1 /*-
2 * SPDX-License-Identifier: BSD-4-Clause
3 *
4 * Copyright (c) 1985 Sun Microsystems, Inc.
5 * Copyright (c) 1976 Board of Trustees of the University of Illinois.
6 * Copyright (c) 1980, 1993
7 * The Regents of the University of California. All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)indent.c 5.17 (Berkeley) 6/7/93
34 * $FreeBSD: head/usr.bin/indent/indent.c 340138 2018-11-04 19:24:49Z oshogbo $
35 */
36
37 #include <sys/param.h>
38 #include <err.h>
39 #include <errno.h>
40 #include <fcntl.h>
41 #include <unistd.h>
42 #include <stdio.h>
43 #include <stdlib.h>
44 #include <string.h>
45 #include <ctype.h>
46 #include "indent_globs.h"
47 #include "indent_codes.h"
48 #include "indent.h"
49
50 /* Globals */
51 FILE *input, *output;
52 char *labbuf, *s_lab, *e_lab, *l_lab;
53 char *codebuf, *s_code, *e_code, *l_code;
54 char *combuf, *s_com, *e_com, *l_com;
55 char *tokenbuf, *s_token, *e_token, *l_token;
56 char *in_buffer, *in_buffer_limit;
57 char *buf_ptr, *buf_end;
58
59 char sc_buf[sc_size];
60
61 char *save_com, *sc_end;
62 char *bp_save;
63 char *be_save;
64
65 struct options opt;
66 int line_no;
67
68 struct parser_state ps;
69 int ifdef_level;
70 struct parser_state state_stack[5];
71 struct parser_state match_state[5];
72
73
74 static void bakcopy(void);
75 static void indent_declaration(int, int);
76
77 const char *in_name = "Standard Input"; /* will always point to name of input
78 * file */
79 const char *out_name = "Standard Output"; /* will always point to name
80 * of output file */
81 const char *simple_backup_suffix = ".BAK"; /* Suffix to use for backup
82 * files */
83 char bakfile[MAXPATHLEN] = "";
84
85 int
main(int argc,char ** argv)86 main(int argc, char **argv)
87 {
88 int dec_ind; /* current indentation for declarations */
89 int di_stack[20]; /* a stack of structure indentation levels */
90 int force_nl; /* when true, code must be broken */
91 int hd_type = 0; /* used to store type of stmt for if (...),
92 * for (...), etc */
93 int i; /* local loop counter */
94 int scase; /* set to true when we see a case, so we will
95 * know what to do with the following colon */
96 int sp_sw; /* when true, we are in the expression of
97 * if(...), while(...), etc. */
98 int squest; /* when this is positive, we have seen a ?
99 * without the matching : in a <c>?<s>:<s>
100 * construct */
101 const char *t_ptr; /* used for copying tokens */
102 int tabs_to_var; /* true if using tabs to indent to var name */
103 int type_code; /* the type of token, returned by lexi */
104
105 int last_else = 0; /* true iff last keyword was an else */
106 const char *profile_name = NULL;
107 const char *envval = NULL;
108 struct parser_state transient_state; /* a copy for lookup */
109
110 /*-----------------------------------------------*\
111 | INITIALIZATION |
112 \*-----------------------------------------------*/
113
114 found_err = 0;
115
116 ps.p_stack[0] = stmt; /* this is the parser's stack */
117 ps.last_nl = true; /* this is true if the last thing scanned was
118 * a newline */
119 ps.last_token = semicolon;
120 combuf = (char *) malloc(bufsize);
121 if (combuf == NULL)
122 err(1, NULL);
123 labbuf = (char *) malloc(bufsize);
124 if (labbuf == NULL)
125 err(1, NULL);
126 codebuf = (char *) malloc(bufsize);
127 if (codebuf == NULL)
128 err(1, NULL);
129 tokenbuf = (char *) malloc(bufsize);
130 if (tokenbuf == NULL)
131 err(1, NULL);
132 alloc_typenames();
133 init_constant_tt();
134 l_com = combuf + bufsize - 5;
135 l_lab = labbuf + bufsize - 5;
136 l_code = codebuf + bufsize - 5;
137 l_token = tokenbuf + bufsize - 5;
138 combuf[0] = codebuf[0] = labbuf[0] = ' '; /* set up code, label, and
139 * comment buffers */
140 combuf[1] = codebuf[1] = labbuf[1] = '\0';
141 opt.else_if = 1; /* Default else-if special processing to on */
142 s_lab = e_lab = labbuf + 1;
143 s_code = e_code = codebuf + 1;
144 s_com = e_com = combuf + 1;
145 s_token = e_token = tokenbuf + 1;
146
147 in_buffer = (char *) malloc(10);
148 if (in_buffer == NULL)
149 err(1, NULL);
150 in_buffer_limit = in_buffer + 8;
151 buf_ptr = buf_end = in_buffer;
152 line_no = 1;
153 had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
154 sp_sw = force_nl = false;
155 ps.in_or_st = false;
156 ps.bl_line = true;
157 dec_ind = 0;
158 di_stack[ps.dec_nest = 0] = 0;
159 ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
160
161 scase = ps.pcase = false;
162 squest = 0;
163 sc_end = NULL;
164 bp_save = NULL;
165 be_save = NULL;
166
167 output = NULL;
168 tabs_to_var = 0;
169
170 envval = getenv("SIMPLE_BACKUP_SUFFIX");
171 if (envval)
172 simple_backup_suffix = envval;
173
174 /*--------------------------------------------------*\
175 | COMMAND LINE SCAN |
176 \*--------------------------------------------------*/
177
178 #ifdef undef
