xref: /freebsd/usr.bin/indent/indent.c (revision d429ea332342fcb98d27a350d0c4944bf9aec3f9)
1 /*
2  * Copyright (c) 1985 Sun Microsystems, Inc.
3  * Copyright (c) 1976 Board of Trustees of the University of Illinois.
4  * Copyright (c) 1980, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 #ifndef lint
37 static const char copyright[] =
38 "@(#) Copyright (c) 1985 Sun Microsystems, Inc.\n\
39 @(#) Copyright (c) 1976 Board of Trustees of the University of Illinois.\n\
40 @(#) Copyright (c) 1980, 1993\n\
41 	The Regents of the University of California.  All rights reserved.\n";
42 #endif /* not lint */
43 
44 #if 0
45 #ifndef lint
46 static char sccsid[] = "@(#)indent.c	5.17 (Berkeley) 6/7/93";
47 #endif /* not lint */
48 #endif
49 
50 #include <sys/cdefs.h>
51 __FBSDID("$FreeBSD$");
52 
53 #include <sys/param.h>
54 #include <err.h>
55 #include <fcntl.h>
56 #include <unistd.h>
57 #include <stdio.h>
58 #include <stdlib.h>
59 #include <string.h>
60 #include <ctype.h>
61 #include "indent_globs.h"
62 #include "indent_codes.h"
63 #include "indent.h"
64 
65 static void bakcopy(void);
66 
67 const char *in_name = "Standard Input";	/* will always point to name of input
68 					 * file */
69 const char *out_name = "Standard Output";	/* will always point to name
70 						 * of output file */
71 char        bakfile[MAXPATHLEN] = "";
72 
73 extern int  found_err;	/* flag set in diagN() on error */
74 
75 int
76 main(int argc, char **argv)
77 {
78 
79     int         dec_ind;	/* current indentation for declarations */
80     int         di_stack[20];	/* a stack of structure indentation levels */
81     int         flushed_nl;	/* used when buffering up comments to remember
82 				 * that a newline was passed over */
83     int         force_nl;	/* when true, code must be broken */
84     int         hd_type = 0;	/* used to store type of stmt for if (...),
85 				 * for (...), etc */
86     int		i;		/* local loop counter */
87     int         scase;		/* set to true when we see a case, so we will
88 				 * know what to do with the following colon */
89     int         sp_sw;		/* when true, we are in the expression of
90 				 * if(...), while(...), etc. */
91     int         squest;		/* when this is positive, we have seen a ?
92 				 * without the matching : in a <c>?<s>:<s>
93 				 * construct */
94     const char *t_ptr;		/* used for copying tokens */
95     int		tabs_to_var;	/* true if using tabs to indent to var name */
96     int         type_code;	/* the type of token, returned by lexi */
97 
98     int         last_else = 0;	/* true iff last keyword was an else */
99 
100 
101     /*-----------------------------------------------*\
102     |		      INITIALIZATION		      |
103     \*-----------------------------------------------*/
104 
105     found_err = 0;
106 
107     ps.p_stack[0] = stmt;	/* this is the parser's stack */
108     ps.last_nl = true;		/* this is true if the last thing scanned was
109 				 * a newline */
110     ps.last_token = semicolon;
111     combuf = (char *) malloc(bufsize);
112     if (combuf == NULL)
113 	err(1, NULL);
114     labbuf = (char *) malloc(bufsize);
115     if (labbuf == NULL)
116 	err(1, NULL);
117     codebuf = (char *) malloc(bufsize);
118     if (codebuf == NULL)
119 	err(1, NULL);
120     tokenbuf = (char *) malloc(bufsize);
121     if (tokenbuf == NULL)
122 	err(1, NULL);
123     l_com = combuf + bufsize - 5;
124     l_lab = labbuf + bufsize - 5;
125     l_code = codebuf + bufsize - 5;
126     l_token = tokenbuf + bufsize - 5;
127     combuf[0] = codebuf[0] = labbuf[0] = ' ';	/* set up code, label, and
128 						 * comment buffers */
129     combuf[1] = codebuf[1] = labbuf[1] = '\0';
130     ps.else_if = 1;		/* Default else-if special processing to on */
131     s_lab = e_lab = labbuf + 1;
132     s_code = e_code = codebuf + 1;
133     s_com = e_com = combuf + 1;
134     s_token = e_token = tokenbuf + 1;
135 
136     in_buffer = (char *) malloc(10);
137     if (in_buffer == NULL)
138 	err(1, NULL);
139     in_buffer_limit = in_buffer + 8;
140     buf_ptr = buf_end = in_buffer;
141     line_no = 1;
142     had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
143     sp_sw = force_nl = false;
144     ps.in_or_st = false;
145     ps.bl_line = true;
146     dec_ind = 0;
147     di_stack[ps.dec_nest = 0] = 0;
148     ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
149 
150     scase = ps.pcase = false;
151     squest = 0;
152     sc_end = 0;
153     bp_save = 0;
154     be_save = 0;
155 
156     output = 0;
157 
158     /*--------------------------------------------------*\
159     |   		COMMAND LINE SCAN		 |
160     \*--------------------------------------------------*/
