xref: /openbsd-src/usr.bin/indent/indent.c (revision 898184e3e61f9129feb5978fad5a8c6865f00b92)
1 /*	$OpenBSD: indent.c,v 1.20 2009/10/27 23:59:39 deraadt Exp $	*/
2 
3 /*
4  * Copyright (c) 1980, 1993
5  *	The Regents of the University of California.
6  * Copyright (c) 1976 Board of Trustees of the University of Illinois.
7  * Copyright (c) 1985 Sun Microsystems, Inc.
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 #include <sys/param.h>
36 #include <fcntl.h>
37 #include <unistd.h>
38 #include <stdio.h>
39 #include <stdlib.h>
40 #include <string.h>
41 #include "indent_globs.h"
42 #include "indent_codes.h"
43 #include <ctype.h>
44 #include <errno.h>
45 #include <err.h>
46 
47 char       *in_name = "Standard Input";	/* will always point to name of input
48 					 * file */
49 char       *out_name = "Standard Output";	/* will always point to name
50 						 * of output file */
51 char        bakfile[MAXPATHLEN] = "";
52 
53 void bakcopy(void);
54 
55 int
56 main(int argc, char **argv)
57 {
58 
59     extern int  found_err;	/* flag set in diag() on error */
60     int         dec_ind;	/* current indentation for declarations */
61     int         di_stack[20];	/* a stack of structure indentation levels */
62     int         flushed_nl;	/* used when buffering up comments to remember
63 				 * that a newline was passed over */
64     int         force_nl;	/* when true, code must be broken */
65     int         hd_type;	/* used to store type of stmt for if (...),
66 				 * for (...), etc */
67     int 	i;		/* local loop counter */
68     int         scase;		/* set to true when we see a case, so we will
69 				 * know what to do with the following colon */
70     int         sp_sw;		/* when true, we are in the expressin of
71 				 * if(...), while(...), etc. */
72     int         squest;		/* when this is positive, we have seen a ?
73 				 * without the matching : in a <c>?<s>:<s>
74 				 * construct */
75     char 	*t_ptr;		/* used for copying tokens */
76     int         type_code;	/* the type of token, returned by lexi */
77 
78     int         last_else = 0;	/* true iff last keyword was an else */
79 
80 
81     /*-----------------------------------------------*\
82     |		      INITIALIZATION		      |
83     \*-----------------------------------------------*/
84 
85 
86     hd_type = 0;
87     ps.p_stack[0] = stmt;	/* this is the parser's stack */
88     ps.last_nl = true;		/* this is true if the last thing scanned was
89 				 * a newline */
90     ps.last_token = semicolon;
91     combuf = (char *) malloc(bufsize);
92     labbuf = (char *) malloc(bufsize);
93     codebuf = (char *) malloc(bufsize);
94     tokenbuf = (char *) malloc(bufsize);
95     if (combuf == NULL || labbuf == NULL || codebuf == NULL ||
96         tokenbuf == NULL)
97 	    err(1, NULL);
98     l_com = combuf + bufsize - 5;
99     l_lab = labbuf + bufsize - 5;
100     l_code = codebuf + bufsize - 5;
101     l_token = tokenbuf + bufsize - 5;
102     combuf[0] = codebuf[0] = labbuf[0] = ' ';	/* set up code, label, and
103 						 * comment buffers */
104     combuf[1] = codebuf[1] = labbuf[1] = '\0';
105     ps.else_if = 1;		/* Default else-if special processing to on */
106     s_lab = e_lab = labbuf + 1;
107     s_code = e_code = codebuf + 1;
108     s_com = e_com = combuf + 1;
109     s_token = e_token = tokenbuf + 1;
110 
111     in_buffer = (char *) malloc(10);
112     if (in_buffer == NULL)
113 	    err(1, NULL);
114     in_buffer_limit = in_buffer + 8;
115     buf_ptr = buf_end = in_buffer;
116     line_no = 1;
117     had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
118     sp_sw = force_nl = false;
119     ps.in_or_st = false;
120     ps.bl_line = true;
121     dec_ind = 0;
122     di_stack[ps.dec_nest = 0] = 0;
123     ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
124 
125 
126     scase = ps.pcase = false;
127     squest = 0;
128     sc_end = 0;
129     bp_save = 0;
130     be_save = 0;
131 
132     output = 0;
133 
134 
135 
136     /*--------------------------------------------------*\
137     |   		COMMAND LINE SCAN		 |
138     \*--------------------------------------------------*/
139 
140 #ifdef undef
141     max_col = 78;		/* -l78 */
142     lineup_to_parens = 1;	/* -lp */
143     ps.ljust_decl = 0;		/* -ndj */
144     ps.com_ind = 33;		/* -c33 */
145     star_comment_cont = 1;	/* -sc */
