static void
911 process_rbrace(int *inout_sp_sw, int *inout_dec_ind, const int *di_stack)
912 {
913 if (ps.p_stack[ps.tos] == decl && !ps.block_init) /* semicolons can be
914 * omitted in declarations */
915 parse(semicolon);
916 if (ps.p_l_follow) { /* check for unclosed if, for, else. */
917 diag(1, "Unbalanced parens");
918 ps.p_l_follow = 0;
919 *inout_sp_sw = false;
920 }
921 ps.just_saw_decl = 0;
922 ps.block_init_level--;
923 if (code.s != code.e && !ps.block_init) { /* '}' must be first on line */
924 if (opt.verbose)
925 diag(0, "Line broken");
926 dump_line();
927 }
928 *code.e++ = '}';
929 ps.want_blank = true;
930 ps.in_stmt = ps.ind_stmt = false;
931 if (ps.dec_nest > 0) { /* we are in multi-level structure declaration */
932 *inout_dec_ind = di_stack[--ps.dec_nest];
933 if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
934 ps.just_saw_decl = 2;
935 ps.in_decl = true;
936 }
937 prefix_blankline_requested = 0;
938 parse(rbrace); /* let parser know about this */
939 ps.search_brace = opt.cuddle_else
940 && ps.p_stack[ps.tos] == if_expr_stmt
941 && ps.il[ps.tos] >= ps.ind_level;
942 if (ps.tos <= 1 && opt.blanklines_after_procs && ps.dec_nest <= 0)
943 postfix_blankline_requested = 1;
944 }
945
946 static void
947 process_keyword_do_else(int *inout_force_nl, int *inout_last_else)
948 {
949 ps.in_stmt = false;
950 if (*token.s == 'e') {
951 if (code.e != code.s && (!opt.cuddle_else || code.e[-1] != '}')) {
952 if (opt.verbose)
953 diag(0, "Line broken");
954 dump_line(); /* make sure this starts a line */
955 ps.want_blank = false;
956 }
957 *inout_force_nl = true;/* also, following stuff must go onto new line */
958 *inout_last_else = 1;
959 parse(keyword_else);
960 } else {
961 if (code.e != code.s) { /* make sure this starts a line */
962 if (opt.verbose)
963 diag(0, "Line broken");
964 dump_line();
965 ps.want_blank = false;
966 }
967 *inout_force_nl = true;/* also, following stuff must go onto new line */
968 *inout_last_else = 0;
969 parse(keyword_do);
970 }
971 }
972
973 static void
974 process_decl(int *out_dec_ind, int *out_tabs_to_var)
975 {
976 parse(decl); /* let parser worry about indentation */
977 if (ps.last_token == rparen && ps.tos <= 1) {
978 if (code.s != code.e) {
979 dump_line();
980 ps.want_blank = 0;
981 }
982 }
983 if (ps.in_parameter_declaration && opt.indent_parameters && ps.dec_nest == 0) {
984 ps.ind_level = ps.i_l_follow = 1;
985 ps.ind_stmt = 0;
986 }
987 ps.in_or_st = true; /* this might be a structure or initialization
988 * declaration */
989 ps.in_decl = ps.decl_on_line = ps.last_token != type_def;
990 if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
991 ps.just_saw_decl = 2;
992 prefix_blankline_requested = 0;
993 int i;
994 for (i = 0; token.s[i++];); /* get length of token */
995
996 if (ps.ind_level == 0 || ps.dec_nest > 0) {
997 /* global variable or struct member in local variable */
998 *out_dec_ind = opt.decl_indent > 0 ? opt.decl_indent : i;
999 *out_tabs_to_var = (opt.use_tabs ? opt.decl_indent > 0 : 0);
1000 } else {
1001 /* local variable */
1002 *out_dec_ind = opt.local_decl_indent > 0 ? opt.local_decl_indent : i;
1003 *out_tabs_to_var = (opt.use_tabs ? opt.local_decl_indent > 0 : 0);
1004 }
1005 }
1006
1007 static void
1008 process_ident(token_type ttype, int dec_ind, int tabs_to_var,
1009 int *inout_sp_sw, int *inout_force_nl, token_type hd_type)
1010 {
1011 if (ps.in_decl) {
1012 if (ttype == funcname) {
