xref: /freebsd/usr.bin/indent/indent.c (revision 40a8ac8f62b535d30349faf28cf47106b7041b83)
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     tabs_to_var = 0;
156 
157     /*--------------------------------------------------*\
158     |   		COMMAND LINE SCAN		 |
159     \*--------------------------------------------------*/
160 
161 #ifdef undef
162     max_col = 78;		/* -l78 */
163     lineup_to_parens = 1;	/* -lp */
164     ps.ljust_decl = 0;		/* -ndj */
165     ps.com_ind = 33;		/* -c33 */
166     star_comment_cont = 1;	/* -sc */
167     ps.ind_size = 8;		/* -i8 */
168     verbose = 0;
169     ps.decl_indent = 16;	/* -di16 */
170     ps.local_decl_indent = -1;	/* if this is not set to some nonnegative value
171 				 * by an arg, we will set this equal to
172 				 * ps.decl_ind */
173     ps.indent_parameters = 1;	/* -ip */
174     ps.decl_com_ind = 0;	/* if this is not set to some positive value
175 				 * by an arg, we will set this equal to
176 				 * ps.com_ind */
177     btype_2 = 1;		/* -br */
178     cuddle_else = 1;		/* -ce */
179     ps.unindent_displace = 0;	/* -d0 */
180     ps.case_indent = 0;		/* -cli0 */
181     format_block_comments = 1;	/* -fcb */
182     format_col1_comments = 1;	/* -fc1 */
183     procnames_start_line = 1;	/* -psl */
184     proc_calls_space = 0;	/* -npcs */
185     comment_delimiter_on_blankline = 1;	/* -cdb */
186     ps.leave_comma = 1;		/* -nbc */
187 #endif
188 
189     for (i = 1; i < argc; ++i)
190 	if (strcmp(argv[i], "-npro") == 0)
191 	    break;
192     set_defaults();
193     if (i >= argc)
194 	set_profile();
195 
196     for (i = 1; i < argc; ++i) {
197 
198 	/*
199 	 * look thru args (if any) for changes to defaults
200 	 */
201 	if (argv[i][0] != '-') {/* no flag on parameter */
202 	    if (input == NULL) {	/* we must have the input file */
203 		in_name = argv[i];	/* remember name of input file */
204 		input = fopen(in_name, "r");
205 		if (input == NULL)	/* check for open error */
206 			err(1, "%s", in_name);
207 		continue;
208 	    }
209 	    else if (output == NULL) {	/* we have the output file */
210 		out_name = argv[i];	/* remember name of output file */
211 		if (strcmp(in_name, out_name) == 0) {	/* attempt to overwrite
212 							 * the file */
213 		    errx(1, "input and output files must be different");
214 		}
215 		output = fopen(out_name, "w");
216 		if (output == NULL)	/* check for create error */
217 			err(1, "%s", out_name);
218 		continue;
219 	    }
220 	    errx(1, "unknown parameter: %s", argv[i]);
221 	}
222 	else
223 	    set_option(argv[i]);
224     }				/* end of for */
225     if (input == NULL)
226 	input = stdin;
227     if (output == NULL) {
228 	if (troff || input == stdin)
229 	    output = stdout;
230 	else {
231 	    out_name = in_name;
232 	    bakcopy();
233 	}
234     }
235     if (ps.com_ind <= 1)
236 	ps.com_ind = 2;		/* dont put normal comments before column 2 */
237     if (troff) {
238 	if (bodyf.font[0] == 0)
239 	    parsefont(&bodyf, "R");
240 	if (scomf.font[0] == 0)
241 	    parsefont(&scomf, "I");
242 	if (blkcomf.font[0] == 0)
243 	    blkcomf = scomf, blkcomf.size += 2;
244 	if (boxcomf.font[0] == 0)
245 	    boxcomf = blkcomf;
246 	if (stringf.font[0] == 0)
247 	    parsefont(&stringf, "L");
248 	if (keywordf.font[0] == 0)
249 	    parsefont(&keywordf, "B");
250 	writefdef(&bodyf, 'B');
251 	writefdef(&scomf, 'C');
252 	writefdef(&blkcomf, 'L');
253 	writefdef(&boxcomf, 'X');
254 	writefdef(&stringf, 'S');
255 	writefdef(&keywordf, 'K');
256     }
257     if (block_comment_max_col <= 0)
258 	block_comment_max_col = max_col;
259     if (ps.local_decl_indent < 0)	/* if not specified by user, set this */
260 	ps.local_decl_indent = ps.decl_indent;
261     if (ps.decl_com_ind <= 0)	/* if not specified by user, set this */
262 	ps.decl_com_ind = ps.ljust_decl ? (ps.com_ind <= 10 ? 2 : ps.com_ind - 8) : ps.com_ind;
263     if (continuation_indent == 0)
264 	continuation_indent = ps.ind_size;
265     fill_buffer();		/* get first batch of stuff into input buffer */
266 
267     parse(semicolon);
268     {
269 	char *p = buf_ptr;
270 	int col = 1;
271 
272 	while (1) {
273 	    if (*p == ' ')
274 		col++;
275 	    else if (*p == '\t')
276 		col = ((col - 1) & ~7) + 9;
277 	    else
278 		break;
279 	    p++;
280 	}
281 	if (col > ps.ind_size)
282 	    ps.ind_level = ps.i_l_follow = col / ps.ind_size;
283     }
284     if (troff) {
285 	const char *p = in_name,
286 	           *beg = in_name;
287 
288 	while (*p)
289 	    if (*p++ == '/')
290 		beg = p;
291 	fprintf(output, ".Fn \"%s\"\n", beg);
292     }
293     /*
294      * START OF MAIN LOOP
295      */
296 
297     while (1) {			/* this is the main loop.  it will go until we
298 				 * reach eof */
299 	int         is_procname;
300 
301 	type_code = lexi();	/* lexi reads one token.  The actual
302 				 * characters read are stored in "token". lexi
303 				 * returns a code indicating the type of token */
304 	is_procname = ps.procname[0];
305 
306 	/*
307 	 * The following code moves everything following an if (), while (),
308 	 * else, etc. up to the start of the following stmt to a buffer. This
309 	 * allows proper handling of both kinds of brace placement.