146     ps.ind_size = 8;		/* -i8 */
147     verbose = 0;
148     ps.decl_indent = 16;	/* -di16 */
149     ps.indent_parameters = 1;	/* -ip */
150     ps.decl_com_ind = 0;	/* if this is not set to some positive value
151 				 * by an arg, we will set this equal to
152 				 * ps.com_ind */
153     btype_2 = 1;		/* -br */
154     cuddle_else = 1;		/* -ce */
155     ps.unindent_displace = 0;	/* -d0 */
156     ps.case_indent = 0;		/* -cli0 */
157     format_col1_comments = 1;	/* -fc1 */
158     procnames_start_line = 1;	/* -psl */
159     proc_calls_space = 0;	/* -npcs */
160     comment_delimiter_on_blankline = 1;	/* -cdb */
161     ps.leave_comma = 1;		/* -nbc */
162 #endif
163 
164     for (i = 1; i < argc; ++i)
165 	if (strcmp(argv[i], "-npro") == 0)
166 	    break;
167     set_defaults();
168     if (i >= argc)
169 	set_profile();
170 
171     for (i = 1; i < argc; ++i) {
172 
173 	/*
174 	 * look thru args (if any) for changes to defaults
175 	 */
176 	if (argv[i][0] != '-') {/* no flag on parameter */
177 	    if (input == 0) {	/* we must have the input file */
178 		in_name = argv[i];	/* remember name of input file */
179 		input = fopen(in_name, "r");
180 		if (input == NULL)		/* check for open error */
181 			err(1, "%s", in_name);
182 		continue;
183 	    }
184 	    else if (output == 0) {	/* we have the output file */
185 		out_name = argv[i];	/* remember name of output file */
186 		if (strcmp(in_name, out_name) == 0)	/* attempt to overwrite
187 							 * the file */
188 			errx(1, "input and output files must be different");
189 		output = fopen(out_name, "w");
190 		if (output == NULL)	/* check for create error */
191 			err(1, "%s", out_name);
192 		continue;
193 	    }
194 	    errx(1, "unknown parameter: %s", argv[i]);
195 	}
196 	else
197 	    set_option(argv[i]);
198     }				/* end of for */
199     if (input == NULL) {
200 	fprintf(stderr, "usage: indent input-file [output-file] [options]\n");
201 	exit(1);
202     }
203     if (output == NULL) {
204 	if (troff)
205 	    output = stdout;
206 	else {
207 	    out_name = in_name;
208 	    bakcopy();
209 	}
210     }
211     if (ps.com_ind <= 1)
212 	ps.com_ind = 2;		/* dont put normal comments before column 2 */
213     if (troff) {
214 	if (bodyf.font[0] == 0)
215 	    parsefont(&bodyf, "R");
216 	if (scomf.font[0] == 0)
217 	    parsefont(&scomf, "I");
218 	if (blkcomf.font[0] == 0)
219 	    blkcomf = scomf, blkcomf.size += 2;
220 	if (boxcomf.font[0] == 0)
221 	    boxcomf = blkcomf;
222 	if (stringf.font[0] == 0)
223 	    parsefont(&stringf, "L");
224 	if (keywordf.font[0] == 0)
225 	    parsefont(&keywordf, "B");
226 	writefdef(&bodyf, 'B');
227 	writefdef(&scomf, 'C');
228 	writefdef(&blkcomf, 'L');
229 	writefdef(&boxcomf, 'X');
230 	writefdef(&stringf, 'S');
231 	writefdef(&keywordf, 'K');
232     }
233     if (block_comment_max_col <= 0)
234 	block_comment_max_col = max_col;
235     if (ps.decl_com_ind <= 0)	/* if not specified by user, set this */
236 	ps.decl_com_ind = ps.ljust_decl ? (ps.com_ind <= 10 ? 2 : ps.com_ind - 8) : ps.com_ind;
237     if (continuation_indent == 0)
238 	continuation_indent = ps.ind_size;
239     fill_buffer();	/* get first batch of stuff into input buffer */
240 
241     parse(semicolon);
242     {
243 	char *p = buf_ptr;
244 	int   col = 1;
245 
246 	while (1) {
247 	    if (*p == ' ')
248 		col++;
249 	    else if (*p == '\t')
250 		col = ((col - 1) & ~7) + 9;
251 	    else
252 		break;
253 	    p++;
254 	}
255 	if (col > ps.ind_size)
256 	    ps.ind_level = ps.i_l_follow = col / ps.ind_size;
257     }
258     if (troff) {
259 	char *p = in_name,
260 	           *beg = in_name;
261 
262 	while (*p)
263 	    if (*p++ == '/')
264 		beg = p;
265 	fprintf(output, ".Fn \"%s\"\n", beg);
266     }
267     /*
268      * START OF MAIN LOOP
269      */
270 
271     while (1) {			/* this is the main loop.  it will go until we
272 				 * reach eof */
273 	int         is_procname;
274 
275 	type_code = lexi();	/* lexi reads one token.  The actual
276 				 * characters read are stored in "token". lexi
277 				 * returns a code indicating the type of token */
278 	is_procname = ps.procname[0];
279 
280 	/*
281 	 * The following code moves everything following an if (), while (),
282 	 * else, etc. up to the start of the following stmt to a buffer. This
283 	 * allows proper handling of both kinds of brace placement.
