Home | History | Annotate | Line # | Download | only in indent
indent.c revision 1.95
      1 /*	$NetBSD: indent.c,v 1.95 2021/09/30 21:38:43 rillig Exp $	*/
      2 
      3 /*-
      4  * SPDX-License-Identifier: BSD-4-Clause
      5  *
      6  * Copyright (c) 1985 Sun Microsystems, Inc.
      7  * Copyright (c) 1976 Board of Trustees of the University of Illinois.
      8  * Copyright (c) 1980, 1993
      9  *	The Regents of the University of California.  All rights reserved.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  * 3. All advertising materials mentioning features or use of this software
     20  *    must display the following acknowledgement:
     21  *	This product includes software developed by the University of
     22  *	California, Berkeley and its contributors.
     23  * 4. Neither the name of the University nor the names of its contributors
     24  *    may be used to endorse or promote products derived from this software
     25  *    without specific prior written permission.
     26  *
     27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     37  * SUCH DAMAGE.
     38  */
     39 
     40 #if 0
     41 static char sccsid[] = "@(#)indent.c	5.17 (Berkeley) 6/7/93";
     42 #endif
     43 
     44 #include <sys/cdefs.h>
     45 #if defined(__NetBSD__)
     46 __RCSID("$NetBSD: indent.c,v 1.95 2021/09/30 21:38:43 rillig Exp $");
     47 #elif defined(__FreeBSD__)
     48 __FBSDID("$FreeBSD: head/usr.bin/indent/indent.c 340138 2018-11-04 19:24:49Z oshogbo $");
     49 #endif
     50 
     51 #include <sys/param.h>
     52 #if HAVE_CAPSICUM
     53 #include <sys/capsicum.h>
     54 #include <capsicum_helpers.h>
     55 #endif
     56 #include <ctype.h>
     57 #include <err.h>
     58 #include <errno.h>
     59 #include <fcntl.h>
     60 #include <stdio.h>
     61 #include <stdlib.h>
     62 #include <string.h>
     63 #include <unistd.h>
     64 
     65 #include "indent.h"
     66 
     67 struct options opt = {
     68     .btype_2 = true,
     69     .comment_delimiter_on_blankline = true,
     70     .cuddle_else = true,
     71     .comment_column = 33,
     72     .decl_indent = 16,
     73     .else_if = true,
     74     .function_brace_split = true,
     75     .format_col1_comments = true,
     76     .format_block_comments = true,
     77     .indent_parameters = true,
     78     .indent_size = 8,
     79     .local_decl_indent = -1,
     80     .lineup_to_parens = true,
     81     .procnames_start_line = true,
     82     .star_comment_cont = true,
     83     .tabsize = 8,
     84     .max_line_length = 78,
     85     .use_tabs = true,
     86 };
     87 
     88 struct parser_state ps;
     89 
     90 struct buffer lab;
     91 struct buffer code;
     92 struct buffer com;
     93 struct buffer token;
     94 
     95 char *in_buffer;
     96 char *in_buffer_limit;
     97 char *buf_ptr;
     98 char *buf_end;
     99 
    100 char sc_buf[sc_size];
    101 char *save_com;
    102 static char *sc_end;		/* pointer into save_com buffer */
    103 
    104 char *bp_save;
    105 char *be_save;
    106 
    107 bool found_err;
    108 int n_real_blanklines;
    109 bool prefix_blankline_requested;
    110 bool postfix_blankline_requested;
    111 bool break_comma;
    112 float case_ind;
    113 bool had_eof;
    114 int line_no;
    115 bool inhibit_formatting;
    116 
    117 static int ifdef_level;
    118 static struct parser_state state_stack[5];
    119 
    120 FILE *input;
    121 FILE *output;
    122 
    123 static void bakcopy(void);
    124 static void indent_declaration(int, bool);
    125 
    126 static const char *in_name = "Standard Input";
    127 static const char *out_name = "Standard Output";
    128 static const char *backup_suffix = ".BAK";
    129 static char bakfile[MAXPATHLEN] = "";
    130 
    131 static void
    132 check_size_code(size_t desired_size)
    133 {
    134     if (code.e + desired_size >= code.l)
    135 	buf_expand(&code, desired_size);
    136 }
    137 
    138 static void
    139 check_size_label(size_t desired_size)
    140 {
    141     if (lab.e + desired_size >= lab.l)
    142 	buf_expand(&lab, desired_size);
    143 }
    144 
    145 #if HAVE_CAPSICUM
    146 static void
    147 init_capsicum(void)
    148 {
    149     cap_rights_t rights;
    150 
    151     /* Restrict input/output descriptors and enter Capsicum sandbox. */
    152     cap_rights_init(&rights, CAP_FSTAT, CAP_WRITE);
    153     if (caph_rights_limit(fileno(output), &rights) < 0)
    154 	err(EXIT_FAILURE, "unable to limit rights for %s", out_name);
    155     cap_rights_init(&rights, CAP_FSTAT, CAP_READ);
    156     if (caph_rights_limit(fileno(input), &rights) < 0)
    157 	err(EXIT_FAILURE, "unable to limit rights for %s", in_name);
    158     if (caph_enter() < 0)
    159 	err(EXIT_FAILURE, "unable to enter capability mode");
    160 }
    161 #endif
    162 
    163 static void
    164 search_brace_newline(bool *inout_force_nl)
    165 {
    166     if (sc_end == NULL) {
    167 	save_com = sc_buf;
    168 	save_com[0] = save_com[1] = ' ';
    169 	sc_end = &save_com[2];
    170     }
    171     *sc_end++ = '\n';
    172 
    173     /*
    174      * We may have inherited a force_nl == true from the previous token (like
    175      * a semicolon). But once we know that a newline has been scanned in this
    176      * loop, force_nl should be false.
    177      *
    178      * However, the force_nl == true must be preserved if newline is never
    179      * scanned in this loop, so this assignment cannot be done earlier.
    180      */
    181     *inout_force_nl = false;
    182 }
    183 
    184 static void
    185 search_brace_comment(bool *inout_comment_buffered)
    186 {
    187     if (sc_end == NULL) {
    188 	/*
    189 	 * Copy everything from the start of the line, because
    190 	 * process_comment() will use that to calculate original indentation
    191 	 * of a boxed comment.
    192 	 */
    193 	memcpy(sc_buf, in_buffer, (size_t)(buf_ptr - in_buffer) - 4);
    194 	save_com = sc_buf + (buf_ptr - in_buffer - 4);
    195 	save_com[0] = save_com[1] = ' ';
    196 	sc_end = &save_com[2];
    197     }
    198     *inout_comment_buffered = true;
    199     *sc_end++ = '/';		/* copy in start of comment */
    200     *sc_end++ = '*';
    201     for (;;) {			/* loop until the end of the comment */
    202 	*sc_end = *buf_ptr++;
    203 	if (buf_ptr >= buf_end)
    204 	    fill_buffer();
    205 	if (*sc_end++ == '*' && *buf_ptr == '/')
    206 	    break;		/* we are at end of comment */
    207 	if (sc_end >= &save_com[sc_size]) {	/* check for temp buffer
    208 						 * overflow */
    209 	    diag(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever");
    210 	    fflush(output);
    211 	    exit(1);
    212 	}
    213     }
    214     *sc_end++ = '/';		/* add ending slash */
    215     if (++buf_ptr >= buf_end)	/* get past / in buffer */
    216 	fill_buffer();
    217 }
    218 
    219 static bool
    220 search_brace_lbrace(void)
    221 {
    222     /*
    223      * Put KNF-style lbraces before the buffered up tokens and jump out of
    224      * this loop in order to avoid copying the token again under the default
    225      * case of the switch below.
