usr.bin/indent/indent.c

   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  * @(#) Copyright (c) 1985 Sun Microsystems, Inc. @(#) Copyright (c) 1976 Board of Trustees of the University of Illinois. @(#) Copyright (c) 1980, 1993 The Regents of the University of California.  All rights reserved.
  36  * @(#)indent.c 5.17 (Berkeley) 6/7/93
  37  * $FreeBSD: src/usr.bin/indent/indent.c,v 1.5.2.6 2001/12/06 19:28:47 schweikh Exp $
  38  * $DragonFly: src/usr.bin/indent/indent.c,v 1.4 2007/05/13 18:33:58 swildner Exp $
  39  */
  40
  41 #include <sys/param.h>
  42 #include <err.h>
  43 #include <fcntl.h>
  44 #include <unistd.h>
  45 #include <stdio.h>
  46 #include <stdlib.h>
  47 #include <string.h>
  48 #include <ctype.h>
  49 #include "indent_globs.h"
  50 #include "indent_codes.h"
  51 #include "indent.h"
  52
  53 static void bakcopy(void);
  54
  55 char       *in_name = "Standard Input"; /* will always point to name of input
  56                                          * file */
  57 char       *out_name = "Standard Output";       /* will always point to name
  58                                                  * of output file */
  59 char        bakfile[MAXPATHLEN] = "";
  60
  61 int
  62 main(int argc, char **argv)
  63 {
  64
  65     extern int  found_err;      /* flag set in diagN() on error */
  66     int         dec_ind;        /* current indentation for declarations */
  67     int         di_stack[20];   /* a stack of structure indentation levels */
  68     int         flushed_nl;     /* used when buffering up comments to remember
  69                                  * that a newline was passed over */
  70     int         force_nl;       /* when true, code must be broken */
  71     int         hd_type = 0;    /* used to store type of stmt for if (...),
  72                                  * for (...), etc */
  73     int i;                      /* local loop counter */
  74     int         scase;          /* set to true when we see a case, so we will
  75                                  * know what to do with the following colon */
  76     int         sp_sw;          /* when true, we are in the expressin of
  77                                  * if(...), while(...), etc. */
  78     int         squest;         /* when this is positive, we have seen a ?
  79                                  * without the matching : in a <c>?<s>:<s>
  80                                  * construct */
  81     char *t_ptr;        /* used for copying tokens */
  82     int         type_code;      /* the type of token, returned by lexi */
  83
  84     int         last_else = 0;  /* true iff last keyword was an else */
  85
  86
  87     /*-----------------------------------------------*\
  88     |                 INITIALIZATION                  |
  89     \*-----------------------------------------------*/
  90
  91
  92     ps.p_stack[0] = stmt;       /* this is the parser's stack */
  93     ps.last_nl = true;          /* this is true if the last thing scanned was
  94                                  * a newline */
  95     ps.last_token = semicolon;
  96     combuf = (char *) malloc(bufsize);
  97     labbuf = (char *) malloc(bufsize);
  98     codebuf = (char *) malloc(bufsize);
  99     tokenbuf = (char *) malloc(bufsize);
 100     l_com = combuf + bufsize - 5;
 101     l_lab = labbuf + bufsize - 5;
 102     l_code = codebuf + bufsize - 5;
 103     l_token = tokenbuf + bufsize - 5;
 104     combuf[0] = codebuf[0] = labbuf[0] = ' ';   /* set up code, label, and
 105                                                  * comment buffers */
 106     combuf[1] = codebuf[1] = labbuf[1] = '\0';
 107     ps.else_if = 1;             /* Default else-if special processing to on */
 108     s_lab = e_lab = labbuf + 1;
 109     s_code = e_code = codebuf + 1;
 110     s_com = e_com = combuf + 1;
 111     s_token = e_token = tokenbuf + 1;
 112
 113     in_buffer = (char *) malloc(10);
 114     in_buffer_limit = in_buffer + 8;
 115     buf_ptr = buf_end = in_buffer;
 116     line_no = 1;
 117     had_eof = ps.in_decl = ps.decl_on_line = break_comma = false;
 118     sp_sw = force_nl = false;
 119     ps.in_or_st = false;
 120     ps.bl_line = true;
 121     dec_ind = 0;
 122     di_stack[ps.dec_nest = 0] = 0;
 123     ps.want_blank = ps.in_stmt = ps.ind_stmt = false;
 124
 125
 126     scase = ps.pcase = false;
 127     squest = 0;
 128     sc_end = 0;
 129     bp_save = 0;
 130     be_save = 0;
 131
 132     output = 0;
 133
 134
 135
 136     /*--------------------------------------------------*\
 137     |                   COMMAND LINE SCAN                |
 138     \*--------------------------------------------------*/
 139
 140 #ifdef undef
 141     max_col = 78;               /* -l78 */
 142     lineup_to_parens = 1;       /* -lp */
 143     ps.ljust_decl = 0;          /* -ndj */
 144     ps.com_ind = 33;            /* -c33 */
 145     star_comment_cont = 1;      /* -sc */
 146     ps.ind_size = 8;            /* -i8 */
 147     verbose = 0;
 148     ps.decl_indent = 16;        /* -di16 */
 149     ps.indent_parameters = 1;   /* -ip */
 150     ps.decl_com_ind = 0;        /* if this is not set to some positive value
 151                                  * by an arg, we will set this equal to