310 	 */
311 
312 	flushed_nl = false;
313 	while (ps.search_brace) {	/* if we scanned an if(), while(),
314 					 * etc., we might need to copy stuff
315 					 * into a buffer we must loop, copying
316 					 * stuff into save_com, until we find
317 					 * the start of the stmt which follows
318 					 * the if, or whatever */
319 	    switch (type_code) {
320 	    case newline:
321 		++line_no;
322 		flushed_nl = true;
323 	    case form_feed:
324 		break;		/* form feeds and newlines found here will be
325 				 * ignored */
326 
327 	    case lbrace:	/* this is a brace that starts the compound
328 				 * stmt */
329 		if (sc_end == 0) {	/* ignore buffering if a comment wasn't
330 					 * stored up */
331 		    ps.search_brace = false;
332 		    goto check_type;
333 		}
334 		if (btype_2) {
335 		    save_com[0] = '{';	/* we either want to put the brace
336 					 * right after the if */
337 		    goto sw_buffer;	/* go to common code to get out of
338 					 * this loop */
339 		}
340 	    case comment:	/* we have a comment, so we must copy it into
341 				 * the buffer */
342 		if (!flushed_nl || sc_end != 0) {
343 		    if (sc_end == 0) {	/* if this is the first comment, we
344 					 * must set up the buffer */
345 			save_com[0] = save_com[1] = ' ';
346 			sc_end = &(save_com[2]);
347 		    }
348 		    else {
349 			*sc_end++ = '\n';	/* add newline between
350 						 * comments */
351 			*sc_end++ = ' ';
352 			--line_no;
353 		    }
354 		    *sc_end++ = '/';	/* copy in start of comment */
355 		    *sc_end++ = '*';
356 
357 		    for (;;) {	/* loop until we get to the end of the comment */
358 			*sc_end = *buf_ptr++;
359 			if (buf_ptr >= buf_end)
360 			    fill_buffer();
361 
362 			if (*sc_end++ == '*' && *buf_ptr == '/')
363 			    break;	/* we are at end of comment */
364 
365 			if (sc_end >= &(save_com[sc_size])) {	/* check for temp buffer
366 								 * overflow */
367 			    diag2(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
368 			    fflush(output);
369 			    exit(1);
370 			}
371 		    }
372 		    *sc_end++ = '/';	/* add ending slash */
373 		    if (++buf_ptr >= buf_end)	/* get past / in buffer */
374 			fill_buffer();
375 		    break;
376 		}
377 	    default:		/* it is the start of a normal statement */
378 		if (flushed_nl)	/* if we flushed a newline, make sure it is
379 				 * put back */
380 		    force_nl = true;
381 		if ((type_code == sp_paren && *token == 'i'
382 			&& last_else && ps.else_if)
383 			|| (type_code == sp_nparen && *token == 'e'
384 			&& e_code != s_code && e_code[-1] == '}'))
385 		    force_nl = false;
386 
387 		if (sc_end == 0) {	/* ignore buffering if comment wasn't
388 					 * saved up */
389 		    ps.search_brace = false;
390 		    goto check_type;
391 		}
392 		if (force_nl) {	/* if we should insert a nl here, put it into
393 				 * the buffer */
394 		    force_nl = false;
395 		    --line_no;	/* this will be re-increased when the nl is
396 				 * read from the buffer */
397 		    *sc_end++ = '\n';
398 		    *sc_end++ = ' ';
399 		    if (verbose && !flushed_nl)	/* print error msg if the line
400 						 * was not already broken */
401 			diag2(0, "Line broken");
402 		    flushed_nl = false;
403 		}
404 		for (t_ptr = token; *t_ptr; ++t_ptr)
405 		    *sc_end++ = *t_ptr;	/* copy token into temp buffer */
406 		ps.procname[0] = 0;
407 
408 	sw_buffer:
409 		ps.search_brace = false;	/* stop looking for start of
410 						 * stmt */
411 		bp_save = buf_ptr;	/* save current input buffer */
412 		be_save = buf_end;
413 		buf_ptr = save_com;	/* fix so that subsequent calls to
414 					 * lexi will take tokens out of
415 					 * save_com */
416 		*sc_end++ = ' ';/* add trailing blank, just in case */
417 		buf_end = sc_end;
418 		sc_end = 0;
419 		break;
420 	    }			/* end of switch */
421 	    if (type_code != 0)	/* we must make this check, just in case there
422 				 * was an unexpected EOF */
423 		type_code = lexi();	/* read another token */
424 	    /* if (ps.search_brace) ps.procname[0] = 0; */
425 	    if ((is_procname = ps.procname[0]) && flushed_nl