    226      */
    227     if (sc_end != NULL && opt.btype_2) {
    228 	save_com[0] = '{';
    229 	/*
    230 	 * Originally the lbrace may have been alone on its own line, but it
    231 	 * will be moved into "the else's line", so if there was a newline
    232 	 * resulting from the "{" before, it must be scanned now and ignored.
    233 	 */
    234 	while (isspace((unsigned char)*buf_ptr)) {
    235 	    if (++buf_ptr >= buf_end)
    236 		fill_buffer();
    237 	    if (*buf_ptr == '\n')
    238 		break;
    239 	}
    240 	return true;
    241     }
    242     return false;
    243 }
    244 
    245 static bool
    246 search_brace_other(token_type ttype, bool *inout_force_nl,
    247     bool comment_buffered, bool last_else)
    248 {
    249     bool remove_newlines;
    250 
    251     remove_newlines =
    252 	    /* "} else" */
    253 	    (ttype == keyword_do_else && *token.s == 'e' &&
    254 	     code.e != code.s && code.e[-1] == '}')
    255 	    /* "else if" */
    256 	    || (ttype == keyword_for_if_while &&
    257 		*token.s == 'i' && last_else && opt.else_if);
    258     if (remove_newlines)
    259 	*inout_force_nl = false;
    260     if (sc_end == NULL) {	/* ignore buffering if comment wasn't saved
    261 				 * up */
    262 	ps.search_brace = false;
    263 	return false;
    264     }
    265     while (sc_end > save_com && isblank((unsigned char)sc_end[-1])) {
    266 	sc_end--;
    267     }
    268     if (opt.swallow_optional_blanklines ||
    269 	(!comment_buffered && remove_newlines)) {
    270 	*inout_force_nl = !remove_newlines;
    271 	while (sc_end > save_com && sc_end[-1] == '\n') {
    272 	    sc_end--;
    273 	}
    274     }
    275     if (*inout_force_nl) {	/* if we should insert a nl here, put it into
    276 				 * the buffer */
    277 	*inout_force_nl = false;
    278 	--line_no;		/* this will be re-increased when the newline
    279 				 * is read from the buffer */
    280 	*sc_end++ = '\n';
    281 	*sc_end++ = ' ';
    282 	if (opt.verbose)	/* print error msg if the line was not already
    283 				 * broken */
    284 	    diag(0, "Line broken");
    285     }
    286     for (const char *t_ptr = token.s; *t_ptr != '\0'; ++t_ptr)
    287 	*sc_end++ = *t_ptr;
    288     return true;
    289 }
    290 
    291 static void
    292 switch_buffer(void)
    293 {
    294     ps.search_brace = false;	/* stop looking for start of stmt */
    295     bp_save = buf_ptr;		/* save current input buffer */
    296     be_save = buf_end;
    297     buf_ptr = save_com;		/* fix so that subsequent calls to lexi will
    298 				 * take tokens out of save_com */
    299     *sc_end++ = ' ';		/* add trailing blank, just in case */
    300     buf_end = sc_end;
    301     sc_end = NULL;
    302     debug_println("switched buf_ptr to save_com");
    303 }
    304 
    305 static void
    306 search_brace_lookahead(token_type *inout_ttype)
    307 {
    308     /*
    309      * We must make this check, just in case there was an unexpected EOF.
    310      */
    311     if (*inout_ttype != end_of_file) {
    312 	/*
    313 	 * The only intended purpose of calling lexi() below is to categorize
    314 	 * the next token in order to decide whether to continue buffering
    315 	 * forthcoming tokens. Once the buffering is over, lexi() will be
    316 	 * called again elsewhere on all of the tokens - this time for normal
    317 	 * processing.
    318 	 *
    319 	 * Calling it for this purpose is a bug, because lexi() also changes
    320 	 * the parser state and discards leading whitespace, which is needed
    321 	 * mostly for comment-related considerations.
    322 	 *
    323 	 * Work around the former problem by giving lexi() a copy of the
    324 	 * current parser state and discard it if the call turned out to be
    325 	 * just a look ahead.
    326 	 *
    327 	 * Work around the latter problem by copying all whitespace characters
    328 	 * into the buffer so that the later lexi() call will read them.
    329 	 */
    330 	if (sc_end != NULL) {
    331 	    while (*buf_ptr == ' ' || *buf_ptr == '\t') {
    332 		*sc_end++ = *buf_ptr++;
    333 		if (sc_end >= &save_com[sc_size]) {
    334 		    errx(1, "input too long");
    335 		}
    336 	    }
    337 	    if (buf_ptr >= buf_end) {
    338 		fill_buffer();
    339 	    }
    340 	}
    341 
    342 	struct parser_state transient_state;
    343 	transient_state = ps;
    344 	*inout_ttype = lexi(&transient_state);	/* read another token */
    345 	if (*inout_ttype != newline && *inout_ttype != form_feed &&
    346 	    *inout_ttype != comment && !transient_state.search_brace) {
    347 	    ps = transient_state;
    348 	}
    349     }
    350 }
    351 
    352 static void
    353 search_brace(token_type *inout_ttype, bool *inout_force_nl,
    354     bool *inout_comment_buffered, bool *inout_last_else)
    355 {
    356     while (ps.search_brace) {
    357 	switch (*inout_ttype) {
    358 	case newline:
    359 	    search_brace_newline(inout_force_nl);
    360 	    break;
    361 	case form_feed:
    362 	    break;
    363 	case comment:
    364 	    search_brace_comment(inout_comment_buffered);
    365 	    break;
    366 	case lbrace:
    367 	    if (search_brace_lbrace())
    368 		goto sw_buffer;
    369 	    /* FALLTHROUGH */
    370 	default:		/* it is the start of a normal statement */
    371 	    if (!search_brace_other(*inout_ttype, inout_force_nl,
    372 		    *inout_comment_buffered, *inout_last_else))
    373 		return;
    374 	sw_buffer:
    375 	    switch_buffer();
    376 	}
    377 	search_brace_lookahead(inout_ttype);
    378     }
    379 
    380     *inout_last_else = false;
    381 }
    382 
    383 static void
    384 buf_init(struct buffer *buf)
    385 {
    386     buf->buf = xmalloc(bufsize);
    387     buf->buf[0] = ' ';		/* allow accessing buf->e[-1] */
    388     buf->buf[1] = '\0';
    389     buf->s = buf->buf + 1;
    390     buf->e = buf->s;
    391     buf->l = buf->buf + bufsize - 5;	/* safety margin, though unreliable */
    392 }
    393 
    394 void
    395 buf_expand(struct buffer *buf, size_t desired_size)
    396 {
    397     size_t nsize = buf->l - buf->s + 400 + desired_size;
    398     size_t code_len = buf->e - buf->s;
    399     buf->buf = xrealloc(buf->buf, nsize);
    400     buf->e = buf->buf + code_len + 1;
    401     buf->l = buf->buf + nsize - 5;
    402     buf->s = buf->buf + 1;
    403 }
    404 
    405 static void
    406 main_init_globals(void)
    407 {
    408     found_err = false;
    409 
    410     ps.p_stack[0] = stmt;	/* this is the parser's stack */
    411     ps.last_nl = true;		/* this is true if the last thing scanned was
    412 				 * a newline */
    413     ps.last_token = semicolon;
    414     buf_init(&com);
    415     buf_init(&lab);
    416     buf_init(&code);
    417     buf_init(&token);
    418     opt.else_if = true;		/* XXX: redundant? */
    419 
    420     in_buffer = xmalloc(10);
    421     in_buffer_limit = in_buffer + 8;
    422     buf_ptr = buf_end = in_buffer;
    423     line_no = 1;
    424     had_eof = ps.in_decl = ps.decl_on_line = (break_comma = false);
    425     ps.in_or_st = false;
    426     ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
    427 
    428     ps.pcase = false;
    429     sc_end = NULL;