161 
162 #ifdef undef
163     max_col = 78;		/* -l78 */
164     lineup_to_parens = 1;	/* -lp */
165     ps.ljust_decl = 0;		/* -ndj */
166     ps.com_ind = 33;		/* -c33 */
167     star_comment_cont = 1;	/* -sc */
168     ps.ind_size = 8;		/* -i8 */
169     verbose = 0;
170     ps.decl_indent = 16;	/* -di16 */
171     ps.local_decl_indent = -1;	/* if this is not set to some nonnegative value
172 				 * by an arg, we will set this equal to
173 				 * ps.decl_ind */
174     ps.indent_parameters = 1;	/* -ip */
175     ps.decl_com_ind = 0;	/* if this is not set to some positive value
176 				 * by an arg, we will set this equal to
177 				 * ps.com_ind */
178     btype_2 = 1;		/* -br */
179     cuddle_else = 1;		/* -ce */
180     ps.unindent_displace = 0;	/* -d0 */
181     ps.case_indent = 0;		/* -cli0 */
182     format_block_comments = 1;	/* -fcb */
183     format_col1_comments = 1;	/* -fc1 */
184     procnames_start_line = 1;	/* -psl */
185     proc_calls_space = 0;	/* -npcs */
186     comment_delimiter_on_blankline = 1;	/* -cdb */
187     ps.leave_comma = 1;		/* -nbc */
188 #endif
189 
190     for (i = 1; i < argc; ++i)
191 	if (strcmp(argv[i], "-npro") == 0)
192 	    break;
193     set_defaults();
194     if (i >= argc)
195 	set_profile();
196 
197     for (i = 1; i < argc; ++i) {
198 
199 	/*
200 	 * look thru args (if any) for changes to defaults
201 	 */
202 	if (argv[i][0] != '-') {/* no flag on parameter */
203 	    if (input == 0) {	/* we must have the input file */
204 		in_name = argv[i];	/* remember name of input file */
205 		input = fopen(in_name, "r");
206 		if (input == 0)		/* check for open error */
207 			err(1, "%s", in_name);
208 		continue;
209 	    }
210 	    else if (output == 0) {	/* we have the output file */
211 		out_name = argv[i];	/* remember name of output file */
212 		if (strcmp(in_name, out_name) == 0) {	/* attempt to overwrite
213 							 * the file */
214 		    errx(1, "input and output files must be different");
215 		}
216 		output = fopen(out_name, "w");
217 		if (output == 0)	/* check for create error */
218 			err(1, "%s", out_name);
219 		continue;
220 	    }
221 	    errx(1, "unknown parameter: %s", argv[i]);
222 	}
223 	else
224 	    set_option(argv[i]);
225     }				/* end of for */
226     if (input == 0)
227 	input = stdin;
228     if (output == 0) {
229 	if (troff || input == stdin)
230 	    output = stdout;
231 	else {
232 	    out_name = in_name;
233 	    bakcopy();
234 	}
235     }
236     if (ps.com_ind <= 1)
237 	ps.com_ind = 2;		/* dont put normal comments before column 2 */
238     if (troff) {
239 	if (bodyf.font[0] == 0)
240 	    parsefont(&bodyf, "R");
241 	if (scomf.font[0] == 0)
242 	    parsefont(&scomf, "I");
243 	if (blkcomf.font[0] == 0)
244 	    blkcomf = scomf, blkcomf.size += 2;
245 	if (boxcomf.font[0] == 0)
246 	    boxcomf = blkcomf;
247 	if (stringf.font[0] == 0)
248 	    parsefont(&stringf, "L");
249 	if (keywordf.font[0] == 0)
250 	    parsefont(&keywordf, "B");
251 	writefdef(&bodyf, 'B');
252 	writefdef(&scomf, 'C');
253 	writefdef(&blkcomf, 'L');
254 	writefdef(&boxcomf, 'X');
255 	writefdef(&stringf, 'S');
256 	writefdef(&keywordf, 'K');
257     }
258     if (block_comment_max_col <= 0)
259 	block_comment_max_col = max_col;
260     if (ps.local_decl_indent < 0)	/* if not specified by user, set this */
261 	ps.local_decl_indent = ps.decl_indent;
262     if (ps.decl_com_ind <= 0)	/* if not specified by user, set this */
263 	ps.decl_com_ind = ps.ljust_decl ? (ps.com_ind <= 10 ? 2 : ps.com_ind - 8) : ps.com_ind;
264     if (continuation_indent == 0)
265 	continuation_indent = ps.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 = ((col - 1) & ~7) + 9;
278 	    else
279 		break;
280 	    p++;
281 	}
282 	if (col > ps.ind_size)
283 	    ps.ind_level = ps.i_l_follow = col / ps.ind_size;
284     }
285     if (troff) {
286 	const char *p = in_name,
287 	           *beg = in_name;
288 
289 	while (*p)
290 	    if (*p++ == '/')
291 		beg = p;
292 	fprintf(output, ".Fn \"%s\"\n", beg);
293     }
294     /*
295      * START OF MAIN LOOP
296      */
297 
298     while (1) {			/* this is the main loop.  it will go until we
299 				 * reach eof */
300 	int         is_procname;
301 
302 	type_code = lexi();	/* lexi reads one token.  The actual
303 				 * characters read are stored in "token". lexi
304 				 * returns a code indicating the type of token */
305 	is_procname = ps.procname[0];
306 
307 	/*
308 	 * The following code moves everything following an if (), while (),
309 	 * else, etc. up to the start of the following stmt to a buffer. This
310 	 * allows proper handling of both kinds of brace placement.
