annotate lib-src/etags.c @ 27893:0f639a575917

(tty_defined_color): Apply xfaces.c change from 02-17.

/* Tags file maker to go with GNU Emacs
   Copyright (C) 1984, 87, 88, 89, 93, 94, 95, 98, 99
   Free Software Foundation, Inc. and Ken Arnold

This file is not considered part of GNU Emacs.

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software Foundation,
Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */

/*
 * Authors:
 *	Ctags originally by Ken Arnold.
 *	Fortran added by Jim Kleckner.
 *	Ed Pelegri-Llopart added C typedefs.
 *	Gnu Emacs TAGS format and modifications by RMS?
 *	Sam Kendall added C++.
 *	Francesco Potorti` reorganised C and C++ based on work by Joe Wells.
 *	Regexp tags by Tom Tromey.
 *
 *	Francesco Potorti` ([e-mail]) is the current maintainer.
 */

char pot_etags_version[] = "@(#) pot revision number is 13.44";

#define	TRUE	1
#define	FALSE	0

#ifndef DEBUG
# define DEBUG FALSE
#endif

#if defined(__STDC__) && (__STDC__ || defined(__SUNPRO_C))
# define P_(proto) proto
#else
# define P_(proto) ()
#endif

#ifdef HAVE_CONFIG_H
# include <config.h>
  /* On some systems, Emacs defines static as nothing for the sake
     of unexec.  We don't want that here since we don't use unexec. */
# undef static
# define ETAGS_REGEXPS		/* use the regexp features */
# define LONG_OPTIONS		/* accept long options */
#endif /* HAVE_CONFIG_H */

#ifndef _GNU_SOURCE
# define _GNU_SOURCE 1		/* enables some compiler checks on GNU */
#endif

#ifdef MSDOS
# undef MSDOS
# define MSDOS TRUE
# include <fcntl.h>
# include <sys/param.h>
# include <io.h>
# ifndef HAVE_CONFIG_H
#   define DOS_NT
#   include <sys/config.h>
# endif
#else
# define MSDOS FALSE
#endif /* MSDOS */

#ifdef WINDOWSNT
# include <stdlib.h>
# include <fcntl.h>
# include <string.h>
# include <direct.h>
# include <io.h>
# define MAXPATHLEN _MAX_PATH
# ifdef HAVE_CONFIG_H
#   undef HAVE_NTGUI
# else
#   define DOS_NT
# endif /* not HAVE_CONFIG_H */
# ifndef HAVE_GETCWD
#   define HAVE_GETCWD
# endif /* undef HAVE_GETCWD */
#else /* !WINDOWSNT */
# ifdef STDC_HEADERS
#  include <stdlib.h>
#  include <string.h>
# else
    extern char *getenv ();
# endif
#endif /* !WINDOWSNT */

#ifdef HAVE_UNISTD_H
# include <unistd.h>
#else
# if defined (HAVE_GETCWD) && !WINDOWSNT
    extern char *getcwd (char *buf, size_t size);
# endif
#endif /* HAVE_UNISTD_H */

#include <stdio.h>
#include <ctype.h>
#include <errno.h>
#ifndef errno
  extern int errno;
#endif
#include <sys/types.h>
#include <sys/stat.h>

#if !defined (S_ISREG) && defined (S_IFREG)
# define S_ISREG(m)	(((m) & S_IFMT) == S_IFREG)
#endif

#ifdef LONG_OPTIONS
# include <getopt.h>
#else
# define getopt_long(argc,argv,optstr,lopts,lind) getopt (argc, argv, optstr)
  extern char *optarg;
  extern int optind, opterr;
#endif /* LONG_OPTIONS */

#ifdef ETAGS_REGEXPS
# include <regex.h>
#endif /* ETAGS_REGEXPS */

/* Define CTAGS to make the program "ctags" compatible with the usual one.
 Leave it undefined to make the program "etags", which makes emacs-style
 tag tables and tags typedefs, #defines and struct/union/enum by default. */
#ifdef CTAGS
# undef  CTAGS
# define CTAGS TRUE
#else
# define CTAGS FALSE
#endif

/* Exit codes for success and failure.  */
#ifdef VMS
# define	GOOD	1
# define	BAD	0
#else
# define	GOOD	0
# define	BAD	1
#endif

/* C extensions. */
#define C_PLPL	0x00001		/* C++ */
#define C_STAR	0x00003		/* C* */
#define C_JAVA	0x00005		/* JAVA */
#define YACC	0x10000		/* yacc file */

#define streq(s,t)	((DEBUG && (s) == NULL && (t) == NULL	\
			  && (abort (), 1)) || !strcmp (s, t))
#define strneq(s,t,n)	((DEBUG && (s) == NULL && (t) == NULL	\
			  && (abort (), 1)) || !strncmp (s, t, n))

#define lowcase(c)	tolower ((char)c)

#define CHARS 256		/* 2^sizeof(char) */
#define CHAR(x)		((unsigned int)x & (CHARS - 1))
#define	iswhite(c)	(_wht[CHAR(c)]) /* c is white */
#define notinname(c)	(_nin[CHAR(c)]) /* c is not in a name */
#define	begtoken(c)	(_btk[CHAR(c)]) /* c can start token */
#define	intoken(c)	(_itk[CHAR(c)]) /* c can be in token */
#define	endtoken(c)	(_etk[CHAR(c)]) /* c ends tokens */