    430     bp_save = NULL;
    431     be_save = NULL;
    432 
    433     output = NULL;
    434 
    435     const char *suffix = getenv("SIMPLE_BACKUP_SUFFIX");
    436     if (suffix != NULL)
    437 	backup_suffix = suffix;
    438 }
    439 
    440 static void
    441 main_parse_command_line(int argc, char **argv)
    442 {
    443     int i;
    444     const char *profile_name = NULL;
    445 
    446     for (i = 1; i < argc; ++i)
    447 	if (strcmp(argv[i], "-npro") == 0)
    448 	    break;
    449 	else if (argv[i][0] == '-' && argv[i][1] == 'P' && argv[i][2] != '\0')
    450 	    profile_name = argv[i];	/* non-empty -P (set profile) */
    451     if (i >= argc)
    452 	set_profile(profile_name);
    453 
    454     for (i = 1; i < argc; ++i) {
    455 	if (argv[i][0] == '-') {
    456 	    set_option(argv[i]);
    457 
    458 	} else if (input == NULL) {
    459 	    in_name = argv[i];
    460 	    input = fopen(in_name, "r");
    461 	    if (input == NULL)
    462 		err(1, "%s", in_name);
    463 
    464 	} else if (output == NULL) {
    465 	    out_name = argv[i];
    466 	    if (strcmp(in_name, out_name) == 0)
    467 		errx(1, "input and output files must be different");
    468 	    output = fopen(out_name, "w");
    469 	    if (output == NULL)
    470 		err(1, "%s", out_name);
    471 
    472 	} else
    473 	    errx(1, "unknown parameter: %s", argv[i]);
    474     }
    475 
    476     if (input == NULL)
    477 	input = stdin;
    478     if (output == NULL) {
    479 	if (input == stdin)
    480 	    output = stdout;
    481 	else {
    482 	    out_name = in_name;
    483 	    bakcopy();
    484 	}
    485     }
    486 
    487     if (opt.comment_column <= 1)
    488 	opt.comment_column = 2;	/* don't put normal comments before column 2 */
    489     if (opt.block_comment_max_line_length <= 0)
    490 	opt.block_comment_max_line_length = opt.max_line_length;
    491     if (opt.local_decl_indent < 0) /* if not specified by user, set this */
    492 	opt.local_decl_indent = opt.decl_indent;
    493     if (opt.decl_comment_column <= 0)	/* if not specified by user, set this */
    494 	opt.decl_comment_column = opt.ljust_decl
    495 	    ? (opt.comment_column <= 10 ? 2 : opt.comment_column - 8)
    496 	    : opt.comment_column;
    497     if (opt.continuation_indent == 0)
    498 	opt.continuation_indent = opt.indent_size;
    499 }
    500 
    501 static void
    502 main_prepare_parsing(void)
    503 {
    504     fill_buffer();		/* get first batch of stuff into input buffer */
    505 
    506     parse(semicolon);
    507 
    508     char *p = buf_ptr;
    509     int col = 1;
    510 
    511     for (;;) {
    512 	if (*p == ' ')
    513 	    col++;
    514 	else if (*p == '\t')
    515 	    col = opt.tabsize * (1 + (col - 1) / opt.tabsize) + 1;
    516 	else
    517 	    break;
    518 	p++;
    519     }
    520     if (col > opt.indent_size)
    521 	ps.ind_level = ps.ind_level_follow = col / opt.indent_size;
    522 }
    523 
    524 static void __attribute__((__noreturn__))
    525 process_end_of_file(void)
    526 {
    527     if (lab.s != lab.e || code.s != code.e || com.s != com.e)
    528 	dump_line();
    529 
    530     if (ps.tos > 1)		/* check for balanced braces */
    531 	diag(1, "Stuff missing from end of file");
    532 
    533     if (opt.verbose) {
    534 	printf("There were %d output lines and %d comments\n",
    535 	    ps.stats.lines, ps.stats.comments);
    536 	printf("(Lines with comments)/(Lines with code): %6.3f\n",
    537 	    (1.0 * ps.stats.comment_lines) / ps.stats.code_lines);
    538     }
    539 
    540     fflush(output);
    541     exit(found_err ? EXIT_FAILURE : EXIT_SUCCESS);
    542 }
    543 
    544 static void
    545 process_comment_in_code(token_type ttype, bool *inout_force_nl)
    546 {
    547     if (*inout_force_nl &&
    548 	ttype != semicolon &&
    549 	(ttype != lbrace || !opt.btype_2)) {
    550 
    551 	/* we should force a broken line here */
    552 	if (opt.verbose)
    553 	    diag(0, "Line broken");
    554 	dump_line();
    555 	ps.want_blank = false;	/* dont insert blank at line start */
    556 	*inout_force_nl = false;
    557     }
    558 
    559     ps.in_stmt = true;		/* turn on flag which causes an extra level of
    560 				 * indentation. this is turned off by a ';' or
    561 				 * '}' */
    562     if (com.s != com.e) {	/* the turkey has embedded a comment in a
    563 				 * line. fix it */
    564 	size_t len = com.e - com.s;
    565 
    566 	check_size_code(len + 3);
    567 	*code.e++ = ' ';
    568 	memcpy(code.e, com.s, len);
    569 	code.e += len;
    570 	*code.e++ = ' ';
    571 	*code.e = '\0';
    572 	ps.want_blank = false;
    573 	com.e = com.s;
    574     }
    575 }
    576 
    577 static void
    578 process_form_feed(void)
    579 {
    580     ps.use_ff = true;		/* a form feed is treated much like a newline */
    581     dump_line();
    582     ps.want_blank = false;
    583 }
    584 
    585 static void
    586 process_newline(void)
    587 {
    588     if (ps.last_token != comma || ps.p_l_follow > 0 || opt.break_after_comma
    589 	|| ps.block_init || !break_comma || com.s != com.e) {
    590 	dump_line();
    591 	ps.want_blank = false;
    592     }
    593     ++line_no;			/* keep track of input line number */
    594 }
    595 
    596 static bool
    597 want_blank_before_lparen(void)
    598 {
    599     if (!ps.want_blank)
    600 	return false;
    601     if (ps.last_token != ident && ps.last_token != funcname)
    602 	return true;
    603     if (opt.proc_calls_space)
    604 	return true;
    605     if (ps.keyword == kw_sizeof)
    606 	return opt.blank_after_sizeof;
    607     return ps.keyword != kw_0 && ps.keyword != kw_offsetof;
    608 }
    609 
    610 static void
    611 process_lparen_or_lbracket(int dec_ind, bool tabs_to_var, bool sp_sw)
    612 {
    613     /* count parens to make Healy happy */
    614     if (++ps.p_l_follow == nitems(ps.paren_indents)) {
    615 	diag(0, "Reached internal limit of %zu unclosed parens",
    616 	    nitems(ps.paren_indents));
    617 	ps.p_l_follow--;
    618     }
    619     if (token.s[0] == '(' && ps.in_decl
    620 	&& !ps.block_init && !ps.dumped_decl_indent &&
    621 	ps.procname[0] == '\0' && ps.paren_level == 0) {
    622 	/* function pointer declarations */
    623 	indent_declaration(dec_ind, tabs_to_var);
    624 	ps.dumped_decl_indent = true;
    625     } else if (want_blank_before_lparen())
    626 	*code.e++ = ' ';
    627     ps.want_blank = false;
    628     *code.e++ = token.s[0];
    629 
    630     ps.paren_indents[ps.p_l_follow - 1] =
    631 	indentation_after_range(0, code.s, code.e);
    632     debug_println("paren_indent[%d] is now %d",
    633 	ps.p_l_follow - 1, ps.paren_indents[ps.p_l_follow - 1]);
    634 
    635     if (sp_sw && ps.p_l_follow == 1 && opt.extra_expression_indent
    636 	    && ps.paren_indents[0] < 2 * opt.indent_size) {
    637 	ps.paren_indents[0] = 2 * opt.indent_size;
    638 	debug_println("paren_indent[0] is now %d", ps.paren_indents[0]);
    639     }
    640     if (ps.in_or_st && *token.s == '(' && ps.tos <= 2) {
    641 	/*
    642 	 * this is a kluge to make sure that declarations will be aligned
    643 	 * right if proc decl has an explicit type on it, i.e. "int a(x) {..."