 152                                  * ps.com_ind */
 153     btype_2 = 1;                /* -br */
 154     cuddle_else = 1;            /* -ce */
 155     ps.unindent_displace = 0;   /* -d0 */
 156     ps.case_indent = 0;         /* -cli0 */
 157     format_block_comments = 1;  /* -fcb */
 158     format_col1_comments = 1;   /* -fc1 */
 159     procnames_start_line = 1;   /* -psl */
 160     proc_calls_space = 0;       /* -npcs */
 161     comment_delimiter_on_blankline = 1; /* -cdb */
 162     ps.leave_comma = 1;         /* -nbc */
 163 #endif
 164
 165     for (i = 1; i < argc; ++i)
 166         if (strcmp(argv[i], "-npro") == 0)
 167             break;
 168     set_defaults();
 169     if (i >= argc)
 170         set_profile();
 171
 172     for (i = 1; i < argc; ++i) {
 173
 174         /*
 175          * look thru args (if any) for changes to defaults
 176          */
 177         if (argv[i][0] != '-') {/* no flag on parameter */
 178             if (input == 0) {   /* we must have the input file */
 179                 in_name = argv[i];      /* remember name of input file */
 180                 input = fopen(in_name, "r");
 181                 if (input == 0)         /* check for open error */
 182                         err(1, "%s", in_name);
 183                 continue;
 184             }
 185             else if (output == 0) {     /* we have the output file */
 186                 out_name = argv[i];     /* remember name of output file */
 187                 if (strcmp(in_name, out_name) == 0) {   /* attempt to overwrite
 188                                                          * the file */
 189                     errx(1, "input and output files must be different");
 190                 }
 191                 output = fopen(out_name, "w");
 192                 if (output == 0)        /* check for create error */
 193                         err(1, "%s", out_name);
 194                 continue;
 195             }
 196             errx(1, "unknown parameter: %s", argv[i]);
 197         }
 198         else
 199             set_option(argv[i]);
 200     }                           /* end of for */
 201     if (input == 0)
 202         input = stdin;
 203     if (output == 0) {
 204         if (troff || input == stdin)
 205             output = stdout;
 206         else {
 207             out_name = in_name;
 208             bakcopy();
 209         }
 210     }
 211     if (ps.com_ind <= 1)
 212         ps.com_ind = 2;         /* dont put normal comments before column 2 */
 213     if (troff) {
 214         if (bodyf.font[0] == 0)
 215             parsefont(&bodyf, "R");
 216         if (scomf.font[0] == 0)
 217             parsefont(&scomf, "I");
 218         if (blkcomf.font[0] == 0)
 219             blkcomf = scomf, blkcomf.size += 2;
 220         if (boxcomf.font[0] == 0)
 221             boxcomf = blkcomf;
 222         if (stringf.font[0] == 0)
 223             parsefont(&stringf, "L");
 224         if (keywordf.font[0] == 0)
 225             parsefont(&keywordf, "B");
 226         writefdef(&bodyf, 'B');
 227         writefdef(&scomf, 'C');
 228         writefdef(&blkcomf, 'L');
 229         writefdef(&boxcomf, 'X');
 230         writefdef(&stringf, 'S');
 231         writefdef(&keywordf, 'K');
 232     }
 233     if (block_comment_max_col <= 0)
 234         block_comment_max_col = max_col;
 235     if (ps.decl_com_ind <= 0)   /* if not specified by user, set this */
 236         ps.decl_com_ind = ps.ljust_decl ? (ps.com_ind <= 10 ? 2 : ps.com_ind - 8) : ps.com_ind;
 237     if (continuation_indent == 0)
 238         continuation_indent = ps.ind_size;
 239     fill_buffer();              /* get first batch of stuff into input buffer */
 240
 241     parse(semicolon);
 242     {
 243         char *p = buf_ptr;
 244         int col = 1;
 245
 246         while (1) {
 247             if (*p == ' ')
 248                 col++;
 249             else if (*p == '\t')
 250                 col = ((col - 1) & ~7) + 9;
 251             else
 252                 break;
 253             p++;
 254         }
 255         if (col > ps.ind_size)
 256             ps.ind_level = ps.i_l_follow = col / ps.ind_size;
 257     }
 258     if (troff) {
 259         char *p = in_name,
 260                    *beg = in_name;
 261
 262         while (*p)
 263             if (*p++ == '/')
 264                 beg = p;
 265         fprintf(output, ".Fn \"%s\"\n", beg);
 266     }
 267     /*
 268      * START OF MAIN LOOP
 269      */
 270
 271     while (1) {                 /* this is the main loop.  it will go until we
 272                                  * reach eof */
 273         int         is_procname;
 274
 275         type_code = lexi();     /* lexi reads one token.  The actual
 276                                  * characters read are stored in "token". lexi
 277                                  * returns a code indicating the type of token */
 278         is_procname = ps.procname[0];
 279
 280         /*
 281          * The following code moves everything following an if (), while (),
 282          * else, etc. up to the start of the following stmt to a buffer. This
 283          * allows proper handling of both kinds of brace placement.