1013 ps.in_decl = false;
1014 if (opt.procnames_start_line && code.s != code.e) {
1015 *code.e = '\0';
1016 dump_line();
1017 } else if (ps.want_blank) {
1018 *code.e++ = ' ';
1019 }
1020 ps.want_blank = false;
1021 } else if (!ps.block_init && !ps.dumped_decl_indent &&
1022 ps.paren_level == 0) { /* if we are in a declaration, we
1023 * must indent identifier */
1024 indent_declaration(dec_ind, tabs_to_var);
1025 ps.dumped_decl_indent = true;
1026 ps.want_blank = false;
1027 }
1028 } else if (*inout_sp_sw && ps.p_l_follow == 0) {
1029 *inout_sp_sw = false;
1030 *inout_force_nl = true;
1031 ps.last_u_d = true;
1032 ps.in_stmt = false;
1033 parse(hd_type);
1034 }
1035 }
1036
1037 static void
1038 copy_id(void)
1039 {
1040 size_t len = token.e - token.s;
1041
1042 check_size_code(len + 1);
1043 if (ps.want_blank)
1044 *code.e++ = ' ';
1045 memcpy(code.e, token.s, len);
1046 code.e += len;
1047 }
1048
1049 static void
1050 process_string_prefix(void)
1051 {
1052 size_t len = token.e - token.s;
1053
1054 check_size_code(len + 1);
1055 if (ps.want_blank)
1056 *code.e++ = ' ';
1057 memcpy(code.e, token.s, len);
1058 code.e += len;
1059
1060 ps.want_blank = false;
1061 }
1062
1063 static void
1064 process_period(void)
1065 {
1066 *code.e++ = '.'; /* move the period into line */
1067 ps.want_blank = false; /* dont put a blank after a period */
1068 }
1069
1070 static void
1071 process_comma(int dec_ind, int tabs_to_var, int *inout_force_nl)
1072 {
1073 ps.want_blank = (code.s != code.e); /* only put blank after comma
1074 * if comma does not start the line */
1075 if (ps.in_decl && ps.procname[0] == '\0' && !ps.block_init &&
1076 !ps.dumped_decl_indent && ps.paren_level == 0) {
1077 /* indent leading commas and not the actual identifiers */
1078 indent_declaration(dec_ind - 1, tabs_to_var);
1079 ps.dumped_decl_indent = true;
1080 }
1081 *code.e++ = ',';
1082 if (ps.p_l_follow == 0) {
1083 if (ps.block_init_level <= 0)
1084 ps.block_init = 0;
1085 if (break_comma && (!opt.leave_comma ||
1086 indentation_after_range(
1087 compute_code_indent(), code.s, code.e)
1088 >= opt.max_line_length - opt.tabsize))
1089 *inout_force_nl = true;
1090 }
1091 }
1092
1093 static void
1094 process_preprocessing(void)
1095 {
1096 if (com.s != com.e || lab.s != lab.e || code.s != code.e)
1097 dump_line();
1098 check_size_label(1);
1099 *lab.e++ = '#'; /* move whole line to 'label' buffer */
1100
1101 {
1102 int in_comment = 0;
1103 int com_start = 0;
1104 char quote = '\0';
1105 int com_end = 0;
1106
1107 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
1108 buf_ptr++;
1109 if (buf_ptr >= buf_end)
1110 fill_buffer();
1111 }
1112 while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
1113 check_size_label(2);
1114 *lab.e = *buf_ptr++;
1115 if (buf_ptr >= buf_end)
1116 fill_buffer();
1117 switch (*lab.e++) {
1118 case '\\':
1119 if (!in_comment) {
1120 *lab.e++ = *buf_ptr++;
1121 if (buf_ptr >= buf_end)
1122 fill_buffer();
1123 }
1124 break;
1125 case '/':
1126 if (*buf_ptr == '*' && !in_comment && quote == '\0') {
1127 in_comment = 1;
1128 *lab.e++ = *buf_ptr++;
1129 com_start = (int)(lab.e - lab.s) - 2;
1130 }
1131 break;
1132 case '"':
1133 if (quote == '"')
1134 quote = '\0';
1135 else if (quote == '\0')
1136 quote = '"';
1137 break;
1138 case '\'':
1139 if (quote == '\'')
1140 quote = '\0';
1141 else if (quote == '\0')
1142 quote = '\'';
1143 break;
1144 case '*':