284 	 */
285 
286 	flushed_nl = false;
287 	while (ps.search_brace) {	/* if we scanned an if(), while(),
288 					 * etc., we might need to copy stuff
289 					 * into a buffer we must loop, copying
290 					 * stuff into save_com, until we find
291 					 * the start of the stmt which follows
292 					 * the if, or whatever */
293 	    switch (type_code) {
294 	    case newline:
295 		++line_no;
296 		flushed_nl = true;
297 	    case form_feed:
298 		break;		/* form feeds and newlines found here will be
299 				 * ignored */
300 
301 	    case lbrace:	/* this is a brace that starts the compound
302 				 * stmt */
303 		if (sc_end == 0) {	/* ignore buffering if a comment wasnt
304 					 * stored up */
305 		    ps.search_brace = false;
306 		    goto check_type;
307 		}
308 		if (btype_2) {
309 		    save_com[0] = '{';	/* we either want to put the brace
310 					 * right after the if */
311 		    goto sw_buffer;	/* go to common code to get out of
312 					 * this loop */
313 		}
314 	    case comment:	/* we have a comment, so we must copy it into
315 				 * the buffer */
316 		if (!flushed_nl || sc_end != 0) {
317 		    if (sc_end == 0) {	/* if this is the first comment, we
318 					 * must set up the buffer */
319 			save_com[0] = save_com[1] = ' ';
320 			sc_end = &(save_com[2]);
321 		    }
322 		    else {
323 			*sc_end++ = '\n';	/* add newline between
324 						 * comments */
325 			*sc_end++ = ' ';
326 			--line_no;
327 		    }
328 		    *sc_end++ = '/';	/* copy in start of comment */
329 		    *sc_end++ = '*';
330 
331 		    for (;;) {	/* loop until we get to the end of the comment */
332 			*sc_end = *buf_ptr++;
333 			if (buf_ptr >= buf_end)
334 			    fill_buffer();
335 
336 			if (*sc_end++ == '*' && *buf_ptr == '/')
337 			    break;	/* we are at end of comment */
338 
339 			if (sc_end >= &(save_com[sc_size])) {	/* check for temp buffer
340 								 * overflow */
341 			    diag(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever.");
342 			    fflush(output);
343 			    exit(1);
344 			}
345 		    }
346 		    *sc_end++ = '/';	/* add ending slash */
347 		    if (++buf_ptr >= buf_end)	/* get past / in buffer */
348 			fill_buffer();
349 		    break;
350 		}
351 	    default:		/* it is the start of a normal statment */
352 		if (flushed_nl)	/* if we flushed a newline, make sure it is
353 				 * put back */
354 		    force_nl = true;
355 		if ((type_code == sp_paren && *token == 'i'
356 		     && last_else && ps.else_if) ||
357 		    (type_code == sp_nparen && *token == 'e'
358 		     && e_code != s_code && e_code[-1] == '}'))
359 			force_nl = false;
360 
361 		if (sc_end == 0) {	/* ignore buffering if comment wasnt
362 					 * saved up */
363 		    ps.search_brace = false;
364 		    goto check_type;
365 		}
366 		if (force_nl) {	/* if we should insert a nl here, put it into
367 				 * the buffer */
368 		    force_nl = false;
369 		    --line_no;	/* this will be re-increased when the nl is
370 				 * read from the buffer */
371 		    *sc_end++ = '\n';
372 		    *sc_end++ = ' ';
373 		    if (verbose && !flushed_nl)	/* print error msg if the line
374 						 * was not already broken */
375 			diag(0, "Line broken");
376 		    flushed_nl = false;
377 		}
378 		for (t_ptr = token; *t_ptr; ++t_ptr)
379 		    *sc_end++ = *t_ptr;	/* copy token into temp buffer */
380 		ps.procname[0] = 0;
381 
382 	sw_buffer:
383 		ps.search_brace = false;	/* stop looking for start of
384 						 * stmt */
385 		bp_save = buf_ptr;	/* save current input buffer */
386 		be_save = buf_end;
387 		buf_ptr = save_com;	/* fix so that subsequent calls to
388 					 * lexi will take tokens out of
389 					 * save_com */
390 		*sc_end++ = ' ';/* add trailing blank, just in case */
391 		buf_end = sc_end;
392 		sc_end = 0;
393 		break;
394 	    }			/* end of switch */
395 	    if (type_code != 0)	/* we must make this check, just in case there
396 				 * was an unexpected EOF */
397 		type_code = lexi();	/* read another token */
398 	    /* if (ps.search_brace) ps.procname[0] = 0; */
399 	    if ((is_procname = ps.procname[0]) && flushed_nl
400 		    && !procnames_start_line && ps.in_decl
401 		    && type_code == ident)
402 		flushed_nl = 0;
403 	}			/* end of while (search_brace) */
404 	last_else = 0;
405 check_type:
406 	if (type_code == 0) {	/* we got eof */
407 	    if (s_lab != e_lab || s_code != e_code
408 		    || s_com != e_com)	/* must dump end of line */
409 		dump_line();
410 	    if (ps.tos > 1)	/* check for balanced braces */
411 		diag(1, "Missing braces at end of file.");
412 
413 	    if (verbose) {
414 		printf("There were %d output lines and %d comments\n",
415 		       ps.out_lines, ps.out_coms);
416 		printf("(Lines with comments)/(Lines with code): %6.3f\n",
417 		       (1.0 * ps.com_lines) / code_lines);
418 	    }
419 	    fflush(output);
420 	    exit(found_err);
421 	}
422 	if (
423 		(type_code != comment) &&