311 	 */
312 
313 	flushed_nl = false;
314 	while (ps.search_brace) {	/* if we scanned an if(), while(),
315 					 * etc., we might need to copy stuff
316 					 * into a buffer we must loop, copying
317 					 * stuff into save_com, until we find
318 					 * the start of the stmt which follows
319 					 * the if, or whatever */
320 	    switch (type_code) {
321 	    case newline:
322 		++line_no;
323 		flushed_nl = true;
324 	    case form_feed:
325 		break;		/* form feeds and newlines found here will be
326 				 * ignored */
327 
328 	    case lbrace:	/* this is a brace that starts the compound
329 				 * stmt */
330 		if (sc_end == 0) {	/* ignore buffering if a comment wasnt
331 					 * stored up */
332 		    ps.search_brace = false;
333 		    goto check_type;
334 		}
335 		if (btype_2) {
336 		    save_com[0] = '{';	/* we either want to put the brace
337 					 * right after the if */
338 		    goto sw_buffer;	/* go to common code to get out of
339 					 * this loop */
340 		}
341 	    case comment:	/* we have a comment, so we must copy it into
342 				 * the buffer */
343 		if (!flushed_nl || sc_end != 0) {
344 		    if (sc_end == 0) {	/* if this is the first comment, we
345 					 * must set up the buffer */
346 			save_com[0] = save_com[1] = ' ';
347 			sc_end = &(save_com[2]);
348 		    }
349 		    else {
350 			*sc_end++ = '\n';	/* add newline between
351 						 * comments */
352 			*sc_end++ = ' ';
353 			--line_no;
354 		    }
355 		    *sc_end++ = '/';	/* copy in start of comment */
356 		    *sc_end++ = '*';
357 
358 		    for (;;) {	/* loop until we get to the end of the comment */
359 			*sc_end = *buf_ptr++;
360 			if (buf_ptr >= buf_end)
361 			    fill_buffer();
362 
363 			if (*sc_end++ == '*' && *buf_ptr == '/')
364 			    break;	/* we are at end of comment */
365 
366 			if (sc_end >= &(save_com[sc_size])) {	/* check for temp buffer
367 								 * overflow */
368 			    diag2(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
369 			    fflush(output);
370 			    exit(1);
371 			}
372 		    }
373 		    *sc_end++ = '/';	/* add ending slash */
374 		    if (++buf_ptr >= buf_end)	/* get past / in buffer */
375 			fill_buffer();
376 		    break;
377 		}
378 	    default:		/* it is the start of a normal statement */
379 		if (flushed_nl)	/* if we flushed a newline, make sure it is
380 				 * put back */
381 		    force_nl = true;
382 		if ((type_code == sp_paren && *token == 'i'
383 			&& last_else && ps.else_if)
384 			|| (type_code == sp_nparen && *token == 'e'
385 			&& e_code != s_code && e_code[-1] == '}'))
386 		    force_nl = false;
387 
388 		if (sc_end == 0) {	/* ignore buffering if comment wasnt
389 					 * saved up */
390 		    ps.search_brace = false;
391 		    goto check_type;
392 		}
393 		if (force_nl) {	/* if we should insert a nl here, put it into
394 				 * the buffer */
395 		    force_nl = false;
396 		    --line_no;	/* this will be re-increased when the nl is
397 				 * read from the buffer */
398 		    *sc_end++ = '\n';
399 		    *sc_end++ = ' ';
400 		    if (verbose && !flushed_nl)	/* print error msg if the line
401 						 * was not already broken */
402 			diag2(0, "Line broken");
403 		    flushed_nl = false;
404 		}
405 		for (t_ptr = token; *t_ptr; ++t_ptr)
406 		    *sc_end++ = *t_ptr;	/* copy token into temp buffer */
407 		ps.procname[0] = 0;
408 
409 	sw_buffer:
410 		ps.search_brace = false;	/* stop looking for start of
411 						 * stmt */
412 		bp_save = buf_ptr;	/* save current input buffer */
413 		be_save = buf_end;
414 		buf_ptr = save_com;	/* fix so that subsequent calls to
415 					 * lexi will take tokens out of
416 					 * save_com */
417 		*sc_end++ = ' ';/* add trailing blank, just in case */
418 		buf_end = sc_end;
419 		sc_end = 0;
420 		break;
421 	    }			/* end of switch */
422 	    if (type_code != 0)	/* we must make this check, just in case there
423 				 * was an unexpected EOF */
424 		type_code = lexi();	/* read another token */
425 	    /* if (ps.search_brace) ps.procname[0] = 0; */
426 	    if ((is_procname = ps.procname[0]) && flushed_nl
427 		    && !procnames_start_line && ps.in_decl
428 		    && type_code == ident)
429 		flushed_nl = 0;
430 	}			/* end of while (search_brace) */
431 	last_else = 0;
432 check_type:
433 	if (type_code == 0) {	/* we got eof */
434 	    if (s_lab != e_lab || s_code != e_code
435 		    || s_com != e_com)	/* must dump end of line */
436 		dump_line();
437 	    if (ps.tos > 1)	/* check for balanced braces */
438 		diag2(1, "Stuff missing from end of file");
439 
440 	    if (verbose) {
441 		printf("There were %d output lines and %d comments\n",
442 		       ps.out_lines, ps.out_coms);
443 		printf("(Lines with comments)/(Lines with code): %6.3f\n",
444 		       (1.0 * ps.com_lines) / code_lines);
445 	    }
446 	    fflush(output);
447 	    exit(found_err);
448 	}
449 	if (
450 		(type_code != comment) &&
451 		(type_code != newline) &&
452 		(type_code != preesc) &&
453 		(type_code != form_feed)) {
454 	    if (force_nl &&
455 		    (type_code != semicolon) &&
456 		    (type_code != lbrace || !btype_2)) {
457 		/* we should force a broken line here */