1145 if (*buf_ptr == '/' && in_comment) {
1146 in_comment = 0;
1147 *lab.e++ = *buf_ptr++;
1148 com_end = (int)(lab.e - lab.s);
1149 }
1150 break;
1151 }
1152 }
1153
1154 while (lab.e > lab.s && (lab.e[-1] == ' ' || lab.e[-1] == '\t'))
1155 lab.e--;
1156 if (lab.e - lab.s == com_end && bp_save == NULL) {
1157 /* comment on preprocessor line */
1158 if (sc_end == NULL) { /* if this is the first comment,
1159 * we must set up the buffer */
1160 save_com = sc_buf;
1161 sc_end = &save_com[0];
1162 } else {
1163 *sc_end++ = '\n'; /* add newline between
1164 * comments */
1165 *sc_end++ = ' ';
1166 --line_no;
1167 }
1168 if (sc_end - save_com + com_end - com_start > sc_size)
1169 errx(1, "input too long");
1170 memmove(sc_end, lab.s + com_start, (size_t)(com_end - com_start));
1171 sc_end += com_end - com_start;
1172 lab.e = lab.s + com_start;
1173 while (lab.e > lab.s && (lab.e[-1] == ' ' || lab.e[-1] == '\t'))
1174 lab.e--;
1175 bp_save = buf_ptr; /* save current input buffer */
1176 be_save = buf_end;
1177 buf_ptr = save_com; /* fix so that subsequent calls to lexi will
1178 * take tokens out of save_com */
1179 *sc_end++ = ' '; /* add trailing blank, just in case */
1180 buf_end = sc_end;
1181 sc_end = NULL;
1182 debug_println("switched buf_ptr to save_com");
1183 }
1184 check_size_label(1);
1185 *lab.e = '\0'; /* null terminate line */
1186 ps.pcase = false;
1187 }
1188
1189 if (strncmp(lab.s, "#if", 3) == 0) { /* also ifdef, ifndef */
1190 if ((size_t)ifdef_level < nitems(state_stack)) {
1191 match_state[ifdef_level].tos = -1;
1192 state_stack[ifdef_level++] = ps;
1193 } else
1194 diag(1, "#if stack overflow");
1195 } else if (strncmp(lab.s, "#el", 3) == 0) { /* else, elif */
1196 if (ifdef_level <= 0)
1197 diag(1, lab.s[3] == 'i' ? "Unmatched #elif" : "Unmatched #else");
1198 else {
1199 match_state[ifdef_level - 1] = ps;
1200 ps = state_stack[ifdef_level - 1];
1201 }
1202 } else if (strncmp(lab.s, "#endif", 6) == 0) {
1203 if (ifdef_level <= 0)
1204 diag(1, "Unmatched #endif");
1205 else
1206 ifdef_level--;
1207 } else {
1208 if (strncmp(lab.s + 1, "pragma", 6) != 0 &&
1209 strncmp(lab.s + 1, "error", 5) != 0 &&
1210 strncmp(lab.s + 1, "line", 4) != 0 &&
1211 strncmp(lab.s + 1, "undef", 5) != 0 &&
1212 strncmp(lab.s + 1, "define", 6) != 0 &&
1213 strncmp(lab.s + 1, "include", 7) != 0) {
1214 diag(1, "Unrecognized cpp directive");
1215 return;
1216 }
1217 }
1218 if (opt.blanklines_around_conditional_compilation) {
1219 postfix_blankline_requested++;
1220 n_real_blanklines = 0;
1221 } else {
1222 postfix_blankline_requested = 0;
1223 prefix_blankline_requested = 0;
1224 }
1225
1226 /*
1227 * subsequent processing of the newline character will cause the line to
1228 * be printed
1229 */
1230 }
1231
1232 static void __attribute__((__noreturn__))
1233 main_loop(void)
1234 {
1235 token_type ttype;
1236 int force_nl; /* when true, code must be broken */
1237 int last_else = false; /* true iff last keyword was an else */
1238 int dec_ind; /* current indentation for declarations */
1239 int di_stack[20]; /* a stack of structure indentation levels */
1240 int tabs_to_var; /* true if using tabs to indent to var name */
1241 int sp_sw; /* when true, we are in the expression of
1242 * if(...), while(...), etc. */
1243 token_type hd_type = end_of_file; /* used to store type of stmt
1244 * for if (...), for (...), etc */
1245 int squest; /* when this is positive, we have seen a ?