424 		(type_code != newline) &&
425 		(type_code != preesc) &&
426 		(type_code != form_feed)) {
427 	    if (force_nl &&
428 		    (type_code != semicolon) &&
429 		    (type_code != lbrace || !btype_2)) {
430 		/* we should force a broken line here */
431 		if (verbose && !flushed_nl)
432 		    diag(0, "Line broken");
433 		flushed_nl = false;
434 		dump_line();
435 		ps.want_blank = false;	/* dont insert blank at line start */
436 		force_nl = false;
437 	    }
438 	    ps.in_stmt = true;	/* turn on flag which causes an extra level of
439 				 * indentation. this is turned off by a ; or
440 				 * '}' */
441 	    if (s_com != e_com) {	/* the turkey has embedded a comment
442 					 * in a line. fix it */
443 		*e_code++ = ' ';
444 		for (t_ptr = s_com; *t_ptr; ++t_ptr) {
445 		    CHECK_SIZE_CODE;
446 		    *e_code++ = *t_ptr;
447 		}
448 		*e_code++ = ' ';
449 		*e_code = '\0';	/* null terminate code sect */
450 		ps.want_blank = false;
451 		e_com = s_com;
452 	    }
453 	}
454 	else if (type_code != comment)	/* preserve force_nl thru a comment */
455 	    force_nl = false;	/* cancel forced newline after newline, form
456 				 * feed, etc */
457 
458 
459 
460 	/*-----------------------------------------------------*\
461 	|	   do switch on type of token scanned		|
462 	\*-----------------------------------------------------*/
463 	CHECK_SIZE_CODE;
464 	switch (type_code) {	/* now, decide what to do with the token */
465 
466 	case form_feed:	/* found a form feed in line */
467 	    ps.use_ff = true;	/* a form feed is treated much like a newline */
468 	    dump_line();
469 	    ps.want_blank = false;
470 	    break;
471 
472 	case newline:
473 	    if (ps.last_token != comma || ps.p_l_follow > 0
474 		    || !ps.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
475 		dump_line();
476 		ps.want_blank = false;
477 	    }
478 	    ++line_no;		/* keep track of input line number */
479 	    break;
480 
481 	case lparen:		/* got a '(' or '[' */
482 	    ++ps.p_l_follow;	/* count parens to make Healy happy */
483 	    if (ps.want_blank && *token != '[' &&
484 		    (ps.last_token != ident || proc_calls_space
485 	      || (ps.its_a_keyword && (!ps.sizeof_keyword || Bill_Shannon))))
486 		*e_code++ = ' ';
487 	    if (ps.in_decl && !ps.block_init)
488 		if (troff && !ps.dumped_decl_indent && !is_procname && ps.last_token == decl) {
489 		    ps.dumped_decl_indent = 1;
490 		    snprintf(e_code, (l_code - e_code) + 5,
491 			"\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
492 		    e_code += strlen(e_code);
493 		    CHECK_SIZE_CODE;
494 		}
495 		else {
496 		    while ((e_code - s_code) < dec_ind) {
497 			CHECK_SIZE_CODE;
498 			*e_code++ = ' ';
499 		    }
500 		    *e_code++ = token[0];
501 		}
502 	    else
503 		*e_code++ = token[0];
504 	    ps.paren_indents[ps.p_l_follow - 1] = e_code - s_code;
505 	    if (sp_sw && ps.p_l_follow == 1 && extra_expression_indent
506 		    && ps.paren_indents[0] < 2 * ps.ind_size)
507 		ps.paren_indents[0] = 2 * ps.ind_size;
508 	    ps.want_blank = false;
509 	    if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
510 		/*
511 		 * this is a kluge to make sure that declarations will be
512 		 * aligned right if proc decl has an explicit type on it, i.e.
513 		 * "int a(x) {..."
514 		 */
515 		parse(semicolon);	/* I said this was a kluge... */
516 		ps.in_or_st = false;	/* turn off flag for structure decl or
517 					 * initialization */
518 	    }
519 	    if (ps.sizeof_keyword)
520 		ps.sizeof_mask |= 1 << ps.p_l_follow;
521 	    break;
522 
523 	case rparen:		/* got a ')' or ']' */
524 	    rparen_count--;
525 	    if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.sizeof_mask) {
526 		ps.last_u_d = true;
527 		ps.cast_mask &= (1 << ps.p_l_follow) - 1;
528 	    }
529 	    ps.sizeof_mask &= (1 << ps.p_l_follow) - 1;
530 	    if (--ps.p_l_follow < 0) {
531 		ps.p_l_follow = 0;
532 		diag(0, "Extra %c", *token);
533 	    }
534 	    if (e_code == s_code)	/* if the paren starts the line */
535 		ps.paren_level = ps.p_l_follow;	/* then indent it */
536 
537 	    *e_code++ = token[0];
538 	    ps.want_blank = true;
539 
540 	    if (sp_sw && (ps.p_l_follow == 0)) {	/* check for end of if
541 							 * (...), or some such */
542 		sp_sw = false;
543 		force_nl = true;/* must force newline after if */
544 		ps.last_u_d = true;	/* inform lexi that a following
545 					 * operator is unary */
546 		ps.in_stmt = false;	/* dont use stmt continuation
547 					 * indentation */
548 
549 		parse(hd_type);	/* let parser worry about if, or whatever */
550 	    }
551 	    ps.search_brace = btype_2;	/* this should insure that constructs
552 					 * such as main(){...} and int[]{...}
553 					 * have their braces put in the right
554 					 * place */
555 	    break;
556 
557 	case unary_op:		/* this could be any unary operation */
558 	    if (ps.want_blank)
559 		*e_code++ = ' ';
560 