458 		if (verbose && !flushed_nl)
459 		    diag2(0, "Line broken");
460 		flushed_nl = false;
461 		dump_line();
462 		ps.want_blank = false;	/* dont insert blank at line start */
463 		force_nl = false;
464 	    }
465 	    ps.in_stmt = true;	/* turn on flag which causes an extra level of
466 				 * indentation. this is turned off by a ; or
467 				 * '}' */
468 	    if (s_com != e_com) {	/* the turkey has embedded a comment
469 					 * in a line. fix it */
470 		*e_code++ = ' ';
471 		for (t_ptr = s_com; *t_ptr; ++t_ptr) {
472 		    CHECK_SIZE_CODE;
473 		    *e_code++ = *t_ptr;
474 		}
475 		*e_code++ = ' ';
476 		*e_code = '\0';	/* null terminate code sect */
477 		ps.want_blank = false;
478 		e_com = s_com;
479 	    }
480 	}
481 	else if (type_code != comment)	/* preserve force_nl thru a comment */
482 	    force_nl = false;	/* cancel forced newline after newline, form
483 				 * feed, etc */
484 
485 
486 
487 	/*-----------------------------------------------------*\
488 	|	   do switch on type of token scanned		|
489 	\*-----------------------------------------------------*/
490 	CHECK_SIZE_CODE;
491 	switch (type_code) {	/* now, decide what to do with the token */
492 
493 	case form_feed:	/* found a form feed in line */
494 	    ps.use_ff = true;	/* a form feed is treated much like a newline */
495 	    dump_line();
496 	    ps.want_blank = false;
497 	    break;
498 
499 	case newline:
500 	    if (ps.last_token != comma || ps.p_l_follow > 0
501 		    || !ps.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
502 		dump_line();
503 		ps.want_blank = false;
504 	    }
505 	    ++line_no;		/* keep track of input line number */
506 	    break;
507 
508 	case lparen:		/* got a '(' or '[' */
509 	    ++ps.p_l_follow;	/* count parens to make Healy happy */
510 	    if (ps.want_blank && *token != '[' &&
511 		    (ps.last_token != ident || proc_calls_space
512 	      || (ps.its_a_keyword && (!ps.sizeof_keyword || Bill_Shannon))))
513 		*e_code++ = ' ';
514 	    if (ps.in_decl && !ps.block_init)
515 		if (troff && !ps.dumped_decl_indent && !is_procname && ps.last_token == decl) {
516 		    ps.dumped_decl_indent = 1;
517 		    sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
518 		    e_code += strlen(e_code);
519 		}
520 		else {
521 		    while ((e_code - s_code) < dec_ind) {
522 			CHECK_SIZE_CODE;
523 			*e_code++ = ' ';
524 		    }
525 		    *e_code++ = token[0];
526 		}
527 	    else
528 		*e_code++ = token[0];
529 	    ps.paren_indents[ps.p_l_follow - 1] = e_code - s_code;
530 	    if (sp_sw && ps.p_l_follow == 1 && extra_expression_indent
531 		    && ps.paren_indents[0] < 2 * ps.ind_size)
532 		ps.paren_indents[0] = 2 * ps.ind_size;
533 	    ps.want_blank = false;
534 	    if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
535 		/*
536 		 * this is a kluge to make sure that declarations will be
537 		 * aligned right if proc decl has an explicit type on it, i.e.
538 		 * "int a(x) {..."
539 		 */
540 		parse(semicolon);	/* I said this was a kluge... */
541 		ps.in_or_st = false;	/* turn off flag for structure decl or
542 					 * initialization */
543 	    }
544 	    if (ps.sizeof_keyword)
545 		ps.sizeof_mask |= 1 << ps.p_l_follow;
546 	    break;
547 
548 	case rparen:		/* got a ')' or ']' */
549 	    rparen_count--;
550 	    if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.sizeof_mask) {
551 		ps.last_u_d = true;
552 		ps.cast_mask &= (1 << ps.p_l_follow) - 1;
553 		ps.want_blank = false;
554 	    } else
555 		ps.want_blank = true;
556 	    ps.sizeof_mask &= (1 << ps.p_l_follow) - 1;
557 	    if (--ps.p_l_follow < 0) {
558 		ps.p_l_follow = 0;
559 		diag3(0, "Extra %c", *token);
560 	    }
561 	    if (e_code == s_code)	/* if the paren starts the line */
562 		ps.paren_level = ps.p_l_follow;	/* then indent it */
563 
564 	    *e_code++ = token[0];
565 
566 	    if (sp_sw && (ps.p_l_follow == 0)) {	/* check for end of if
567 							 * (...), or some such */
568 		sp_sw = false;
569 		force_nl = true;/* must force newline after if */
570 		ps.last_u_d = true;	/* inform lexi that a following
571 					 * operator is unary */
572 		ps.in_stmt = false;	/* dont use stmt continuation
573 					 * indentation */
574 
575 		parse(hd_type);	/* let parser worry about if, or whatever */
576 	    }
577 	    ps.search_brace = btype_2;	/* this should insure that constructs
578 					 * such as main(){...} and int[]{...}
579 					 * have their braces put in the right
580 					 * place */
581 	    break;
582 
583 	case unary_op:		/* this could be any unary operation */
584 	    if (ps.want_blank)
585 		*e_code++ = ' ';
586 
587 	    if (troff && !ps.dumped_decl_indent && ps.in_decl && !is_procname) {
588 		sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
589 		ps.dumped_decl_indent = 1;
590 		e_code += strlen(e_code);
591 	    }
592 	    else {
593 		const char *res = token;
594 
595 		if (ps.in_decl && !ps.block_init) {	/* if this is a unary op
596 							 * in a declaration, we
597 							 * should indent this
598 							 * token */
599 		    for (i = 0; token[i]; ++i);	/* find length of token */
