xref: /freebsd/usr.bin/indent/indent.c (revision 87569f75a91f298c52a71823c04d41cf53c88889)
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 int
74 main(int argc, char **argv)
75 {
76 
77     int         dec_ind;	/* current indentation for declarations */
78     int         di_stack[20];	/* a stack of structure indentation levels */
79     int         flushed_nl;	/* used when buffering up comments to remember
80 				 * that a newline was passed over */
81     int         force_nl;	/* when true, code must be broken */
82     int         hd_type = 0;	/* used to store type of stmt for if (...),
83 				 * for (...), etc */
84     int		i;		/* local loop counter */
85     int         scase;		/* set to true when we see a case, so we will
86 				 * know what to do with the following colon */
87     int         sp_sw;		/* when true, we are in the expression of
88 				 * if(...), while(...), etc. */
89     int         squest;		/* when this is positive, we have seen a ?
90 				 * without the matching : in a <c>?<s>:<s>
91 				 * construct */
92     const char *t_ptr;		/* used for copying tokens */
93     int		tabs_to_var;	/* true if using tabs to indent to var name */
94     int         type_code;	/* the type of token, returned by lexi */
95 
96     int         last_else = 0;	/* true iff last keyword was an else */
97 
98 
99     /*-----------------------------------------------*\
100     |		      INITIALIZATION		      |
101     \*-----------------------------------------------*/
102 
103     found_err = 0;
104 
105     ps.p_stack[0] = stmt;	/* this is the parser's stack */
106     ps.last_nl = true;		/* this is true if the last thing scanned was
107 				 * a newline */
108     ps.last_token = semicolon;
109     combuf = (char *) malloc(bufsize);
110     if (combuf == NULL)
111 	err(1, NULL);
112     labbuf = (char *) malloc(bufsize);
113     if (labbuf == NULL)
114 	err(1, NULL);
115     codebuf = (char *) malloc(bufsize);
116     if (codebuf == NULL)
117 	err(1, NULL);
118     tokenbuf = (char *) malloc(bufsize);
119     if (tokenbuf == NULL)
120 	err(1, NULL);
121     l_com = combuf + bufsize - 5;
122     l_lab = labbuf + bufsize - 5;
123     l_code = codebuf + bufsize - 5;
124     l_token = tokenbuf + bufsize - 5;
125     combuf[0] = codebuf[0] = labbuf[0] = ' ';	/* set up code, label, and
126 						 * comment buffers */
127     combuf[1] = codebuf[1] = labbuf[1] = '\0';
128     ps.else_if = 1;		/* Default else-if special processing to on */
129     s_lab = e_lab = labbuf + 1;
130     s_code = e_code = codebuf + 1;
131     s_com = e_com = combuf + 1;
132     s_token = e_token = tokenbuf + 1;
133 
134     in_buffer = (char *) malloc(10);
135     if (in_buffer == NULL)
136 	err(1, NULL);
137     in_buffer_limit = in_buffer + 8;
138     buf_ptr = buf_end = in_buffer;
139     line_no = 1;
140     had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
141     sp_sw = force_nl = false;
142     ps.in_or_st = false;
143     ps.bl_line = true;
144     dec_ind = 0;
145     di_stack[ps.dec_nest = 0] = 0;
146     ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
147 
148     scase = ps.pcase = false;
149     squest = 0;
150     sc_end = 0;
151     bp_save = 0;
152     be_save = 0;
153 
154     output = 0;
155 
156     /*--------------------------------------------------*\
157     |   		COMMAND LINE SCAN		 |
158     \*--------------------------------------------------*/
159 
160 #ifdef undef
161     max_col = 78;		/* -l78 */
162     lineup_to_parens = 1;	/* -lp */
163     ps.ljust_decl = 0;		/* -ndj */
164     ps.com_ind = 33;		/* -c33 */
165     star_comment_cont = 1;	/* -sc */
166     ps.ind_size = 8;		/* -i8 */
167     verbose = 0;
168     ps.decl_indent = 16;	/* -di16 */
169     ps.local_decl_indent = -1;	/* if this is not set to some nonnegative value
170 				 * by an arg, we will set this equal to
171 				 * ps.decl_ind */
172     ps.indent_parameters = 1;	/* -ip */
173     ps.decl_com_ind = 0;	/* if this is not set to some positive value
174 				 * by an arg, we will set this equal to
175 				 * ps.com_ind */
176     btype_2 = 1;		/* -br */
177     cuddle_else = 1;		/* -ce */
178     ps.unindent_displace = 0;	/* -d0 */
179     ps.case_indent = 0;		/* -cli0 */
180     format_block_comments = 1;	/* -fcb */
181     format_col1_comments = 1;	/* -fc1 */
182     procnames_start_line = 1;	/* -psl */
183     proc_calls_space = 0;	/* -npcs */
184     comment_delimiter_on_blankline = 1;	/* -cdb */
185     ps.leave_comma = 1;		/* -nbc */
186 #endif
187 
188     for (i = 1; i < argc; ++i)
189 	if (strcmp(argv[i], "-npro") == 0)
190 	    break;
191     set_defaults();
192     if (i >= argc)
193 	set_profile();
194 
195     for (i = 1; i < argc; ++i) {
196 
197 	/*
198 	 * look thru args (if any) for changes to defaults
199 	 */
200 	if (argv[i][0] != '-') {/* no flag on parameter */
201 	    if (input == 0) {	/* we must have the input file */
202 		in_name = argv[i];	/* remember name of input file */
203 		input = fopen(in_name, "r");
204 		if (input == 0)		/* check for open error */
205 			err(1, "%s", in_name);
206 		continue;
207 	    }
208 	    else if (output == 0) {	/* we have the output file */
209 		out_name = argv[i];	/* remember name of output file */
210 		if (strcmp(in_name, out_name) == 0) {	/* attempt to overwrite
211 							 * the file */
212 		    errx(1, "input and output files must be different");
213 		}
214 		output = fopen(out_name, "w");