561 	    if (troff && !ps.dumped_decl_indent && ps.in_decl && !is_procname) {
562 		snprintf(e_code, (l_code - e_code) + 5,
563 		    "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
564 		ps.dumped_decl_indent = 1;
565 		e_code += strlen(e_code);
566 		CHECK_SIZE_CODE;
567 	    }
568 	    else {
569 		char       *res = token;
570 
571 		if (ps.in_decl && !ps.block_init) {	/* if this is a unary op
572 							 * in a declaration, we
573 							 * should indent this
574 							 * token */
575 		    for (i = 0; token[i]; ++i);	/* find length of token */
576 		    while ((e_code - s_code) < (dec_ind - i)) {
577 			CHECK_SIZE_CODE;
578 			*e_code++ = ' ';	/* pad it */
579 		    }
580 		}
581 		if (troff && token[0] == '-' && token[1] == '>')
582 		    res = "\\(->";
583 		for (t_ptr = res; *t_ptr; ++t_ptr) {
584 		    CHECK_SIZE_CODE;
585 		    *e_code++ = *t_ptr;
586 		}
587 	    }
588 	    ps.want_blank = false;
589 	    break;
590 
591 	case binary_op:	/* any binary operation */
592 	    if (ps.want_blank)
593 		*e_code++ = ' ';
594 	    {
595 		char       *res = token;
596 
597 		if (troff)
598 		    switch (token[0]) {
599 		    case '<':
600 			if (token[1] == '=')
601 			    res = "\\(<=";
602 			break;
603 		    case '>':
604 			if (token[1] == '=')
605 			    res = "\\(>=";
606 			break;
607 		    case '!':
608 			if (token[1] == '=')
609 			    res = "\\(!=";
610 			break;
611 		    case '|':
612 			if (token[1] == '|')
613 			    res = "\\(br\\(br";
614 			else if (token[1] == 0)
615 			    res = "\\(br";
616 			break;
617 		    }
618 		for (t_ptr = res; *t_ptr; ++t_ptr) {
619 		    CHECK_SIZE_CODE;
620 		    *e_code++ = *t_ptr;	/* move the operator */
621 		}
622 	    }
623 	    ps.want_blank = true;
624 	    break;
625 
626 	case postop:		/* got a trailing ++ or -- */
627 	    *e_code++ = token[0];
628 	    *e_code++ = token[1];
629 	    ps.want_blank = true;
630 	    break;
631 
632 	case question:		/* got a ? */
633 	    squest++;		/* this will be used when a later colon
634 				 * appears so we can distinguish the
635 				 * <c>?<n>:<n> construct */
636 	    if (ps.want_blank)
637 		*e_code++ = ' ';
638 	    *e_code++ = '?';
639 	    ps.want_blank = true;
640 	    break;
641 
642 	case casestmt:		/* got word 'case' or 'default' */
643 	    scase = true;	/* so we can process the later colon properly */
644 	    goto copy_id;
645 
646 	case colon:		/* got a ':' */
647 	    if (squest > 0) {	/* it is part of the <c>?<n>: <n> construct */
648 		--squest;
649 		if (ps.want_blank)
650 		    *e_code++ = ' ';
651 		*e_code++ = ':';
652 		ps.want_blank = true;
653 		break;
654 	    }
655 	    if (ps.in_decl) {
656 		*e_code++ = ':';
657 		ps.want_blank = false;
658 		break;
659 	    }
660 	    ps.in_stmt = false;	/* seeing a label does not imply we are in a
661 				 * stmt */
662 	    for (t_ptr = s_code; *t_ptr; ++t_ptr)
663 		*e_lab++ = *t_ptr;	/* turn everything so far into a label */
664 	    e_code = s_code;
665 	    *e_lab++ = ':';
666 	    *e_lab++ = ' ';
667 	    *e_lab = '\0';
668 
669 	    force_nl = ps.pcase = scase;	/* ps.pcase will be used by
670 						 * dump_line to decide how to
671 						 * indent the label. force_nl
672 						 * will force a case n: to be
673 						 * on a line by itself */
674 	    scase = false;
675 	    ps.want_blank = false;
676 	    break;
677 
678 	case semicolon:	/* got a ';' */
679 	    ps.in_or_st = false;/* we are not in an initialization or
680 				 * structure declaration */
681 	    scase = false;	/* these will only need resetting in a error */
682 	    squest = 0;
683 	    if (ps.last_token == rparen && rparen_count == 0)
684 		ps.in_parameter_declaration = 0;
685 	    ps.cast_mask = 0;
686 	    ps.sizeof_mask = 0;
687 	    ps.block_init = 0;
688 	    ps.block_init_level = 0;
689 	    ps.just_saw_decl--;
690 
691 	    if (ps.in_decl && s_code == e_code && !ps.block_init)
692 		while ((e_code - s_code) < (dec_ind - 1)) {
693 		    CHECK_SIZE_CODE;
694 		    *e_code++ = ' ';
695 		}
696 
697 	    ps.in_decl = (ps.dec_nest > 0);	/* if we were in a first level
698 						 * structure declaration, we
699 						 * arent any more */
700 
701 	    if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
702 
703 		/*
704 		 * This should be true iff there were unbalanced parens in the
705 		 * stmt.  It is a bit complicated, because the semicolon might
706 		 * be in a for stmt
707 		 */
708 		diag(1, "Unbalanced parens");
709 		ps.p_l_follow = 0;
710 		if (sp_sw) {	/* this is a check for a if, while, etc. with
711 				 * unbalanced parens */
712 		    sp_sw = false;
713 		    parse(hd_type);	/* dont lose the if, or whatever */
714 		}
715 	    }
716 	    *e_code++ = ';';
717 	    ps.want_blank = true;
718 	    ps.in_stmt = (ps.p_l_follow > 0);	/* we are no longer in the
719 						 * middle of a stmt */
720 
721 	    if (!sp_sw) {	/* if not if for (;;) */