/*
 *	xnew, xrnew -- allocate, reallocate storage
 *
 * SYNOPSIS:	Type *xnew (int n, Type);
 *		Type *xrnew (OldPointer, int n, Type);
 */
#ifdef chkmalloc
# include "chkmalloc.h"
# define xnew(n,Type)	  ((Type *) trace_malloc (__FILE__, __LINE__, \
						  (n) * sizeof (Type)))
# define xrnew(op,n,Type) ((Type *) trace_realloc (__FILE__, __LINE__, \
						   (op), (n) * sizeof (Type)))
#else
# define xnew(n,Type)	  ((Type *) xmalloc ((n) * sizeof (Type)))
# define xrnew(op,n,Type) ((Type *) xrealloc ((op), (n) * sizeof (Type)))
#endif

typedef int bool;

typedef void Lang_function P_((FILE *));

typedef struct
{
  char *suffix;
  char *command;		/* Takes one arg and decompresses to stdout */
} compressor;

typedef struct
{
  char *name;
  Lang_function *function;
  char **suffixes;
  char **interpreters;
} language;

typedef struct node_st
{				/* sorting structure		*/
  char *name;			/* function or type name	*/
  char *file;			/* file name			*/
  bool is_func;			/* use pattern or line no	*/
  bool been_warned;		/* set if noticed dup		*/
  int lno;			/* line number tag is on	*/
  long cno;			/* character number line starts on */
  char *pat;			/* search pattern		*/
  struct node_st *left, *right;	/* left and right sons		*/
} node;

/*
 * A `linebuffer' is a structure which holds a line of text.
 * `readline_internal' reads a line from a stream into a linebuffer
 * and works regardless of the length of the line.
 * SIZE is the size of BUFFER, LEN is the length of the string in
 * BUFFER after readline reads it.
 */
typedef struct
{
  long size;
  int len;
  char *buffer;
} linebuffer;

/* Many compilers barf on this:
	Lang_function Ada_funcs;
   so let's write it this way */
static void Ada_funcs P_((FILE *));
static void Asm_labels P_((FILE *));
static void C_entries P_((int c_ext, FILE *));
static void default_C_entries P_((FILE *));
static void plain_C_entries P_((FILE *));
static void Cjava_entries P_((FILE *));
static void Cobol_paragraphs P_((FILE *));
static void Cplusplus_entries P_((FILE *));
static void Cstar_entries P_((FILE *));
static void Erlang_functions P_((FILE *));
static void Fortran_functions P_((FILE *));
static void Yacc_entries P_((FILE *));
static void Lisp_functions P_((FILE *));
static void Pascal_functions P_((FILE *));
static void Perl_functions P_((FILE *));
static void Postscript_functions P_((FILE *));
static void Prolog_functions P_((FILE *));
static void Python_functions P_((FILE *));
static void Scheme_functions P_((FILE *));
static void TeX_functions P_((FILE *));
static void just_read_file P_((FILE *));

static void print_language_names P_((void));
static void print_version P_((void));
static void print_help P_((void));
int main P_((int, char **));
static int number_len P_((long));

static compressor *get_compressor_from_suffix P_((char *, char **));
static language *get_language_from_name P_((char *));
static language *get_language_from_interpreter P_((char *));
static language *get_language_from_suffix P_((char *));
static int total_size_of_entries P_((node *));
static long readline P_((linebuffer *, FILE *));
static long readline_internal P_((linebuffer *, FILE *));
static void get_tag P_((char *));

#ifdef ETAGS_REGEXPS
static void analyse_regex P_((char *, bool));
static void add_regex P_((char *, bool, language *));
static void free_patterns P_((void));
#endif /* ETAGS_REGEXPS */
static void error P_((const char *, const char *));
static void suggest_asking_for_help P_((void));
static void fatal P_((char *, char *));
static void pfatal P_((char *));
static void add_node P_((node *, node **));

static void init P_((void));
static void initbuffer P_((linebuffer *));
static void find_entries P_((char *, FILE *));
static void free_tree P_((node *));
static void pfnote P_((char *, bool, char *, int, int, long));
static void new_pfnote P_((char *, int, bool, char *, int, int, long));
static void process_file P_((char *));
static void put_entries P_((node *));
static void takeprec P_((void));

static char *concat P_((char *, char *, char *));
static char *skip_spaces P_((char *));
static char *skip_non_spaces P_((char *));
static char *savenstr P_((char *, int));
static char *savestr P_((char *));
static char *etags_strchr P_((const char *, int));
static char *etags_strrchr P_((const char *, int));
static char *etags_getcwd P_((void));
static char *relative_filename P_((char *, char *));
static char *absolute_filename P_((char *, char *));
static char *absolute_dirname P_((char *, char *));
static bool filename_is_absolute P_((char *f));
static void canonicalize_filename P_((char *));
static void grow_linebuffer P_((linebuffer *, int));
static long *xmalloc P_((unsigned int));
static long *xrealloc P_((char *, unsigned int));


char searchar = '/';		/* use /.../ searches */

char *tagfile;			/* output file */
char *progname;			/* name this program was invoked with */
char *cwd;			/* current working directory */
char *tagfiledir;		/* directory of tagfile */
FILE *tagf;			/* ioptr for tags file */

char *curfile;			/* current input file name */
language *curlang;		/* current language */

int lineno;			/* line number of current line */
long charno;			/* current character number */
long linecharno;		/* charno of start of current line */
char *dbp;			/* pointer to start of current tag */

node *head;			/* the head of the binary tree of tags */

linebuffer lb;			/* the current line */
linebuffer token_name;		/* used by C_entries as a temporary area */
struct
{
  long linepos;
  linebuffer lb;		/* used by C_entries instead of lb */
} lbs[2];