215 		if (output == 0)	/* check for create error */
216 			err(1, "%s", out_name);
217 		continue;
218 	    }
219 	    errx(1, "unknown parameter: %s", argv[i]);
220 	}
221 	else
222 	    set_option(argv[i]);
223     }				/* end of for */
224     if (input == 0)
225 	input = stdin;
226     if (output == 0) {
227 	if (troff || input == stdin)
228 	    output = stdout;
229 	else {
230 	    out_name = in_name;
231 	    bakcopy();
232 	}
233     }
234     if (ps.com_ind <= 1)
235 	ps.com_ind = 2;		/* dont put normal comments before column 2 */
236     if (troff) {
237 	if (bodyf.font[0] == 0)
238 	    parsefont(&bodyf, "R");
239 	if (scomf.font[0] == 0)
240 	    parsefont(&scomf, "I");
241 	if (blkcomf.font[0] == 0)
242 	    blkcomf = scomf, blkcomf.size += 2;
243 	if (boxcomf.font[0] == 0)
244 	    boxcomf = blkcomf;
245 	if (stringf.font[0] == 0)
246 	    parsefont(&stringf, "L");
247 	if (keywordf.font[0] == 0)
248 	    parsefont(&keywordf, "B");
249 	writefdef(&bodyf, 'B');
250 	writefdef(&scomf, 'C');
251 	writefdef(&blkcomf, 'L');
252 	writefdef(&boxcomf, 'X');
253 	writefdef(&stringf, 'S');
254 	writefdef(&keywordf, 'K');
255     }
256     if (block_comment_max_col <= 0)
257 	block_comment_max_col = max_col;
258     if (ps.local_decl_indent < 0)	/* if not specified by user, set this */
259 	ps.local_decl_indent = ps.decl_indent;
260     if (ps.decl_com_ind <= 0)	/* if not specified by user, set this */
261 	ps.decl_com_ind = ps.ljust_decl ? (ps.com_ind <= 10 ? 2 : ps.com_ind - 8) : ps.com_ind;
262     if (continuation_indent == 0)
263 	continuation_indent = ps.ind_size;
264     fill_buffer();		/* get first batch of stuff into input buffer */
265 
266     parse(semicolon);
267     {
268 	char *p = buf_ptr;
269 	int col = 1;
270 
271 	while (1) {
272 	    if (*p == ' ')
273 		col++;
274 	    else if (*p == '\t')
275 		col = ((col - 1) & ~7) + 9;
276 	    else
277 		break;
278 	    p++;
279 	}
280 	if (col > ps.ind_size)
281 	    ps.ind_level = ps.i_l_follow = col / ps.ind_size;
282     }
283     if (troff) {
284 	const char *p = in_name,
285 	           *beg = in_name;
286 
287 	while (*p)
288 	    if (*p++ == '/')
289 		beg = p;
290 	fprintf(output, ".Fn \"%s\"\n", beg);
291     }
292     /*
293      * START OF MAIN LOOP
294      */
295 
296     while (1) {			/* this is the main loop.  it will go until we
297 				 * reach eof */
298 	int         is_procname;
299 
300 	type_code = lexi();	/* lexi reads one token.  The actual
301 				 * characters read are stored in "token". lexi
302 				 * returns a code indicating the type of token */
303 	is_procname = ps.procname[0];
304 
305 	/*
306 	 * The following code moves everything following an if (), while (),
307 	 * else, etc. up to the start of the following stmt to a buffer. This
308 	 * allows proper handling of both kinds of brace placement.
309 	 */
310 
311 	flushed_nl = false;
312 	while (ps.search_brace) {	/* if we scanned an if(), while(),
313 					 * etc., we might need to copy stuff
314 					 * into a buffer we must loop, copying
315 					 * stuff into save_com, until we find
316 					 * the start of the stmt which follows
317 					 * the if, or whatever */
318 	    switch (type_code) {
319 	    case newline:
320 		++line_no;
321 		flushed_nl = true;
322 	    case form_feed:
323 		break;		/* form feeds and newlines found here will be
324 				 * ignored */
325 
326 	    case lbrace:	/* this is a brace that starts the compound
327 				 * stmt */
328 		if (sc_end == 0) {	/* ignore buffering if a comment wasnt
329 					 * stored up */
330 		    ps.search_brace = false;
331 		    goto check_type;
332 		}
333 		if (btype_2) {
334 		    save_com[0] = '{';	/* we either want to put the brace
335 					 * right after the if */
336 		    goto sw_buffer;	/* go to common code to get out of
337 					 * this loop */
338 		}
339 	    case comment:	/* we have a comment, so we must copy it into
340 				 * the buffer */
341 		if (!flushed_nl || sc_end != 0) {
342 		    if (sc_end == 0) {	/* if this is the first comment, we
343 					 * must set up the buffer */
344 			save_com[0] = save_com[1] = ' ';
345 			sc_end = &(save_com[2]);
346 		    }
347 		    else {
348 			*sc_end++ = '\n';	/* add newline between
349 						 * comments */
350 			*sc_end++ = ' ';
351 			--line_no;
352 		    }
353 		    *sc_end++ = '/';	/* copy in start of comment */
354 		    *sc_end++ = '*';
355 
356 		    for (;;) {	/* loop until we get to the end of the comment */
357 			*sc_end = *buf_ptr++;
358 			if (buf_ptr >= buf_end)
359 			    fill_buffer();
360 
361 			if (*sc_end++ == '*' && *buf_ptr == '/')
362 			    break;	/* we are at end of comment */
363 
364 			if (sc_end >= &(save_com[sc_size])) {	/* check for temp buffer
365 								 * overflow */
366 			    diag2(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
367 			    fflush(output);
368 			    exit(1);
369 			}
370 		    }
371 		    *sc_end++ = '/';	/* add ending slash */
372 		    if (++buf_ptr >= buf_end)	/* get past / in buffer */
373 			fill_buffer();
374 		    break;
375 		}
376 	    default:		/* it is the start of a normal statement */
377 		if (flushed_nl)	/* if we flushed a newline, make sure it is
378 				 * put back */
379 		    force_nl = true;
380 		if ((type_code == sp_paren && *token == 'i'
381 			&& last_else && ps.else_if)
382 			|| (type_code == sp_nparen && *token == 'e'
383 			&& e_code != s_code && e_code[-1] == '}'))
384 		    force_nl = false;
385 