179 max_col = 78; /* -l78 */
180 lineup_to_parens = 1; /* -lp */
181 lineup_to_parens_always = 0; /* -nlpl */
182 ps.ljust_decl = 0; /* -ndj */
183 ps.com_ind = 33; /* -c33 */
184 star_comment_cont = 1; /* -sc */
185 ps.ind_size = 8; /* -i8 */
186 verbose = 0;
187 ps.decl_indent = 16; /* -di16 */
188 ps.local_decl_indent = -1; /* if this is not set to some nonnegative value
189 * by an arg, we will set this equal to
190 * ps.decl_ind */
191 ps.indent_parameters = 1; /* -ip */
192 ps.decl_com_ind = 0; /* if this is not set to some positive value
193 * by an arg, we will set this equal to
194 * ps.com_ind */
195 btype_2 = 1; /* -br */
196 cuddle_else = 1; /* -ce */
197 ps.unindent_displace = 0; /* -d0 */
198 ps.case_indent = 0; /* -cli0 */
199 format_block_comments = 1; /* -fcb */
200 format_col1_comments = 1; /* -fc1 */
201 procnames_start_line = 1; /* -psl */
202 proc_calls_space = 0; /* -npcs */
203 comment_delimiter_on_blankline = 1; /* -cdb */
204 ps.leave_comma = 1; /* -nbc */
205 #endif
206
207 for (i = 1; i < argc; ++i)
208 if (strcmp(argv[i], "-npro") == 0)
209 break;
210 else if (argv[i][0] == '-' && argv[i][1] == 'P' && argv[i][2] != '\0')
211 profile_name = argv[i]; /* non-empty -P (set profile) */
212 set_defaults();
213 if (i >= argc)
214 set_profile(profile_name);
215
216 for (i = 1; i < argc; ++i) {
217
218 /*
219 * look thru args (if any) for changes to defaults
220 */
221 if (argv[i][0] != '-') { /* no flag on parameter */
222 if (input == NULL) { /* we must have the input file */
223 in_name = argv[i]; /* remember name of input file */
224 input = fopen(in_name, "r");
225 if (input == NULL) /* check for open error */
226 err(1, "%s", in_name);
227 continue;
228 }
229 else if (output == NULL) { /* we have the output file */
230 out_name = argv[i]; /* remember name of output file */
231 if (strcmp(in_name, out_name) == 0) { /* attempt to overwrite
232 * the file */
233 errx(1, "input and output files must be different");
234 }
235 output = fopen(out_name, "w");
236 if (output == NULL) /* check for create error */
237 err(1, "%s", out_name);
238 continue;
239 }
240 errx(1, "unknown parameter: %s", argv[i]);
241 }
242 else
243 set_option(argv[i]);
244 } /* end of for */
245 if (input == NULL)
246 input = stdin;
247 if (output == NULL) {
248 if (input == stdin)
249 output = stdout;
250 else {
251 out_name = in_name;
252 bakcopy();
253 }
254 }
255
256 if (opt.com_ind <= 1)
257 opt.com_ind = 2; /* don't put normal comments before column 2 */
258 if (opt.block_comment_max_col <= 0)
259 opt.block_comment_max_col = opt.max_col;
260 if (opt.local_decl_indent < 0) /* if not specified by user, set this */
261 opt.local_decl_indent = opt.decl_indent;
262 if (opt.decl_com_ind <= 0) /* if not specified by user, set this */
263 opt.decl_com_ind = opt.ljust_decl ? (opt.com_ind <= 10 ? 2 : opt.com_ind - 8) : opt.com_ind;
264 if (opt.continuation_indent == 0)
265 opt.continuation_indent = opt.ind_size;
266 fill_buffer(); /* get first batch of stuff into input buffer */
267
268 parse(semicolon);
269 {
270 char *p = buf_ptr;
271 int col = 1;
272
273 while (1) {
274 if (*p == ' ')
275 col++;
276 else if (*p == '\t')
277 col = opt.tabsize * (1 + (col - 1) / opt.tabsize) + 1;
278 else
279 break;
280 p++;
281 }
282 if (col > opt.ind_size)
283 ps.ind_level = ps.i_l_follow = col / opt.ind_size;
284 }
285
286 /*
287 * START OF MAIN LOOP
288 */
289
290 while (1) { /* this is the main loop. it will go until we
291 * reach eof */
292 int comment_buffered = false;
293
294 type_code = lexi(&ps); /* lexi reads one token. The actual
295 * characters read are stored in "token". lexi
296 * returns a code indicating the type of token */
297
298 /*
299 * The following code moves newlines and comments following an if (),
300 * while (), else, etc. up to the start of the following stmt to
301 * a buffer. This allows proper handling of both kinds of brace
302 * placement (-br, -bl) and cuddling "else" (-ce).
303 */
304
305 while (ps.search_brace) {
306 switch (type_code) {
307 case newline:
308 if (sc_end == NULL) {
309 save_com = sc_buf;
310 save_com[0] = save_com[1] = ' ';
311 sc_end = &save_com[2];
312 }
313 *sc_end++ = '\n';
314 /*
315 * We may have inherited a force_nl == true from the previous
316 * token (like a semicolon). But once we know that a newline
317 * has been scanned in this loop, force_nl should be false.
318 *
319 * However, the force_nl == true must be preserved if newline
320 * is never scanned in this loop, so this assignment cannot be
321 * done earlier.
322 */
323 force_nl = false;
324 case form_feed:
325 break;
326 case comment:
327 if (sc_end == NULL) {
328 /*
329 * Copy everything from the start of the line, because
330 * pr_comment() will use that to calculate original
331 * indentation of a boxed comment.