722 		parse(semicolon);	/* let parser know about end of stmt */
723 		force_nl = true;/* force newline after a end of stmt */
724 	    }
725 	    break;
726 
727 	case lbrace:		/* got a '{' */
728 	    ps.in_stmt = false;	/* dont indent the {} */
729 	    if (!ps.block_init)
730 		force_nl = true;/* force other stuff on same line as '{' onto
731 				 * new line */
732 	    else if (ps.block_init_level <= 0)
733 		ps.block_init_level = 1;
734 	    else
735 		ps.block_init_level++;
736 
737 	    if (s_code != e_code && !ps.block_init) {
738 		if (!btype_2) {
739 		    dump_line();
740 		    ps.want_blank = false;
741 		}
742 		else if (ps.in_parameter_declaration && !ps.in_or_st) {
743 		    ps.i_l_follow = 0;
744 		    dump_line();
745 		    ps.want_blank = false;
746 		}
747 	    }
748 	    if (ps.in_parameter_declaration)
749 		prefix_blankline_requested = 0;
750 
751 	    if (ps.p_l_follow > 0) {	/* check for preceding unbalanced
752 					 * parens */
753 		diag(1, "Unbalanced parens");
754 		ps.p_l_follow = 0;
755 		if (sp_sw) {	/* check for unclosed if, for, etc. */
756 		    sp_sw = false;
757 		    parse(hd_type);
758 		    ps.ind_level = ps.i_l_follow;
759 		}
760 	    }
761 	    if (s_code == e_code)
762 		ps.ind_stmt = false;	/* dont put extra indentation on line
763 					 * with '{' */
764 	    if (ps.in_decl && ps.in_or_st) {	/* this is either a structure
765 						 * declaration or an init */
766 		di_stack[ps.dec_nest++] = dec_ind;
767 		/* ?		dec_ind = 0; */
768 	    }
769 	    else {
770 		ps.decl_on_line = false;
771 		/* we can't be in the middle of a declaration, so don't do
772 		 * special indentation of comments */
773 		if (blanklines_after_declarations_at_proctop
774 			&& ps.in_parameter_declaration)
775 		    postfix_blankline_requested = 1;
776 		ps.in_parameter_declaration = 0;
777 	    }
778 	    dec_ind = 0;
779 	    parse(lbrace);	/* let parser know about this */
780 	    if (ps.want_blank)	/* put a blank before '{' if '{' is not at
781 				 * start of line */
782 		*e_code++ = ' ';
783 	    ps.want_blank = false;
784 	    *e_code++ = '{';
785 	    ps.just_saw_decl = 0;
786 	    break;
787 
788 	case rbrace:		/* got a '}' */
789 	    if (ps.p_stack[ps.tos] == decl && !ps.block_init)	/* semicolons can be
790 								 * omitted in
791 								 * declarations */
792 		parse(semicolon);
793 	    if (ps.p_l_follow) {/* check for unclosed if, for, else. */
794 		diag(1, "Unbalanced parens");
795 		ps.p_l_follow = 0;
796 		sp_sw = false;
797 	    }
798 	    ps.just_saw_decl = 0;
799 	    ps.block_init_level--;
800 	    if (s_code != e_code && !ps.block_init) {	/* '}' must be first on
801 							 * line */
802 		if (verbose)
803 		    diag(0, "Line broken");
804 		dump_line();
805 	    }
806 	    *e_code++ = '}';
807 	    ps.want_blank = true;
808 	    ps.in_stmt = ps.ind_stmt = false;
809 	    if (ps.dec_nest > 0) {	/* we are in multi-level structure
810 					 * declaration */
811 		dec_ind = di_stack[--ps.dec_nest];
812 		if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
813 		    ps.just_saw_decl = 2;
814 		ps.in_decl = true;
815 	    }
816 	    prefix_blankline_requested = 0;
817 	    parse(rbrace);	/* let parser know about this */
818 	    ps.search_brace = cuddle_else && ps.p_stack[ps.tos] == ifhead
819 		&& ps.il[ps.tos] >= ps.ind_level;
820 	    if (ps.tos <= 1 && blanklines_after_procs && ps.dec_nest <= 0)
821 		postfix_blankline_requested = 1;
822 	    break;
823 
824 	case swstmt:		/* got keyword "switch" */
825 	    sp_sw = true;
826 	    hd_type = swstmt;	/* keep this for when we have seen the
827 				 * expression */
828 	    goto copy_id;	/* go move the token into buffer */
829 
830 	case sp_paren:		/* token is if, while, for */
831 	    sp_sw = true;	/* the interesting stuff is done after the
832 				 * expression is scanned */
833 	    hd_type = (*token == 'i' ? ifstmt :
834 		       (*token == 'w' ? whilestmt : forstmt));
835 
836 	    /*
837 	     * remember the type of header for later use by parser
838 	     */
839 	    goto copy_id;	/* copy the token into line */
840 
841 	case sp_nparen:	/* got else, do */
842 	    ps.in_stmt = false;
843 	    if (*token == 'e') {
844 		if (e_code != s_code && (!cuddle_else || e_code[-1] != '}')) {
845 		    if (verbose)
846 			diag(0, "Line broken");
847 		    dump_line();/* make sure this starts a line */
848 		    ps.want_blank = false;
849 		}
850 		force_nl = true;/* also, following stuff must go onto new line */
851 		last_else = 1;
852 		parse(elselit);
853 	    }
854 	    else {
855 		if (e_code != s_code) {	/* make sure this starts a line */
856 		    if (verbose)
857 			diag(0, "Line broken");
858 		    dump_line();
859 		    ps.want_blank = false;
860 		}
861 		force_nl = true;/* also, following stuff must go onto new line */
862 		last_else = 0;
863 		parse(dolit);
864 	    }
865 	    goto copy_id;	/* move the token into line */
866 