426 		    && !procnames_start_line && ps.in_decl
427 		    && type_code == ident)
428 		flushed_nl = 0;
429 	}			/* end of while (search_brace) */
430 	last_else = 0;
431 check_type:
432 	if (type_code == 0) {	/* we got eof */
433 	    if (s_lab != e_lab || s_code != e_code
434 		    || s_com != e_com)	/* must dump end of line */
435 		dump_line();
436 	    if (ps.tos > 1)	/* check for balanced braces */
437 		diag2(1, "Stuff missing from end of file");
438 
439 	    if (verbose) {
440 		printf("There were %d output lines and %d comments\n",
441 		       ps.out_lines, ps.out_coms);
442 		printf("(Lines with comments)/(Lines with code): %6.3f\n",
443 		       (1.0 * ps.com_lines) / code_lines);
444 	    }
445 	    fflush(output);
446 	    exit(found_err);
447 	}
448 	if (
449 		(type_code != comment) &&
450 		(type_code != newline) &&
451 		(type_code != preesc) &&
452 		(type_code != form_feed)) {
453 	    if (force_nl &&
454 		    (type_code != semicolon) &&
455 		    (type_code != lbrace || !btype_2)) {
456 		/* we should force a broken line here */
457 		if (verbose && !flushed_nl)
458 		    diag2(0, "Line broken");
459 		flushed_nl = false;
460 		dump_line();
461 		ps.want_blank = false;	/* dont insert blank at line start */
462 		force_nl = false;
463 	    }
464 	    ps.in_stmt = true;	/* turn on flag which causes an extra level of
465 				 * indentation. this is turned off by a ; or
466 				 * '}' */
467 	    if (s_com != e_com) {	/* the turkey has embedded a comment
468 					 * in a line. fix it */
469 		*e_code++ = ' ';
470 		for (t_ptr = s_com; *t_ptr; ++t_ptr) {
471 		    CHECK_SIZE_CODE;
472 		    *e_code++ = *t_ptr;
473 		}
474 		*e_code++ = ' ';
475 		*e_code = '\0';	/* null terminate code sect */
476 		ps.want_blank = false;
477 		e_com = s_com;
478 	    }
479 	}
480 	else if (type_code != comment)	/* preserve force_nl thru a comment */
481 	    force_nl = false;	/* cancel forced newline after newline, form
482 				 * feed, etc */
483 
484 
485 
486 	/*-----------------------------------------------------*\
487 	|	   do switch on type of token scanned		|
488 	\*-----------------------------------------------------*/
489 	CHECK_SIZE_CODE;
490 	switch (type_code) {	/* now, decide what to do with the token */
491 
492 	case form_feed:	/* found a form feed in line */
493 	    ps.use_ff = true;	/* a form feed is treated much like a newline */
494 	    dump_line();
495 	    ps.want_blank = false;
496 	    break;
497 
498 	case newline:
499 	    if (ps.last_token != comma || ps.p_l_follow > 0
500 		    || !ps.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
501 		dump_line();
502 		ps.want_blank = false;
503 	    }
504 	    ++line_no;		/* keep track of input line number */
505 	    break;
506 
507 	case lparen:		/* got a '(' or '[' */
508 	    ++ps.p_l_follow;	/* count parens to make Healy happy */
509 	    if (ps.want_blank && *token != '[' &&
510 		    (ps.last_token != ident || proc_calls_space
511 	      || (ps.its_a_keyword && (!ps.sizeof_keyword || Bill_Shannon))))
512 		*e_code++ = ' ';
513 	    if (ps.in_decl && !ps.block_init)
514 		if (troff && !ps.dumped_decl_indent && !is_procname && ps.last_token == decl) {
515 		    ps.dumped_decl_indent = 1;
516 		    sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
517 		    e_code += strlen(e_code);
518 		}
519 		else {
520 		    while ((e_code - s_code) < dec_ind) {
521 			CHECK_SIZE_CODE;
522 			*e_code++ = ' ';
523 		    }
524 		    *e_code++ = token[0];
525 		}
526 	    else
527 		*e_code++ = token[0];
528 	    ps.paren_indents[ps.p_l_follow - 1] = e_code - s_code;
529 	    if (sp_sw && ps.p_l_follow == 1 && extra_expression_indent
530 		    && ps.paren_indents[0] < 2 * ps.ind_size)
531 		ps.paren_indents[0] = 2 * ps.ind_size;
532 	    ps.want_blank = false;
533 	    if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
534 		/*
535 		 * this is a kluge to make sure that declarations will be
536 		 * aligned right if proc decl has an explicit type on it, i.e.
537 		 * "int a(x) {..."