 284          */
 285
 286         flushed_nl = false;
 287         while (ps.search_brace) {       /* if we scanned an if(), while(),
 288                                          * etc., we might need to copy stuff
 289                                          * into a buffer we must loop, copying
 290                                          * stuff into save_com, until we find
 291                                          * the start of the stmt which follows
 292                                          * the if, or whatever */
 293             switch (type_code) {
 294             case newline:
 295                 ++line_no;
 296                 flushed_nl = true;
 297             case form_feed:
 298                 break;          /* form feeds and newlines found here will be
 299                                  * ignored */
 300
 301             case lbrace:        /* this is a brace that starts the compound
 302                                  * stmt */
 303                 if (sc_end == 0) {      /* ignore buffering if a comment wasnt
 304                                          * stored up */
 305                     ps.search_brace = false;
 306                     goto check_type;
 307                 }
 308                 if (btype_2) {
 309                     save_com[0] = '{';  /* we either want to put the brace
 310                                          * right after the if */
 311                     goto sw_buffer;     /* go to common code to get out of
 312                                          * this loop */
 313                 }
 314             case comment:       /* we have a comment, so we must copy it into
 315                                  * the buffer */
 316                 if (!flushed_nl || sc_end != 0) {
 317                     if (sc_end == 0) {  /* if this is the first comment, we
 318                                          * must set up the buffer */
 319                         save_com[0] = save_com[1] = ' ';
 320                         sc_end = &(save_com[2]);
 321                     }
 322                     else {
 323                         *sc_end++ = '\n';       /* add newline between
 324                                                  * comments */
 325                         *sc_end++ = ' ';
 326                         --line_no;
 327                     }
 328                     *sc_end++ = '/';    /* copy in start of comment */
 329                     *sc_end++ = '*';
 330
 331                     for (;;) {  /* loop until we get to the end of the comment */
 332                         *sc_end = *buf_ptr++;
 333                         if (buf_ptr >= buf_end)
 334                             fill_buffer();
 335
 336                         if (*sc_end++ == '*' && *buf_ptr == '/')
 337                             break;      /* we are at end of comment */
 338
 339                         if (sc_end >= &(save_com[sc_size])) {   /* check for temp buffer
 340                                                                  * overflow */
 341                             diag2(1, "Internal buffer overflow - Move big comment from right after if, while, or whatever.");
 342                             fflush(output);
 343                             exit(1);
 344                         }
 345                     }
 346                     *sc_end++ = '/';    /* add ending slash */
 347                     if (++buf_ptr >= buf_end)   /* get past / in buffer */
 348                         fill_buffer();
 349                     break;
 350                 }
 351             default:            /* it is the start of a normal statment */
 352                 if (flushed_nl) /* if we flushed a newline, make sure it is
 353                                  * put back */
 354                     force_nl = true;
 355                 if ((type_code == sp_paren && *token == 'i'
 356                         && last_else && ps.else_if)
 357                         || (type_code == sp_nparen && *token == 'e'
 358                         && e_code != s_code && e_code[-1] == '}'))
 359                     force_nl = false;
 360
 361                 if (sc_end == 0) {      /* ignore buffering if comment wasnt
 362                                          * saved up */
 363                     ps.search_brace = false;
 364                     goto check_type;
 365                 }
 366                 if (force_nl) { /* if we should insert a nl here, put it into
 367                                  * the buffer */
 368                     force_nl = false;
 369                     --line_no;  /* this will be re-increased when the nl is
 370                                  * read from the buffer */
 371                     *sc_end++ = '\n';
 372                     *sc_end++ = ' ';
 373                     if (verbose && !flushed_nl) /* print error msg if the line
 374                                                  * was not already broken */
 375                         diag2(0, "Line broken");
 376                     flushed_nl = false;
 377                 }
 378                 for (t_ptr = token; *t_ptr; ++t_ptr)
 379                     *sc_end++ = *t_ptr; /* copy token into temp buffer */
 380                 ps.procname[0] = 0;
 381
 382         sw_buffer:
 383                 ps.search_brace = false;        /* stop looking for start of
 384                                                  * stmt */
 385                 bp_save = buf_ptr;      /* save current input buffer */
 386                 be_save = buf_end;
 387                 buf_ptr = save_com;     /* fix so that subsequent calls to
 388                                          * lexi will take tokens out of
 389                                          * save_com */
 390                 *sc_end++ = ' ';/* add trailing blank, just in case */
 391                 buf_end = sc_end;
 392                 sc_end = 0;
 393                 break;
 394             }                   /* end of switch */
 395             if (type_code != 0) /* we must make this check, just in case there
 396                                  * was an unexpected EOF */
 397                 type_code = lexi();     /* read another token */
 398             /* if (ps.search_brace) ps.procname[0] = 0; */
 399             if ((is_procname = ps.procname[0]) && flushed_nl
 400                     && !procnames_start_line && ps.in_decl
 401                     && type_code == ident)
 402                 flushed_nl = 0;
 403         }                       /* end of while (search_brace) */
 404         last_else = 0;
 405 check_type:
 406         if (type_code == 0) {   /* we got eof */
 407             if (s_lab != e_lab || s_code != e_code
 408                     || s_com != e_com)  /* must dump end of line */
 409                 dump_line();
 410             if (ps.tos > 1)     /* check for balanced braces */
 411                 diag2(1, "Stuff missing from end of file.");
 412