    644 	 */
    645 	parse(semicolon);	/* I said this was a kluge... */
    646 	ps.in_or_st = false;	/* turn off flag for structure decl or
    647 				 * initialization */
    648     }
    649     /* parenthesized type following sizeof or offsetof is not a cast */
    650     if (ps.keyword == kw_offsetof || ps.keyword == kw_sizeof)
    651 	ps.not_cast_mask |= 1 << ps.p_l_follow;
    652 }
    653 
    654 static void
    655 process_rparen_or_rbracket(bool *inout_sp_sw, bool *inout_force_nl,
    656     token_type hd_type)
    657 {
    658     if ((ps.cast_mask & (1 << ps.p_l_follow) & ~ps.not_cast_mask) != 0) {
    659 	ps.last_u_d = true;
    660 	ps.cast_mask &= (1 << ps.p_l_follow) - 1;
    661 	ps.want_blank = opt.space_after_cast;
    662     } else
    663 	ps.want_blank = true;
    664     ps.not_cast_mask &= (1 << ps.p_l_follow) - 1;
    665 
    666     if (--ps.p_l_follow < 0) {
    667 	ps.p_l_follow = 0;
    668 	diag(0, "Extra %c", *token.s);
    669     }
    670 
    671     if (code.e == code.s)	/* if the paren starts the line */
    672 	ps.paren_level = ps.p_l_follow;	/* then indent it */
    673 
    674     *code.e++ = token.s[0];
    675 
    676     if (*inout_sp_sw && (ps.p_l_follow == 0)) {	/* check for end of if (...),
    677 						 * or some such */
    678 	*inout_sp_sw = false;
    679 	*inout_force_nl = true;	/* must force newline after if */
    680 	ps.last_u_d = true;	/* inform lexi that a following operator is
    681 				 * unary */
    682 	ps.in_stmt = false;	/* dont use stmt continuation indentation */
    683 
    684 	parse(hd_type);		/* let parser worry about if, or whatever */
    685     }
    686     ps.search_brace = opt.btype_2;	/* this should ensure that constructs
    687 					 * such as main(){...} and int[]{...}
    688 					 * have their braces put in the right
    689 					 * place */
    690 }
    691 
    692 static void
    693 process_unary_op(int dec_ind, bool tabs_to_var)
    694 {
    695     if (!ps.dumped_decl_indent && ps.in_decl && !ps.block_init &&
    696 	ps.procname[0] == '\0' && ps.paren_level == 0) {
    697 	/* pointer declarations */
    698 	indent_declaration(dec_ind - (int)strlen(token.s), tabs_to_var);
    699 	ps.dumped_decl_indent = true;
    700     } else if (ps.want_blank)
    701 	*code.e++ = ' ';
    702 
    703     {
    704 	size_t len = token.e - token.s;
    705 
    706 	check_size_code(len);
    707 	memcpy(code.e, token.s, len);
    708 	code.e += len;
    709     }
    710     ps.want_blank = false;
    711 }
    712 
    713 static void
    714 process_binary_op(void)
    715 {
    716     size_t len = token.e - token.s;
    717 
    718     check_size_code(len + 1);
    719     if (ps.want_blank)
    720 	*code.e++ = ' ';
    721     memcpy(code.e, token.s, len);
    722     code.e += len;
    723 
    724     ps.want_blank = true;
    725 }
    726 
    727 static void
    728 process_postfix_op(void)
    729 {
    730     *code.e++ = token.s[0];
    731     *code.e++ = token.s[1];
    732     ps.want_blank = true;
    733 }
    734 
    735 static void
    736 process_question(int *inout_squest)
    737 {
    738     (*inout_squest)++;		/* this will be used when a later colon
    739 				 * appears so we can distinguish the
    740 				 * <c>?<n>:<n> construct */
    741     if (ps.want_blank)
    742 	*code.e++ = ' ';
    743     *code.e++ = '?';
    744     ps.want_blank = true;
    745 }
    746 
    747 static void
    748 process_colon(int *inout_squest, bool *inout_force_nl, bool *inout_scase)
    749 {
    750     if (*inout_squest > 0) {	/* it is part of the <c>?<n>: <n> construct */
    751 	--*inout_squest;
    752 	if (ps.want_blank)
    753 	    *code.e++ = ' ';
    754 	*code.e++ = ':';
    755 	ps.want_blank = true;
    756 	return;
    757     }
    758     if (ps.in_or_st) {
    759 	*code.e++ = ':';
    760 	ps.want_blank = false;
    761 	return;
    762     }
    763     ps.in_stmt = false;		/* seeing a label does not imply we are in a
    764 				 * stmt */
    765     /*
    766      * turn everything so far into a label
    767      */
    768     {
    769 	size_t len = code.e - code.s;
    770 
    771 	check_size_label(len + 3);
    772 	memcpy(lab.e, code.s, len);
    773 	lab.e += len;
    774 	*lab.e++ = ':';
    775 	*lab.e = '\0';
    776 	code.e = code.s;
    777     }
    778     *inout_force_nl = ps.pcase = *inout_scase;	/* ps.pcase will be used by
    779 						 * dump_line to decide how to
    780 						 * indent the label. force_nl
    781 						 * will force a case n: to be
    782 						 * on a line by itself */
    783     *inout_scase = false;
    784     ps.want_blank = false;
    785 }
    786 
    787 static void
    788 process_semicolon(bool *inout_scase, int *inout_squest, int dec_ind,
    789     bool tabs_to_var, bool *inout_sp_sw,
    790     token_type hd_type,
    791     bool *inout_force_nl)
    792 {
    793     if (ps.decl_nest == 0)
    794 	ps.in_or_st = false;	/* we are not in an initialization or
    795 				 * structure declaration */
    796     *inout_scase = false;	/* these will only need resetting in an error */
    797     *inout_squest = 0;
    798     if (ps.last_token == rparen)
    799 	ps.in_parameter_declaration = false;
    800     ps.cast_mask = 0;
    801     ps.not_cast_mask = 0;
    802     ps.block_init = false;
    803     ps.block_init_level = 0;
    804     ps.just_saw_decl--;
    805 
    806     if (ps.in_decl && code.s == code.e && !ps.block_init &&
    807 	!ps.dumped_decl_indent && ps.paren_level == 0) {
    808 	/* indent stray semicolons in declarations */
    809 	indent_declaration(dec_ind - 1, tabs_to_var);
    810 	ps.dumped_decl_indent = true;
    811     }
    812 
    813     ps.in_decl = (ps.decl_nest > 0);	/* if we were in a first level
    814 					 * structure declaration, we arent any
    815 					 * more */
    816 
    817     if ((!*inout_sp_sw || hd_type != for_exprs) && ps.p_l_follow > 0) {
    818 
    819 	/*
    820 	 * This should be true iff there were unbalanced parens in the stmt.