1246 * without the matching : in a <c>?<s>:<s>
1247 * construct */
1248 int scase; /* set to true when we see a case, so we will
1249 * know what to do with the following colon */
1250
1251 sp_sw = force_nl = false;
1252 dec_ind = 0;
1253 di_stack[ps.dec_nest = 0] = 0;
1254 scase = false;
1255 squest = 0;
1256 tabs_to_var = 0;
1257
1258 for (;;) { /* this is the main loop. it will go until we
1259 * reach eof */
1260 int comment_buffered = false;
1261
1262 ttype = lexi(&ps); /* Read the next token. The actual characters
1263 * read are stored in "token". */
1264
1265 /*
1266 * The following code moves newlines and comments following an if (),
1267 * while (), else, etc. up to the start of the following stmt to
1268 * a buffer. This allows proper handling of both kinds of brace
1269 * placement (-br, -bl) and cuddling "else" (-ce).
1270 */
1271 search_brace(&ttype, &force_nl, &comment_buffered, &last_else);
1272
1273 if (ttype == end_of_file) {
1274 process_end_of_file();
1275 /* NOTREACHED */
1276 }
1277
1278 if (
1279 ttype != comment &&
1280 ttype != newline &&
1281 ttype != preprocessing &&
1282 ttype != form_feed) {
1283 process_comment_in_code(ttype, &force_nl);
1284
1285 } else if (ttype != comment) /* preserve force_nl through a comment */
1286 force_nl = false; /* cancel forced newline after newline, form
1287 * feed, etc */
1288
1289
1290
1291 /*-----------------------------------------------------*\
1292 | do switch on type of token scanned |
1293 \*-----------------------------------------------------*/
1294 check_size_code(3); /* maximum number of increments of code.e
1295 * before the next check_size_code or
1296 * dump_line() is 2. After that there's the
1297 * final increment for the null character. */
1298 switch (ttype) {
1299
1300 case form_feed:
1301 process_form_feed();
1302 break;
1303
1304 case newline:
1305 process_newline();
1306 break;
1307
1308 case lparen: /* got a '(' or '[' */
1309 process_lparen_or_lbracket(dec_ind, tabs_to_var, sp_sw);
1310 break;
1311
1312 case rparen: /* got a ')' or ']' */
1313 process_rparen_or_rbracket(&sp_sw, &force_nl, hd_type);
1314 break;
1315
1316 case unary_op: /* this could be any unary operation */
1317 process_unary_op(dec_ind, tabs_to_var);
1318 break;
1319
1320 case binary_op: /* any binary operation */
1321 process_binary_op();
1322 break;
1323
1324 case postfix_op: /* got a trailing ++ or -- */
1325 process_postfix_op();
1326 break;
1327
1328 case question: /* got a ? */
1329 process_question(&squest);
1330 break;
1331
1332 case case_label: /* got word 'case' or 'default' */
1333 scase = true; /* so we can process the later colon properly */
1334 goto copy_id;
1335
1336 case colon: /* got a ':' */
1337 process_colon(&squest, &force_nl, &scase);
1338 break;
1339
1340 case semicolon: /* got a ';' */
1341 process_semicolon(&scase, &squest, dec_ind, tabs_to_var, &sp_sw,
1342 hd_type, &force_nl);
1343 break;
1344
1345 case lbrace: /* got a '{' */