538 		 */
539 		parse(semicolon);	/* I said this was a kluge... */
540 		ps.in_or_st = false;	/* turn off flag for structure decl or
541 					 * initialization */
542 	    }
543 	    if (ps.sizeof_keyword)
544 		ps.sizeof_mask |= 1 << ps.p_l_follow;
545 	    break;
546 
547 	case rparen:		/* got a ')' or ']' */
548 	    rparen_count--;
549 	    if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.sizeof_mask) {
550 		ps.last_u_d = true;
551 		ps.cast_mask &= (1 << ps.p_l_follow) - 1;
552 		ps.want_blank = false;
553 	    } else
554 		ps.want_blank = true;
555 	    ps.sizeof_mask &= (1 << ps.p_l_follow) - 1;
556 	    if (--ps.p_l_follow < 0) {
557 		ps.p_l_follow = 0;
558 		diag3(0, "Extra %c", *token);
559 	    }
560 	    if (e_code == s_code)	/* if the paren starts the line */
561 		ps.paren_level = ps.p_l_follow;	/* then indent it */
562 
563 	    *e_code++ = token[0];
564 
565 	    if (sp_sw && (ps.p_l_follow == 0)) {	/* check for end of if
566 							 * (...), or some such */
567 		sp_sw = false;
568 		force_nl = true;/* must force newline after if */
569 		ps.last_u_d = true;	/* inform lexi that a following
570 					 * operator is unary */
571 		ps.in_stmt = false;	/* dont use stmt continuation
572 					 * indentation */
573 
574 		parse(hd_type);	/* let parser worry about if, or whatever */
575 	    }
576 	    ps.search_brace = btype_2;	/* this should insure that constructs
577 					 * such as main(){...} and int[]{...}
578 					 * have their braces put in the right
579 					 * place */
580 	    break;
581 
582 	case unary_op:		/* this could be any unary operation */
583 	    if (ps.want_blank)
584 		*e_code++ = ' ';
585 
586 	    if (troff && !ps.dumped_decl_indent && ps.in_decl && !is_procname) {
587 		sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
588 		ps.dumped_decl_indent = 1;
589 		e_code += strlen(e_code);
590 	    }
591 	    else {
592 		const char *res = token;
593 
594 		if (ps.in_decl && !ps.block_init) {	/* if this is a unary op
595 							 * in a declaration, we
596 							 * should indent this
597 							 * token */
598 		    for (i = 0; token[i]; ++i);	/* find length of token */
599 		    while ((e_code - s_code) < (dec_ind - i)) {
600 			CHECK_SIZE_CODE;
601 			*e_code++ = ' ';	/* pad it */
602 		    }
603 		}
604 		if (troff && token[0] == '-' && token[1] == '>')
605 		    res = "\\(->";
606 		for (t_ptr = res; *t_ptr; ++t_ptr) {
607 		    CHECK_SIZE_CODE;
608 		    *e_code++ = *t_ptr;
609 		}
610 	    }
611 	    ps.want_blank = false;
612 	    break;
613 
614 	case binary_op:	/* any binary operation */
615 	    if (ps.want_blank)
616 		*e_code++ = ' ';
617 	    {
618 		const char *res = token;
619 
620 		if (troff)
621 		    switch (token[0]) {
622 		    case '<':
623 			if (token[1] == '=')
624 			    res = "\\(<=";
625 			break;
626 		    case '>':
627 			if (token[1] == '=')
628 			    res = "\\(>=";
629 			break;
630 		    case '!':
631 			if (token[1] == '=')
632 			    res = "\\(!=";
633 			break;
634 		    case '|':
635 			if (token[1] == '|')
636 			    res = "\\(br\\(br";
637 			else if (token[1] == 0)
638 			    res = "\\(br";
639 			break;
640 		    }
641 		for (t_ptr = res; *t_ptr; ++t_ptr) {
642 		    CHECK_SIZE_CODE;
643 		    *e_code++ = *t_ptr;	/* move the operator */
644 		}
645 	    }
646 	    ps.want_blank = true;
647 	    break;
648 
649 	case postop:		/* got a trailing ++ or -- */
650 	    *e_code++ = token[0];
651 	    *e_code++ = token[1];
652 	    ps.want_blank = true;
653 	    break;
654 
655 	case question:		/* got a ? */
656 	    squest++;		/* this will be used when a later colon
657 				 * appears so we can distinguish the
658 				 * <c>?<n>:<n> construct */
659 	    if (ps.want_blank)
660 		*e_code++ = ' ';
661 	    *e_code++ = '?';
662 	    ps.want_blank = true;
663 	    break;
664 
665 	case casestmt:		/* got word 'case' or 'default' */
666 	    scase = true;	/* so we can process the later colon properly */
667 	    goto copy_id;
668 
669 	case colon:		/* got a ':' */
670 	    if (squest > 0) {	/* it is part of the <c>?<n>: <n> construct */
671 		--squest;
672 		if (ps.want_blank)
673 		    *e_code++ = ' ';
674 		*e_code++ = ':';
675 		ps.want_blank = true;
676 		break;
677 	    }
678 	    if (ps.in_or_st) {
679 		*e_code++ = ':';
680 		ps.want_blank = false;
681 		break;
682 	    }
683 	    ps.in_stmt = false;	/* seeing a label does not imply we are in a
684 				 * stmt */
685 	    for (t_ptr = s_code; *t_ptr; ++t_ptr)
686 		*e_lab++ = *t_ptr;	/* turn everything so far into a label */
687 	    e_code = s_code;