386 		if (sc_end == 0) {	/* ignore buffering if comment wasnt
387 					 * saved up */
388 		    ps.search_brace = false;
389 		    goto check_type;
390 		}
391 		if (force_nl) {	/* if we should insert a nl here, put it into
392 				 * the buffer */
393 		    force_nl = false;
394 		    --line_no;	/* this will be re-increased when the nl is
395 				 * read from the buffer */
396 		    *sc_end++ = '\n';
397 		    *sc_end++ = ' ';
398 		    if (verbose && !flushed_nl)	/* print error msg if the line
399 						 * was not already broken */
400 			diag2(0, "Line broken");
401 		    flushed_nl = false;
402 		}
403 		for (t_ptr = token; *t_ptr; ++t_ptr)
404 		    *sc_end++ = *t_ptr;	/* copy token into temp buffer */
405 		ps.procname[0] = 0;
406 
407 	sw_buffer:
408 		ps.search_brace = false;	/* stop looking for start of
409 						 * stmt */
410 		bp_save = buf_ptr;	/* save current input buffer */
411 		be_save = buf_end;
412 		buf_ptr = save_com;	/* fix so that subsequent calls to
413 					 * lexi will take tokens out of
414 					 * save_com */
415 		*sc_end++ = ' ';/* add trailing blank, just in case */
416 		buf_end = sc_end;
417 		sc_end = 0;
418 		break;
419 	    }			/* end of switch */
420 	    if (type_code != 0)	/* we must make this check, just in case there
421 				 * was an unexpected EOF */
422 		type_code = lexi();	/* read another token */
423 	    /* if (ps.search_brace) ps.procname[0] = 0; */
424 	    if ((is_procname = ps.procname[0]) && flushed_nl
425 		    && !procnames_start_line && ps.in_decl
426 		    && type_code == ident)
427 		flushed_nl = 0;
428 	}			/* end of while (search_brace) */
429 	last_else = 0;
430 check_type:
431 	if (type_code == 0) {	/* we got eof */
432 	    if (s_lab != e_lab || s_code != e_code
433 		    || s_com != e_com)	/* must dump end of line */
434 		dump_line();
435 	    if (ps.tos > 1)	/* check for balanced braces */
436 		diag2(1, "Stuff missing from end of file");
437 
438 	    if (verbose) {
439 		printf("There were %d output lines and %d comments\n",
440 		       ps.out_lines, ps.out_coms);
441 		printf("(Lines with comments)/(Lines with code): %6.3f\n",
442 		       (1.0 * ps.com_lines) / code_lines);
443 	    }
444 	    fflush(output);
445 	    exit(found_err);
446 	}
447 	if (
448 		(type_code != comment) &&
449 		(type_code != newline) &&
450 		(type_code != preesc) &&
451 		(type_code != form_feed)) {
452 	    if (force_nl &&
453 		    (type_code != semicolon) &&
454 		    (type_code != lbrace || !btype_2)) {
455 		/* we should force a broken line here */
456 		if (verbose && !flushed_nl)
457 		    diag2(0, "Line broken");
458 		flushed_nl = false;
459 		dump_line();
460 		ps.want_blank = false;	/* dont insert blank at line start */
461 		force_nl = false;
462 	    }
463 	    ps.in_stmt = true;	/* turn on flag which causes an extra level of
464 				 * indentation. this is turned off by a ; or
465 				 * '}' */
466 	    if (s_com != e_com) {	/* the turkey has embedded a comment
467 					 * in a line. fix it */
468 		*e_code++ = ' ';
469 		for (t_ptr = s_com; *t_ptr; ++t_ptr) {
470 		    CHECK_SIZE_CODE;
471 		    *e_code++ = *t_ptr;
472 		}
473 		*e_code++ = ' ';
474 		*e_code = '\0';	/* null terminate code sect */
475 		ps.want_blank = false;
476 		e_com = s_com;
477 	    }
478 	}
479 	else if (type_code != comment)	/* preserve force_nl thru a comment */
480 	    force_nl = false;	/* cancel forced newline after newline, form
481 				 * feed, etc */
482 
483 
484 
485 	/*-----------------------------------------------------*\
486 	|	   do switch on type of token scanned		|
487 	\*-----------------------------------------------------*/
488 	CHECK_SIZE_CODE;
489 	switch (type_code) {	/* now, decide what to do with the token */
490 
491 	case form_feed:	/* found a form feed in line */
492 	    ps.use_ff = true;	/* a form feed is treated much like a newline */
493 	    dump_line();
494 	    ps.want_blank = false;
495 	    break;
496 
497 	case newline:
498 	    if (ps.last_token != comma || ps.p_l_follow > 0
499 		    || !ps.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
500 		dump_line();
501 		ps.want_blank = false;
502 	    }
503 	    ++line_no;		/* keep track of input line number */
504 	    break;
505 
506 	case lparen:		/* got a '(' or '[' */
507 	    ++ps.p_l_follow;	/* count parens to make Healy happy */
508 	    if (ps.want_blank && *token != '[' &&
509 		    (ps.last_token != ident || proc_calls_space
510 	      || (ps.its_a_keyword && (!ps.sizeof_keyword || Bill_Shannon))))
511 		*e_code++ = ' ';
512 	    if (ps.in_decl && !ps.block_init)
513 		if (troff && !ps.dumped_decl_indent && !is_procname && ps.last_token == decl) {
514 		    ps.dumped_decl_indent = 1;
515 		    sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
516 		    e_code += strlen(e_code);
517 		}
518 		else {
519 		    while ((e_code - s_code) < dec_ind) {
520 			CHECK_SIZE_CODE;
521 			*e_code++ = ' ';
522 		    }
523 		    *e_code++ = token[0];
524 		}
525 	    else
526 		*e_code++ = token[0];
527 	    ps.paren_indents[ps.p_l_follow - 1] = e_code - s_code;
528 	    if (sp_sw && ps.p_l_follow == 1 && extra_expression_indent
529 		    && ps.paren_indents[0] < 2 * ps.ind_size)
530 		ps.paren_indents[0] = 2 * ps.ind_size;
531 	    ps.want_blank = false;
532 	    if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
533 		/*
534 		 * this is a kluge to make sure that declarations will be
535 		 * aligned right if proc decl has an explicit type on it, i.e.