/* boolean "functions" (see init)	*/
bool _wht[CHARS], _nin[CHARS], _itk[CHARS], _btk[CHARS], _etk[CHARS];
char
  /* white chars */
  *white = " \f\t\n\r\v",
  /* not in a name */
  *nonam = " \f\t\n\r(=,[;",
  /* token ending chars */
  *endtk = " \t\n\r\"'#()[]{}=-+%*/&|^~!<>;,.:?",
  /* token starting chars */
  *begtk = "ABCDEFGHIJKLMNOPQRSTUVWXYZ_abcdefghijklmnopqrstuvwxyz$~@",
  /* valid in-token chars */
  *midtk = "ABCDEFGHIJKLMNOPQRSTUVWXYZ_abcdefghijklmnopqrstuvwxyz$0123456789";

bool append_to_tagfile;		/* -a: append to tags */
/* The following four default to TRUE for etags, but to FALSE for ctags.  */
bool typedefs;			/* -t: create tags for C and Ada typedefs */
bool typedefs_and_cplusplus;	/* -T: create tags for C typedefs, level */
				/* 0 struct/enum/union decls, and C++ */
				/* member functions. */
bool constantypedefs;		/* -d: create tags for C #define, enum */
				/* constants and variables. */
				/* -D: opposite of -d.  Default under ctags. */
bool declarations;		/* --declarations: tag them and extern in C&Co*/
bool globals;			/* create tags for global variables */
bool members;			/* create tags for C member variables */
bool update;			/* -u: update tags */
bool vgrind_style;		/* -v: create vgrind style index output */
bool no_warnings;		/* -w: suppress warnings */
bool cxref_style;		/* -x: create cxref style output */
bool cplusplus;			/* .[hc] means C++, not C */
bool noindentypedefs;		/* -I: ignore indentation in C */
bool packages_only;		/* --packages-only: in Ada, only tag packages*/

#ifdef LONG_OPTIONS
struct option longopts[] =
{
  { "packages-only",      no_argument,	     &packages_only, TRUE  },
  { "append",		  no_argument,	     NULL,	     'a'   },
  { "backward-search",	  no_argument,	     NULL,	     'B'   },
  { "c++",		  no_argument,	     NULL,	     'C'   },
  { "cxref",		  no_argument,	     NULL,	     'x'   },
  { "defines",		  no_argument,	     NULL,	     'd'   },
  { "declarations",	  no_argument,	     &declarations,  TRUE  },
  { "no-defines",	  no_argument,	     NULL,	     'D'   },
  { "globals",		  no_argument,	     &globals, 	     TRUE  },
  { "no-globals",	  no_argument,	     &globals, 	     FALSE },
  { "help",		  no_argument,	     NULL,     	     'h'   },
  { "help",		  no_argument,	     NULL,     	     'H'   },
  { "ignore-indentation", no_argument,	     NULL,     	     'I'   },
  { "include",		  required_argument, NULL,     	     'i'   },
  { "language",           required_argument, NULL,     	     'l'   },
  { "members",		  no_argument,	     &members, 	     TRUE  },
  { "no-members",	  no_argument,	     &members, 	     FALSE },
  { "no-warn",		  no_argument,	     NULL,	     'w'   },
  { "output",		  required_argument, NULL,	     'o'   },
#ifdef ETAGS_REGEXPS
  { "regex",		  required_argument, NULL,	     'r'   },
  { "no-regex",		  no_argument,	     NULL,	     'R'   },
  { "ignore-case-regex",  required_argument, NULL,	     'c'   },
#endif /* ETAGS_REGEXPS */
  { "typedefs",		  no_argument,	     NULL,	     't'   },
  { "typedefs-and-c++",	  no_argument,	     NULL,     	     'T'   },
  { "update",		  no_argument,	     NULL,     	     'u'   },
  { "version",		  no_argument,	     NULL,     	     'V'   },
  { "vgrind",		  no_argument,	     NULL,     	     'v'   },
  { NULL }
};
#endif /* LONG_OPTIONS */

#ifdef ETAGS_REGEXPS
/* Structure defining a regular expression.  Elements are
   the compiled pattern, and the name string. */
typedef struct pattern
{
  struct pattern *p_next;
  language *language;
  char *regex;
  struct re_pattern_buffer *pattern;
  struct re_registers regs;
  char *name_pattern;
  bool error_signaled;
} pattern;

/* List of all regexps. */
pattern *p_head = NULL;

/* How many characters in the character set.  (From regex.c.)  */
#define CHAR_SET_SIZE 256
/* Translation table for case-insensitive matching. */
char lc_trans[CHAR_SET_SIZE];
#endif /* ETAGS_REGEXPS */

compressor compressors[] =
{
  { "z", "gzip -d -c"},
  { "Z", "gzip -d -c"},
  { "gz", "gzip -d -c"},
  { "GZ", "gzip -d -c"},
  { "bz2", "bzip2 -d -c" },
  { NULL }
};

/*
 * Language stuff.
 */

/* Non-NULL if language fixed. */
language *forced_lang = NULL;