332 */
333 memcpy(sc_buf, in_buffer, buf_ptr - in_buffer - 4);
334 save_com = sc_buf + (buf_ptr - in_buffer - 4);
335 save_com[0] = save_com[1] = ' ';
336 sc_end = &save_com[2];
337 }
338 comment_buffered = true;
339 *sc_end++ = '/'; /* copy in start of comment */
340 *sc_end++ = '*';
341 for (;;) { /* loop until we get to the end of the comment */
342 *sc_end = *buf_ptr++;
343 if (buf_ptr >= buf_end)
344 fill_buffer();
345 if (*sc_end++ == '*' && *buf_ptr == '/')
346 break; /* we are at end of comment */
347 if (sc_end >= &save_com[sc_size]) { /* check for temp buffer
348 * overflow */
349 diag2(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
350 fflush(output);
351 exit(1);
352 }
353 }
354 /* FALLTHROUGH */
355 *sc_end++ = '/'; /* add ending slash */
356 if (++buf_ptr >= buf_end) /* get past / in buffer */
357 fill_buffer();
358 break;
359 case lbrace:
360 /*
361 * Put KNF-style lbraces before the buffered up tokens and
362 * jump out of this loop in order to avoid copying the token
363 * again under the default case of the switch below.
364 */
365 if (sc_end != NULL && opt.btype_2) {
366 save_com[0] = '{';
367 /*
368 * Originally the lbrace may have been alone on its own
369 * line, but it will be moved into "the else's line", so
370 * if there was a newline resulting from the "{" before,
371 * it must be scanned now and ignored.
372 */
373 while (isspace((unsigned char)*buf_ptr)) {
374 if (++buf_ptr >= buf_end)
375 fill_buffer();
376 if (*buf_ptr == '\n')
377 break;
378 }
379 goto sw_buffer;
380 }
381 /* FALLTHROUGH */
382 default: /* it is the start of a normal statement */
383 {
384 int remove_newlines;
385
386 remove_newlines =
387 /* "} else" */
388 (type_code == sp_nparen && *token == 'e' &&
389 e_code != s_code && e_code[-1] == '}')
390 /* "else if" */
391 || (type_code == sp_paren && *token == 'i' &&
392 last_else && opt.else_if);
393 if (remove_newlines)
394 force_nl = false;
395 if (sc_end == NULL) { /* ignore buffering if
396 * comment wasn't saved up */
397 ps.search_brace = false;
398 goto check_type;
399 }
400 while (sc_end > save_com && isblank((unsigned char)sc_end[-1])) {
401 sc_end--;
402 }
403 if (opt.swallow_optional_blanklines ||
404 (!comment_buffered && remove_newlines)) {
405 force_nl = !remove_newlines;
406 while (sc_end > save_com && sc_end[-1] == '\n') {
407 sc_end--;
408 }
409 }
410 if (force_nl) { /* if we should insert a nl here, put
411 * it into the buffer */
412 force_nl = false;
413 --line_no; /* this will be re-increased when the
414 * newline is read from the buffer */
415 *sc_end++ = '\n';
416 *sc_end++ = ' ';
417 if (opt.verbose) /* print error msg if the line was
418 * not already broken */
419 diag2(0, "Line broken");
420 }
421 for (t_ptr = token; *t_ptr; ++t_ptr)
422 *sc_end++ = *t_ptr;
423
424 sw_buffer:
425 ps.search_brace = false; /* stop looking for start of
426 * stmt */
427 bp_save = buf_ptr; /* save current input buffer */
428 be_save = buf_end;
429 buf_ptr = save_com; /* fix so that subsequent calls to
430 * lexi will take tokens out of
431 * save_com */
432 *sc_end++ = ' ';/* add trailing blank, just in case */
433 buf_end = sc_end;
434 sc_end = NULL;
435 break;
436 }
437 } /* end of switch */
438 /*
439 * We must make this check, just in case there was an unexpected
440 * EOF.
441 */
442 if (type_code != 0) {
443 /*
444 * The only intended purpose of calling lexi() below is to
445 * categorize the next token in order to decide whether to
446 * continue buffering forthcoming tokens. Once the buffering
447 * is over, lexi() will be called again elsewhere on all of
448 * the tokens - this time for normal processing.
449 *
450 * Calling it for this purpose is a bug, because lexi() also
451 * changes the parser state and discards leading whitespace,
452 * which is needed mostly for comment-related considerations.
453 *
454 * Work around the former problem by giving lexi() a copy of
455 * the current parser state and discard it if the call turned
456 * out to be just a look ahead.
457 *
458 * Work around the latter problem by copying all whitespace
459 * characters into the buffer so that the later lexi() call
460 * will read them.