536 		 * "int a(x) {..."
537 		 */
538 		parse(semicolon);	/* I said this was a kluge... */
539 		ps.in_or_st = false;	/* turn off flag for structure decl or
540 					 * initialization */
541 	    }
542 	    if (ps.sizeof_keyword)
543 		ps.sizeof_mask |= 1 << ps.p_l_follow;
544 	    break;
545 
546 	case rparen:		/* got a ')' or ']' */
547 	    rparen_count--;
548 	    if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.sizeof_mask) {
549 		ps.last_u_d = true;
550 		ps.cast_mask &= (1 << ps.p_l_follow) - 1;
551 		ps.want_blank = false;
552 	    } else
553 		ps.want_blank = true;
554 	    ps.sizeof_mask &= (1 << ps.p_l_follow) - 1;
555 	    if (--ps.p_l_follow < 0) {
556 		ps.p_l_follow = 0;
557 		diag3(0, "Extra %c", *token);
558 	    }
559 	    if (e_code == s_code)	/* if the paren starts the line */
560 		ps.paren_level = ps.p_l_follow;	/* then indent it */
561 
562 	    *e_code++ = token[0];
563 
564 	    if (sp_sw && (ps.p_l_follow == 0)) {	/* check for end of if
565 							 * (...), or some such */
566 		sp_sw = false;
567 		force_nl = true;/* must force newline after if */
568 		ps.last_u_d = true;	/* inform lexi that a following
569 					 * operator is unary */
570 		ps.in_stmt = false;	/* dont use stmt continuation
571 					 * indentation */
572 
573 		parse(hd_type);	/* let parser worry about if, or whatever */
574 	    }
575 	    ps.search_brace = btype_2;	/* this should insure that constructs
576 					 * such as main(){...} and int[]{...}
577 					 * have their braces put in the right
578 					 * place */
579 	    break;
580 
581 	case unary_op:		/* this could be any unary operation */
582 	    if (ps.want_blank)
583 		*e_code++ = ' ';
584 
585 	    if (troff && !ps.dumped_decl_indent && ps.in_decl && !is_procname) {
586 		sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
587 		ps.dumped_decl_indent = 1;
588 		e_code += strlen(e_code);
589 	    }
590 	    else {
591 		const char *res = token;
592 
593 		if (ps.in_decl && !ps.block_init) {	/* if this is a unary op
594 							 * in a declaration, we
595 							 * should indent this
596 							 * token */
597 		    for (i = 0; token[i]; ++i);	/* find length of token */
598 		    while ((e_code - s_code) < (dec_ind - i)) {
599 			CHECK_SIZE_CODE;
600 			*e_code++ = ' ';	/* pad it */
601 		    }
602 		}
603 		if (troff && token[0] == '-' && token[1] == '>')
604 		    res = "\\(->";
605 		for (t_ptr = res; *t_ptr; ++t_ptr) {
606 		    CHECK_SIZE_CODE;
607 		    *e_code++ = *t_ptr;
608 		}
609 	    }
610 	    ps.want_blank = false;
611 	    break;
612 
613 	case binary_op:	/* any binary operation */
614 	    if (ps.want_blank)
615 		*e_code++ = ' ';
616 	    {
617 		const char *res = token;
618 
619 		if (troff)
620 		    switch (token[0]) {
621 		    case '<':
622 			if (token[1] == '=')
623 			    res = "\\(<=";
624 			break;
625 		    case '>':
626 			if (token[1] == '=')
627 			    res = "\\(>=";
628 			break;
629 		    case '!':
630 			if (token[1] == '=')
631 			    res = "\\(!=";
632 			break;
633 		    case '|':
634 			if (token[1] == '|')
635 			    res = "\\(br\\(br";
636 			else if (token[1] == 0)
637 			    res = "\\(br";
638 			break;
639 		    }
640 		for (t_ptr = res; *t_ptr; ++t_ptr) {
641 		    CHECK_SIZE_CODE;
642 		    *e_code++ = *t_ptr;	/* move the operator */
643 		}
644 	    }
645 	    ps.want_blank = true;
646 	    break;
647 
648 	case postop:		/* got a trailing ++ or -- */
649 	    *e_code++ = token[0];
650 	    *e_code++ = token[1];
651 	    ps.want_blank = true;
652 	    break;
653 
654 	case question:		/* got a ? */
655 	    squest++;		/* this will be used when a later colon
656 				 * appears so we can distinguish the
657 				 * <c>?<n>:<n> construct */
658 	    if (ps.want_blank)
659 		*e_code++ = ' ';
660 	    *e_code++ = '?';
661 	    ps.want_blank = true;
662 	    break;
663 
664 	case casestmt:		/* got word 'case' or 'default' */
665 	    scase = true;	/* so we can process the later colon properly */
666 	    goto copy_id;
667 
668 	case colon:		/* got a ':' */
669 	    if (squest > 0) {	/* it is part of the <c>?<n>: <n> construct */
670 		--squest;
671 		if (ps.want_blank)
672 		    *e_code++ = ' ';
673 		*e_code++ = ':';
674 		ps.want_blank = true;
675 		break;
676 	    }
677 	    if (ps.in_decl) {
678 		*e_code++ = ':';
679 		ps.want_blank = false;
680 		break;
681 	    }
682 	    ps.in_stmt = false;	/* seeing a label does not imply we are in a
683 				 * stmt */
684 	    for (t_ptr = s_code; *t_ptr; ++t_ptr)
685 		*e_lab++ = *t_ptr;	/* turn everything so far into a label */