/* Ada code */
char *Ada_suffixes [] =
  { "ads", "adb", "ada", NULL };

/* Assembly code */
char *Asm_suffixes [] = { "a",	/* Unix assembler */
			  "asm", /* Microcontroller assembly */
			  "def", /* BSO/Tasking definition includes  */
			  "inc", /* Microcontroller include files */
			  "ins", /* Microcontroller include files */
			  "s", "sa", /* Unix assembler */
			  "S",   /* cpp-processed Unix assembler */
			  "src", /* BSO/Tasking C compiler output */
			  NULL
			};

/* Note that .c and .h can be considered C++, if the --c++ flag was
   given.  That is why default_C_entries is called here. */
char *default_C_suffixes [] =
  { "c", "h", NULL };

char *Cplusplus_suffixes [] =
  { "C", "H", "c++", "cc", "cpp", "cxx", "h++", "hh", "hpp", "hxx",
    "M",			/* Objective C++ */
    "pdb",			/* Postscript with C syntax */
    NULL };

char *Cjava_suffixes [] =
  { "java", NULL };

char *Cobol_suffixes [] =
  { "COB", "cob", NULL };

char *Cstar_suffixes [] =
  { "cs", "hs", NULL };

char *Erlang_suffixes [] =
  { "erl", "hrl", NULL };

char *Fortran_suffixes [] =
  { "F", "f", "f90", "for", NULL };

char *Lisp_suffixes [] =
  { "cl", "clisp", "el", "l", "lisp", "lsp", "ml", NULL };

char *Pascal_suffixes [] =
  { "p", "pas", NULL };

char *Perl_suffixes [] =
  { "pl", "pm", NULL };
char *Perl_interpreters [] =
  { "perl", "@PERL@", NULL };

char *plain_C_suffixes [] =
  { "pc",			/* Pro*C file */
    "m",			/* Objective C file */
    "lm",			/* Objective lex file */
     NULL };

char *Postscript_suffixes [] =
  { "ps", "psw", NULL };	/* .psw is for PSWrap */

char *Prolog_suffixes [] =
  { "prolog", NULL };

char *Python_suffixes [] =
  { "py", NULL };

/* Can't do the `SCM' or `scm' prefix with a version number. */
char *Scheme_suffixes [] =
  { "SCM", "SM", "oak", "sch", "scheme", "scm", "sm", "ss", "t", NULL };

char *TeX_suffixes [] =
  { "TeX", "bib", "clo", "cls", "ltx", "sty", "tex", NULL };

char *Yacc_suffixes [] =
  { "y", "ym", "yy", "yxx", "y++", NULL }; /* .ym is Objective yacc file */

/*
 * Table of languages.
 *
 * It is ok for a given function to be listed under more than one
 * name.  I just didn't.
 */

language lang_names [] =
{
  { "ada",     Ada_funcs,           Ada_suffixes,         NULL              },
  { "asm",     Asm_labels,          Asm_suffixes,         NULL              },
  { "c",       default_C_entries,   default_C_suffixes,   NULL              },
  { "c++",     Cplusplus_entries,   Cplusplus_suffixes,   NULL              },
  { "c*",      Cstar_entries,       Cstar_suffixes,       NULL              },
  { "cobol",   Cobol_paragraphs,    Cobol_suffixes,       NULL              },
  { "erlang",  Erlang_functions,    Erlang_suffixes,      NULL              },
  { "fortran", Fortran_functions,   Fortran_suffixes,     NULL              },
  { "java",    Cjava_entries,       Cjava_suffixes,       NULL              },
  { "lisp",    Lisp_functions,      Lisp_suffixes,        NULL              },
  { "pascal",  Pascal_functions,    Pascal_suffixes,      NULL              },
  { "perl",    Perl_functions,      Perl_suffixes,        Perl_interpreters },
  { "postscript", Postscript_functions, Postscript_suffixes, NULL           },
  { "proc",    plain_C_entries,     plain_C_suffixes,     NULL              },
  { "prolog",  Prolog_functions,    Prolog_suffixes,      NULL              },
  { "python",  Python_functions,    Python_suffixes,      NULL              },
  { "scheme",  Scheme_functions,    Scheme_suffixes,      NULL              },
  { "tex",     TeX_functions,       TeX_suffixes,         NULL              },
  { "yacc",    Yacc_entries,        Yacc_suffixes,        NULL              },
  { "auto", NULL },             /* default guessing scheme */
  { "none", just_read_file },   /* regexp matching only */
  { NULL, NULL }                /* end of list */
};

static void
print_language_names ()
{
  language *lang;
  char **ext;

  puts ("\nThese are the currently supported languages, along with the\n\
default file name suffixes:");
  for (lang = lang_names; lang->name != NULL; lang++)
    {
      printf ("\t%s\t", lang->name);
      if (lang->suffixes != NULL)
	for (ext = lang->suffixes; *ext != NULL; ext++)
	  printf (" .%s", *ext);
      puts ("");
    }
  puts ("Where `auto' means use default language for files based on file\n\
name suffix, and `none' means only do regexp processing on files.\n\
If no language is specified and no matching suffix is found,\n\
the first line of the file is read for a sharp-bang (#!) sequence\n\
followed by the name of an interpreter.  If no such sequence is found,\n\
Fortran is tried first; if no tags are found, C is tried next.\n\
Compressed files are supported using gzip and bzip2.");
}