    821 	 * It is a bit complicated, because the semicolon might be in a for
    822 	 * stmt
    823 	 */
    824 	diag(1, "Unbalanced parens");
    825 	ps.p_l_follow = 0;
    826 	if (*inout_sp_sw) {	/* this is a check for an if, while, etc. with
    827 				 * unbalanced parens */
    828 	    *inout_sp_sw = false;
    829 	    parse(hd_type);	/* dont lose the if, or whatever */
    830 	}
    831     }
    832     *code.e++ = ';';
    833     ps.want_blank = true;
    834     ps.in_stmt = (ps.p_l_follow > 0);	/* we are no longer in the middle of a
    835 					 * stmt */
    836 
    837     if (!*inout_sp_sw) {	/* if not if for (;;) */
    838 	parse(semicolon);	/* let parser know about end of stmt */
    839 	*inout_force_nl = true;	/* force newline after an end of stmt */
    840     }
    841 }
    842 
    843 static void
    844 process_lbrace(bool *inout_force_nl, bool *inout_sp_sw, token_type hd_type,
    845     int *di_stack, int di_stack_cap, int *inout_dec_ind)
    846 {
    847     ps.in_stmt = false;		/* dont indent the {} */
    848     if (!ps.block_init)
    849 	*inout_force_nl = true;	/* force other stuff on same line as '{' onto
    850 				 * new line */
    851     else if (ps.block_init_level <= 0)
    852 	ps.block_init_level = 1;
    853     else
    854 	ps.block_init_level++;
    855 
    856     if (code.s != code.e && !ps.block_init) {
    857 	if (!opt.btype_2) {
    858 	    dump_line();
    859 	    ps.want_blank = false;
    860 	} else if (ps.in_parameter_declaration && !ps.in_or_st) {
    861 	    ps.ind_level_follow = 0;
    862 	    if (opt.function_brace_split) {	/* dump the line prior to the
    863 						 * brace ... */
    864 		dump_line();
    865 		ps.want_blank = false;
    866 	    } else		/* add a space between the decl and brace */
    867 		ps.want_blank = true;
    868 	}
    869     }
    870     if (ps.in_parameter_declaration)
    871 	prefix_blankline_requested = false;
    872 
    873     if (ps.p_l_follow > 0) {	/* check for preceding unbalanced parens */
    874 	diag(1, "Unbalanced parens");
    875 	ps.p_l_follow = 0;
    876 	if (*inout_sp_sw) {	/* check for unclosed if, for, etc. */
    877 	    *inout_sp_sw = false;
    878 	    parse(hd_type);
    879 	    ps.ind_level = ps.ind_level_follow;
    880 	}
    881     }
    882     if (code.s == code.e)
    883 	ps.ind_stmt = false;	/* dont put extra indentation on line
    884 				 * with '{' */
    885     if (ps.in_decl && ps.in_or_st) {	/* this is either a structure
    886 					 * declaration or an init */
    887 	di_stack[ps.decl_nest] = *inout_dec_ind;
    888 	if (++ps.decl_nest == di_stack_cap) {
    889 	    diag(0, "Reached internal limit of %d struct levels",
    890 		di_stack_cap);
    891 	    ps.decl_nest--;
    892 	}
    893 	/* ?		dec_ind = 0; */
    894     } else {
    895 	ps.decl_on_line = false;	/* we can't be in the middle of a
    896 					 * declaration, so don't do special
    897 					 * indentation of comments */
    898 	if (opt.blanklines_after_declarations_at_proctop
    899 	    && ps.in_parameter_declaration)
    900 	    postfix_blankline_requested = true;
    901 	ps.in_parameter_declaration = false;
    902 	ps.in_decl = false;
    903     }
    904     *inout_dec_ind = 0;
    905     parse(lbrace);		/* let parser know about this */
    906     if (ps.want_blank)		/* put a blank before '{' if '{' is not at
    907 				 * start of line */
    908 	*code.e++ = ' ';
    909     ps.want_blank = false;
    910     *code.e++ = '{';
    911     ps.just_saw_decl = 0;
    912 }
    913 
    914 static void
    915 process_rbrace(bool *inout_sp_sw, int *inout_dec_ind, const int *di_stack)
    916 {
    917     if (ps.p_stack[ps.tos] == decl && !ps.block_init)	/* semicolons can be
    918 							 * omitted in
    919 							 * declarations */
    920 	parse(semicolon);
    921     if (ps.p_l_follow != 0) {	/* check for unclosed if, for, else. */
    922 	diag(1, "Unbalanced parens");
    923 	ps.p_l_follow = 0;
    924 	*inout_sp_sw = false;
    925     }
    926     ps.just_saw_decl = 0;
    927     ps.block_init_level--;
    928     if (code.s != code.e && !ps.block_init) {	/* '}' must be first on line */
    929 	if (opt.verbose)
    930 	    diag(0, "Line broken");
    931 	dump_line();
    932     }
    933     *code.e++ = '}';
    934     ps.want_blank = true;
    935     ps.in_stmt = ps.ind_stmt = false;
    936     if (ps.decl_nest > 0) { /* we are in multi-level structure declaration */
    937 	*inout_dec_ind = di_stack[--ps.decl_nest];
    938 	if (ps.decl_nest == 0 && !ps.in_parameter_declaration)
    939 	    ps.just_saw_decl = 2;
    940 	ps.in_decl = true;
    941     }
    942     prefix_blankline_requested = false;
    943     parse(rbrace);		/* let parser know about this */
    944     ps.search_brace = opt.cuddle_else
    945 	&& ps.p_stack[ps.tos] == if_expr_stmt
    946 	&& ps.il[ps.tos] >= ps.ind_level;
    947     if (ps.tos <= 1 && opt.blanklines_after_procs && ps.decl_nest <= 0)
    948 	postfix_blankline_requested = true;
    949 }
    950 
    951 static void
    952 process_keyword_do_else(bool *inout_force_nl, bool *inout_last_else)
    953 {
    954     ps.in_stmt = false;
    955     if (*token.s == 'e') {
    956 	if (code.e != code.s && (!opt.cuddle_else || code.e[-1] != '}')) {
    957 	    if (opt.verbose)
    958 		diag(0, "Line broken");
    959 	    dump_line();	/* make sure this starts a line */
    960 	    ps.want_blank = false;
    961 	}
    962 	*inout_force_nl = true;/* also, following stuff must go onto new line */
    963 	*inout_last_else = true;
    964 	parse(keyword_else);
    965     } else {
    966 	if (code.e != code.s) {	/* make sure this starts a line */
    967 	    if (opt.verbose)
    968 		diag(0, "Line broken");
    969 	    dump_line();
    970 	    ps.want_blank = false;
    971 	}
    972 	*inout_force_nl = true;/* also, following stuff must go onto new line */
    973 	*inout_last_else = false;
    974 	parse(keyword_do);
    975     }
    976 }
    977 
    978 static void
    979 process_decl(int *out_dec_ind, bool *out_tabs_to_var)
    980 {
    981     parse(decl);		/* let parser worry about indentation */
    982     if (ps.last_token == rparen && ps.tos <= 1) {
    983 	if (code.s != code.e) {
    984 	    dump_line();
    985 	    ps.want_blank = false;
    986 	}
    987     }
    988     if (ps.in_parameter_declaration && opt.indent_parameters &&
    989 	ps.decl_nest == 0) {
    990 	ps.ind_level = ps.ind_level_follow = 1;
    991 	ps.ind_stmt = false;
    992     }