688 	    *e_lab++ = ':';
689 	    *e_lab++ = ' ';
690 	    *e_lab = '\0';
691 
692 	    force_nl = ps.pcase = scase;	/* ps.pcase will be used by
693 						 * dump_line to decide how to
694 						 * indent the label. force_nl
695 						 * will force a case n: to be
696 						 * on a line by itself */
697 	    scase = false;
698 	    ps.want_blank = false;
699 	    break;
700 
701 	case semicolon:	/* got a ';' */
702 	    ps.in_or_st = false;/* we are not in an initialization or
703 				 * structure declaration */
704 	    scase = false;	/* these will only need resetting in an error */
705 	    squest = 0;
706 	    if (ps.last_token == rparen && rparen_count == 0)
707 		ps.in_parameter_declaration = 0;
708 	    ps.cast_mask = 0;
709 	    ps.sizeof_mask = 0;
710 	    ps.block_init = 0;
711 	    ps.block_init_level = 0;
712 	    ps.just_saw_decl--;
713 
714 	    if (ps.in_decl && s_code == e_code && !ps.block_init)
715 		while ((e_code - s_code) < (dec_ind - 1)) {
716 		    CHECK_SIZE_CODE;
717 		    *e_code++ = ' ';
718 		}
719 
720 	    ps.in_decl = (ps.dec_nest > 0);	/* if we were in a first level
721 						 * structure declaration, we
722 						 * arent any more */
723 
724 	    if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
725 
726 		/*
727 		 * This should be true iff there were unbalanced parens in the
728 		 * stmt.  It is a bit complicated, because the semicolon might
729 		 * be in a for stmt
730 		 */
731 		diag2(1, "Unbalanced parens");
732 		ps.p_l_follow = 0;
733 		if (sp_sw) {	/* this is a check for an if, while, etc. with
734 				 * unbalanced parens */
735 		    sp_sw = false;
736 		    parse(hd_type);	/* dont lose the if, or whatever */
737 		}
738 	    }
739 	    *e_code++ = ';';
740 	    ps.want_blank = true;
741 	    ps.in_stmt = (ps.p_l_follow > 0);	/* we are no longer in the
742 						 * middle of a stmt */
743 
744 	    if (!sp_sw) {	/* if not if for (;;) */
745 		parse(semicolon);	/* let parser know about end of stmt */
746 		force_nl = true;/* force newline after an end of stmt */
747 	    }
748 	    break;
749 
750 	case lbrace:		/* got a '{' */
751 	    ps.in_stmt = false;	/* dont indent the {} */
752 	    if (!ps.block_init)
753 		force_nl = true;/* force other stuff on same line as '{' onto
754 				 * new line */
755 	    else if (ps.block_init_level <= 0)
756 		ps.block_init_level = 1;
757 	    else
758 		ps.block_init_level++;
759 
760 	    if (s_code != e_code && !ps.block_init) {
761 		if (!btype_2) {
762 		    dump_line();
763 		    ps.want_blank = false;
764 		}
765 		else if (ps.in_parameter_declaration && !ps.in_or_st) {
766 		    ps.i_l_follow = 0;
767 		    if (function_brace_split) {	/* dump the line prior to the
768 						 * brace ... */
769 			dump_line();
770 			ps.want_blank = false;
771 		    } else	/* add a space between the decl and brace */
772 			ps.want_blank = true;
773 		}
774 	    }
775 	    if (ps.in_parameter_declaration)
776 		prefix_blankline_requested = 0;
777 
778 	    if (ps.p_l_follow > 0) {	/* check for preceding unbalanced
779 					 * parens */
780 		diag2(1, "Unbalanced parens");
781 		ps.p_l_follow = 0;
782 		if (sp_sw) {	/* check for unclosed if, for, etc. */
783 		    sp_sw = false;
784 		    parse(hd_type);
785 		    ps.ind_level = ps.i_l_follow;
786 		}
787 	    }
788 	    if (s_code == e_code)
789 		ps.ind_stmt = false;	/* dont put extra indentation on line
790 					 * with '{' */
791 	    if (ps.in_decl && ps.in_or_st) {	/* this is either a structure
792 						 * declaration or an init */
793 		di_stack[ps.dec_nest++] = dec_ind;
794 		/* ?		dec_ind = 0; */
795 	    }
796 	    else {
797 		ps.decl_on_line = false;	/* we can't be in the middle of
798 						 * a declaration, so don't do
799 						 * special indentation of
800 						 * comments */
801 		if (blanklines_after_declarations_at_proctop
802 			&& ps.in_parameter_declaration)
803 		    postfix_blankline_requested = 1;
804 		ps.in_parameter_declaration = 0;
805 	    }
806 	    dec_ind = 0;
807 	    parse(lbrace);	/* let parser know about this */
808 	    if (ps.want_blank)	/* put a blank before '{' if '{' is not at
809 				 * start of line */
810 		*e_code++ = ' ';
811 	    ps.want_blank = false;
812 	    *e_code++ = '{';
813 	    ps.just_saw_decl = 0;
814 	    break;
815 
816 	case rbrace:		/* got a '}' */
817 	    if (ps.p_stack[ps.tos] == decl && !ps.block_init)	/* semicolons can be
818 								 * omitted in
819 								 * declarations */
820 		parse(semicolon);
821 	    if (ps.p_l_follow) {/* check for unclosed if, for, else. */