461 */
462 if (sc_end != NULL) {
463 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
464 *sc_end++ = *buf_ptr++;
465 if (sc_end >= &save_com[sc_size]) {
466 errx(1, "input too long");
467 }
468 }
469 if (buf_ptr >= buf_end) {
470 fill_buffer();
471 }
472 }
473 transient_state = ps;
474 type_code = lexi(&transient_state); /* read another token */
475 if (type_code != newline && type_code != form_feed &&
476 type_code != comment && !transient_state.search_brace) {
477 ps = transient_state;
478 }
479 }
480 } /* end of while (search_brace) */
481 last_else = 0;
482 check_type:
483 if (type_code == 0) { /* we got eof */
484 if (s_lab != e_lab || s_code != e_code
485 || s_com != e_com) /* must dump end of line */
486 dump_line();
487 if (ps.tos > 1) /* check for balanced braces */
488 diag2(1, "Stuff missing from end of file");
489
490 if (opt.verbose) {
491 printf("There were %d output lines and %d comments\n",
492 ps.out_lines, ps.out_coms);
493 printf("(Lines with comments)/(Lines with code): %6.3f\n",
494 (1.0 * ps.com_lines) / code_lines);
495 }
496 fflush(output);
497 exit(found_err);
498 }
499 if (
500 (type_code != comment) &&
501 (type_code != newline) &&
502 (type_code != preesc) &&
503 (type_code != form_feed)) {
504 if (force_nl &&
505 (type_code != semicolon) &&
506 (type_code != lbrace || !opt.btype_2)) {
507 /* we should force a broken line here */
508 if (opt.verbose)
509 diag2(0, "Line broken");
510 dump_line();
511 ps.want_blank = false; /* dont insert blank at line start */
512 force_nl = false;
513 }
514 ps.in_stmt = true; /* turn on flag which causes an extra level of
515 * indentation. this is turned off by a ; or
516 * '}' */
517 if (s_com != e_com) { /* the turkey has embedded a comment
518 * in a line. fix it */
519 int len = e_com - s_com;
520
521 CHECK_SIZE_CODE(len + 3);
522 *e_code++ = ' ';
523 memcpy(e_code, s_com, len);
524 e_code += len;
525 *e_code++ = ' ';
526 *e_code = '\0'; /* null terminate code sect */
527 ps.want_blank = false;
528 e_com = s_com;
529 }
530 }
531 else if (type_code != comment) /* preserve force_nl thru a comment */
532 force_nl = false; /* cancel forced newline after newline, form
533 * feed, etc */
534
535
536
537 /*-----------------------------------------------------*\
538 | do switch on type of token scanned |
539 \*-----------------------------------------------------*/
540 CHECK_SIZE_CODE(3); /* maximum number of increments of e_code
541 * before the next CHECK_SIZE_CODE or
542 * dump_line() is 2. After that there's the
543 * final increment for the null character. */
544 switch (type_code) { /* now, decide what to do with the token */
545
546 case form_feed: /* found a form feed in line */
547 ps.use_ff = true; /* a form feed is treated much like a newline */
548 dump_line();
549 ps.want_blank = false;
550 break;
551
552 case newline:
553 if (ps.last_token != comma || ps.p_l_follow > 0
554 || !opt.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
555 dump_line();
556 ps.want_blank = false;
557 }
558 ++line_no; /* keep track of input line number */
559 break;
560
561 case lparen: /* got a '(' or '[' */
562 /* count parens to make Healy happy */
563 if (++ps.p_l_follow == nitems(ps.paren_indents)) {
564 diag3(0, "Reached internal limit of %d unclosed parens",
565 nitems(ps.paren_indents));
566 ps.p_l_follow--;
567 }
568 if (*token == '[')
569 /* not a function pointer declaration or a function call */;
570 else if (ps.in_decl && !ps.block_init && !ps.dumped_decl_indent &&
571 ps.procname[0] == '\0' && ps.paren_level == 0) {
572 /* function pointer declarations */
573 indent_declaration(dec_ind, tabs_to_var);
574 ps.dumped_decl_indent = true;
575 }
576 else if (ps.want_blank &&
577 ((ps.last_token != ident && ps.last_token != funcname) ||
578 opt.proc_calls_space ||
579 /* offsetof (1) is never allowed a space; sizeof (2) gets
580 * one iff -bs; all other keywords (>2) always get a space
581 * before lparen */
582 ps.keyword + opt.Bill_Shannon > 2))
583 *e_code++ = ' ';
584 ps.want_blank = false;
585 *e_code++ = token[0];
586 ps.paren_indents[ps.p_l_follow - 1] = count_spaces_until(1, s_code, e_code) - 1;
587 if (sp_sw && ps.p_l_follow == 1 && opt.extra_expression_indent
588 && ps.paren_indents[0] < 2 * opt.ind_size)
589 ps.paren_indents[0] = 2 * opt.ind_size;
590 if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
591 /*
592 * this is a kluge to make sure that declarations will be
593 * aligned right if proc decl has an explicit type on it, i.e.
594 * "int a(x) {..."
595 */
596 parse(semicolon); /* I said this was a kluge... */
597 ps.in_or_st = false; /* turn off flag for structure decl or
598 * initialization */
599 }
600 /* parenthesized type following sizeof or offsetof is not a cast */
601 if (ps.keyword == 1 || ps.keyword == 2)
602 ps.not_cast_mask |= 1 << ps.p_l_follow;
603 break;
604
605 case rparen: /* got a ')' or ']' */