686 	    e_code = s_code;
687 	    *e_lab++ = ':';
688 	    *e_lab++ = ' ';
689 	    *e_lab = '\0';
690 
691 	    force_nl = ps.pcase = scase;	/* ps.pcase will be used by
692 						 * dump_line to decide how to
693 						 * indent the label. force_nl
694 						 * will force a case n: to be
695 						 * on a line by itself */
696 	    scase = false;
697 	    ps.want_blank = false;
698 	    break;
699 
700 	case semicolon:	/* got a ';' */
701 	    ps.in_or_st = false;/* we are not in an initialization or
702 				 * structure declaration */
703 	    scase = false;	/* these will only need resetting in an error */
704 	    squest = 0;
705 	    if (ps.last_token == rparen && rparen_count == 0)
706 		ps.in_parameter_declaration = 0;
707 	    ps.cast_mask = 0;
708 	    ps.sizeof_mask = 0;
709 	    ps.block_init = 0;
710 	    ps.block_init_level = 0;
711 	    ps.just_saw_decl--;
712 
713 	    if (ps.in_decl && s_code == e_code && !ps.block_init)
714 		while ((e_code - s_code) < (dec_ind - 1)) {
715 		    CHECK_SIZE_CODE;
716 		    *e_code++ = ' ';
717 		}
718 
719 	    ps.in_decl = (ps.dec_nest > 0);	/* if we were in a first level
720 						 * structure declaration, we
721 						 * arent any more */
722 
723 	    if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
724 
725 		/*
726 		 * This should be true iff there were unbalanced parens in the
727 		 * stmt.  It is a bit complicated, because the semicolon might
728 		 * be in a for stmt
729 		 */
730 		diag2(1, "Unbalanced parens");
731 		ps.p_l_follow = 0;
732 		if (sp_sw) {	/* this is a check for an if, while, etc. with
733 				 * unbalanced parens */
734 		    sp_sw = false;
735 		    parse(hd_type);	/* dont lose the if, or whatever */
736 		}
737 	    }
738 	    *e_code++ = ';';
739 	    ps.want_blank = true;
740 	    ps.in_stmt = (ps.p_l_follow > 0);	/* we are no longer in the
741 						 * middle of a stmt */
742 
743 	    if (!sp_sw) {	/* if not if for (;;) */
744 		parse(semicolon);	/* let parser know about end of stmt */
745 		force_nl = true;/* force newline after an end of stmt */
746 	    }
747 	    break;
748 
749 	case lbrace:		/* got a '{' */
750 	    ps.in_stmt = false;	/* dont indent the {} */
751 	    if (!ps.block_init)
752 		force_nl = true;/* force other stuff on same line as '{' onto
753 				 * new line */
754 	    else if (ps.block_init_level <= 0)
755 		ps.block_init_level = 1;
756 	    else
757 		ps.block_init_level++;
758 
759 	    if (s_code != e_code && !ps.block_init) {
760 		if (!btype_2) {
761 		    dump_line();
762 		    ps.want_blank = false;
763 		}
764 		else if (ps.in_parameter_declaration && !ps.in_or_st) {
765 		    ps.i_l_follow = 0;
766 		    if (function_brace_split) {	/* dump the line prior to the
767 						 * brace ... */
768 			dump_line();
769 			ps.want_blank = false;
770 		    } else	/* add a space between the decl and brace */
771 			ps.want_blank = true;
772 		}
773 	    }
774 	    if (ps.in_parameter_declaration)
775 		prefix_blankline_requested = 0;
776 
777 	    if (ps.p_l_follow > 0) {	/* check for preceding unbalanced
778 					 * parens */
779 		diag2(1, "Unbalanced parens");
780 		ps.p_l_follow = 0;
781 		if (sp_sw) {	/* check for unclosed if, for, etc. */
782 		    sp_sw = false;
783 		    parse(hd_type);
784 		    ps.ind_level = ps.i_l_follow;
785 		}
786 	    }
787 	    if (s_code == e_code)
788 		ps.ind_stmt = false;	/* dont put extra indentation on line
789 					 * with '{' */
790 	    if (ps.in_decl && ps.in_or_st) {	/* this is either a structure
791 						 * declaration or an init */
792 		di_stack[ps.dec_nest++] = dec_ind;
793 		/* ?		dec_ind = 0; */
794 	    }
795 	    else {
796 		ps.decl_on_line = false;	/* we cant be in the middle of
797 						 * a declaration, so dont do
798 						 * special indentation of
799 						 * comments */
800 		if (blanklines_after_declarations_at_proctop
801 			&& ps.in_parameter_declaration)
802 		    postfix_blankline_requested = 1;
803 		ps.in_parameter_declaration = 0;
804 	    }
805 	    dec_ind = 0;
806 	    parse(lbrace);	/* let parser know about this */
807 	    if (ps.want_blank)	/* put a blank before '{' if '{' is not at
808 				 * start of line */
809 		*e_code++ = ' ';
810 	    ps.want_blank = false;
811 	    *e_code++ = '{';
812 	    ps.just_saw_decl = 0;
813 	    break;
814 
815 	case rbrace:		/* got a '}' */
816 	    if (ps.p_stack[ps.tos] == decl && !ps.block_init)	/* semicolons can be
817 								 * omitted in
818 								 * declarations */
819 		parse(semicolon);
820 	    if (ps.p_l_follow) {/* check for unclosed if, for, else. */