867 	case decl:		/* we have a declaration type (int, register,
868 				 * etc.) */
869 	    parse(decl);	/* let parser worry about indentation */
870 	    if (ps.last_token == rparen && ps.tos <= 1) {
871 		ps.in_parameter_declaration = 1;
872 		if (s_code != e_code) {
873 		    dump_line();
874 		    ps.want_blank = 0;
875 		}
876 	    }
877 	    if (ps.in_parameter_declaration && ps.indent_parameters && ps.dec_nest == 0) {
878 		ps.ind_level = ps.i_l_follow = 1;
879 		ps.ind_stmt = 0;
880 	    }
881 	    ps.in_or_st = true;	/* this might be a structure or initialization
882 				 * declaration */
883 	    ps.in_decl = ps.decl_on_line = true;
884 	    if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
885 		ps.just_saw_decl = 2;
886 	    prefix_blankline_requested = 0;
887 	    for (i = 0; token[i++];);	/* get length of token */
888 
889 	    /*
890 	     * dec_ind = e_code - s_code + (ps.decl_indent>i ? ps.decl_indent
891 	     * : i);
892 	     */
893 	    dec_ind = ps.decl_indent > 0 ? ps.decl_indent : i;
894 	    goto copy_id;
895 
896 	case ident:		/* got an identifier or constant */
897 	    if (ps.in_decl) {	/* if we are in a declaration, we must indent
898 				 * identifier */
899 		if (ps.want_blank)
900 		    *e_code++ = ' ';
901 		ps.want_blank = false;
902 		if (is_procname == 0 || !procnames_start_line) {
903 		    if (!ps.block_init) {
904 			if (troff && !ps.dumped_decl_indent) {
905 			    snprintf(e_code, (l_code - e_code) + 5,
906 				"\n.De %dp+\200p\n", dec_ind * 7);
907 			    ps.dumped_decl_indent = 1;
908 			    e_code += strlen(e_code);
909 			    CHECK_SIZE_CODE;
910 			}
911 			else
912 			    while ((e_code - s_code) < dec_ind) {
913 				CHECK_SIZE_CODE;
914 				*e_code++ = ' ';
915 			    }
916 		    }
917 		}
918 		else {
919 		    if (dec_ind && s_code != e_code)
920 			dump_line();
921 		    dec_ind = 0;
922 		    ps.want_blank = false;
923 		}
924 	    }
925 	    else if (sp_sw && ps.p_l_follow == 0) {
926 		sp_sw = false;
927 		force_nl = true;
928 		ps.last_u_d = true;
929 		ps.in_stmt = false;
930 		parse(hd_type);
931 	    }
932     copy_id:
933 	    if (ps.want_blank)
934 		*e_code++ = ' ';
935 	    if (troff && ps.its_a_keyword) {
936 		e_code = chfont(&bodyf, &keywordf, e_code);
937 		for (t_ptr = token; *t_ptr; ++t_ptr) {
938 		    CHECK_SIZE_CODE;
939 		    *e_code++ = keywordf.allcaps && islower(*t_ptr)
940 			? toupper(*t_ptr) : *t_ptr;
941 		}
942 		e_code = chfont(&keywordf, &bodyf, e_code);
943 	    }
944 	    else
945 		for (t_ptr = token; *t_ptr; ++t_ptr) {
946 		    CHECK_SIZE_CODE;
947 		    *e_code++ = *t_ptr;
948 		}
949 	    ps.want_blank = true;
950 	    break;
951 
952 	case period:		/* treat a period kind of like a binary
953 				 * operation */
954 	    *e_code++ = '.';	/* move the period into line */
955 	    ps.want_blank = false;	/* dont put a blank after a period */
956 	    break;
957 
958 	case comma:
959 	    ps.want_blank = (s_code != e_code);	/* only put blank after comma
960 						 * if comma does not start the
961 						 * line */
962 	    if (ps.in_decl && is_procname == 0 && !ps.block_init)
963 		while ((e_code - s_code) < (dec_ind - 1)) {
964 		    CHECK_SIZE_CODE;
965 		    *e_code++ = ' ';
966 		}
967 
968 	    *e_code++ = ',';
969 	    if (ps.p_l_follow == 0) {
970 		if (ps.block_init_level <= 0)
971 		    ps.block_init = 0;
972 		if (break_comma && (!ps.leave_comma || compute_code_target() + (e_code - s_code) > max_col - 8))
973 		    force_nl = true;
974 	    }
975 	    break;
976 
977 	case preesc:		/* got the character '#' */
978 	    if ((s_com != e_com) ||
979 		    (s_lab != e_lab) ||
980 		    (s_code != e_code))
981 		dump_line();
982 	    *e_lab++ = '#';	/* move whole line to 'label' buffer */
983 	    {
984 		int         in_comment = 0;
985 		int         com_start = 0;
986 		char        quote = 0;
987 		int         com_end = 0;
988 
989 		while (*buf_ptr == ' ' || *buf_ptr == '\t') {
990 		    buf_ptr++;
991 		    if (buf_ptr >= buf_end)
992 			fill_buffer();
993 		}
994 		while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
995 		    CHECK_SIZE_LAB;
996 		    *e_lab = *buf_ptr++;
997 		    if (buf_ptr >= buf_end)
998 			fill_buffer();
999 		    switch (*e_lab++) {
1000 		    case BACKSLASH:
1001 			if (troff)
1002 			    *e_lab++ = BACKSLASH;
1003 			if (!in_comment) {
1004 			    *e_lab++ = *buf_ptr++;
1005 			    if (buf_ptr >= buf_end)
1006 				fill_buffer();
1007 			}
1008 			break;
1009 		    case '/':
1010 			if (*buf_ptr == '*' && !in_comment && !quote) {
1011 			    in_comment = 1;
1012 			    *e_lab++ = *buf_ptr++;
1013 			    com_start = e_lab - s_lab - 2;
1014 			}
1015 			break;
1016 		    case '"':
1017 			if (quote == '"')
1018 			    quote = 0;
1019 			break;
1020 		    case '\'':
1021 			if (quote == '\'')
1022 			    quote = 0;
1023 			break;
1024 		    case '*':
1025 			if (*buf_ptr == '/' && in_comment) {
1026 			    in_comment = 0;