#ifndef EMACS_NAME
# define EMACS_NAME "GNU Emacs"
#endif
#ifndef VERSION
# define VERSION "21"
#endif
static void
print_version ()
{
  printf ("%s (%s %s)\n", (CTAGS) ? "ctags" : "etags", EMACS_NAME, VERSION);
  puts ("Copyright (C) 1999 Free Software Foundation, Inc. and Ken Arnold");
  puts ("This program is distributed under the same terms as Emacs");

  exit (GOOD);
}

static void
print_help ()
{
  printf ("Usage: %s [options] [[regex-option ...] file-name] ...\n\
\n\
These are the options accepted by %s.\n", progname, progname);
#ifdef LONG_OPTIONS
  puts ("You may use unambiguous abbreviations for the long option names.");
#else
  puts ("Long option names do not work with this executable, as it is not\n\
linked with GNU getopt.");
#endif /* LONG_OPTIONS */
  puts ("A - as file name means read names from stdin (one per line).");
  if (!CTAGS)
    printf ("  Absolute names are stored in the output file as they are.\n\
Relative ones are stored relative to the output file's directory.");
  puts ("\n");

  puts ("-a, --append\n\
        Append tag entries to existing tags file.");

  puts ("--packages-only\n\
        For Ada files, only generate tags for packages .");

  if (CTAGS)
    puts ("-B, --backward-search\n\
        Write the search commands for the tag entries using '?', the\n\
        backward-search command instead of '/', the forward-search command.");

  puts ("-C, --c++\n\
        Treat files whose name suffix defaults to C language as C++ files.");

  puts ("--declarations\n\
	In C and derived languages, create tags for function declarations,");
  if (CTAGS)
    puts ("\tand create tags for extern variables if --globals is used.");
  else
    puts
      ("\tand create tags for extern variables unless --no-globals is used.");

  if (CTAGS)
    puts ("-d, --defines\n\
        Create tag entries for C #define constants and enum constants, too.");
  else
    puts ("-D, --no-defines\n\
        Don't create tag entries for C #define constants and enum constants.\n\
	This makes the tags file smaller.");

  if (!CTAGS)
    {
      puts ("-i FILE, --include=FILE\n\
        Include a note in tag file indicating that, when searching for\n\
        a tag, one should also consult the tags file FILE after\n\
        checking the current file.");
      puts ("-l LANG, --language=LANG\n\
        Force the following files to be considered as written in the\n\
	named language up to the next --language=LANG option.");
    }

  if (CTAGS)
    puts ("--globals\n\
	Create tag entries for global variables in some languages.");
  else
    puts ("--no-globals\n\
	Do not create tag entries for global variables in some\n\
	languages.  This makes the tags file smaller.");
  puts ("--members\n\
	Create tag entries for member variables in C and derived languages.");

#ifdef ETAGS_REGEXPS
  puts ("-r /REGEXP/, --regex=/REGEXP/ or --regex=@regexfile\n\
        Make a tag for each line matching pattern REGEXP in the following\n\
 	files.  {LANGUAGE}/REGEXP/ uses REGEXP for LANGUAGE files only.\n\
	regexfile is a file containing one REGEXP per line.\n\
	REGEXP is anchored (as if preceded by ^).\n\
	The form /REGEXP/NAME/ creates a named tag.\n\
	For example Tcl named tags can be created with:\n\
	--regex=/proc[ \\t]+\\([^ \\t]+\\)/\\1/.");
  puts ("-c /REGEXP/, --ignore-case-regex=/REGEXP/ or --ignore-case-regex=@regexfile\n\
        Like -r, --regex but ignore case when matching expressions.");
  puts ("-R, --no-regex\n\
        Don't create tags from regexps for the following files.");
#endif /* ETAGS_REGEXPS */
  puts ("-o FILE, --output=FILE\n\
        Write the tags to FILE.");
  puts ("-I, --ignore-indentation\n\
        Don't rely on indentation quite as much as normal.  Currently,\n\
        this means not to assume that a closing brace in the first\n\
        column is the final brace of a function or structure\n\
        definition in C and C++.");

  if (CTAGS)
    {
      puts ("-t, --typedefs\n\
        Generate tag entries for C and Ada typedefs.");
      puts ("-T, --typedefs-and-c++\n\
        Generate tag entries for C typedefs, C struct/enum/union tags,\n\
        and C++ member functions.");
      puts ("-u, --update\n\
        Update the tag entries for the given files, leaving tag\n\
        entries for other files in place.  Currently, this is\n\
        implemented by deleting the existing entries for the given\n\
        files and then rewriting the new entries at the end of the\n\
        tags file.  It is often faster to simply rebuild the entire\n\
        tag file than to use this.");
      puts ("-v, --vgrind\n\
        Generates an index of items intended for human consumption,\n\
        similar to the output of vgrind.  The index is sorted, and\n\
        gives the page number of each item.");
      puts ("-w, --no-warn\n\
        Suppress warning messages about entries defined in multiple\n\
        files.");
      puts ("-x, --cxref\n\
        Like --vgrind, but in the style of cxref, rather than vgrind.\n\
        The output uses line numbers instead of page numbers, but\n\
        beyond that the differences are cosmetic; try both to see\n\
        which you like.");
    }

  puts ("-V, --version\n\
        Print the version of the program.\n\
-h, --help\n\
        Print this help message.");

  print_language_names ();

  puts ("");
  puts ("Report bugs to [e-mail]");

  exit (GOOD);
}


enum argument_type
{
  at_language,
  at_regexp,
  at_filename,
  at_icregexp
};