 413             if (verbose) {
 414                 printf("There were %d output lines and %d comments\n",
 415                        ps.out_lines, ps.out_coms);
 416                 printf("(Lines with comments)/(Lines with code): %6.3f\n",
 417                        (1.0 * ps.com_lines) / code_lines);
 418             }
 419             fflush(output);
 420             exit(found_err);
 421         }
 422         if (
 423                 (type_code != comment) &&
 424                 (type_code != newline) &&
 425                 (type_code != preesc) &&
 426                 (type_code != form_feed)) {
 427             if (force_nl &&
 428                     (type_code != semicolon) &&
 429                     (type_code != lbrace || !btype_2)) {
 430                 /* we should force a broken line here */
 431                 if (verbose && !flushed_nl)
 432                     diag2(0, "Line broken");
 433                 flushed_nl = false;
 434                 dump_line();
 435                 ps.want_blank = false;  /* dont insert blank at line start */
 436                 force_nl = false;
 437             }
 438             ps.in_stmt = true;  /* turn on flag which causes an extra level of
 439                                  * indentation. this is turned off by a ; or
 440                                  * '}' */
 441             if (s_com != e_com) {       /* the turkey has embedded a comment
 442                                          * in a line. fix it */
 443                 *e_code++ = ' ';
 444                 for (t_ptr = s_com; *t_ptr; ++t_ptr) {
 445                     CHECK_SIZE_CODE;
 446                     *e_code++ = *t_ptr;
 447                 }
 448                 *e_code++ = ' ';
 449                 *e_code = '\0'; /* null terminate code sect */
 450                 ps.want_blank = false;
 451                 e_com = s_com;
 452             }
 453         }
 454         else if (type_code != comment)  /* preserve force_nl thru a comment */
 455             force_nl = false;   /* cancel forced newline after newline, form
 456                                  * feed, etc */
 457
 458
 459
 460         /*-----------------------------------------------------*\
 461         |          do switch on type of token scanned           |
 462         \*-----------------------------------------------------*/
 463         CHECK_SIZE_CODE;
 464         switch (type_code) {    /* now, decide what to do with the token */
 465
 466         case form_feed: /* found a form feed in line */
 467             ps.use_ff = true;   /* a form feed is treated much like a newline */
 468             dump_line();
 469             ps.want_blank = false;
 470             break;
 471
 472         case newline:
 473             if (ps.last_token != comma || ps.p_l_follow > 0
 474                     || !ps.leave_comma || ps.block_init || !break_comma || s_com != e_com) {
 475                 dump_line();
 476                 ps.want_blank = false;
 477             }
 478             ++line_no;          /* keep track of input line number */
 479             break;
 480
 481         case lparen:            /* got a '(' or '[' */
 482             ++ps.p_l_follow;    /* count parens to make Healy happy */
 483             if (ps.want_blank && *token != '[' &&
 484                     (ps.last_token != ident || proc_calls_space
 485               || (ps.its_a_keyword && (!ps.sizeof_keyword || Bill_Shannon))))
 486                 *e_code++ = ' ';
 487             if (ps.in_decl && !ps.block_init)
 488                 if (troff && !ps.dumped_decl_indent && !is_procname && ps.last_token == decl) {
 489                     ps.dumped_decl_indent = 1;
 490                     sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
 491                     e_code += strlen(e_code);
 492                 }
 493                 else {
 494                     while ((e_code - s_code) < dec_ind) {
 495                         CHECK_SIZE_CODE;
 496                         *e_code++ = ' ';
 497                     }
 498                     *e_code++ = token[0];
 499                 }
 500             else
 501                 *e_code++ = token[0];
 502             ps.paren_indents[ps.p_l_follow - 1] = e_code - s_code;
 503             if (sp_sw && ps.p_l_follow == 1 && extra_expression_indent
 504                     && ps.paren_indents[0] < 2 * ps.ind_size)
 505                 ps.paren_indents[0] = 2 * ps.ind_size;
 506             ps.want_blank = false;
 507             if (ps.in_or_st && *token == '(' && ps.tos <= 2) {
 508                 /*
 509                  * this is a kluge to make sure that declarations will be
 510                  * aligned right if proc decl has an explicit type on it, i.e.
 511                  * "int a(x) {..."
 512                  */
 513                 parse(semicolon);       /* I said this was a kluge... */
 514                 ps.in_or_st = false;    /* turn off flag for structure decl or
 515                                          * initialization */
 516             }
 517             if (ps.sizeof_keyword)
 518                 ps.sizeof_mask |= 1 << ps.p_l_follow;
 519             break;
 520
 521         case rparen:            /* got a ')' or ']' */
 522             rparen_count--;
 523             if (ps.cast_mask & (1 << ps.p_l_follow) & ~ps.sizeof_mask) {
 524                 ps.last_u_d = true;
 525                 ps.cast_mask &= (1 << ps.p_l_follow) - 1;
 526                 ps.want_blank = false;
 527             } else
 528                 ps.want_blank = true;
 529             ps.sizeof_mask &= (1 << ps.p_l_follow) - 1;
 530             if (--ps.p_l_follow < 0) {
 531                 ps.p_l_follow = 0;
 532                 diag3(0, "Extra %c", *token);
 533             }
 534             if (e_code == s_code)       /* if the paren starts the line */
 535                 ps.paren_level = ps.p_l_follow; /* then indent it */
 536
 537             *e_code++ = token[0];
 538
 539             if (sp_sw && (ps.p_l_follow == 0)) {        /* check for end of if
 540                                                          * (...), or some such */
 541                 sp_sw = false;
 542                 force_nl = true;/* must force newline after if */
 543                 ps.last_u_d = true;     /* inform lexi that a following
 544                                          * operator is unary */
 545                 ps.in_stmt = false;     /* dont use stmt continuation
 546                                          * indentation */
 547
 548                 parse(hd_type); /* let parser worry about if, or whatever */
 549             }
 550             ps.search_brace = btype_2;  /* this should insure that constructs
 551                                          * such as main(){...} and int[]{...}
 552                                          * have their braces put in the right
 553                                          * place */