606 if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.not_cast_mask) {
607 ps.last_u_d = true;
608 ps.cast_mask &= (1 << ps.p_l_follow) - 1;
609 ps.want_blank = opt.space_after_cast;
610 } else
611 ps.want_blank = true;
612 ps.not_cast_mask &= (1 << ps.p_l_follow) - 1;
613 if (--ps.p_l_follow < 0) {
614 ps.p_l_follow = 0;
615 diag3(0, "Extra %c", *token);
616 }
617 if (e_code == s_code) /* if the paren starts the line */
618 ps.paren_level = ps.p_l_follow; /* then indent it */
619
620 *e_code++ = token[0];
621
622 if (sp_sw && (ps.p_l_follow == 0)) { /* check for end of if
623 * (...), or some such */
624 sp_sw = false;
625 force_nl = true;/* must force newline after if */
626 ps.last_u_d = true; /* inform lexi that a following
627 * operator is unary */
628 ps.in_stmt = false; /* dont use stmt continuation
629 * indentation */
630
631 parse(hd_type); /* let parser worry about if, or whatever */
632 }
633 ps.search_brace = opt.btype_2; /* this should ensure that
634 * constructs such as main(){...}
635 * and int[]{...} have their braces
636 * put in the right place */
637 break;
638
639 case unary_op: /* this could be any unary operation */
640 if (!ps.dumped_decl_indent && ps.in_decl && !ps.block_init &&
641 ps.procname[0] == '\0' && ps.paren_level == 0) {
642 /* pointer declarations */
643
644 /*
645 * if this is a unary op in a declaration, we should indent
646 * this token
647 */
648 for (i = 0; token[i]; ++i)
649 /* find length of token */;
650 indent_declaration(dec_ind - i, tabs_to_var);
651 ps.dumped_decl_indent = true;
652 }
653 else if (ps.want_blank)
654 *e_code++ = ' ';
655
656 {
657 int len = e_token - s_token;
658
659 CHECK_SIZE_CODE(len);
660 memcpy(e_code, token, len);
661 e_code += len;
662 }
663 ps.want_blank = false;
664 break;
665
666 case binary_op: /* any binary operation */
667 {
668 int len = e_token - s_token;
669
670 CHECK_SIZE_CODE(len + 1);
671 if (ps.want_blank)
672 *e_code++ = ' ';
673 memcpy(e_code, token, len);
674 e_code += len;
675 }
676 ps.want_blank = true;
677 break;
678
679 case postop: /* got a trailing ++ or -- */
680 *e_code++ = token[0];
681 *e_code++ = token[1];
682 ps.want_blank = true;
683 break;
684
685 case question: /* got a ? */
686 squest++; /* this will be used when a later colon
687 * appears so we can distinguish the
688 * <c>?<n>:<n> construct */
689 if (ps.want_blank)
690 *e_code++ = ' ';
691 *e_code++ = '?';
692 ps.want_blank = true;
693 break;
694
695 case casestmt: /* got word 'case' or 'default' */
696 scase = true; /* so we can process the later colon properly */
697 goto copy_id;
698
699 case colon: /* got a ':' */
700 if (squest > 0) { /* it is part of the <c>?<n>: <n> construct */
701 --squest;
702 if (ps.want_blank)
703 *e_code++ = ' ';
704 *e_code++ = ':';
705 ps.want_blank = true;
706 break;
707 }
708 if (ps.in_or_st) {
709 *e_code++ = ':';
710 ps.want_blank = false;
711 break;
712 }
713 ps.in_stmt = false; /* seeing a label does not imply we are in a
714 * stmt */
715 /*
716 * turn everything so far into a label
717 */
718 {
719 int len = e_code - s_code;
720
721 CHECK_SIZE_LAB(len + 3);
722 memcpy(e_lab, s_code, len);
723 e_lab += len;
724 *e_lab++ = ':';
725 *e_lab = '\0';
726 e_code = s_code;
727 }
728 force_nl = ps.pcase = scase; /* ps.pcase will be used by
729 * dump_line to decide how to
730 * indent the label. force_nl
731 * will force a case n: to be
732 * on a line by itself */
733 scase = false;
734 ps.want_blank = false;
735 break;
736
737 case semicolon: /* got a ';' */
738 if (ps.dec_nest == 0)
739 ps.in_or_st = false;/* we are not in an initialization or
740 * structure declaration */
741 scase = false; /* these will only need resetting in an error */
742 squest = 0;
743 if (ps.last_token == rparen)
744 ps.in_parameter_declaration = 0;
745 ps.cast_mask = 0;
746 ps.not_cast_mask = 0;
747 ps.block_init = 0;
748 ps.block_init_level = 0;
749 ps.just_saw_decl--;
750
751 if (ps.in_decl && s_code == e_code && !ps.block_init &&
752 !ps.dumped_decl_indent && ps.paren_level == 0) {
753 /* indent stray semicolons in declarations */
754 indent_declaration(dec_ind - 1, tabs_to_var);
755 ps.dumped_decl_indent = true;
756 }
757
758 ps.in_decl = (ps.dec_nest > 0); /* if we were in a first level
759 * structure declaration, we
760 * arent any more */
761
762 if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
763
764 /*
765 * This should be true iff there were unbalanced parens in the
766 * stmt. It is a bit complicated, because the semicolon might
767 * be in a for stmt
768 */
769 diag2(1, "Unbalanced parens");
770 ps.p_l_follow = 0;
771 if (sp_sw) { /* this is a check for an if, while, etc. with
772 * unbalanced parens */
773 sp_sw = false;
774 parse(hd_type); /* dont lose the if, or whatever */
775 }
776 }
777 *e_code++ = ';';
778 ps.want_blank = true;
779 ps.in_stmt = (ps.p_l_follow > 0); /* we are no longer in the
780 * middle of a stmt */
781
782 if (!sp_sw) { /* if not if for (;;) */