822 		diag2(1, "Unbalanced parens");
823 		ps.p_l_follow = 0;
824 		sp_sw = false;
825 	    }
826 	    ps.just_saw_decl = 0;
827 	    ps.block_init_level--;
828 	    if (s_code != e_code && !ps.block_init) {	/* '}' must be first on
829 							 * line */
830 		if (verbose)
831 		    diag2(0, "Line broken");
832 		dump_line();
833 	    }
834 	    *e_code++ = '}';
835 	    ps.want_blank = true;
836 	    ps.in_stmt = ps.ind_stmt = false;
837 	    if (ps.dec_nest > 0) {	/* we are in multi-level structure
838 					 * declaration */
839 		dec_ind = di_stack[--ps.dec_nest];
840 		if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
841 		    ps.just_saw_decl = 2;
842 		ps.in_decl = true;
843 	    }
844 	    prefix_blankline_requested = 0;
845 	    parse(rbrace);	/* let parser know about this */
846 	    ps.search_brace = cuddle_else && ps.p_stack[ps.tos] == ifhead
847 		&& ps.il[ps.tos] >= ps.ind_level;
848 	    if (ps.tos <= 1 && blanklines_after_procs && ps.dec_nest <= 0)
849 		postfix_blankline_requested = 1;
850 	    break;
851 
852 	case swstmt:		/* got keyword "switch" */
853 	    sp_sw = true;
854 	    hd_type = swstmt;	/* keep this for when we have seen the
855 				 * expression */
856 	    goto copy_id;	/* go move the token into buffer */
857 
858 	case sp_paren:		/* token is if, while, for */
859 	    sp_sw = true;	/* the interesting stuff is done after the
860 				 * expression is scanned */
861 	    hd_type = (*token == 'i' ? ifstmt :
862 		       (*token == 'w' ? whilestmt : forstmt));
863 
864 	    /*
865 	     * remember the type of header for later use by parser
866 	     */
867 	    goto copy_id;	/* copy the token into line */
868 
869 	case sp_nparen:	/* got else, do */
870 	    ps.in_stmt = false;
871 	    if (*token == 'e') {
872 		if (e_code != s_code && (!cuddle_else || e_code[-1] != '}')) {
873 		    if (verbose)
874 			diag2(0, "Line broken");
875 		    dump_line();/* make sure this starts a line */
876 		    ps.want_blank = false;
877 		}
878 		force_nl = true;/* also, following stuff must go onto new line */
879 		last_else = 1;
880 		parse(elselit);
881 	    }
882 	    else {
883 		if (e_code != s_code) {	/* make sure this starts a line */
884 		    if (verbose)
885 			diag2(0, "Line broken");
886 		    dump_line();
887 		    ps.want_blank = false;
888 		}
889 		force_nl = true;/* also, following stuff must go onto new line */
890 		last_else = 0;
891 		parse(dolit);
892 	    }
893 	    goto copy_id;	/* move the token into line */
894 
895 	case decl:		/* we have a declaration type (int, register,
896 				 * etc.) */
897 	    parse(decl);	/* let parser worry about indentation */
898 	    if (ps.last_token == rparen && ps.tos <= 1) {
899 		ps.in_parameter_declaration = 1;
900 		if (s_code != e_code) {
901 		    dump_line();
902 		    ps.want_blank = 0;
903 		}
904 	    }
905 	    if (ps.in_parameter_declaration && ps.indent_parameters && ps.dec_nest == 0) {
906 		ps.ind_level = ps.i_l_follow = 1;
907 		ps.ind_stmt = 0;
908 	    }
909 	    ps.in_or_st = true;	/* this might be a structure or initialization
910 				 * declaration */
911 	    ps.in_decl = ps.decl_on_line = true;
912 	    if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
913 		ps.just_saw_decl = 2;
914 	    prefix_blankline_requested = 0;
915 	    for (i = 0; token[i++];);	/* get length of token */
916 
917 	    if (ps.ind_level == 0 || ps.dec_nest > 0) {
918 		/* global variable or struct member in local variable */
919 		dec_ind = ps.decl_indent > 0 ? ps.decl_indent : i;
920 		tabs_to_var = (use_tabs ? ps.decl_indent > 0 : 0);
921 	    } else {
922 		/* local variable */
923 		dec_ind = ps.local_decl_indent > 0 ? ps.local_decl_indent : i;
924 		tabs_to_var = (use_tabs ? ps.local_decl_indent > 0 : 0);
925 	    }
926 	    goto copy_id;
927 
928 	case ident:		/* got an identifier or constant */
929 	    if (ps.in_decl) {	/* if we are in a declaration, we must indent
930 				 * identifier */
931 		if (is_procname == 0 || !procnames_start_line) {
932 		    if (!ps.block_init) {
933 			if (troff && !ps.dumped_decl_indent) {
934 			    if (ps.want_blank)
935 				*e_code++ = ' ';
936 			    ps.want_blank = false;
937 			    sprintf(e_code, "\n.De %dp+\200p\n", dec_ind * 7);
938 			    ps.dumped_decl_indent = 1;
939 			    e_code += strlen(e_code);
940 			} else {
941 			    int cur_dec_ind;
942 			    int pos, startpos;
943 
944 			    /*
945 			     * in order to get the tab math right for
946 			     * indentations that are not multiples of 8 we
947 			     * need to modify both startpos and dec_ind