600 		    while ((e_code - s_code) < (dec_ind - i)) {
601 			CHECK_SIZE_CODE;
602 			*e_code++ = ' ';	/* pad it */
603 		    }
604 		}
605 		if (troff && token[0] == '-' && token[1] == '>')
606 		    res = "\\(->";
607 		for (t_ptr = res; *t_ptr; ++t_ptr) {
608 		    CHECK_SIZE_CODE;
609 		    *e_code++ = *t_ptr;
610 		}
611 	    }
612 	    ps.want_blank = false;
613 	    break;
614 
615 	case binary_op:	/* any binary operation */
616 	    if (ps.want_blank)
617 		*e_code++ = ' ';
618 	    {
619 		const char *res = token;
620 
621 		if (troff)
622 		    switch (token[0]) {
623 		    case '<':
624 			if (token[1] == '=')
625 			    res = "\\(<=";
626 			break;
627 		    case '>':
628 			if (token[1] == '=')
629 			    res = "\\(>=";
630 			break;
631 		    case '!':
632 			if (token[1] == '=')
633 			    res = "\\(!=";
634 			break;
635 		    case '|':
636 			if (token[1] == '|')
637 			    res = "\\(br\\(br";
638 			else if (token[1] == 0)
639 			    res = "\\(br";
640 			break;
641 		    }
642 		for (t_ptr = res; *t_ptr; ++t_ptr) {
643 		    CHECK_SIZE_CODE;
644 		    *e_code++ = *t_ptr;	/* move the operator */
645 		}
646 	    }
647 	    ps.want_blank = true;
648 	    break;
649 
650 	case postop:		/* got a trailing ++ or -- */
651 	    *e_code++ = token[0];
652 	    *e_code++ = token[1];
653 	    ps.want_blank = true;
654 	    break;
655 
656 	case question:		/* got a ? */
657 	    squest++;		/* this will be used when a later colon
658 				 * appears so we can distinguish the
659 				 * <c>?<n>:<n> construct */
660 	    if (ps.want_blank)
661 		*e_code++ = ' ';
662 	    *e_code++ = '?';
663 	    ps.want_blank = true;
664 	    break;
665 
666 	case casestmt:		/* got word 'case' or 'default' */
667 	    scase = true;	/* so we can process the later colon properly */
668 	    goto copy_id;
669 
670 	case colon:		/* got a ':' */
671 	    if (squest > 0) {	/* it is part of the <c>?<n>: <n> construct */
672 		--squest;
673 		if (ps.want_blank)
674 		    *e_code++ = ' ';
675 		*e_code++ = ':';
676 		ps.want_blank = true;
677 		break;
678 	    }
679 	    if (ps.in_decl) {
680 		*e_code++ = ':';
681 		ps.want_blank = false;
682 		break;
683 	    }
684 	    ps.in_stmt = false;	/* seeing a label does not imply we are in a
685 				 * stmt */
686 	    for (t_ptr = s_code; *t_ptr; ++t_ptr)
687 		*e_lab++ = *t_ptr;	/* turn everything so far into a label */
688 	    e_code = s_code;
689 	    *e_lab++ = ':';
690 	    *e_lab++ = ' ';
691 	    *e_lab = '\0';
692 
693 	    force_nl = ps.pcase = scase;	/* ps.pcase will be used by
694 						 * dump_line to decide how to
695 						 * indent the label. force_nl
696 						 * will force a case n: to be
697 						 * on a line by itself */
698 	    scase = false;
699 	    ps.want_blank = false;
700 	    break;
701 
702 	case semicolon:	/* got a ';' */
703 	    ps.in_or_st = false;/* we are not in an initialization or
704 				 * structure declaration */
705 	    scase = false;	/* these will only need resetting in an error */
706 	    squest = 0;
707 	    if (ps.last_token == rparen && rparen_count == 0)
708 		ps.in_parameter_declaration = 0;
709 	    ps.cast_mask = 0;
710 	    ps.sizeof_mask = 0;
711 	    ps.block_init = 0;
712 	    ps.block_init_level = 0;
713 	    ps.just_saw_decl--;
714 
715 	    if (ps.in_decl && s_code == e_code && !ps.block_init)
716 		while ((e_code - s_code) < (dec_ind - 1)) {
717 		    CHECK_SIZE_CODE;
718 		    *e_code++ = ' ';
719 		}
720 
721 	    ps.in_decl = (ps.dec_nest > 0);	/* if we were in a first level
722 						 * structure declaration, we
723 						 * arent any more */
724 
725 	    if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
726 
727 		/*
728 		 * This should be true iff there were unbalanced parens in the
729 		 * stmt.  It is a bit complicated, because the semicolon might
730 		 * be in a for stmt
731 		 */
732 		diag2(1, "Unbalanced parens");
733 		ps.p_l_follow = 0;
734 		if (sp_sw) {	/* this is a check for an if, while, etc. with
735 				 * unbalanced parens */
736 		    sp_sw = false;
737 		    parse(hd_type);	/* dont lose the if, or whatever */
738 		}
739 	    }
740 	    *e_code++ = ';';
741 	    ps.want_blank = true;
742 	    ps.in_stmt = (ps.p_l_follow > 0);	/* we are no longer in the
743 						 * middle of a stmt */
744 
745 	    if (!sp_sw) {	/* if not if for (;;) */
746 		parse(semicolon);	/* let parser know about end of stmt */
747 		force_nl = true;/* force newline after an end of stmt */
748 	    }
749 	    break;
750 
751 	case lbrace:		/* got a '{' */
752 	    ps.in_stmt = false;	/* dont indent the {} */
753 	    if (!ps.block_init)
754 		force_nl = true;/* force other stuff on same line as '{' onto
755 				 * new line */
756 	    else if (ps.block_init_level <= 0)
757 		ps.block_init_level = 1;
758 	    else
759 		ps.block_init_level++;
760 
761 	    if (s_code != e_code && !ps.block_init) {
762 		if (!btype_2) {
763 		    dump_line();
764 		    ps.want_blank = false;
765 		}
766 		else if (ps.in_parameter_declaration && !ps.in_or_st) {
767 		    ps.i_l_follow = 0;
768 		    if (function_brace_split) {	/* dump the line prior to the
769 						 * brace ... */
770 			dump_line();