821 		diag2(1, "Unbalanced parens");
822 		ps.p_l_follow = 0;
823 		sp_sw = false;
824 	    }
825 	    ps.just_saw_decl = 0;
826 	    ps.block_init_level--;
827 	    if (s_code != e_code && !ps.block_init) {	/* '}' must be first on
828 							 * line */
829 		if (verbose)
830 		    diag2(0, "Line broken");
831 		dump_line();
832 	    }
833 	    *e_code++ = '}';
834 	    ps.want_blank = true;
835 	    ps.in_stmt = ps.ind_stmt = false;
836 	    if (ps.dec_nest > 0) {	/* we are in multi-level structure
837 					 * declaration */
838 		dec_ind = di_stack[--ps.dec_nest];
839 		if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
840 		    ps.just_saw_decl = 2;
841 		ps.in_decl = true;
842 	    }
843 	    prefix_blankline_requested = 0;
844 	    parse(rbrace);	/* let parser know about this */
845 	    ps.search_brace = cuddle_else && ps.p_stack[ps.tos] == ifhead
846 		&& ps.il[ps.tos] >= ps.ind_level;
847 	    if (ps.tos <= 1 && blanklines_after_procs && ps.dec_nest <= 0)
848 		postfix_blankline_requested = 1;
849 	    break;
850 
851 	case swstmt:		/* got keyword "switch" */
852 	    sp_sw = true;
853 	    hd_type = swstmt;	/* keep this for when we have seen the
854 				 * expression */
855 	    goto copy_id;	/* go move the token into buffer */
856 
857 	case sp_paren:		/* token is if, while, for */
858 	    sp_sw = true;	/* the interesting stuff is done after the
859 				 * expression is scanned */
860 	    hd_type = (*token == 'i' ? ifstmt :
861 		       (*token == 'w' ? whilestmt : forstmt));
862 
863 	    /*
864 	     * remember the type of header for later use by parser
865 	     */
866 	    goto copy_id;	/* copy the token into line */
867 
868 	case sp_nparen:	/* got else, do */
869 	    ps.in_stmt = false;
870 	    if (*token == 'e') {
871 		if (e_code != s_code && (!cuddle_else || e_code[-1] != '}')) {
872 		    if (verbose)
873 			diag2(0, "Line broken");
874 		    dump_line();/* make sure this starts a line */
875 		    ps.want_blank = false;
876 		}
877 		force_nl = true;/* also, following stuff must go onto new line */
878 		last_else = 1;
879 		parse(elselit);
880 	    }
881 	    else {
882 		if (e_code != s_code) {	/* make sure this starts a line */
883 		    if (verbose)
884 			diag2(0, "Line broken");
885 		    dump_line();
886 		    ps.want_blank = false;
887 		}
888 		force_nl = true;/* also, following stuff must go onto new line */
889 		last_else = 0;
890 		parse(dolit);
891 	    }
892 	    goto copy_id;	/* move the token into line */
893 
894 	case decl:		/* we have a declaration type (int, register,
895 				 * etc.) */
896 	    parse(decl);	/* let parser worry about indentation */
897 	    if (ps.last_token == rparen && ps.tos <= 1) {
898 		ps.in_parameter_declaration = 1;
899 		if (s_code != e_code) {
900 		    dump_line();
901 		    ps.want_blank = 0;
902 		}
903 	    }
904 	    if (ps.in_parameter_declaration && ps.indent_parameters && ps.dec_nest == 0) {
905 		ps.ind_level = ps.i_l_follow = 1;
906 		ps.ind_stmt = 0;
907 	    }
908 	    ps.in_or_st = true;	/* this might be a structure or initialization
909 				 * declaration */
910 	    ps.in_decl = ps.decl_on_line = true;
911 	    if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
912 		ps.just_saw_decl = 2;
913 	    prefix_blankline_requested = 0;
914 	    for (i = 0; token[i++];);	/* get length of token */
915 
916 	    if (ps.ind_level == 0 || ps.dec_nest > 0) {
917 		/* global variable or struct member in local variable */
918 		dec_ind = ps.decl_indent > 0 ? ps.decl_indent : i;
919 		tabs_to_var = (use_tabs ? ps.decl_indent > 0 : 0);
920 	    } else {
921 		/* local variable */
922 		dec_ind = ps.local_decl_indent > 0 ? ps.local_decl_indent : i;
923 		tabs_to_var = (use_tabs ? ps.local_decl_indent > 0 : 0);
924 	    }
925 	    goto copy_id;
926 
927 	case ident:		/* got an identifier or constant */
928 	    if (ps.in_decl) {	/* if we are in a declaration, we must indent
929 				 * identifier */
930 		if (is_procname == 0 || !procnames_start_line) {
931 		    if (!ps.block_init) {
932 			if (troff && !ps.dumped_decl_indent) {
933 			    if (ps.want_blank)
934 				*e_code++ = ' ';
935 			    ps.want_blank = false;
936 			    sprintf(e_code, "\n.De %dp+\200p\n", dec_ind * 7);
937 			    ps.dumped_decl_indent = 1;
938 			    e_code += strlen(e_code);
939 			} else {
940 			    int cur_dec_ind;
941 			    int pos, startpos;
942 
943 			    /*
944 			     * in order to get the tab math right for
945 			     * indentations that are not multiples of 8 we