 554             break;
 555
 556         case unary_op:          /* this could be any unary operation */
 557             if (ps.want_blank)
 558                 *e_code++ = ' ';
 559
 560             if (troff && !ps.dumped_decl_indent && ps.in_decl && !is_procname) {
 561                 sprintf(e_code, "\n.Du %dp+\200p \"%s\"\n", dec_ind * 7, token);
 562                 ps.dumped_decl_indent = 1;
 563                 e_code += strlen(e_code);
 564             }
 565             else {
 566                 char       *res = token;
 567
 568                 if (ps.in_decl && !ps.block_init) {     /* if this is a unary op
 569                                                          * in a declaration, we
 570                                                          * should indent this
 571                                                          * token */
 572                     for (i = 0; token[i]; ++i); /* find length of token */
 573                     while ((e_code - s_code) < (dec_ind - i)) {
 574                         CHECK_SIZE_CODE;
 575                         *e_code++ = ' ';        /* pad it */
 576                     }
 577                 }
 578                 if (troff && token[0] == '-' && token[1] == '>')
 579                     res = "\\(->";
 580                 for (t_ptr = res; *t_ptr; ++t_ptr) {
 581                     CHECK_SIZE_CODE;
 582                     *e_code++ = *t_ptr;
 583                 }
 584             }
 585             ps.want_blank = false;
 586             break;
 587
 588         case binary_op: /* any binary operation */
 589             if (ps.want_blank)
 590                 *e_code++ = ' ';
 591             {
 592                 char       *res = token;
 593
 594                 if (troff)
 595                     switch (token[0]) {
 596                     case '<':
 597                         if (token[1] == '=')
 598                             res = "\\(<=";
 599                         break;
 600                     case '>':
 601                         if (token[1] == '=')
 602                             res = "\\(>=";
 603                         break;
 604                     case '!':
 605                         if (token[1] == '=')
 606                             res = "\\(!=";
 607                         break;
 608                     case '|':
 609                         if (token[1] == '|')
 610                             res = "\\(br\\(br";
 611                         else if (token[1] == 0)
 612                             res = "\\(br";
 613                         break;
 614                     }
 615                 for (t_ptr = res; *t_ptr; ++t_ptr) {
 616                     CHECK_SIZE_CODE;
 617                     *e_code++ = *t_ptr; /* move the operator */
 618                 }
 619             }
 620             ps.want_blank = true;
 621             break;
 622
 623         case postop:            /* got a trailing ++ or -- */
 624             *e_code++ = token[0];
 625             *e_code++ = token[1];
 626             ps.want_blank = true;
 627             break;
 628
 629         case question:          /* got a ? */
 630             squest++;           /* this will be used when a later colon
 631                                  * appears so we can distinguish the
 632                                  * <c>?<n>:<n> construct */
 633             if (ps.want_blank)
 634                 *e_code++ = ' ';
 635             *e_code++ = '?';
 636             ps.want_blank = true;
 637             break;
 638
 639         case casestmt:          /* got word 'case' or 'default' */
 640             scase = true;       /* so we can process the later colon properly */
 641             goto copy_id;
 642
 643         case colon:             /* got a ':' */
 644             if (squest > 0) {   /* it is part of the <c>?<n>: <n> construct */
 645                 --squest;
 646                 if (ps.want_blank)
 647                     *e_code++ = ' ';
 648                 *e_code++ = ':';
 649                 ps.want_blank = true;
 650                 break;
 651             }
 652             if (ps.in_decl) {
 653                 *e_code++ = ':';
 654                 ps.want_blank = false;
 655                 break;
 656             }
 657             ps.in_stmt = false; /* seeing a label does not imply we are in a
 658                                  * stmt */
 659             for (t_ptr = s_code; *t_ptr; ++t_ptr)
 660                 *e_lab++ = *t_ptr;      /* turn everything so far into a label */
 661             e_code = s_code;
 662             *e_lab++ = ':';
 663             *e_lab++ = ' ';
 664             *e_lab = '\0';
 665
 666             force_nl = ps.pcase = scase;        /* ps.pcase will be used by
 667                                                  * dump_line to decide how to
 668                                                  * indent the label. force_nl
 669                                                  * will force a case n: to be
 670                                                  * on a line by itself */
 671             scase = false;
 672             ps.want_blank = false;
 673             break;
 674
 675         case semicolon: /* got a ';' */
 676             ps.in_or_st = false;/* we are not in an initialization or
 677                                  * structure declaration */
 678             scase = false;      /* these will only need resetting in a error */
 679             squest = 0;
 680             if (ps.last_token == rparen && rparen_count == 0)
 681                 ps.in_parameter_declaration = 0;
 682             ps.cast_mask = 0;
 683             ps.sizeof_mask = 0;
 684             ps.block_init = 0;
 685             ps.block_init_level = 0;
 686             ps.just_saw_decl--;
 687
 688             if (ps.in_decl && s_code == e_code && !ps.block_init)
 689                 while ((e_code - s_code) < (dec_ind - 1)) {
 690                     CHECK_SIZE_CODE;
 691                     *e_code++ = ' ';
 692                 }
 693
 694             ps.in_decl = (ps.dec_nest > 0);     /* if we were in a first level
 695                                                  * structure declaration, we
 696                                                  * arent any more */
 697
 698             if ((!sp_sw || hd_type != forstmt) && ps.p_l_follow > 0) {
 699
 700                 /*
 701                  * This should be true iff there were unbalanced parens in the
 702                  * stmt.  It is a bit complicated, because the semicolon might
 703                  * be in a for stmt
 704                  */
 705                 diag2(1, "Unbalanced parens");
 706                 ps.p_l_follow = 0;
 707                 if (sp_sw) {    /* this is a check for a if, while, etc. with
 708                                  * unbalanced parens */
 709                     sp_sw = false;
 710                     parse(hd_type);     /* dont lose the if, or whatever */