/* This structure helps us allow mixing of --lang and file names. */
typedef struct
{
  enum argument_type arg_type;
  char *what;
  language *lang;		/* language of the regexp */
} argument;

#ifdef VMS			/* VMS specific functions */

#define	EOS	'\0'

/* This is a BUG!  ANY arbitrary limit is a BUG!
   Won't someone please fix this?  */
#define	MAX_FILE_SPEC_LEN	255
typedef struct	{
  short   curlen;
  char    body[MAX_FILE_SPEC_LEN + 1];
} vspec;

/*
 v1.05 nmm 26-Jun-86 fn_exp - expand specification of list of file names
 returning in each successive call the next file name matching the input
 spec. The function expects that each in_spec passed
 to it will be processed to completion; in particular, up to and
 including the call following that in which the last matching name
 is returned, the function ignores the value of in_spec, and will
 only start processing a new spec with the following call.
 If an error occurs, on return out_spec contains the value
 of in_spec when the error occurred.

 With each successive file name returned in out_spec, the
 function's return value is one. When there are no more matching
 names the function returns zero. If on the first call no file
 matches in_spec, or there is any other error, -1 is returned.
*/

#include	<rmsdef.h>
#include	<descrip.h>
#define		OUTSIZE	MAX_FILE_SPEC_LEN
static short
fn_exp (out, in)
     vspec *out;
     char *in;
{
  static long context = 0;
  static struct dsc$descriptor_s o;
  static struct dsc$descriptor_s i;
  static bool pass1 = TRUE;
  long status;
  short retval;

  if (pass1)
    {
      pass1 = FALSE;
      o.dsc$a_pointer = (char *) out;
      o.dsc$w_length = (short)OUTSIZE;
      i.dsc$a_pointer = in;
      i.dsc$w_length = (short)strlen(in);
      i.dsc$b_dtype = DSC$K_DTYPE_T;
      i.dsc$b_class = DSC$K_CLASS_S;
      o.dsc$b_dtype = DSC$K_DTYPE_VT;
      o.dsc$b_class = DSC$K_CLASS_VS;
    }
  if ((status = lib$find_file(&i, &o, &context, 0, 0)) == RMS$_NORMAL)
    {
      out->body[out->curlen] = EOS;
      return 1;
    }
  else if (status == RMS$_NMF)
    retval = 0;
  else
    {
      strcpy(out->body, in);
      retval = -1;
    }
  lib$find_file_end(&context);
  pass1 = TRUE;
  return retval;
}

/*
  v1.01 nmm 19-Aug-85 gfnames - return in successive calls the
  name of each file specified by the provided arg expanding wildcards.
*/
static char *
gfnames (arg, p_error)
     char *arg;
     bool *p_error;
{
  static vspec filename = {MAX_FILE_SPEC_LEN, "\0"};

  switch (fn_exp (&filename, arg))
    {
    case 1:
      *p_error = FALSE;
      return filename.body;
    case 0:
      *p_error = FALSE;
      return NULL;
    default:
      *p_error = TRUE;
      return filename.body;
    }
}

#ifndef OLD  /* Newer versions of VMS do provide `system'.  */
system (cmd)
     char *cmd;
{
  error ("%s", "system() function not implemented under VMS");
}
#endif

#define	VERSION_DELIM	';'
char *massage_name (s)
     char *s;
{
  char *start = s;

  for ( ; *s; s++)
    if (*s == VERSION_DELIM)
      {
	*s = EOS;
	break;
      }
    else
      *s = lowcase (*s);
  return start;
}
#endif /* VMS */


int
main (argc, argv)
     int argc;
     char *argv[];
{
  int i;
  unsigned int nincluded_files;
  char **included_files;
  char *this_file;
  argument *argbuffer;
  int current_arg, file_count;
  linebuffer filename_lb;
#ifdef VMS
  bool got_err;
#endif

#ifdef DOS_NT
  _fmode = O_BINARY;   /* all of files are treated as binary files */
#endif /* DOS_NT */

  progname = argv[0];
  nincluded_files = 0;
  included_files = xnew (argc, char *);
  current_arg = 0;
  file_count = 0;

  /* Allocate enough no matter what happens.  Overkill, but each one
     is small. */
  argbuffer = xnew (argc, argument);

#ifdef ETAGS_REGEXPS
  /* Set syntax for regular expression routines. */
  re_set_syntax (RE_SYNTAX_EMACS | RE_INTERVALS);
  /* Translation table for case-insensitive search. */
  for (i = 0; i < CHAR_SET_SIZE; i++)
    lc_trans[i] = lowcase (i);
#endif /* ETAGS_REGEXPS */

  /*
   * If etags, always find typedefs and structure tags.  Why not?
   * Also default is to find macro constants, enum constants and
   * global variables.
   */
  if (!CTAGS)
    {
      typedefs = typedefs_and_cplusplus = constantypedefs = TRUE;
      globals = TRUE;
      members = FALSE;
    }

  while (1)
    {
      int opt;
      char *optstring;

#ifdef ETAGS_REGEXPS
      optstring = "-aCdDf:Il:o:r:c:RStTi:BuvxwVhH";
#else
      optstring = "-aCdDf:Il:o:StTi:BuvxwVhH";
#endif /* ETAGS_REGEXPS */