946 			     * need to modify both startpos and dec_ind
947 			     * (cur_dec_ind) here by eight minus the
948 			     * remainder of the current starting column
949 			     * divided by eight. This seems to be a
950 			     * properly working fix
951 			     */
952 			    startpos = e_code - s_code;
953 			    cur_dec_ind = dec_ind;
954 			    pos = startpos;
955 			    if ((ps.ind_level * ps.ind_size) % 8 != 0) {
956 				pos += (ps.ind_level * ps.ind_size) % 8;
957 				cur_dec_ind += (ps.ind_level * ps.ind_size) % 8;
958 			    }
959 
960 			    if (tabs_to_var) {
961 				while ((pos & ~7) + 8 <= cur_dec_ind) {
962 				    CHECK_SIZE_CODE;
963 				    *e_code++ = '\t';
964 				    pos = (pos & ~7) + 8;
965 				}
966 			    }
967 			    while (pos < cur_dec_ind) {
968 				CHECK_SIZE_CODE;
969 				*e_code++ = ' ';
970 				pos++;
971 			    }
972 			    if (ps.want_blank && e_code - s_code == startpos)
973 				*e_code++ = ' ';
974 			    ps.want_blank = false;
975 			}
976 		    }
977 		} else {
978 		    if (ps.want_blank)
979 			*e_code++ = ' ';
980 		    ps.want_blank = false;
981 		    if (dec_ind && s_code != e_code)
982 			dump_line();
983 		    dec_ind = 0;
984 		}
985 	    }
986 	    else if (sp_sw && ps.p_l_follow == 0) {
987 		sp_sw = false;
988 		force_nl = true;
989 		ps.last_u_d = true;
990 		ps.in_stmt = false;
991 		parse(hd_type);
992 	    }
993     copy_id:
994 	    if (ps.want_blank)
995 		*e_code++ = ' ';
996 	    if (troff && ps.its_a_keyword) {
997 		e_code = chfont(&bodyf, &keywordf, e_code);
998 		for (t_ptr = token; *t_ptr; ++t_ptr) {
999 		    CHECK_SIZE_CODE;
1000 		    *e_code++ = keywordf.allcaps && islower(*t_ptr)
1001 			? toupper(*t_ptr) : *t_ptr;
1002 		}
1003 		e_code = chfont(&keywordf, &bodyf, e_code);
1004 	    }
1005 	    else
1006 		for (t_ptr = token; *t_ptr; ++t_ptr) {
1007 		    CHECK_SIZE_CODE;
1008 		    *e_code++ = *t_ptr;
1009 		}
1010 	    ps.want_blank = true;
1011 	    break;
1012 
1013 	case period:		/* treat a period kind of like a binary
1014 				 * operation */
1015 	    *e_code++ = '.';	/* move the period into line */
1016 	    ps.want_blank = false;	/* dont put a blank after a period */
1017 	    break;
1018 
1019 	case comma:
1020 	    ps.want_blank = (s_code != e_code);	/* only put blank after comma
1021 						 * if comma does not start the
1022 						 * line */
1023 	    if (ps.in_decl && is_procname == 0 && !ps.block_init)
1024 		while ((e_code - s_code) < (dec_ind - 1)) {
1025 		    CHECK_SIZE_CODE;
1026 		    *e_code++ = ' ';
1027 		}
1028 
1029 	    *e_code++ = ',';
1030 	    if (ps.p_l_follow == 0) {
1031 		if (ps.block_init_level <= 0)
1032 		    ps.block_init = 0;
1033 		if (break_comma && (!ps.leave_comma || compute_code_target() + (e_code - s_code) > max_col - 8))
1034 		    force_nl = true;
1035 	    }
1036 	    break;
1037 
1038 	case preesc:		/* got the character '#' */
1039 	    if ((s_com != e_com) ||
1040 		    (s_lab != e_lab) ||
1041 		    (s_code != e_code))
1042 		dump_line();
1043 	    *e_lab++ = '#';	/* move whole line to 'label' buffer */
1044 	    {
1045 		int         in_comment = 0;
1046 		int         com_start = 0;
1047 		char        quote = 0;
1048 		int         com_end = 0;
1049 
1050 		while (*buf_ptr == ' ' || *buf_ptr == '\t') {
1051 		    buf_ptr++;
1052 		    if (buf_ptr >= buf_end)
1053 			fill_buffer();
1054 		}
1055 		while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
1056 		    CHECK_SIZE_LAB;
1057 		    *e_lab = *buf_ptr++;
1058 		    if (buf_ptr >= buf_end)
1059 			fill_buffer();
1060 		    switch (*e_lab++) {
1061 		    case BACKSLASH:
1062 			if (troff)
1063 			    *e_lab++ = BACKSLASH;
1064 			if (!in_comment) {
1065 			    *e_lab++ = *buf_ptr++;
1066 			    if (buf_ptr >= buf_end)
1067 				fill_buffer();
1068 			}
1069 			break;
1070 		    case '/':
1071 			if (*buf_ptr == '*' && !in_comment && !quote) {
1072 			    in_comment = 1;
1073 			    *e_lab++ = *buf_ptr++;
1074 			    com_start = e_lab - s_lab - 2;
1075 			}
1076 			break;
1077 		    case '"':
1078 			if (quote == '"')
1079 			    quote = 0;
1080 			break;
1081 		    case '\'':
1082 			if (quote == '\'')
1083 			    quote = 0;
1084 			break;
1085 		    case '*':
1086 			if (*buf_ptr == '/' && in_comment) {
1087 			    in_comment = 0;
1088 			    *e_lab++ = *buf_ptr++;
1089 			    com_end = e_lab - s_lab;
1090 			}
1091 			break;
1092 		    }
1093 		}
1094 
1095 		while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1096 		    e_lab--;
1097 		if (e_lab - s_lab == com_end && bp_save == 0) {	/* comment on
1098 								 * preprocessor line */