783 parse(semicolon); /* let parser know about end of stmt */
784 force_nl = true;/* force newline after an end of stmt */
785 }
786 break;
787
788 case lbrace: /* got a '{' */
789 ps.in_stmt = false; /* dont indent the {} */
790 if (!ps.block_init)
791 force_nl = true;/* force other stuff on same line as '{' onto
792 * new line */
793 else if (ps.block_init_level <= 0)
794 ps.block_init_level = 1;
795 else
796 ps.block_init_level++;
797
798 if (s_code != e_code && !ps.block_init) {
799 if (!opt.btype_2) {
800 dump_line();
801 ps.want_blank = false;
802 }
803 else if (ps.in_parameter_declaration && !ps.in_or_st) {
804 ps.i_l_follow = 0;
805 if (opt.function_brace_split) { /* dump the line prior
806 * to the brace ... */
807 dump_line();
808 ps.want_blank = false;
809 } else /* add a space between the decl and brace */
810 ps.want_blank = true;
811 }
812 }
813 if (ps.in_parameter_declaration)
814 prefix_blankline_requested = 0;
815
816 if (ps.p_l_follow > 0) { /* check for preceding unbalanced
817 * parens */
818 diag2(1, "Unbalanced parens");
819 ps.p_l_follow = 0;
820 if (sp_sw) { /* check for unclosed if, for, etc. */
821 sp_sw = false;
822 parse(hd_type);
823 ps.ind_level = ps.i_l_follow;
824 }
825 }
826 if (s_code == e_code)
827 ps.ind_stmt = false; /* dont put extra indentation on line
828 * with '{' */
829 if (ps.in_decl && ps.in_or_st) { /* this is either a structure
830 * declaration or an init */
831 di_stack[ps.dec_nest] = dec_ind;
832 if (++ps.dec_nest == nitems(di_stack)) {
833 diag3(0, "Reached internal limit of %d struct levels",
834 nitems(di_stack));
835 ps.dec_nest--;
836 }
837 /* ? dec_ind = 0; */
838 }
839 else {
840 ps.decl_on_line = false; /* we can't be in the middle of
841 * a declaration, so don't do
842 * special indentation of
843 * comments */
844 if (opt.blanklines_after_declarations_at_proctop
845 && ps.in_parameter_declaration)
846 postfix_blankline_requested = 1;
847 ps.in_parameter_declaration = 0;
848 ps.in_decl = false;
849 }
850 dec_ind = 0;
851 parse(lbrace); /* let parser know about this */
852 if (ps.want_blank) /* put a blank before '{' if '{' is not at
853 * start of line */
854 *e_code++ = ' ';
855 ps.want_blank = false;
856 *e_code++ = '{';
857 ps.just_saw_decl = 0;
858 break;
859
860 case rbrace: /* got a '}' */
861 if (ps.p_stack[ps.tos] == decl && !ps.block_init) /* semicolons can be
862 * omitted in
863 * declarations */
864 parse(semicolon);
865 if (ps.p_l_follow) {/* check for unclosed if, for, else. */
866 diag2(1, "Unbalanced parens");
867 ps.p_l_follow = 0;
868 sp_sw = false;
869 }
870 ps.just_saw_decl = 0;
871 ps.block_init_level--;
872 if (s_code != e_code && !ps.block_init) { /* '}' must be first on
873 * line */
874 if (opt.verbose)
875 diag2(0, "Line broken");
876 dump_line();
877 }
878 *e_code++ = '}';
879 ps.want_blank = true;
880 ps.in_stmt = ps.ind_stmt = false;
881 if (ps.dec_nest > 0) { /* we are in multi-level structure
882 * declaration */
883 dec_ind = di_stack[--ps.dec_nest];
884 if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
885 ps.just_saw_decl = 2;
886 ps.in_decl = true;
887 }
888 prefix_blankline_requested = 0;
889 parse(rbrace); /* let parser know about this */
890 ps.search_brace = opt.cuddle_else && ps.p_stack[ps.tos] == ifhead
891 && ps.il[ps.tos] >= ps.ind_level;
892 if (ps.tos <= 1 && opt.blanklines_after_procs && ps.dec_nest <= 0)
893 postfix_blankline_requested = 1;
894 break;
895
896 case swstmt: /* got keyword "switch" */
897 sp_sw = true;
898 hd_type = swstmt; /* keep this for when we have seen the
899 * expression */
900 goto copy_id; /* go move the token into buffer */
901
902 case sp_paren: /* token is if, while, for */
903 sp_sw = true; /* the interesting stuff is done after the
904 * expression is scanned */
905 hd_type = (*token == 'i' ? ifstmt :
906 (*token == 'w' ? whilestmt : forstmt));
907
908 /*
909 * remember the type of header for later use by parser
910 */
911 goto copy_id; /* copy the token into line */
912
913 case sp_nparen: /* got else, do */
914 ps.in_stmt = false;
915 if (*token == 'e') {
916 if (e_code != s_code && (!opt.cuddle_else || e_code[-1] != '}')) {
917 if (opt.verbose)
918 diag2(0, "Line broken");
919 dump_line();/* make sure this starts a line */
920 ps.want_blank = false;
921 }
922 force_nl = true;/* also, following stuff must go onto new line */
923 last_else = 1;
924 parse(elselit);
925 }
926 else {
927 if (e_code != s_code) { /* make sure this starts a line */
928 if (opt.verbose)
929 diag2(0, "Line broken");
930 dump_line();
931 ps.want_blank = false;
932 }
933 force_nl = true;/* also, following stuff must go onto new line */
934 last_else = 0;
935 parse(dolit);
936 }
937 goto copy_id; /* move the token into line */
938
939 case type_def:
940 case storage:
941 prefix_blankline_requested = 0;
942 goto copy_id;
943
944 case structure:
945 if (ps.p_l_follow > 0)
946 goto copy_id;
947 /* FALLTHROUGH */