1346 process_lbrace(&force_nl, &sp_sw, hd_type, di_stack,
1347 (int)nitems(di_stack), &dec_ind);
1348 break;
1349
1350 case rbrace: /* got a '}' */
1351 process_rbrace(&sp_sw, &dec_ind, di_stack);
1352 break;
1353
1354 case switch_expr: /* got keyword "switch" */
1355 sp_sw = true;
1356 hd_type = switch_expr; /* keep this for when we have seen the
1357 * expression */
1358 goto copy_id; /* go move the token into buffer */
1359
1360 case keyword_for_if_while:
1361 sp_sw = true; /* the interesting stuff is done after the
1362 * expression is scanned */
1363 hd_type = (*token.s == 'i' ? if_expr :
1364 (*token.s == 'w' ? while_expr : for_exprs));
1365
1366 /* remember the type of header for later use by parser */
1367 goto copy_id; /* copy the token into line */
1368
1369 case keyword_do_else:
1370 process_keyword_do_else(&force_nl, &last_else);
1371 goto copy_id; /* move the token into line */
1372
1373 case type_def:
1374 case storage_class:
1375 prefix_blankline_requested = 0;
1376 goto copy_id;
1377
1378 case keyword_struct_union_enum:
1379 if (ps.p_l_follow > 0)
1380 goto copy_id;
1381 /* FALLTHROUGH */
1382 case decl: /* we have a declaration type (int, etc.) */
1383 process_decl(&dec_ind, &tabs_to_var);
1384 goto copy_id;
1385
1386 case funcname:
1387 case ident: /* got an identifier or constant */
1388 process_ident(ttype, dec_ind, tabs_to_var, &sp_sw, &force_nl,
1389 hd_type);
1390 copy_id:
1391 copy_id();
1392 if (ttype != funcname)
1393 ps.want_blank = true;
1394 break;
1395
1396 case string_prefix:
1397 process_string_prefix();
1398 break;
1399
1400 case period:
1401 process_period();
1402 break;
1403
1404 case comma:
1405 process_comma(dec_ind, tabs_to_var, &force_nl);
1406 break;
1407
1408 case preprocessing: /* '#' */
1409 process_preprocessing();
1410 break;
1411 case comment: /* the initial '/' '*' or '//' of a comment */
1412 process_comment();
1413 break;
1414
1415 default:
1416 break;
1417 }
1418
1419 *code.e = '\0';
1420 if (ttype != comment &&
1421 ttype != newline &&
1422 ttype != preprocessing)
1423 ps.last_token = ttype;
1424 }
1425 }
1426
1427 int
1428 main(int argc, char **argv)
1429 {
1430 main_init_globals();
1431 main_parse_command_line(argc, argv);
1432 #if HAVE_CAPSICUM
1433 init_capsicum();
1434 #endif
1435 main_prepare_parsing();
1436 main_loop();
1437 }
1438
1439 /*
1440 * copy input file to backup file if in_name is /blah/blah/blah/file, then
1441 * backup file will be ".Bfile" then make the backup file the input and
1442 * original input file the output
1443 */
1444 static void
1445 bakcopy(void)
1446 {
1447 ssize_t n;
1448 int bakchn;
1449 char buff[8 * 1024];
1450 const char *p;
1451
1452 /* construct file name .Bfile */
1453 for (p = in_name; *p; p++); /* skip to end of string */
1454 while (p > in_name && *p != '/') /* find last '/' */
1455 p--;
1456 if (*p == '/')
1457 p++;
1458 sprintf(bakfile, "%s%s", p, simple_backup_suffix);
1459
1460 /* copy in_name to backup file */