948 			     * (cur_dec_ind) here by eight minus the
949 			     * remainder of the current starting column
950 			     * divided by eight. This seems to be a
951 			     * properly working fix
952 			     */
953 			    startpos = e_code - s_code;
954 			    cur_dec_ind = dec_ind;
955 			    pos = startpos;
956 			    if ((ps.ind_level * ps.ind_size) % 8 != 0) {
957 				pos += (ps.ind_level * ps.ind_size) % 8;
958 				cur_dec_ind += (ps.ind_level * ps.ind_size) % 8;
959 			    }
960 
961 			    if (tabs_to_var) {
962 				while ((pos & ~7) + 8 <= cur_dec_ind) {
963 				    CHECK_SIZE_CODE;
964 				    *e_code++ = '\t';
965 				    pos = (pos & ~7) + 8;
966 				}
967 			    }
968 			    while (pos < cur_dec_ind) {
969 				CHECK_SIZE_CODE;
970 				*e_code++ = ' ';
971 				pos++;
972 			    }
973 			    if (ps.want_blank && e_code - s_code == startpos)
974 				*e_code++ = ' ';
975 			    ps.want_blank = false;
976 			}
977 		    }
978 		} else {
979 		    if (ps.want_blank)
980 			*e_code++ = ' ';
981 		    ps.want_blank = false;
982 		    if (dec_ind && s_code != e_code)
983 			dump_line();
984 		    dec_ind = 0;
985 		}
986 	    }
987 	    else if (sp_sw && ps.p_l_follow == 0) {
988 		sp_sw = false;
989 		force_nl = true;
990 		ps.last_u_d = true;
991 		ps.in_stmt = false;
992 		parse(hd_type);
993 	    }
994     copy_id:
995 	    if (ps.want_blank)
996 		*e_code++ = ' ';
997 	    if (troff && ps.its_a_keyword) {
998 		e_code = chfont(&bodyf, &keywordf, e_code);
999 		for (t_ptr = token; *t_ptr; ++t_ptr) {
1000 		    CHECK_SIZE_CODE;
1001 		    *e_code++ = keywordf.allcaps && islower(*t_ptr)
1002 			? toupper(*t_ptr) : *t_ptr;
1003 		}
1004 		e_code = chfont(&keywordf, &bodyf, e_code);
1005 	    }
1006 	    else
1007 		for (t_ptr = token; *t_ptr; ++t_ptr) {
1008 		    CHECK_SIZE_CODE;
1009 		    *e_code++ = *t_ptr;
1010 		}
1011 	    ps.want_blank = true;
1012 	    break;
1013 
1014 	case period:		/* treat a period kind of like a binary
1015 				 * operation */
1016 	    *e_code++ = '.';	/* move the period into line */
1017 	    ps.want_blank = false;	/* dont put a blank after a period */
1018 	    break;
1019 
1020 	case comma:
1021 	    ps.want_blank = (s_code != e_code);	/* only put blank after comma
1022 						 * if comma does not start the
1023 						 * line */
1024 	    if (ps.in_decl && is_procname == 0 && !ps.block_init)
1025 		while ((e_code - s_code) < (dec_ind - 1)) {
1026 		    CHECK_SIZE_CODE;
1027 		    *e_code++ = ' ';
1028 		}
1029 
1030 	    *e_code++ = ',';
1031 	    if (ps.p_l_follow == 0) {
1032 		if (ps.block_init_level <= 0)
1033 		    ps.block_init = 0;
1034 		if (break_comma && (!ps.leave_comma || compute_code_target() + (e_code - s_code) > max_col - 8))
1035 		    force_nl = true;
1036 	    }
1037 	    break;
1038 
1039 	case preesc:		/* got the character '#' */
1040 	    if ((s_com != e_com) ||
1041 		    (s_lab != e_lab) ||
1042 		    (s_code != e_code))
1043 		dump_line();
1044 	    *e_lab++ = '#';	/* move whole line to 'label' buffer */
1045 	    {
1046 		int         in_comment = 0;
1047 		int         com_start = 0;
1048 		char        quote = 0;
1049 		int         com_end = 0;
1050 
1051 		while (*buf_ptr == ' ' || *buf_ptr == '\t') {
1052 		    buf_ptr++;
1053 		    if (buf_ptr >= buf_end)
1054 			fill_buffer();
1055 		}
1056 		while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
1057 		    CHECK_SIZE_LAB;
1058 		    *e_lab = *buf_ptr++;
1059 		    if (buf_ptr >= buf_end)
1060 			fill_buffer();
1061 		    switch (*e_lab++) {
1062 		    case BACKSLASH:
1063 			if (troff)
1064 			    *e_lab++ = BACKSLASH;
1065 			if (!in_comment) {
1066 			    *e_lab++ = *buf_ptr++;
1067 			    if (buf_ptr >= buf_end)
1068 				fill_buffer();
1069 			}
1070 			break;
1071 		    case '/':
1072 			if (*buf_ptr == '*' && !in_comment && !quote) {
1073 			    in_comment = 1;
1074 			    *e_lab++ = *buf_ptr++;
1075 			    com_start = e_lab - s_lab - 2;
1076 			}
1077 			break;
1078 		    case '"':
1079 			if (quote == '"')
1080 			    quote = 0;
1081 			break;
1082 		    case '\'':
1083 			if (quote == '\'')
1084 			    quote = 0;
1085 			break;
1086 		    case '*':
1087 			if (*buf_ptr == '/' && in_comment) {
1088 			    in_comment = 0;
1089 			    *e_lab++ = *buf_ptr++;
1090 			    com_end = e_lab - s_lab;
1091 			}
1092 			break;
1093 		    }
1094 		}
1095 
1096 		while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1097 		    e_lab--;
1098 		if (e_lab - s_lab == com_end && bp_save == 0) {	/* comment on
1099 								 * preprocessor line */