#ifndef LONG_OPTIONS
      optstring = optstring + 1;
#endif /* LONG_OPTIONS */

      opt = getopt_long (argc, argv, optstring, longopts, 0);
      if (opt == EOF)
	break;

      switch (opt)
	{
	case 0:
	  /* If getopt returns 0, then it has already processed a
	     long-named option.  We should do nothing.  */
	  break;

	case 1:
	  /* This means that a file name has been seen.  Record it. */
	  argbuffer[current_arg].arg_type = at_filename;
	  argbuffer[current_arg].what = optarg;
	  ++current_arg;
	  ++file_count;
	  break;

	  /* Common options. */
	case 'a': append_to_tagfile = TRUE;	break;
	case 'C': cplusplus = TRUE;		break;
	case 'd': constantypedefs = TRUE;	break;
	case 'D': constantypedefs = FALSE;	break;
	case 'f':		/* for compatibility with old makefiles */
	case 'o':
	  if (tagfile)
	    {
	      error ("-o option may only be given once.", (char *)NULL);
	      suggest_asking_for_help ();
	    }
	  tagfile = optarg;
	  break;
	case 'I':
	case 'S':		/* for backward compatibility */
	  noindentypedefs = TRUE;
	  break;
	case 'l':
	  {
	    language *lang = get_language_from_name (optarg);
	    if (lang != NULL)
	      {
		argbuffer[current_arg].lang = lang;
		argbuffer[current_arg].arg_type = at_language;
		++current_arg;
	      }
	  }
	  break;
#ifdef ETAGS_REGEXPS
	case 'r':
	  argbuffer[current_arg].arg_type = at_regexp;
	  argbuffer[current_arg].what = optarg;
	  ++current_arg;
	  break;
	case 'R':
	  argbuffer[current_arg].arg_type = at_regexp;
	  argbuffer[current_arg].what = NULL;
	  ++current_arg;
	  break;
        case 'c':
	  argbuffer[current_arg].arg_type = at_icregexp;
	  argbuffer[current_arg].what = optarg;
	  ++current_arg;
	  break;
#endif /* ETAGS_REGEXPS */
	case 'V':
	  print_version ();
	  break;
	case 'h':
	case 'H':
	  print_help ();
	  break;
	case 't':
	  typedefs = TRUE;
	  break;
	case 'T':
	  typedefs = typedefs_and_cplusplus = TRUE;
	  break;
#if (!CTAGS)
	  /* Etags options */
	case 'i':
	  included_files[nincluded_files++] = optarg;
	  break;
#else /* CTAGS */
	  /* Ctags options. */
	case 'B': searchar = '?';	break;
	case 'u': update = TRUE;	break;
	case 'v': vgrind_style = TRUE;	/*FALLTHRU*/
	case 'x': cxref_style = TRUE;	break;
	case 'w': no_warnings = TRUE;	break;
#endif /* CTAGS */
	default:
	  suggest_asking_for_help ();
	}
    }

  for (; optind < argc; ++optind)
    {
      argbuffer[current_arg].arg_type = at_filename;
      argbuffer[current_arg].what = argv[optind];
      ++current_arg;
      ++file_count;
    }

  if (nincluded_files == 0 && file_count == 0)
    {
      error ("no input files specified.", (char *)NULL);
      suggest_asking_for_help ();
    }

  if (tagfile == NULL)
    tagfile = CTAGS ? "tags" : "TAGS";
  cwd = etags_getcwd ();	/* the current working directory */
  if (cwd[strlen (cwd) - 1] != '/')
    {
      char *oldcwd = cwd;
      cwd = concat (oldcwd, "/", "");
      free (oldcwd);
    }
  if (streq (tagfile, "-"))
    tagfiledir = cwd;
  else
    tagfiledir = absolute_dirname (tagfile, cwd);

  init ();			/* set up boolean "functions" */

  initbuffer (&lb);
  initbuffer (&token_name);
  initbuffer (&lbs[0].lb);
  initbuffer (&lbs[1].lb);
  initbuffer (&filename_lb);

  if (!CTAGS)
    {
      if (streq (tagfile, "-"))
	{
	  tagf = stdout;
#ifdef DOS_NT
	  /* Switch redirected `stdout' to binary mode (setting `_fmode'
	     doesn't take effect until after `stdout' is already open). */
	  if (!isatty (fileno (stdout)))
	    setmode (fileno (stdout), O_BINARY);
#endif /* DOS_NT */
	}
      else
	tagf = fopen (tagfile, append_to_tagfile ? "a" : "w");
      if (tagf == NULL)
	pfatal (tagfile);
    }

  /*
   * Loop through files finding functions.
   */
  for (i = 0; i < current_arg; ++i)
    {
      switch (argbuffer[i].arg_type)
	{
	case at_language:
	  forced_lang = argbuffer[i].lang;
	  break;
#ifdef ETAGS_REGEXPS
	case at_regexp:
	  analyse_regex (argbuffer[i].what, FALSE);
	  break;
	case at_icregexp:
	  analyse_regex (argbuffer[i].what, TRUE);
	  break;
#endif
	case at_filename:
#ifdef VMS
	  while ((this_file = gfnames (argbuffer[i].what, &got_err)) != NULL)
	    {
	      if (got_err)
		{
		  error ("can't find file %s\n", this_file);
		  argc--, argv++;
		}
	      else
		{
		  this_file = massage_name (this_file);
		}
#else
	      this_file = argbuffer[i].what;
#endif
	      /* Input file named "-" means read file names from stdin
		 (one per line) and use them. */
	      if (streq (this_file, "-"))
		while (readline_internal (&filename_lb, stdin) > 0)
		  process_file (filename_lb.buffer);
	      else
		process_file (this_file);
#ifdef VMS
	    }
#endif
	  break;
	}
    }