1461 bakchn = creat(bakfile, 0600);
1462 if (bakchn < 0)
1463 err(1, "%s", bakfile);
1464 while ((n = read(fileno(input), buff, sizeof(buff))) > 0)
1465 if (write(bakchn, buff, (size_t)n) != n)
1466 err(1, "%s", bakfile);
1467 if (n < 0)
1468 err(1, "%s", in_name);
1469 close(bakchn);
1470 fclose(input);
1471
1472 /* re-open backup file as the input file */
1473 input = fopen(bakfile, "r");
1474 if (input == NULL)
1475 err(1, "%s", bakfile);
1476 /* now the original input file will be the output */
1477 output = fopen(in_name, "w");
1478 if (output == NULL) {
1479 unlink(bakfile);
1480 err(1, "%s", in_name);
1481 }
1482 }
1483
1484 static void
1485 indent_declaration(int cur_dec_ind, int tabs_to_var)
1486 {
1487 int pos = (int)(code.e - code.s);
1488 char *startpos = code.e;
1489
1490 /*
1491 * get the tab math right for indentations that are not multiples of tabsize
1492 */
1493 if ((ps.ind_level * opt.indent_size) % opt.tabsize != 0) {
1494 pos += (ps.ind_level * opt.indent_size) % opt.tabsize;
1495 cur_dec_ind += (ps.ind_level * opt.indent_size) % opt.tabsize;
1496 }
1497 if (tabs_to_var) {
1498 int tpos;
1499
1500 check_size_code((size_t)(cur_dec_ind / opt.tabsize));
1501 while ((tpos = opt.tabsize * (1 + pos / opt.tabsize)) <= cur_dec_ind) {
1502 *code.e++ = '\t';
1503 pos = tpos;
1504 }
1505 }
1506 check_size_code((size_t)(cur_dec_ind - pos + 1));
1507 while (pos < cur_dec_ind) {
1508 *code.e++ = ' ';
1509 pos++;
1510 }
1511 if (code.e == startpos && ps.want_blank) {
1512 *code.e++ = ' ';
1513 ps.want_blank = false;
1514 }
1515 }
1516
1517 #ifdef debug
1518 void
1519 debug_printf(const char *fmt, ...)
1520 {
1521 FILE *f = output == stdout ? stderr : stdout;
1522 va_list ap;
1523
1524 va_start(ap, fmt);
1525 vfprintf(f, fmt, ap);
1526 va_end(ap);
1527 }
1528
1529 void
1530 debug_println(const char *fmt, ...)
1531 {
1532 FILE *f = output == stdout ? stderr : stdout;
1533 va_list ap;
1534
1535 va_start(ap, fmt);
1536 vfprintf(f, fmt, ap);
1537 va_end(ap);
1538 fprintf(f, "\n");
1539 }
1540
1541 void
1542 debug_vis_range(const char *prefix, const char *s, const char *e,
1543 const char *suffix)
1544 {
1545 debug_printf("%s", prefix);
1546 for (const char *p = s; p < e; p++) {
1547 if (isprint((unsigned char)*p) && *p != '\\' && *p != '"')
1548 debug_printf("%c", *p);
1549 else if (*p == '\n')
1550 debug_printf("\\n");
1551 else if (*p == '\t')
1552 debug_printf("\\t");
1553 else
1554 debug_printf("\\x%02x", *p);
1555 }
1556 debug_printf("%s", suffix);
1557 }
1558 #endif
1559
1560 static void *
1561 nonnull(void *p)
1562 {
1563 if (p == NULL)
1564 err(EXIT_FAILURE, NULL);
1565 return p;
1566 }
1567
1568 void *
1569 xmalloc(size_t size)
1570 {
1571 return nonnull(malloc(size));
1572 }
1573
1574 void *
1575 xrealloc(void *p, size_t new_size)
1576 {
1577 return nonnull(realloc(p, new_size));
1578 }
1579
1580 char *
1581 xstrdup(const char *s)
1582 {
1583 return nonnull(strdup(s));
1584 }
1585