 711                 }
 712             }
 713             *e_code++ = ';';
 714             ps.want_blank = true;
 715             ps.in_stmt = (ps.p_l_follow > 0);   /* we are no longer in the
 716                                                  * middle of a stmt */
 717
 718             if (!sp_sw) {       /* if not if for (;;) */
 719                 parse(semicolon);       /* let parser know about end of stmt */
 720                 force_nl = true;/* force newline after a end of stmt */
 721             }
 722             break;
 723
 724         case lbrace:            /* got a '{' */
 725             ps.in_stmt = false; /* dont indent the {} */
 726             if (!ps.block_init)
 727                 force_nl = true;/* force other stuff on same line as '{' onto
 728                                  * new line */
 729             else if (ps.block_init_level <= 0)
 730                 ps.block_init_level = 1;
 731             else
 732                 ps.block_init_level++;
 733
 734             if (s_code != e_code && !ps.block_init) {
 735                 if (!btype_2) {
 736                     dump_line();
 737                     ps.want_blank = false;
 738                 }
 739                 else if (ps.in_parameter_declaration && !ps.in_or_st) {
 740                     ps.i_l_follow = 0;
 741                     dump_line();
 742                     ps.want_blank = false;
 743                 }
 744             }
 745             if (ps.in_parameter_declaration)
 746                 prefix_blankline_requested = 0;
 747
 748             if (ps.p_l_follow > 0) {    /* check for preceding unbalanced
 749                                          * parens */
 750                 diag2(1, "Unbalanced parens");
 751                 ps.p_l_follow = 0;
 752                 if (sp_sw) {    /* check for unclosed if, for, etc. */
 753                     sp_sw = false;
 754                     parse(hd_type);
 755                     ps.ind_level = ps.i_l_follow;
 756                 }
 757             }
 758             if (s_code == e_code)
 759                 ps.ind_stmt = false;    /* dont put extra indentation on line
 760                                          * with '{' */
 761             if (ps.in_decl && ps.in_or_st) {    /* this is either a structure
 762                                                  * declaration or an init */
 763                 di_stack[ps.dec_nest++] = dec_ind;
 764                 /* ?            dec_ind = 0; */
 765             }
 766             else {
 767                 ps.decl_on_line = false;        /* we cant be in the middle of
 768                                                  * a declaration, so dont do
 769                                                  * special indentation of
 770                                                  * comments */
 771                 if (blanklines_after_declarations_at_proctop
 772                         && ps.in_parameter_declaration)
 773                     postfix_blankline_requested = 1;
 774                 ps.in_parameter_declaration = 0;
 775             }
 776             dec_ind = 0;
 777             parse(lbrace);      /* let parser know about this */
 778             if (ps.want_blank)  /* put a blank before '{' if '{' is not at
 779                                  * start of line */
 780                 *e_code++ = ' ';
 781             ps.want_blank = false;
 782             *e_code++ = '{';
 783             ps.just_saw_decl = 0;
 784             break;
 785
 786         case rbrace:            /* got a '}' */
 787             if (ps.p_stack[ps.tos] == decl && !ps.block_init)   /* semicolons can be
 788                                                                  * omitted in
 789                                                                  * declarations */
 790                 parse(semicolon);
 791             if (ps.p_l_follow) {/* check for unclosed if, for, else. */
 792                 diag2(1, "Unbalanced parens");
 793                 ps.p_l_follow = 0;
 794                 sp_sw = false;
 795             }
 796             ps.just_saw_decl = 0;
 797             ps.block_init_level--;
 798             if (s_code != e_code && !ps.block_init) {   /* '}' must be first on
 799                                                          * line */
 800                 if (verbose)
 801                     diag2(0, "Line broken");
 802                 dump_line();
 803             }
 804             *e_code++ = '}';
 805             ps.want_blank = true;
 806             ps.in_stmt = ps.ind_stmt = false;
 807             if (ps.dec_nest > 0) {      /* we are in multi-level structure
 808                                          * declaration */
 809                 dec_ind = di_stack[--ps.dec_nest];
 810                 if (ps.dec_nest == 0 && !ps.in_parameter_declaration)
 811                     ps.just_saw_decl = 2;
 812                 ps.in_decl = true;
 813             }
 814             prefix_blankline_requested = 0;
 815             parse(rbrace);      /* let parser know about this */
 816             ps.search_brace = cuddle_else && ps.p_stack[ps.tos] == ifhead
 817                 && ps.il[ps.tos] >= ps.ind_level;
 818             if (ps.tos <= 1 && blanklines_after_procs && ps.dec_nest <= 0)
 819                 postfix_blankline_requested = 1;
 820             break;
 821
 822         case swstmt:            /* got keyword "switch" */
 823             sp_sw = true;
 824             hd_type = swstmt;   /* keep this for when we have seen the
 825                                  * expression */
 826             goto copy_id;       /* go move the token into buffer */
 827
 828         case sp_paren:          /* token is if, while, for */
 829             sp_sw = true;       /* the interesting stuff is done after the
 830                                  * expression is scanned */
 831             hd_type = (*token == 'i' ? ifstmt :
 832                        (*token == 'w' ? whilestmt : forstmt));
 833
 834             /*
 835              * remember the type of header for later use by parser
 836              */
 837             goto copy_id;       /* copy the token into line */
 838
 839         case sp_nparen: /* got else, do */
 840             ps.in_stmt = false;
 841             if (*token == 'e') {
 842                 if (e_code != s_code && (!cuddle_else || e_code[-1] != '}')) {
 843                     if (verbose)
 844                         diag2(0, "Line broken");
 845                     dump_line();/* make sure this starts a line */
 846                     ps.want_blank = false;
 847                 }
 848                 force_nl = true;/* also, following stuff must go onto new line */
 849                 last_else = 1;
 850                 parse(elselit);
 851             }
 852             else {
 853                 if (e_code != s_code) { /* make sure this starts a line */
 854                     if (verbose)