1027 			    *e_lab++ = *buf_ptr++;
1028 			    com_end = e_lab - s_lab;
1029 			}
1030 			break;
1031 		    }
1032 		}
1033 
1034 		while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1035 		    e_lab--;
1036 		if (e_lab - s_lab == com_end && bp_save == 0) {	/* comment on
1037 								 * preprocessor line */
1038 		    if (sc_end == 0)	/* if this is the first comment, we
1039 					 * must set up the buffer */
1040 			sc_end = &(save_com[0]);
1041 		    else {
1042 			*sc_end++ = '\n';	/* add newline between
1043 						 * comments */
1044 			*sc_end++ = ' ';
1045 			--line_no;
1046 		    }
1047 		    bcopy(s_lab + com_start, sc_end, com_end - com_start);
1048 		    sc_end += com_end - com_start;
1049 		    if (sc_end >= &save_com[sc_size])
1050 			abort();
1051 		    e_lab = s_lab + com_start;
1052 		    while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1053 			e_lab--;
1054 		    bp_save = buf_ptr;	/* save current input buffer */
1055 		    be_save = buf_end;
1056 		    buf_ptr = save_com;	/* fix so that subsequent calls to
1057 					 * lexi will take tokens out of
1058 					 * save_com */
1059 		    *sc_end++ = ' ';	/* add trailing blank, just in case */
1060 		    buf_end = sc_end;
1061 		    sc_end = 0;
1062 		}
1063 		*e_lab = '\0';	/* null terminate line */
1064 		ps.pcase = false;
1065 	    }
1066 
1067 	    if (strncmp(s_lab, "#if", 3) == 0) {
1068 		if (blanklines_around_conditional_compilation) {
1069 		    int    c;
1070 		    prefix_blankline_requested++;
1071 		    while ((c = getc(input)) == '\n');
1072 		    ungetc(c, input);
1073 		}
1074 		if (ifdef_level < sizeof state_stack / sizeof state_stack[0]) {
1075 		    match_state[ifdef_level].tos = -1;
1076 		    state_stack[ifdef_level++] = ps;
1077 		}
1078 		else
1079 		    diag(1, "#if stack overflow");
1080 	    }
1081 	    else if (strncmp(s_lab, "#else", 5) == 0)
1082 		if (ifdef_level <= 0)
1083 		    diag(1, "Unmatched #else");
1084 		else {
1085 		    match_state[ifdef_level - 1] = ps;
1086 		    ps = state_stack[ifdef_level - 1];
1087 		}
1088 	    else if (strncmp(s_lab, "#endif", 6) == 0) {
1089 		if (ifdef_level <= 0)
1090 		    diag(1, "Unmatched #endif");
1091 		else {
1092 		    ifdef_level--;
1093 
1094 #ifdef undef
1095 		    /*
1096 		     * This match needs to be more intelligent before the
1097 		     * message is useful
1098 		     */
1099 		    if (match_state[ifdef_level].tos >= 0
1100 			  && bcmp(&ps, &match_state[ifdef_level], sizeof ps))
1101 			diag(0, "Syntactically inconsistent #ifdef alternatives.");
1102 #endif
1103 		}
1104 		if (blanklines_around_conditional_compilation) {
1105 		    postfix_blankline_requested++;
1106 		    n_real_blanklines = 0;
1107 		}
1108 	    }
1109 	    break;		/* subsequent processing of the newline
1110 				 * character will cause the line to be printed */
1111 
1112 	case comment:		/* we have gotten a comment this is a biggie */
1113 	    if (flushed_nl) {	/* we should force a broken line here */
1114 		flushed_nl = false;
1115 		dump_line();
1116 		ps.want_blank = false;	/* dont insert blank at line start */
1117 		force_nl = false;
1118 	    }
1119 	    pr_comment();
1120 	    break;
1121 	}			/* end of big switch stmt */
1122 
1123 	*e_code = '\0';		/* make sure code section is null terminated */
1124 	if (type_code != comment && type_code != newline && type_code != preesc)
1125 	    ps.last_token = type_code;
1126     }				/* end of main while (1) loop */
1127 }
1128 
1129 /*
1130  * copy input file to backup file if in_name is /blah/blah/blah/file, then
1131  * backup file will be ".Bfile" then make the backup file the input and
1132  * original input file the output
1133  */
1134 void
1135 bakcopy(void)
1136 {
1137     int         n,
1138                 bakchn;
1139     char        buff[8 * 1024];
1140     char       *p;
1141 
1142     /* construct file name .Bfile */
1143     for (p = in_name; *p; p++);	/* skip to end of string */
1144     while (p > in_name && *p != '/')	/* find last '/' */
1145 	p--;
1146     if (*p == '/')
1147 	p++;
1148     if (snprintf(bakfile, MAXPATHLEN, "%s.BAK", p) >= MAXPATHLEN)
1149 	    errx(1, "%s.BAK: %s", p, strerror(ENAMETOOLONG));
1150 
1151     /* copy in_name to backup file */
1152     bakchn = creat(bakfile, 0600);
1153     if (bakchn < 0)
1154 	err(1, "%s", bakfile);
1155     while ((n = read(fileno(input), buff, sizeof buff)) > 0)
1156 	if (write(bakchn, buff, n) != n)
1157 	    err(1, "%s", bakfile);
1158     if (n < 0)
1159 	err(1, "%s", in_name);
1160     close(bakchn);
1161     fclose(input);
1162 
1163     /* re-open backup file as the input file */
1164     input = fopen(bakfile, "r");
1165     if (input == NULL)
1166 	err(1, "%s", bakfile);
1167     /* now the original input file will be the output */
1168     output = fopen(in_name, "w");
1169     if (output == NULL) {
1170 	unlink(bakfile);
1171 	err(1, "%s", in_name);
1172     }
1173 }
1174