948 case decl: /* we have a declaration type (int, etc.) */
949 parse(decl); /* let parser worry about indentation */
950 if (ps.last_token == rparen && ps.tos <= 1) {
951 if (s_code != e_code) {
952 dump_line();
953 ps.want_blank = 0;
954 }
955 }
956 if (ps.in_parameter_declaration && opt.indent_parameters && ps.dec_nest == 0) {
957 ps.ind_level = ps.i_l_follow = 1;
958 ps.ind_stmt = 0;
959 }
960 ps.in_or_st = true; /* this might be a structure or initialization
961 * declaration */
962 ps.in_decl = ps.decl_on_line = ps.last_token != type_def;
963 if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
964 ps.just_saw_decl = 2;
965 prefix_blankline_requested = 0;
966 for (i = 0; token[i++];); /* get length of token */
967
968 if (ps.ind_level == 0 || ps.dec_nest > 0) {
969 /* global variable or struct member in local variable */
970 dec_ind = opt.decl_indent > 0 ? opt.decl_indent : i;
971 tabs_to_var = (opt.use_tabs ? opt.decl_indent > 0 : 0);
972 } else {
973 /* local variable */
974 dec_ind = opt.local_decl_indent > 0 ? opt.local_decl_indent : i;
975 tabs_to_var = (opt.use_tabs ? opt.local_decl_indent > 0 : 0);
976 }
977 goto copy_id;
978
979 case funcname:
980 case ident: /* got an identifier or constant */
981 if (ps.in_decl) {
982 if (type_code == funcname) {
983 ps.in_decl = false;
984 if (opt.procnames_start_line && s_code != e_code) {
985 *e_code = '\0';
986 dump_line();
987 }
988 else if (ps.want_blank) {
989 *e_code++ = ' ';
990 }
991 ps.want_blank = false;
992 }
993 else if (!ps.block_init && !ps.dumped_decl_indent &&
994 ps.paren_level == 0) { /* if we are in a declaration, we
995 * must indent identifier */
996 indent_declaration(dec_ind, tabs_to_var);
997 ps.dumped_decl_indent = true;
998 ps.want_blank = false;
999 }
1000 }
1001 else if (sp_sw && ps.p_l_follow == 0) {
1002 sp_sw = false;
1003 force_nl = true;
1004 ps.last_u_d = true;
1005 ps.in_stmt = false;
1006 parse(hd_type);
1007 }
1008 copy_id:
1009 {
1010 int len = e_token - s_token;
1011
1012 CHECK_SIZE_CODE(len + 1);
1013 if (ps.want_blank)
1014 *e_code++ = ' ';
1015 memcpy(e_code, s_token, len);
1016 e_code += len;
1017 }
1018 if (type_code != funcname)
1019 ps.want_blank = true;
1020 break;
1021
1022 case strpfx:
1023 {
1024 int len = e_token - s_token;
1025
1026 CHECK_SIZE_CODE(len + 1);
1027 if (ps.want_blank)
1028 *e_code++ = ' ';
1029 memcpy(e_code, token, len);
1030 e_code += len;
1031 }
1032 ps.want_blank = false;
1033 break;
1034
1035 case period: /* treat a period kind of like a binary
1036 * operation */
1037 *e_code++ = '.'; /* move the period into line */
1038 ps.want_blank = false; /* dont put a blank after a period */
1039 break;
1040
1041 case comma:
1042 ps.want_blank = (s_code != e_code); /* only put blank after comma
1043 * if comma does not start the
1044 * line */
1045 if (ps.in_decl && ps.procname[0] == '\0' && !ps.block_init &&
1046 !ps.dumped_decl_indent && ps.paren_level == 0) {
1047 /* indent leading commas and not the actual identifiers */
1048 indent_declaration(dec_ind - 1, tabs_to_var);
1049 ps.dumped_decl_indent = true;
1050 }
1051 *e_code++ = ',';
1052 if (ps.p_l_follow == 0) {
1053 if (ps.block_init_level <= 0)
1054 ps.block_init = 0;
1055 if (break_comma && (!opt.leave_comma ||
1056 count_spaces_until(compute_code_target(), s_code, e_code) >
1057 opt.max_col - opt.tabsize))
1058 force_nl = true;
1059 }
1060 break;
1061
1062 case preesc: /* got the character '#' */
1063 if ((s_com != e_com) ||
1064 (s_lab != e_lab) ||
1065 (s_code != e_code))
1066 dump_line();
1067 CHECK_SIZE_LAB(1);
1068 *e_lab++ = '#'; /* move whole line to 'label' buffer */
1069 {
1070 int in_comment = 0;
1071 int com_start = 0;
1072 char quote = 0;
1073 int com_end = 0;
1074
1075 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
1076 buf_ptr++;
1077 if (buf_ptr >= buf_end)
1078 fill_buffer();
1079 }
1080 while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
1081 CHECK_SIZE_LAB(2);
1082 *e_lab = *buf_ptr++;
1083 if (buf_ptr >= buf_end)
1084 fill_buffer();
1085 switch (*e_lab++) {
1086 case BACKSLASH:
1087 if (!in_comment) {
1088 *e_lab++ = *buf_ptr++;
1089 if (buf_ptr >= buf_end)
1090 fill_buffer();
1091 }
1092 break;
1093 case '/':
1094 if (*buf_ptr == '*' && !in_comment && !quote) {
1095 in_comment = 1;
1096 *e_lab++ = *buf_ptr++;
1097 com_start = e_lab - s_lab - 2;
1098 }
1099 break;
1100 case '"':
1101 if (quote == '"')
1102 quote = 0;
1103 break;
1104 case '\'':
1105 if (quote == '\'')
1106 quote = 0;
1107 break;
1108 case '*':
1109 if (*buf_ptr == '/' && in_comment) {
1110 in_comment = 0;
1111 *e_lab++ = *buf_ptr++;
1112 com_end = e_lab - s_lab;
1113 }
1114 break;
1115 }
1116 }
1117
1118 while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1119 e_lab--;
1120 if (e_lab - s_lab == com_end && bp_save == NULL) {
1121 /* comment on preprocessor line */
1122 if (sc_end == NULL) { /* if this is the first comment,
1123 * we must set up the buffer */
1124 save_com = sc_buf;