 855                         diag2(0, "Line broken");
 856                     dump_line();
 857                     ps.want_blank = false;
 858                 }
 859                 force_nl = true;/* also, following stuff must go onto new line */
 860                 last_else = 0;
 861                 parse(dolit);
 862             }
 863             goto copy_id;       /* move the token into line */
 864
 865         case decl:              /* we have a declaration type (int, register,
 866                                  * etc.) */
 867             parse(decl);        /* let parser worry about indentation */
 868             if (ps.last_token == rparen && ps.tos <= 1) {
 869                 ps.in_parameter_declaration = 1;
 870                 if (s_code != e_code) {
 871                     dump_line();
 872                     ps.want_blank = 0;
 873                 }
 874             }
 875             if (ps.in_parameter_declaration && ps.indent_parameters && ps.dec_nest == 0) {
 876                 ps.ind_level = ps.i_l_follow = 1;
 877                 ps.ind_stmt = 0;
 878             }
 879             ps.in_or_st = true; /* this might be a structure or initialization
 880                                  * declaration */
 881             ps.in_decl = ps.decl_on_line = true;
 882             if ( /* !ps.in_or_st && */ ps.dec_nest <= 0)
 883                 ps.just_saw_decl = 2;
 884             prefix_blankline_requested = 0;
 885             for (i = 0; token[i++];);   /* get length of token */
 886
 887             /*
 888              * dec_ind = e_code - s_code + (ps.decl_indent>i ? ps.decl_indent
 889              * : i);
 890              */
 891             dec_ind = ps.decl_indent > 0 ? ps.decl_indent : i;
 892             goto copy_id;
 893
 894         case ident:             /* got an identifier or constant */
 895             if (ps.in_decl) {   /* if we are in a declaration, we must indent
 896                                  * identifier */
 897                 if (ps.want_blank)
 898                     *e_code++ = ' ';
 899                 ps.want_blank = false;
 900                 if (is_procname == 0 || !procnames_start_line) {
 901                     if (!ps.block_init) {
 902                         if (troff && !ps.dumped_decl_indent) {
 903                             sprintf(e_code, "\n.De %dp+\200p\n", dec_ind * 7);
 904                             ps.dumped_decl_indent = 1;
 905                             e_code += strlen(e_code);
 906                         } else {
 907                             while ((e_code - s_code) < dec_ind) {
 908                                 CHECK_SIZE_CODE;
 909                                 *e_code++ = ' ';
 910                             }
 911                         }
 912                     }
 913                 } else {
 914                     if (dec_ind && s_code != e_code)
 915                         dump_line();
 916                     dec_ind = 0;
 917                     ps.want_blank = false;
 918                 }
 919             }
 920             else if (sp_sw && ps.p_l_follow == 0) {
 921                 sp_sw = false;
 922                 force_nl = true;
 923                 ps.last_u_d = true;
 924                 ps.in_stmt = false;
 925                 parse(hd_type);
 926             }
 927     copy_id:
 928             if (ps.want_blank)
 929                 *e_code++ = ' ';
 930             if (troff && ps.its_a_keyword) {
 931                 e_code = chfont(&bodyf, &keywordf, e_code);
 932                 for (t_ptr = token; *t_ptr; ++t_ptr) {
 933                     CHECK_SIZE_CODE;
 934                     *e_code++ = keywordf.allcaps && islower(*t_ptr)
 935                         ? toupper(*t_ptr) : *t_ptr;
 936                 }
 937                 e_code = chfont(&keywordf, &bodyf, e_code);
 938             }
 939             else
 940                 for (t_ptr = token; *t_ptr; ++t_ptr) {
 941                     CHECK_SIZE_CODE;
 942                     *e_code++ = *t_ptr;
 943                 }
 944             ps.want_blank = true;
 945             break;
 946
 947         case period:            /* treat a period kind of like a binary
 948                                  * operation */
 949             *e_code++ = '.';    /* move the period into line */
 950             ps.want_blank = false;      /* dont put a blank after a period */
 951             break;
 952
 953         case comma:
 954             ps.want_blank = (s_code != e_code); /* only put blank after comma
 955                                                  * if comma does not start the
 956                                                  * line */
 957             if (ps.in_decl && is_procname == 0 && !ps.block_init)
 958                 while ((e_code - s_code) < (dec_ind - 1)) {
 959                     CHECK_SIZE_CODE;
 960                     *e_code++ = ' ';
 961                 }
 962
 963             *e_code++ = ',';
 964             if (ps.p_l_follow == 0) {
 965                 if (ps.block_init_level <= 0)
 966                     ps.block_init = 0;
 967                 if (break_comma && (!ps.leave_comma || compute_code_target() + (e_code - s_code) > max_col - 8))
 968                     force_nl = true;
 969             }
 970             break;
 971
 972         case preesc:            /* got the character '#' */
 973             if ((s_com != e_com) ||
 974                     (s_lab != e_lab) ||
 975                     (s_code != e_code))
 976                 dump_line();
 977             *e_lab++ = '#';     /* move whole line to 'label' buffer */
 978             {
 979                 int         in_comment = 0;
 980                 int         com_start = 0;
 981                 char        quote = 0;
 982                 int         com_end = 0;
 983
 984                 while (*buf_ptr == ' ' || *buf_ptr == '\t') {
 985                     buf_ptr++;
 986                     if (buf_ptr >= buf_end)
 987                         fill_buffer();
 988                 }
 989                 while (*buf_ptr != '\n' || (in_comment && !had_eof)) {
 990                     CHECK_SIZE_LAB;
 991                     *e_lab = *buf_ptr++;
 992                     if (buf_ptr >= buf_end)
 993                         fill_buffer();
 994                     switch (*e_lab++) {
 995                     case BACKSLASH:
 996                         if (troff)
 997                             *e_lab++ = BACKSLASH;
 998                         if (!in_comment) {
 999                             *e_lab++ = *buf_ptr++;
1000                             if (buf_ptr >= buf_end)
1001                                 fill_buffer();
1002                         }
1003                         break;
1004                     case '/':
1005                         if (*buf_ptr == '*' && !in_comment && !quote) {
1006                             in_comment = 1;
1007                             *e_lab++ = *buf_ptr++;