1099 		    if (sc_end == 0)	/* if this is the first comment, we
1100 					 * must set up the buffer */
1101 			sc_end = &(save_com[0]);
1102 		    else {
1103 			*sc_end++ = '\n';	/* add newline between
1104 						 * comments */
1105 			*sc_end++ = ' ';
1106 			--line_no;
1107 		    }
1108 		    bcopy(s_lab + com_start, sc_end, com_end - com_start);
1109 		    sc_end += com_end - com_start;
1110 		    if (sc_end >= &save_com[sc_size])
1111 			abort();
1112 		    e_lab = s_lab + com_start;
1113 		    while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1114 			e_lab--;
1115 		    bp_save = buf_ptr;	/* save current input buffer */
1116 		    be_save = buf_end;
1117 		    buf_ptr = save_com;	/* fix so that subsequent calls to
1118 					 * lexi will take tokens out of
1119 					 * save_com */
1120 		    *sc_end++ = ' ';	/* add trailing blank, just in case */
1121 		    buf_end = sc_end;
1122 		    sc_end = 0;
1123 		}
1124 		*e_lab = '\0';	/* null terminate line */
1125 		ps.pcase = false;
1126 	    }
1127 
1128 	    if (strncmp(s_lab, "#if", 3) == 0) {
1129 		if (blanklines_around_conditional_compilation) {
1130 		    int c;
1131 		    prefix_blankline_requested++;
1132 		    while ((c = getc(input)) == '\n');
1133 		    ungetc(c, input);
1134 		}
1135 		if ((size_t)ifdef_level < sizeof(state_stack)/sizeof(state_stack[0])) {
1136 		    match_state[ifdef_level].tos = -1;
1137 		    state_stack[ifdef_level++] = ps;
1138 		}
1139 		else
1140 		    diag2(1, "#if stack overflow");
1141 	    }
1142 	    else if (strncmp(s_lab, "#else", 5) == 0)
1143 		if (ifdef_level <= 0)
1144 		    diag2(1, "Unmatched #else");
1145 		else {
1146 		    match_state[ifdef_level - 1] = ps;
1147 		    ps = state_stack[ifdef_level - 1];
1148 		}
1149 	    else if (strncmp(s_lab, "#endif", 6) == 0) {
1150 		if (ifdef_level <= 0)
1151 		    diag2(1, "Unmatched #endif");
1152 		else {
1153 		    ifdef_level--;
1154 
1155 #ifdef undef
1156 		    /*
1157 		     * This match needs to be more intelligent before the
1158 		     * message is useful
1159 		     */
1160 		    if (match_state[ifdef_level].tos >= 0
1161 			  && bcmp(&ps, &match_state[ifdef_level], sizeof ps))
1162 			diag2(0, "Syntactically inconsistent #ifdef alternatives");
1163 #endif
1164 		}
1165 		if (blanklines_around_conditional_compilation) {
1166 		    postfix_blankline_requested++;
1167 		    n_real_blanklines = 0;
1168 		}
1169 	    }
1170 	    break;		/* subsequent processing of the newline
1171 				 * character will cause the line to be printed */
1172 
1173 	case comment:		/* we have gotten a / followed by * this is a biggie */
1174 	    if (flushed_nl) {	/* we should force a broken line here */
1175 		flushed_nl = false;
1176 		dump_line();
1177 		ps.want_blank = false;	/* dont insert blank at line start */
1178 		force_nl = false;
1179 	    }
1180 	    pr_comment();
1181 	    break;
1182 	}			/* end of big switch stmt */
1183 
1184 	*e_code = '\0';		/* make sure code section is null terminated */
1185 	if (type_code != comment && type_code != newline && type_code != preesc)
1186 	    ps.last_token = type_code;
1187     }				/* end of main while (1) loop */
1188 }
1189 
1190 /*
1191  * copy input file to backup file if in_name is /blah/blah/blah/file, then
1192  * backup file will be ".Bfile" then make the backup file the input and
1193  * original input file the output
1194  */
1195 static void
1196 bakcopy(void)
1197 {
1198     int         n,
1199                 bakchn;
1200     char        buff[8 * 1024];
1201     const char *p;
1202 
1203     /* construct file name .Bfile */
1204     for (p = in_name; *p; p++);	/* skip to end of string */
1205     while (p > in_name && *p != '/')	/* find last '/' */
1206 	p--;
1207     if (*p == '/')
1208 	p++;
1209     sprintf(bakfile, "%s.BAK", p);
1210 
1211     /* copy in_name to backup file */
1212     bakchn = creat(bakfile, 0600);
1213     if (bakchn < 0)
1214 	err(1, "%s", bakfile);
1215     while ((n = read(fileno(input), buff, sizeof buff)) != 0)
1216 	if (write(bakchn, buff, n) != n)
1217 	    err(1, "%s", bakfile);
1218     if (n < 0)
1219 	err(1, "%s", in_name);
1220     close(bakchn);
1221     fclose(input);
1222 
1223     /* re-open backup file as the input file */
1224     input = fopen(bakfile, "r");
1225     if (input == 0)
1226 	err(1, "%s", bakfile);
1227     /* now the original input file will be the output */
1228     output = fopen(in_name, "w");
1229     if (output == 0) {
1230 	unlink(bakfile);
1231 	err(1, "%s", in_name);
1232     }
1233 }
1234