1100 		    if (sc_end == 0)	/* if this is the first comment, we
1101 					 * must set up the buffer */
1102 			sc_end = &(save_com[0]);
1103 		    else {
1104 			*sc_end++ = '\n';	/* add newline between
1105 						 * comments */
1106 			*sc_end++ = ' ';
1107 			--line_no;
1108 		    }
1109 		    bcopy(s_lab + com_start, sc_end, com_end - com_start);
1110 		    sc_end += com_end - com_start;
1111 		    if (sc_end >= &save_com[sc_size])
1112 			abort();
1113 		    e_lab = s_lab + com_start;
1114 		    while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1115 			e_lab--;
1116 		    bp_save = buf_ptr;	/* save current input buffer */
1117 		    be_save = buf_end;
1118 		    buf_ptr = save_com;	/* fix so that subsequent calls to
1119 					 * lexi will take tokens out of
1120 					 * save_com */
1121 		    *sc_end++ = ' ';	/* add trailing blank, just in case */
1122 		    buf_end = sc_end;
1123 		    sc_end = 0;
1124 		}
1125 		*e_lab = '\0';	/* null terminate line */
1126 		ps.pcase = false;
1127 	    }
1128 
1129 	    if (strncmp(s_lab, "#if", 3) == 0) {
1130 		if (blanklines_around_conditional_compilation) {
1131 		    int c;
1132 		    prefix_blankline_requested++;
1133 		    while ((c = getc(input)) == '\n');
1134 		    ungetc(c, input);
1135 		}
1136 		if ((size_t)ifdef_level < sizeof(state_stack)/sizeof(state_stack[0])) {
1137 		    match_state[ifdef_level].tos = -1;
1138 		    state_stack[ifdef_level++] = ps;
1139 		}
1140 		else
1141 		    diag2(1, "#if stack overflow");
1142 	    }
1143 	    else if (strncmp(s_lab, "#else", 5) == 0)
1144 		if (ifdef_level <= 0)
1145 		    diag2(1, "Unmatched #else");
1146 		else {
1147 		    match_state[ifdef_level - 1] = ps;
1148 		    ps = state_stack[ifdef_level - 1];
1149 		}
1150 	    else if (strncmp(s_lab, "#endif", 6) == 0) {
1151 		if (ifdef_level <= 0)
1152 		    diag2(1, "Unmatched #endif");
1153 		else {
1154 		    ifdef_level--;
1155 
1156 #ifdef undef
1157 		    /*
1158 		     * This match needs to be more intelligent before the
1159 		     * message is useful
1160 		     */
1161 		    if (match_state[ifdef_level].tos >= 0
1162 			  && bcmp(&ps, &match_state[ifdef_level], sizeof ps))
1163 			diag2(0, "Syntactically inconsistent #ifdef alternatives");
1164 #endif
1165 		}
1166 		if (blanklines_around_conditional_compilation) {
1167 		    postfix_blankline_requested++;
1168 		    n_real_blanklines = 0;
1169 		}
1170 	    }
1171 	    break;		/* subsequent processing of the newline
1172 				 * character will cause the line to be printed */
1173 
1174 	case comment:		/* we have gotten a / followed by * this is a biggie */
1175 	    if (flushed_nl) {	/* we should force a broken line here */
1176 		flushed_nl = false;
1177 		dump_line();
1178 		ps.want_blank = false;	/* dont insert blank at line start */
1179 		force_nl = false;
1180 	    }
1181 	    pr_comment();
1182 	    break;
1183 	}			/* end of big switch stmt */
1184 
1185 	*e_code = '\0';		/* make sure code section is null terminated */
1186 	if (type_code != comment && type_code != newline && type_code != preesc)
1187 	    ps.last_token = type_code;
1188     }				/* end of main while (1) loop */
1189 }
1190 
1191 /*
1192  * copy input file to backup file if in_name is /blah/blah/blah/file, then
1193  * backup file will be ".Bfile" then make the backup file the input and
1194  * original input file the output
1195  */
1196 static void
1197 bakcopy(void)
1198 {
1199     int         n,
1200                 bakchn;
1201     char        buff[8 * 1024];
1202     const char *p;
1203 
1204     /* construct file name .Bfile */
1205     for (p = in_name; *p; p++);	/* skip to end of string */
1206     while (p > in_name && *p != '/')	/* find last '/' */
1207 	p--;
1208     if (*p == '/')
1209 	p++;
1210     sprintf(bakfile, "%s.BAK", p);
1211 
1212     /* copy in_name to backup file */
1213     bakchn = creat(bakfile, 0600);
1214     if (bakchn < 0)
1215 	err(1, "%s", bakfile);
1216     while ((n = read(fileno(input), buff, sizeof buff)) != 0)
1217 	if (write(bakchn, buff, n) != n)
1218 	    err(1, "%s", bakfile);
1219     if (n < 0)
1220 	err(1, "%s", in_name);
1221     close(bakchn);
1222     fclose(input);
1223 
1224     /* re-open backup file as the input file */
1225     input = fopen(bakfile, "r");
1226     if (input == NULL)
1227 	err(1, "%s", bakfile);
1228     /* now the original input file will be the output */
1229     output = fopen(in_name, "w");
1230     if (output == NULL) {
1231 	unlink(bakfile);
1232 	err(1, "%s", in_name);
1233     }
1234 }
1235