771 			ps.want_blank = false;
772 		    } else	/* add a space between the decl and brace */
773 			ps.want_blank = true;
774 		}
775 	    }
776 	    if (ps.in_parameter_declaration)
777 		prefix_blankline_requested = 0;
778 
779 	    if (ps.p_l_follow > 0) {	/* check for preceding unbalanced
780 					 * parens */
781 		diag2(1, "Unbalanced parens");
782 		ps.p_l_follow = 0;
783 		if (sp_sw) {	/* check for unclosed if, for, etc. */
784 		    sp_sw = false;
785 		    parse(hd_type);
786 		    ps.ind_level = ps.i_l_follow;
787 		}
788 	    }
789 	    if (s_code == e_code)
790 		ps.ind_stmt = false;	/* dont put extra indentation on line
791 					 * with '{' */
792 	    if (ps.in_decl && ps.in_or_st) {	/* this is either a structure
793 						 * declaration or an init */
794 		di_stack[ps.dec_nest++] = dec_ind;
795 		/* ?		dec_ind = 0; */
796 	    }
797 	    else {
798 		ps.decl_on_line = false;	/* we cant be in the middle of
799 						 * a declaration, so dont do
800 						 * special indentation of
801 						 * comments */
802 		if (blanklines_after_declarations_at_proctop
803 			&& ps.in_parameter_declaration)
804 		    postfix_blankline_requested = 1;
805 		ps.in_parameter_declaration = 0;
806 	    }
807 	    dec_ind = 0;
808 	    parse(lbrace);	/* let parser know about this */
809 	    if (ps.want_blank)	/* put a blank before '{' if '{' is not at
810 				 * start of line */
811 		*e_code++ = ' ';
812 	    ps.want_blank = false;
813 	    *e_code++ = '{';
814 	    ps.just_saw_decl = 0;
815 	    break;
816 
817 	case rbrace:		/* got a '}' */
818 	    if (ps.p_stack[ps.tos] == decl && !ps.block_init)	/* semicolons can be
819 								 * omitted in
820 								 * declarations */
821 		parse(semicolon);
822 	    if (ps.p_l_follow) {/* check for unclosed if, for, else. */
823 		diag2(1, "Unbalanced parens");
824 		ps.p_l_follow = 0;
825 		sp_sw = false;
826 	    }
827 	    ps.just_saw_decl = 0;
828 	    ps.block_init_level--;
829 	    if (s_code != e_code && !ps.block_init) {	/* '}' must be first on
830 							 * line */
831 		if (verbose)
832 		    diag2(0, "Line broken");
833 		dump_line();
834 	    }
835 	    *e_code++ = '}';
836 	    ps.want_blank = true;
837 	    ps.in_stmt = ps.ind_stmt = false;
838 	    if (ps.dec_nest > 0) {	/* we are in multi-level structure
839 					 * declaration */
840 		dec_ind = di_stack[--ps.dec_nest];
841 		if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
842 		    ps.just_saw_decl = 2;
843 		ps.in_decl = true;
844 	    }
845 	    prefix_blankline_requested = 0;
846 	    parse(rbrace);	/* let parser know about this */
847 	    ps.search_brace = cuddle_else && ps.p_stack[ps.tos] == ifhead
848 		&& ps.il[ps.tos] >= ps.ind_level;
849 	    if (ps.tos <= 1 && blanklines_after_procs && ps.dec_nest <= 0)
850 		postfix_blankline_requested = 1;
851 	    break;
852 
853 	case swstmt:		/* got keyword "switch" */
854 	    sp_sw = true;
855 	    hd_type = swstmt;	/* keep this for when we have seen the
856 				 * expression */
857 	    goto copy_id;	/* go move the token into buffer */
858 
859 	case sp_paren:		/* token is if, while, for */
860 	    sp_sw = true;	/* the interesting stuff is done after the
861 				 * expression is scanned */
862 	    hd_type = (*token == 'i' ? ifstmt :
863 		       (*token == 'w' ? whilestmt : forstmt));
864 
865 	    /*
866 	     * remember the type of header for later use by parser
867 	     */
868 	    goto copy_id;	/* copy the token into line */
869 
870 	case sp_nparen:	/* got else, do */
871 	    ps.in_stmt = false;
872 	    if (*token == 'e') {
873 		if (e_code != s_code && (!cuddle_else || e_code[-1] != '}')) {
874 		    if (verbose)
875 			diag2(0, "Line broken");
876 		    dump_line();/* make sure this starts a line */
877 		    ps.want_blank = false;
878 		}
879 		force_nl = true;/* also, following stuff must go onto new line */
880 		last_else = 1;
881 		parse(elselit);
882 	    }
883 	    else {
884 		if (e_code != s_code) {	/* make sure this starts a line */
885 		    if (verbose)
886 			diag2(0, "Line broken");
887 		    dump_line();
888 		    ps.want_blank = false;
889 		}
890 		force_nl = true;/* also, following stuff must go onto new line */
891 		last_else = 0;
892 		parse(dolit);
893 	    }
894 	    goto copy_id;	/* move the token into line */
895 
896 	case decl:		/* we have a declaration type (int, register,
897 				 * etc.) */
898 	    parse(decl);	/* let parser worry about indentation */
899 	    if (ps.last_token == rparen && ps.tos <= 1) {
900 		ps.in_parameter_declaration = 1;
901 		if (s_code != e_code) {
902 		    dump_line();
903 		    ps.want_blank = 0;
904 		}
905 	    }
906 	    if (ps.in_parameter_declaration && ps.indent_parameters && ps.dec_nest == 0) {
907 		ps.ind_level = ps.i_l_follow = 1;
908 		ps.ind_stmt = 0;
909 	    }
910 	    ps.in_or_st = true;	/* this might be a structure or initialization
911 				 * declaration */
912 	    ps.in_decl = ps.decl_on_line = true;
913 	    if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
914 		ps.just_saw_decl = 2;
915 	    prefix_blankline_requested = 0;
916 	    for (i = 0; token[i++];);	/* get length of token */
917 