    993     ps.in_or_st = true;		/* this might be a structure or initialization
    994 				 * declaration */
    995     ps.in_decl = ps.decl_on_line = ps.last_token != type_def;
    996     if ( /* !ps.in_or_st && */ ps.decl_nest <= 0)
    997 	ps.just_saw_decl = 2;
    998     prefix_blankline_requested = false;
    999 
   1000     int len = (int)strlen(token.s) + 1;
   1001     int ind = ps.ind_level == 0 || ps.decl_nest > 0
   1002 	    ? opt.decl_indent		/* global variable or local member */
   1003 	    : opt.local_decl_indent;	/* local variable */
   1004     *out_dec_ind = ind > 0 ? ind : len;
   1005     *out_tabs_to_var = opt.use_tabs ? ind > 0 : false;
   1006 }
   1007 
   1008 static void
   1009 process_ident(token_type ttype, int dec_ind, bool tabs_to_var,
   1010     bool *inout_sp_sw, bool *inout_force_nl, token_type hd_type)
   1011 {
   1012     if (ps.in_decl) {
   1013 	if (ttype == funcname) {
   1014 	    ps.in_decl = false;
   1015 	    if (opt.procnames_start_line && code.s != code.e) {
   1016 		*code.e = '\0';
   1017 		dump_line();
   1018 	    } else if (ps.want_blank) {
   1019 		*code.e++ = ' ';
   1020 	    }
   1021 	    ps.want_blank = false;
   1022 	} else if (!ps.block_init && !ps.dumped_decl_indent &&
   1023 	    ps.paren_level == 0) {	/* if we are in a declaration, we must
   1024 					 * indent identifier */
   1025 	    indent_declaration(dec_ind, tabs_to_var);
   1026 	    ps.dumped_decl_indent = true;
   1027 	    ps.want_blank = false;
   1028 	}
   1029     } else if (*inout_sp_sw && ps.p_l_follow == 0) {
   1030 	*inout_sp_sw = false;
   1031 	*inout_force_nl = true;
   1032 	ps.last_u_d = true;
   1033 	ps.in_stmt = false;
   1034 	parse(hd_type);
   1035     }
   1036 }
   1037 
   1038 static void
   1039 copy_id(void)
   1040 {
   1041     size_t len = token.e - token.s;
   1042 
   1043     check_size_code(len + 1);
   1044     if (ps.want_blank)
   1045 	*code.e++ = ' ';
   1046     memcpy(code.e, token.s, len);
   1047     code.e += len;
   1048 }
   1049 
   1050 static void
   1051 process_string_prefix(void)
   1052 {
   1053     size_t len = token.e - token.s;
   1054 
   1055     check_size_code(len + 1);
   1056     if (ps.want_blank)
   1057 	*code.e++ = ' ';
   1058     memcpy(code.e, token.s, len);
   1059     code.e += len;
   1060 
   1061     ps.want_blank = false;
   1062 }
   1063 
   1064 static void
   1065 process_period(void)
   1066 {
   1067     if (code.e[-1] == ',')
   1068 	*code.e++ = ' ';
   1069     *code.e++ = '.';		/* move the period into line */
   1070     ps.want_blank = false;	/* dont put a blank after a period */
   1071 }
   1072 
   1073 static void
   1074 process_comma(int dec_ind, bool tabs_to_var, bool *inout_force_nl)
   1075 {
   1076     ps.want_blank = (code.s != code.e);	/* only put blank after comma if comma
   1077 					 * does not start the line */
   1078     if (ps.in_decl && ps.procname[0] == '\0' && !ps.block_init &&
   1079 	!ps.dumped_decl_indent && ps.paren_level == 0) {
   1080 	/* indent leading commas and not the actual identifiers */
   1081 	indent_declaration(dec_ind - 1, tabs_to_var);
   1082 	ps.dumped_decl_indent = true;
   1083     }
   1084     *code.e++ = ',';
   1085     if (ps.p_l_follow == 0) {
   1086 	if (ps.block_init_level <= 0)
   1087 	    ps.block_init = false;
   1088 	if (break_comma && (opt.break_after_comma ||
   1089 			    indentation_after_range(
   1090 				    compute_code_indent(), code.s, code.e)
   1091 			    >= opt.max_line_length - opt.tabsize))
   1092 	    *inout_force_nl = true;
   1093     }
   1094 }
   1095 
   1096 static void
   1097 process_preprocessing(void)
   1098 {
   1099     if (com.s != com.e || lab.s != lab.e || code.s != code.e)
   1100 	dump_line();
   1101     check_size_label(1);
   1102     *lab.e++ = '#';		/* move whole line to 'label' buffer */
   1103 
   1104     {
   1105 	bool in_comment = false;
   1106 	int com_start = 0;
   1107 	char quote = '\0';
   1108 	int com_end = 0;
   1109 
   1110 	while (*buf_ptr == ' ' || *buf_ptr == '\t') {
   1111 	    buf_ptr++;
   1112 	    if (buf_ptr >= buf_end)
   1113 		fill_buffer();
   1114 	}
   1115 	while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
   1116 	    check_size_label(2);
   1117 	    *lab.e = *buf_ptr++;
   1118 	    if (buf_ptr >= buf_end)
   1119 		fill_buffer();
   1120 	    switch (*lab.e++) {
   1121 	    case '\\':
   1122 		if (!in_comment) {
   1123 		    *lab.e++ = *buf_ptr++;
   1124 		    if (buf_ptr >= buf_end)
   1125 			fill_buffer();
   1126 		}
   1127 		break;
   1128 	    case '/':
   1129 		if (*buf_ptr == '*' && !in_comment && quote == '\0') {
   1130 		    in_comment = true;
   1131 		    *lab.e++ = *buf_ptr++;
   1132 		    com_start = (int)(lab.e - lab.s) - 2;
   1133 		}
   1134 		break;
   1135 	    case '"':
   1136 		if (quote == '"')
   1137 		    quote = '\0';
   1138 		else if (quote == '\0')
   1139 		    quote = '"';
   1140 		break;
   1141 	    case '\'':
   1142 		if (quote == '\'')
   1143 		    quote = '\0';
   1144 		else if (quote == '\0')
   1145 		    quote = '\'';
   1146 		break;
   1147 	    case '*':
   1148 		if (*buf_ptr == '/' && in_comment) {
   1149 		    in_comment = false;
   1150 		    *lab.e++ = *buf_ptr++;
   1151 		    com_end = (int)(lab.e - lab.s);
   1152 		}
   1153 		break;
   1154 	    }
   1155 	}
   1156 
   1157 	while (lab.e > lab.s && (lab.e[-1] == ' ' || lab.e[-1] == '\t'))
   1158 	    lab.e--;
   1159 	if (lab.e - lab.s == com_end && bp_save == NULL) {
   1160 	    /* comment on preprocessor line */
   1161 	    if (sc_end == NULL) {	/* if this is the first comment, we
   1162 					 * must set up the buffer */
   1163 		save_com = sc_buf;
   1164 		sc_end = &save_com[0];
   1165 	    } else {
   1166 		*sc_end++ = '\n';	/* add newline between comments */
   1167 		*sc_end++ = ' ';
   1168 		--line_no;
   1169 	    }
   1170 	    if (sc_end - save_com + com_end - com_start > sc_size)
   1171 		errx(1, "input too long");
   1172 	    memmove(sc_end, lab.s + com_start, (size_t)(com_end - com_start));
   1173 	    sc_end += com_end - com_start;
   1174 	    lab.e = lab.s + com_start;
   1175 	    while (lab.e > lab.s && (lab.e[-1] == ' ' || lab.e[-1] == '\t'))
   1176 		lab.e--;
   1177 	    bp_save = buf_ptr;	/* save current input buffer */
   1178 	    be_save = buf_end;
   1179 	    buf_ptr = save_com;	/* fix so that subsequent calls to lexi will
   1180 				 * take tokens out of save_com */
   1181 	    *sc_end++ = ' ';	/* add trailing blank, just in case */
   1182 	    buf_end = sc_end;
   1183 	    sc_end = NULL;