1008                             com_start = e_lab - s_lab - 2;
1009                         }
1010                         break;
1011                     case '"':
1012                         if (quote == '"')
1013                             quote = 0;
1014                         break;
1015                     case '\'':
1016                         if (quote == '\'')
1017                             quote = 0;
1018                         break;
1019                     case '*':
1020                         if (*buf_ptr == '/' && in_comment) {
1021                             in_comment = 0;
1022                             *e_lab++ = *buf_ptr++;
1023                             com_end = e_lab - s_lab;
1024                         }
1025                         break;
1026                     }
1027                 }
1028
1029                 while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1030                     e_lab--;
1031                 if (e_lab - s_lab == com_end && bp_save == 0) { /* comment on
1032                                                                  * preprocessor line */
1033                     if (sc_end == 0)    /* if this is the first comment, we
1034                                          * must set up the buffer */
1035                         sc_end = &(save_com[0]);
1036                     else {
1037                         *sc_end++ = '\n';       /* add newline between
1038                                                  * comments */
1039                         *sc_end++ = ' ';
1040                         --line_no;
1041                     }
1042                     bcopy(s_lab + com_start, sc_end, com_end - com_start);
1043                     sc_end += com_end - com_start;
1044                     if (sc_end >= &save_com[sc_size])
1045                         abort();
1046                     e_lab = s_lab + com_start;
1047                     while (e_lab > s_lab && (e_lab[-1] == ' ' || e_lab[-1] == '\t'))
1048                         e_lab--;
1049                     bp_save = buf_ptr;  /* save current input buffer */
1050                     be_save = buf_end;
1051                     buf_ptr = save_com; /* fix so that subsequent calls to
1052                                          * lexi will take tokens out of
1053                                          * save_com */
1054                     *sc_end++ = ' ';    /* add trailing blank, just in case */
1055                     buf_end = sc_end;
1056                     sc_end = 0;
1057                 }
1058                 *e_lab = '\0';  /* null terminate line */
1059                 ps.pcase = false;
1060             }
1061
1062             if (strncmp(s_lab, "#if", 3) == 0) {
1063                 if (blanklines_around_conditional_compilation) {
1064                     int c;
1065                     prefix_blankline_requested++;
1066                     while ((c = getc(input)) == '\n');
1067                     ungetc(c, input);
1068                 }
1069                 if (ifdef_level < sizeof state_stack / sizeof state_stack[0]) {
1070                     match_state[ifdef_level].tos = -1;
1071                     state_stack[ifdef_level++] = ps;
1072                 }
1073                 else
1074                     diag2(1, "#if stack overflow");
1075             }
1076             else if (strncmp(s_lab, "#else", 5) == 0)
1077                 if (ifdef_level <= 0)
1078                     diag2(1, "Unmatched #else");
1079                 else {
1080                     match_state[ifdef_level - 1] = ps;
1081                     ps = state_stack[ifdef_level - 1];
1082                 }
1083             else if (strncmp(s_lab, "#endif", 6) == 0) {
1084                 if (ifdef_level <= 0)
1085                     diag2(1, "Unmatched #endif");
1086                 else {
1087                     ifdef_level--;
1088
1089 #ifdef undef
1090                     /*
1091                      * This match needs to be more intelligent before the
1092                      * message is useful
1093                      */
1094                     if (match_state[ifdef_level].tos >= 0
1095                           && bcmp(&ps, &match_state[ifdef_level], sizeof ps))
1096                         diag2(0, "Syntactically inconsistent #ifdef alternatives.");
1097 #endif
1098                 }
1099                 if (blanklines_around_conditional_compilation) {
1100                     postfix_blankline_requested++;
1101                     n_real_blanklines = 0;
1102                 }
1103             }
1104             break;              /* subsequent processing of the newline
1105                                  * character will cause the line to be printed */
1106
1107         case comment:           /* we have gotten a / followed by * this is a biggie */
1108             if (flushed_nl) {   /* we should force a broken line here */
1109                 flushed_nl = false;
1110                 dump_line();
1111                 ps.want_blank = false;  /* dont insert blank at line start */
1112                 force_nl = false;
1113             }
1114             pr_comment();
1115             break;
1116         }                       /* end of big switch stmt */
1117
1118         *e_code = '\0';         /* make sure code section is null terminated */
1119         if (type_code != comment && type_code != newline && type_code != preesc)
1120             ps.last_token = type_code;
1121     }                           /* end of main while (1) loop */
1122 }
1123
1124 /*
1125  * copy input file to backup file if in_name is /blah/blah/blah/file, then
1126  * backup file will be ".Bfile" then make the backup file the input and
1127  * original input file the output
1128  */
1129 static void
1130 bakcopy(void)
1131 {
1132     int         n,
1133                 bakchn;
1134     char        buff[8 * 1024];
1135     char *p;
1136
1137     /* construct file name .Bfile */
1138     for (p = in_name; *p; p++); /* skip to end of string */
1139     while (p > in_name && *p != '/')    /* find last '/' */
1140         p--;
1141     if (*p == '/')
1142         p++;
1143     sprintf(bakfile, "%s.BAK", p);
1144
1145     /* copy in_name to backup file */
1146     bakchn = creat(bakfile, 0600);
1147     if (bakchn < 0)
1148         err(1, "%s", bakfile);
1149     while ((n = read(fileno(input), buff, sizeof buff)) != 0)
1150         if (write(bakchn, buff, n) != n)
1151             err(1, "%s", bakfile);
1152     if (n < 0)
1153         err(1, "%s", in_name);
1154     close(bakchn);
1155     fclose(input);
1156
1157     /* re-open backup file as the input file */
1158     input = fopen(bakfile, "r");
1159     if (input == 0)
1160         err(1, "%s", bakfile);
1161     /* now the original input file will be the output */
1162     output = fopen(in_name, "w");
1163     if (output == 0) {
1164         unlink(bakfile);
1165         err(1, "%s", in_name);
1166     }
1167 }