918 	    if (ps.ind_level == 0 || ps.dec_nest > 0) {
919 		/* global variable or struct member in local variable */
920 		dec_ind = ps.decl_indent > 0 ? ps.decl_indent : i;
921 		tabs_to_var = (use_tabs ? ps.decl_indent > 0 : 0);
922 	    } else {
923 		/* local variable */
924 		dec_ind = ps.local_decl_indent > 0 ? ps.local_decl_indent : i;
925 		tabs_to_var = (use_tabs ? ps.local_decl_indent > 0 : 0);
926 	    }
927 	    goto copy_id;
928 
929 	case ident:		/* got an identifier or constant */
930 	    if (ps.in_decl) {	/* if we are in a declaration, we must indent
931 				 * identifier */
932 		if (is_procname == 0 || !procnames_start_line) {
933 		    if (!ps.block_init) {
934 			if (troff && !ps.dumped_decl_indent) {
935 			    if (ps.want_blank)
936 				*e_code++ = ' ';
937 			    ps.want_blank = false;
938 			    sprintf(e_code, "\n.De %dp+\200p\n", dec_ind * 7);
939 			    ps.dumped_decl_indent = 1;
940 			    e_code += strlen(e_code);
941 			} else {
942 			    int cur_dec_ind;
943 			    int pos, startpos;
944 
945 			    /*
946 			     * in order to get the tab math right for
947 			     * indentations that are not multiples of 8 we
948 			     * need to modify both startpos and dec_ind
949 			     * (cur_dec_ind) here by eight minus the
950 			     * remainder of the current starting column
951 			     * divided by eight. This seems to be a
952 			     * properly working fix
953 			     */
954 			    startpos = e_code - s_code;
955 			    cur_dec_ind = dec_ind;
956 			    pos = startpos;
957 			    if ((ps.ind_level * ps.ind_size) % 8 != 0) {
958 				pos += (ps.ind_level * ps.ind_size) % 8;
959 				cur_dec_ind += (ps.ind_level * ps.ind_size) % 8;
960 			    }
961 
962 			    if (tabs_to_var) {
963 				while ((pos & ~7) + 8 <= cur_dec_ind) {
964 				    CHECK_SIZE_CODE;
965 				    *e_code++ = '\t';
966 				    pos = (pos & ~7) + 8;
967 				}
968 			    }
969 			    while (pos < cur_dec_ind) {
970 				CHECK_SIZE_CODE;
971 				*e_code++ = ' ';
972 				pos++;
973 			    }
974 			    if (ps.want_blank && e_code - s_code == startpos)
975 				*e_code++ = ' ';
976 			    ps.want_blank = false;
977 			}
978 		    }
979 		} else {
980 		    if (ps.want_blank)
981 			*e_code++ = ' ';
982 		    ps.want_blank = false;
983 		    if (dec_ind && s_code != e_code)
984 			dump_line();
985 		    dec_ind = 0;
986 		}
987 	    }
988 	    else if (sp_sw && ps.p_l_follow == 0) {
989 		sp_sw = false;
990 		force_nl = true;
991 		ps.last_u_d = true;
992 		ps.in_stmt = false;
993 		parse(hd_type);
994 	    }
995     copy_id:
996 	    if (ps.want_blank)
997 		*e_code++ = ' ';
998 	    if (troff && ps.its_a_keyword) {
999 		e_code = chfont(&bodyf, &keywordf, e_code);
1000 		for (t_ptr = token; *t_ptr; ++t_ptr) {
1001 		    CHECK_SIZE_CODE;
1002 		    *e_code++ = keywordf.allcaps && islower(*t_ptr)
1003 			? toupper(*t_ptr) : *t_ptr;
1004 		}
1005 		e_code = chfont(&keywordf, &bodyf, e_code);
1006 	    }
1007 	    else
1008 		for (t_ptr = token; *t_ptr; ++t_ptr) {
1009 		    CHECK_SIZE_CODE;
1010 		    *e_code++ = *t_ptr;
1011 		}
1012 	    ps.want_blank = true;
1013 	    break;
1014 
1015 	case period:		/* treat a period kind of like a binary
1016 				 * operation */
1017 	    *e_code++ = '.';	/* move the period into line */
1018 	    ps.want_blank = false;	/* dont put a blank after a period */
1019 	    break;
1020 
1021 	case comma:
1022 	    ps.want_blank = (s_code != e_code);	/* only put blank after comma
1023 						 * if comma does not start the
1024 						 * line */
1025 	    if (ps.in_decl && is_procname == 0 && !ps.block_init)
1026 		while ((e_code - s_code) < (dec_ind - 1)) {
1027 		    CHECK_SIZE_CODE;
1028 		    *e_code++ = ' ';
1029 		}
1030 
1031 	    *e_code++ = ',';
1032 	    if (ps.p_l_follow == 0) {
1033 		if (ps.block_init_level <= 0)
1034 		    ps.block_init = 0;
1035 		if (break_comma && (!ps.leave_comma || compute_code_target() + (e_code - s_code) > max_col - 8))
1036 		    force_nl = true;
1037 	    }
1038 	    break;
1039 
1040 	case preesc:		/* got the character '#' */
1041 	    if ((s_com != e_com) ||
1042 		    (s_lab != e_lab) ||
1043 		    (s_code != e_code))
1044 		dump_line();
1045 	    *e_lab++ = '#';	/* move whole line to 'label' buffer */
1046 	    {
1047 		int         in_comment = 0;
1048 		int         com_start = 0;
1049 		char        quote = 0;
1050 		int         com_end = 0;
1051 
1052 		while (*buf_ptr == ' ' || *buf_ptr == '\t') {
1053 		    buf_ptr++;
1054 		    if (buf_ptr >= buf_end)
1055 			fill_buffer();
1056 		}
1057 		while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
1058 		    CHECK_SIZE_LAB;
1059 		    *e_lab = *buf_ptr++;
1060 		    if (buf_ptr >= buf_end)
1061 			fill_buffer();
1062 		    switch (*e_lab++) {
1063 		    case BACKSLASH:
1064 			if (troff)
1065 			    *e_lab++ = BACKSLASH;
1066 			if (!in_comment) {
1067 			    *e_lab++ = *buf_ptr++;
1068 			    if (buf_ptr >= buf_end)
1069 				fill_buffer();
1070 			}
1071 			break;
1072 		    case '/':
1073 			if (*buf_ptr == '*' && !in_comment && !quote) {
1074 			    in_comment = 1;
1075 			    *e_lab++ = *buf_ptr++;
1076 			    com_start = e_lab - s_lab - 2;