1125 sc_end = &save_com[0];
1126 }
1127 else {
1128 *sc_end++ = '\n'; /* add newline between
1129 * comments */
1130 *sc_end++ = ' ';
1131 --line_no;
1132 }
1133 if (sc_end - save_com + com_end - com_start > sc_size)
1134 errx(1, "input too long");
1135 memmove(sc_end, s_lab + com_start, com_end - com_start);
1136 sc_end += com_end - com_start;
1137 e_lab = s_lab + com_start;
1138 while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1139 e_lab--;
1140 bp_save = buf_ptr; /* save current input buffer */
1141 be_save = buf_end;
1142 buf_ptr = save_com; /* fix so that subsequent calls to
1143 * lexi will take tokens out of
1144 * save_com */
1145 *sc_end++ = ' '; /* add trailing blank, just in case */
1146 buf_end = sc_end;
1147 sc_end = NULL;
1148 }
1149 CHECK_SIZE_LAB(1);
1150 *e_lab = '\0'; /* null terminate line */
1151 ps.pcase = false;
1152 }
1153
1154 if (strncmp(s_lab, "#if", 3) == 0) { /* also ifdef, ifndef */
1155 if ((size_t)ifdef_level < nitems(state_stack)) {
1156 match_state[ifdef_level].tos = -1;
1157 state_stack[ifdef_level++] = ps;
1158 }
1159 else
1160 diag2(1, "#if stack overflow");
1161 }
1162 else if (strncmp(s_lab, "#el", 3) == 0) { /* else, elif */
1163 if (ifdef_level <= 0)
1164 diag2(1, s_lab[3] == 'i' ? "Unmatched #elif" : "Unmatched #else");
1165 else {
1166 match_state[ifdef_level - 1] = ps;
1167 ps = state_stack[ifdef_level - 1];
1168 }
1169 }
1170 else if (strncmp(s_lab, "#endif", 6) == 0) {
1171 if (ifdef_level <= 0)
1172 diag2(1, "Unmatched #endif");
1173 else
1174 ifdef_level--;
1175 } else {
1176 struct directives {
1177 int size;
1178 const char *string;
1179 }
1180 recognized[] = {
1181 {7, "include"},
1182 {6, "define"},
1183 {5, "undef"},
1184 {4, "line"},
1185 {5, "error"},
1186 {6, "pragma"}
1187 };
1188 int d = nitems(recognized);
1189 while (--d >= 0)
1190 if (strncmp(s_lab + 1, recognized[d].string, recognized[d].size) == 0)
1191 break;
1192 if (d < 0) {
1193 diag2(1, "Unrecognized cpp directive");
1194 break;
1195 }
1196 }
1197 if (opt.blanklines_around_conditional_compilation) {
1198 postfix_blankline_requested++;
1199 n_real_blanklines = 0;
1200 }
1201 else {
1202 postfix_blankline_requested = 0;
1203 prefix_blankline_requested = 0;
1204 }
1205 break; /* subsequent processing of the newline
1206 * character will cause the line to be printed */
1207
1208 case comment: /* we have gotten a / followed by * this is a biggie */
1209 pr_comment();
1210 break;
1211 } /* end of big switch stmt */
1212
1213 *e_code = '\0'; /* make sure code section is null terminated */
1214 if (type_code != comment && type_code != newline && type_code != preesc)
1215 ps.last_token = type_code;
1216 } /* end of main while (1) loop */
1217 }
1218
1219 /*
1220 * copy input file to backup file if in_name is /blah/blah/blah/file, then
1221 * backup file will be ".Bfile" then make the backup file the input and
1222 * original input file the output
1223 */
1224 static void
bakcopy(void)1225 bakcopy(void)
1226 {
1227 int n,
1228 bakchn;
1229 char buff[8 * 1024];
1230 const char *p;
1231
1232 /* construct file name .Bfile */
1233 for (p = in_name; *p; p++); /* skip to end of string */
1234 while (p > in_name && *p != '/') /* find last '/' */
1235 p--;
1236 if (*p == '/')
1237 p++;
1238 sprintf(bakfile, "%s%s", p, simple_backup_suffix);
1239
1240 /* copy in_name to backup file */
1241 bakchn = creat(bakfile, 0600);
1242 if (bakchn < 0)
1243 err(1, "%s", bakfile);
1244 while ((n = read(fileno(input), buff, sizeof(buff))) > 0)
1245 if (write(bakchn, buff, n) != n)
1246 err(1, "%s", bakfile);
1247 if (n < 0)
1248 err(1, "%s", in_name);
1249 close(bakchn);
1250 fclose(input);
1251
1252 /* re-open backup file as the input file */
1253 input = fopen(bakfile, "r");
1254 if (input == NULL)
1255 err(1, "%s", bakfile);
1256 /* now the original input file will be the output */
1257 output = fopen(in_name, "w");
1258 if (output == NULL) {
1259 unlink(bakfile);
1260 err(1, "%s", in_name);
1261 }
1262 }
1263
1264 static void
indent_declaration(int cur_dec_ind,int tabs_to_var)1265 indent_declaration(int cur_dec_ind, int tabs_to_var)
1266 {
1267 int pos = e_code - s_code;
1268 char *startpos = e_code;
1269
1270 /*
1271 * get the tab math right for indentations that are not multiples of tabsize
1272 */
1273 if ((ps.ind_level * opt.ind_size) % opt.tabsize != 0) {
1274 pos += (ps.ind_level * opt.ind_size) % opt.tabsize;
1275 cur_dec_ind += (ps.ind_level * opt.ind_size) % opt.tabsize;
1276 }
1277 if (tabs_to_var) {
1278 int tpos;
1279
1280 CHECK_SIZE_CODE(cur_dec_ind / opt.tabsize);
1281 while ((tpos = opt.tabsize * (1 + pos / opt.tabsize)) <= cur_dec_ind) {
1282 *e_code++ = '\t';
1283 pos = tpos;
1284 }
1285 }
1286 CHECK_SIZE_CODE(cur_dec_ind - pos + 1);
1287 while (pos < cur_dec_ind) {
1288 *e_code++ = ' ';
1289 pos++;
1290 }
1291 if (e_code == startpos && ps.want_blank) {
1292 *e_code++ = ' ';
1293 ps.want_blank = false;
1294 }
1295 }
1296