   1184 	    debug_println("switched buf_ptr to save_com");
   1185 	}
   1186 	check_size_label(1);
   1187 	*lab.e = '\0';		/* null terminate line */
   1188 	ps.pcase = false;
   1189     }
   1190 
   1191     if (strncmp(lab.s, "#if", 3) == 0) {	/* also ifdef, ifndef */
   1192 	if ((size_t)ifdef_level < nitems(state_stack))
   1193 	    state_stack[ifdef_level++] = ps;
   1194 	else
   1195 	    diag(1, "#if stack overflow");
   1196     } else if (strncmp(lab.s, "#el", 3) == 0) {	/* else, elif */
   1197 	if (ifdef_level <= 0)
   1198 	    diag(1, lab.s[3] == 'i' ? "Unmatched #elif" : "Unmatched #else");
   1199 	else
   1200 	    ps = state_stack[ifdef_level - 1];
   1201     } else if (strncmp(lab.s, "#endif", 6) == 0) {
   1202 	if (ifdef_level <= 0)
   1203 	    diag(1, "Unmatched #endif");
   1204 	else
   1205 	    ifdef_level--;
   1206     } else {
   1207 	if (strncmp(lab.s + 1, "pragma", 6) != 0 &&
   1208 	    strncmp(lab.s + 1, "error", 5) != 0 &&
   1209 	    strncmp(lab.s + 1, "line", 4) != 0 &&
   1210 	    strncmp(lab.s + 1, "undef", 5) != 0 &&
   1211 	    strncmp(lab.s + 1, "define", 6) != 0 &&
   1212 	    strncmp(lab.s + 1, "include", 7) != 0) {
   1213 	    diag(1, "Unrecognized cpp directive");
   1214 	    return;
   1215 	}
   1216     }
   1217     if (opt.blanklines_around_conditional_compilation) {
   1218 	postfix_blankline_requested = true;
   1219 	n_real_blanklines = 0;
   1220     } else {
   1221 	postfix_blankline_requested = false;
   1222 	prefix_blankline_requested = false;
   1223     }
   1224 
   1225     /*
   1226      * subsequent processing of the newline character will cause the line to
   1227      * be printed
   1228      */
   1229 }
   1230 
   1231 static void __attribute__((__noreturn__))
   1232 main_loop(void)
   1233 {
   1234     token_type ttype;
   1235     bool force_nl;		/* when true, code must be broken */
   1236     bool last_else = false;	/* true iff last keyword was an else */
   1237     int dec_ind;		/* current indentation for declarations */
   1238     int di_stack[20];		/* a stack of structure indentation levels */
   1239     bool tabs_to_var;		/* true if using tabs to indent to var name */
   1240     bool sp_sw;			/* when true, we are in the expression of
   1241 				 * if(...), while(...), etc. */
   1242     token_type hd_type = end_of_file;	/* used to store type of stmt for if
   1243 					 * (...), for (...), etc */
   1244     int squest;			/* when this is positive, we have seen a '?'
   1245 				 * without the matching ':' in a <c>?<s>:<s>
   1246 				 * construct */
   1247     bool scase;			/* set to true when we see a 'case', so we
   1248 				 * know what to do with the following colon */
   1249 
   1250     sp_sw = force_nl = false;
   1251     dec_ind = 0;
   1252     di_stack[ps.decl_nest = 0] = 0;
   1253     scase = false;
   1254     squest = 0;
   1255     tabs_to_var = false;
   1256 
   1257     for (;;) {			/* this is the main loop.  it will go until we
   1258 				 * reach eof */
   1259 	bool comment_buffered = false;
   1260 
   1261 	ttype = lexi(&ps);	/* Read the next token.  The actual characters
   1262 				 * read are stored in "token". */
   1263 
   1264 	/*
   1265 	 * The following code moves newlines and comments following an if (),
   1266 	 * while (), else, etc. up to the start of the following stmt to a
   1267 	 * buffer. This allows proper handling of both kinds of brace
   1268 	 * placement (-br, -bl) and cuddling "else" (-ce).
   1269 	 */
   1270 	search_brace(&ttype, &force_nl, &comment_buffered, &last_else);
   1271 
   1272 	if (ttype == end_of_file) {
   1273 	    process_end_of_file();
   1274 	    /* NOTREACHED */
   1275 	}
   1276 
   1277 	if (
   1278 		ttype != comment &&
   1279 		ttype != newline &&
   1280 		ttype != preprocessing &&
   1281 		ttype != form_feed) {
   1282 	    process_comment_in_code(ttype, &force_nl);
   1283 
   1284 	} else if (ttype != comment)	/* preserve force_nl through a comment */
   1285 	    force_nl = false;	/* cancel forced newline after newline, form
   1286 				 * feed, etc */
   1287 
   1288 
   1289 
   1290 	/*-----------------------------------------------------*\
   1291 	|	   do switch on type of token scanned		|
   1292 	\*-----------------------------------------------------*/
   1293 	check_size_code(3);	/* maximum number of increments of code.e
   1294 				 * before the next check_size_code or
   1295 				 * dump_line() is 2. After that there's the
   1296 				 * final increment for the null character. */
   1297 	switch (ttype) {
   1298 
   1299 	case form_feed:
   1300 	    process_form_feed();
   1301 	    break;
   1302 
   1303 	case newline:
   1304 	    process_newline();
   1305 	    break;
   1306 
   1307 	case lparen:		/* got a '(' or '[' */
   1308 	    process_lparen_or_lbracket(dec_ind, tabs_to_var, sp_sw);
   1309 	    break;
   1310 
   1311 	case rparen:		/* got a ')' or ']' */
   1312 	    process_rparen_or_rbracket(&sp_sw, &force_nl, hd_type);
   1313 	    break;
   1314 
   1315 	case unary_op:		/* this could be any unary operation */
   1316 	    process_unary_op(dec_ind, tabs_to_var);
   1317 	    break;
   1318 
   1319 	case binary_op:		/* any binary operation */
   1320 	    process_binary_op();
   1321 	    break;
   1322 
   1323 	case postfix_op:	/* got a trailing ++ or -- */
   1324 	    process_postfix_op();
   1325 	    break;
   1326 
   1327 	case question:		/* got a ? */
   1328 	    process_question(&squest);
   1329 	    break;
   1330 
   1331 	case case_label:	/* got word 'case' or 'default' */
   1332 	    scase = true;	/* so we can process the later colon properly */
   1333 	    goto copy_id;
   1334 
   1335 	case colon:		/* got a ':' */
   1336 	    process_colon(&squest, &force_nl, &scase);
   1337 	    break;
   1338 
   1339 	case semicolon:		/* got a ';' */
   1340 	    process_semicolon(&scase, &squest, dec_ind, tabs_to_var, &sp_sw,
   1341 		hd_type, &force_nl);
   1342 	    break;
   1343 
   1344 	case lbrace:		/* got a '{' */
   1345 	    process_lbrace(&force_nl, &sp_sw, hd_type, di_stack,
   1346 		(int)nitems(di_stack), &dec_ind);
   1347 	    break;
   1348 
   1349 	case rbrace:		/* got a '}' */
   1350 	    process_rbrace(&sp_sw, &dec_ind, di_stack);
   1351 	    break;
   1352 
   1353 	case switch_expr:	/* got keyword "switch" */
   1354 	    sp_sw = true;
   1355 	    hd_type = switch_expr;	/* keep this for when we have seen the
   1356 					 * expression */
   1357 	    goto copy_id;	/* go move the token into buffer */
   1358 
   1359 	case keyword_for_if_while:
   1360 	    sp_sw = true;	/* the interesting stuff is done after the
   1361 				 * expression is scanned */
   1362 	    hd_type = (*token.s == 'i' ? if_expr :
   1363 		(*token.s == 'w' ? while_expr : for_exprs));
   1364 
   1365 	    /* remember the type of header for later use by parser */
   1366 	    goto copy_id;	/* copy the token into line */
   1367 
   1368 	case keyword_do_else:
   1369 	    process_keyword_do_else(&force_nl, &last_else);
   1370 	    goto copy_id;	/* move the token into line */
   1371 
   1372 	case type_def:
   1373 	case storage_class:
   1374 	    prefix_blankline_requested = false;
   1375 	    goto copy_id;
   1376 
   1377 	case keyword_struct_union_enum:
   1378 	    if (ps.p_l_follow > 0)
   1379 		goto copy_id;
   1380 	    /* FALLTHROUGH */
   1381 	case decl:		/* we have a declaration type (int, etc.) */
   1382 	    process_decl(&dec_ind, &tabs_to_var);
   1383 	    goto copy_id;
   1384 
   1385 	case funcname:
   1386 	case ident:		/* got an identifier or constant */
   1387 	    process_ident(ttype, dec_ind, tabs_to_var, &sp_sw, &force_nl,
   1388 		hd_type);
   1389     copy_id:
   1390 	    copy_id();
   1391 	    if (ttype != funcname)
   1392 		ps.want_blank = true;
   1393 	    break;
   1394 
   1395 	case string_prefix:
   1396 	    process_string_prefix();
   1397 	    break;
   1398 
   1399 	case period:
   1400 	    process_period();
   1401 	    break;
   1402 
   1403 	case comma:
   1404 	    process_comma(dec_ind, tabs_to_var, &force_nl);
   1405 	    break;
   1406 
   1407 	case preprocessing:	/* '#' */
   1408 	    process_preprocessing();
   1409 	    break;
   1410 	case comment:		/* the initial '/' '*' or '//' of a comment */
   1411 	    process_comment();
   1412 	    break;
   1413 
   1414 	default:
   1415 	    break;
   1416 	}
   1417 
   1418 	*code.e = '\0';
   1419 	if (ttype != comment &&
   1420 	    ttype != newline &&
   1421 	    ttype != preprocessing)
   1422 	    ps.last_token = ttype;
   1423     }
   1424 }
   1425 
   1426 int
   1427 main(int argc, char **argv)
   1428 {
   1429     main_init_globals();
   1430     main_parse_command_line(argc, argv);
   1431 #if HAVE_CAPSICUM
   1432     init_capsicum();
   1433 #endif
   1434     main_prepare_parsing();
   1435     main_loop();
   1436 }
   1437 
   1438 /*
   1439  * copy input file to backup file if in_name is /blah/blah/blah/file, then
   1440  * backup file will be ".Bfile" then make the backup file the input and
   1441  * original input file the output
   1442  */
   1443 static void
   1444 bakcopy(void)
   1445 {
   1446     ssize_t n;
   1447     int bakchn;
   1448     char buff[8 * 1024];
   1449     const char *p;
   1450 
   1451     /* construct file name .Bfile */
   1452     for (p = in_name; *p != '\0'; p++);	/* skip to end of string */
   1453     while (p > in_name && *p != '/')	/* find last '/' */
   1454 	p--;
   1455     if (*p == '/')
   1456 	p++;
   1457     sprintf(bakfile, "%s%s", p, backup_suffix);
   1458 
   1459     /* copy in_name to backup file */
   1460     bakchn = creat(bakfile, 0600);
   1461     if (bakchn < 0)
   1462 	err(1, "%s", bakfile);
   1463     while ((n = read(fileno(input), buff, sizeof(buff))) > 0)
   1464 	if (write(bakchn, buff, (size_t)n) != n)
   1465 	    err(1, "%s", bakfile);
   1466     if (n < 0)
   1467 	err(1, "%s", in_name);
   1468     close(bakchn);
   1469     fclose(input);
   1470 
   1471     /* re-open backup file as the input file */
   1472     input = fopen(bakfile, "r");
   1473     if (input == NULL)
   1474 	err(1, "%s", bakfile);
   1475     /* now the original input file will be the output */
   1476     output = fopen(in_name, "w");
   1477     if (output == NULL) {
   1478 	unlink(bakfile);
   1479 	err(1, "%s", in_name);
   1480     }
   1481 }
   1482 
   1483 static void
   1484 indent_declaration(int cur_dec_ind, bool tabs_to_var)
   1485 {
   1486     int pos = (int)(code.e - code.s);
   1487     char *startpos = code.e;
   1488 
   1489     /*
   1490      * get the tab math right for indentations that are not multiples of
   1491      * tabsize
   1492      */
   1493     if ((ps.ind_level * opt.indent_size) % opt.tabsize != 0) {
   1494 	pos += (ps.ind_level * opt.indent_size) % opt.tabsize;
   1495 	cur_dec_ind += (ps.ind_level * opt.indent_size) % opt.tabsize;
   1496     }
   1497     if (tabs_to_var) {
   1498 	int tpos;
   1499 
   1500 	check_size_code((size_t)(cur_dec_ind / opt.tabsize));
   1501 	while ((tpos = opt.tabsize * (1 + pos / opt.tabsize)) <= cur_dec_ind) {
   1502 	    *code.e++ = '\t';
   1503 	    pos = tpos;
   1504 	}
   1505     }
   1506     check_size_code((size_t)(cur_dec_ind - pos + 1));
   1507     while (pos < cur_dec_ind) {
   1508 	*code.e++ = ' ';
   1509 	pos++;
   1510     }
   1511     if (code.e == startpos && ps.want_blank) {
   1512 	*code.e++ = ' ';
   1513 	ps.want_blank = false;
   1514     }
   1515 }
   1516 
   1517 #ifdef debug
   1518 void
   1519 debug_printf(const char *fmt, ...)
   1520 {
   1521     FILE *f = output == stdout ? stderr : stdout;
   1522     va_list ap;
   1523 
   1524     va_start(ap, fmt);
   1525     vfprintf(f, fmt, ap);
   1526     va_end(ap);
   1527 }
   1528 
   1529 void
   1530 debug_println(const char *fmt, ...)
   1531 {
   1532     FILE *f = output == stdout ? stderr : stdout;
   1533     va_list ap;
   1534 
   1535     va_start(ap, fmt);
   1536     vfprintf(f, fmt, ap);
   1537     va_end(ap);
   1538     fprintf(f, "\n");
   1539 }
   1540 
   1541 void
   1542 debug_vis_range(const char *prefix, const char *s, const char *e,
   1543     const char *suffix)
   1544 {
   1545     debug_printf("%s", prefix);
   1546     for (const char *p = s; p < e; p++) {
   1547 	if (isprint((unsigned char)*p) && *p != '\\' && *p != '"')
   1548 	    debug_printf("%c", *p);
   1549 	else if (*p == '\n')
   1550 	    debug_printf("\\n");
   1551 	else if (*p == '\t')
   1552 	    debug_printf("\\t");
   1553 	else
   1554 	    debug_printf("\\x%02x", *p);
   1555     }
   1556     debug_printf("%s", suffix);
   1557 }
   1558 #endif
   1559 
   1560 static void *
   1561 nonnull(void *p)
   1562 {
   1563     if (p == NULL)
   1564 	err(EXIT_FAILURE, NULL);
   1565     return p;
   1566 }
   1567 
   1568 void *
   1569 xmalloc(size_t size)
   1570 {
   1571     return nonnull(malloc(size));
   1572 }
   1573 
   1574 void *
   1575 xrealloc(void *p, size_t new_size)
   1576 {
   1577     return nonnull(realloc(p, new_size));
   1578 }
   1579 
   1580 char *
   1581 xstrdup(const char *s)
   1582 {
   1583     return nonnull(strdup(s));
   1584 }
   1585