1077 			}
1078 			break;
1079 		    case '"':
1080 			if (quote == '"')
1081 			    quote = 0;
1082 			break;
1083 		    case '\'':
1084 			if (quote == '\'')
1085 			    quote = 0;
1086 			break;
1087 		    case '*':
1088 			if (*buf_ptr == '/' && in_comment) {
1089 			    in_comment = 0;
1090 			    *e_lab++ = *buf_ptr++;
1091 			    com_end = e_lab - s_lab;
1092 			}
1093 			break;
1094 		    }
1095 		}
1096 
1097 		while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1098 		    e_lab--;
1099 		if (e_lab - s_lab == com_end && bp_save == 0) {	/* comment on
1100 								 * preprocessor line */
1101 		    if (sc_end == 0)	/* if this is the first comment, we
1102 					 * must set up the buffer */
1103 			sc_end = &(save_com[0]);
1104 		    else {
1105 			*sc_end++ = '\n';	/* add newline between
1106 						 * comments */
1107 			*sc_end++ = ' ';
1108 			--line_no;
1109 		    }
1110 		    bcopy(s_lab + com_start, sc_end, com_end - com_start);
1111 		    sc_end += com_end - com_start;
1112 		    if (sc_end >= &save_com[sc_size])
1113 			abort();
1114 		    e_lab = s_lab + com_start;
1115 		    while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1116 			e_lab--;
1117 		    bp_save = buf_ptr;	/* save current input buffer */
1118 		    be_save = buf_end;
1119 		    buf_ptr = save_com;	/* fix so that subsequent calls to
1120 					 * lexi will take tokens out of
1121 					 * save_com */
1122 		    *sc_end++ = ' ';	/* add trailing blank, just in case */
1123 		    buf_end = sc_end;
1124 		    sc_end = 0;
1125 		}
1126 		*e_lab = '\0';	/* null terminate line */
1127 		ps.pcase = false;
1128 	    }
1129 
1130 	    if (strncmp(s_lab, "#if", 3) == 0) {
1131 		if (blanklines_around_conditional_compilation) {
1132 		    int c;
1133 		    prefix_blankline_requested++;
1134 		    while ((c = getc(input)) == '\n');
1135 		    ungetc(c, input);
1136 		}
1137 		if ((size_t)ifdef_level < sizeof(state_stack)/sizeof(state_stack[0])) {
1138 		    match_state[ifdef_level].tos = -1;
1139 		    state_stack[ifdef_level++] = ps;
1140 		}
1141 		else
1142 		    diag2(1, "#if stack overflow");
1143 	    }
1144 	    else if (strncmp(s_lab, "#else", 5) == 0)
1145 		if (ifdef_level <= 0)
1146 		    diag2(1, "Unmatched #else");
1147 		else {
1148 		    match_state[ifdef_level - 1] = ps;
1149 		    ps = state_stack[ifdef_level - 1];
1150 		}
1151 	    else if (strncmp(s_lab, "#endif", 6) == 0) {
1152 		if (ifdef_level <= 0)
1153 		    diag2(1, "Unmatched #endif");
1154 		else {
1155 		    ifdef_level--;
1156 
1157 #ifdef undef
1158 		    /*
1159 		     * This match needs to be more intelligent before the
1160 		     * message is useful
1161 		     */
1162 		    if (match_state[ifdef_level].tos >= 0
1163 			  && bcmp(&ps, &match_state[ifdef_level], sizeof ps))
1164 			diag2(0, "Syntactically inconsistent #ifdef alternatives");
1165 #endif
1166 		}
1167 		if (blanklines_around_conditional_compilation) {
1168 		    postfix_blankline_requested++;
1169 		    n_real_blanklines = 0;
1170 		}
1171 	    }
1172 	    break;		/* subsequent processing of the newline
1173 				 * character will cause the line to be printed */
1174 
1175 	case comment:		/* we have gotten a / followed by * this is a biggie */
1176 	    if (flushed_nl) {	/* we should force a broken line here */
1177 		flushed_nl = false;
1178 		dump_line();
1179 		ps.want_blank = false;	/* dont insert blank at line start */
1180 		force_nl = false;
1181 	    }
1182 	    pr_comment();
1183 	    break;
1184 	}			/* end of big switch stmt */
1185 
1186 	*e_code = '\0';		/* make sure code section is null terminated */
1187 	if (type_code != comment && type_code != newline && type_code != preesc)
1188 	    ps.last_token = type_code;
1189     }				/* end of main while (1) loop */
1190 }
1191 
1192 /*
1193  * copy input file to backup file if in_name is /blah/blah/blah/file, then
1194  * backup file will be ".Bfile" then make the backup file the input and
1195  * original input file the output
1196  */
1197 static void
1198 bakcopy(void)
1199 {
1200     int         n,
1201                 bakchn;
1202     char        buff[8 * 1024];
1203     const char *p;
1204 
1205     /* construct file name .Bfile */
1206     for (p = in_name; *p; p++);	/* skip to end of string */
1207     while (p > in_name && *p != '/')	/* find last '/' */
1208 	p--;
1209     if (*p == '/')
1210 	p++;
1211     sprintf(bakfile, "%s.BAK", p);
1212 
1213     /* copy in_name to backup file */
1214     bakchn = creat(bakfile, 0600);
1215     if (bakchn < 0)
1216 	err(1, "%s", bakfile);
1217     while ((n = read(fileno(input), buff, sizeof buff)) != 0)
1218 	if (write(bakchn, buff, n) != n)
1219 	    err(1, "%s", bakfile);
1220     if (n < 0)
1221 	err(1, "%s", in_name);
1222     close(bakchn);
1223     fclose(input);
1224 
1225     /* re-open backup file as the input file */
1226     input = fopen(bakfile, "r");
1227     if (input == 0)
1228 	err(1, "%s", bakfile);
1229     /* now the original input file will be the output */
1230     output = fopen(in_name, "w");
1231     if (output == 0) {
1232 	unlink(bakfile);
1233 	err(1, "%s", in_name);
1234     }
1235 }
1236