#ifdef ETAGS_REGEXPS
  free_patterns ();
#endif /* ETAGS_REGEXPS */

  if (!CTAGS)
    {
      while (nincluded_files-- > 0)
	fprintf (tagf, "\f\n%s,include\n", *included_files++);

      fclose (tagf);
      exit (GOOD);
    }

  /* If CTAGS, we are here.  process_file did not write the tags yet,
     because we want them ordered.  Let's do it now. */
  if (cxref_style)
    {
      put_entries (head);
      free_tree (head);
      head = NULL;
      exit (GOOD);
    }

  if (update)
    {
      char cmd[BUFSIZ];
      for (i = 0; i < current_arg; ++i)
	{
	  if (argbuffer[i].arg_type != at_filename)
	    continue;
	  sprintf (cmd,
		   "mv %s OTAGS;fgrep -v '\t%s\t' OTAGS >%s;rm OTAGS",
		   tagfile, argbuffer[i].what, tagfile);
	  if (system (cmd) != GOOD)
	    fatal ("failed to execute shell command", (char *)NULL);
	}
      append_to_tagfile = TRUE;
    }

  tagf = fopen (tagfile, append_to_tagfile ? "a" : "w");
  if (tagf == NULL)
    pfatal (tagfile);
  put_entries (head);
  free_tree (head);
  head = NULL;
  fclose (tagf);

  if (update)
    {
      char cmd[BUFSIZ];
      sprintf (cmd, "sort %s -o %s", tagfile, tagfile);
      exit (system (cmd));
    }
  return GOOD;
}



/*
 * Return a compressor given the file name.  If EXTPTR is non-zero,
 * return a pointer into FILE where the compressor-specific
 * extension begins.  If no compressor is found, NULL is returned
 * and EXTPTR is not significant.
 * Idea by Vladimir Alexiev <[e-mail]>
 */
static compressor *
get_compressor_from_suffix (file, extptr)
     char *file;
     char **extptr;
{
  compressor *compr;
  char *slash, *suffix;

  /* This relies on FN to be after canonicalize_filename,
     so we don't need to consider backslashes on DOS_NT.  */
  slash = etags_strrchr (file, '/');
  suffix = etags_strrchr (file, '.');
  if (suffix == NULL || suffix < slash)
    return NULL;
  if (extptr != NULL)
    *extptr = suffix;
  suffix += 1;
  /* Let those poor souls who live with DOS 8+3 file name limits get
     some solace by treating foo.cgz as if it were foo.c.gz, etc.
     Only the first do loop is run if not MSDOS */
  do
    {
      for (compr = compressors; compr->suffix != NULL; compr++)
	if (streq (compr->suffix, suffix))
	  return compr;
      if (!MSDOS)
	break;			/* do it only once: not really a loop */
      if (extptr != NULL)
	*extptr = ++suffix;
    } while (*suffix != '\0');
  return NULL;
}



/*
 * Return a language given the name.
 */
static language *
get_language_from_name (name)
     char *name;
{
  language *lang;

  if (name == NULL)
    error ("empty language name", (char *)NULL);
  else
    {
      for (lang = lang_names; lang->name != NULL; lang++)
	if (streq (name, lang->name))
	  return lang;
      error ("unknown language \"%s\"", name);
    }

  return NULL;
}


/*
 * Return a language given the interpreter name.
 */
static language *
get_language_from_interpreter (interpreter)
     char *interpreter;
{
  language *lang;
  char **iname;

  if (interpreter == NULL)
    return NULL;
  for (lang = lang_names; lang->name != NULL; lang++)
    if (lang->interpreters != NULL)
      for (iname = lang->interpreters; *iname != NULL; iname++)
	if (streq (*iname, interpreter))
	    return lang;

  return NULL;
}



/*
 * Return a language given the file name.
 */
static language *
get_language_from_suffix (file)
     char *file;
{
  language *lang;
  char **ext, *suffix;

  suffix = etags_strrchr (file, '.');
  if (suffix == NULL)
    return NULL;
  suffix += 1;
  for (lang = lang_names; lang->name != NULL; lang++)
    if (lang->suffixes != NULL)
      for (ext = lang->suffixes; *ext != NULL; ext++)
	if (streq (*ext, suffix))
	  return lang;
  return NULL;
}



/*
 * This routine is called on each file argument.
 */
static void
process_file (file)
     char *file;
{
  struct stat stat_buf;
  FILE *inf;
  compressor *compr;
  char *compressed_name, *uncompressed_name;
  char *ext, *real_name;


  canonicalize_filename (file);
  if (streq (file, tagfile) && !streq (tagfile, "-"))
    {
      error ("skipping inclusion of %s in self.", file);
      return;
    }
  if ((compr = get_compressor_from_suffix (file, &ext)) == NULL)
    {
      compressed_name = NULL;
      real_name = uncompressed_name = savestr (file);
    }
  else
    {
      real_name = compressed_name = savestr (file);
      uncompressed_name = savenstr (file, ext - file);
    }

  /* If the canonicalised uncompressed name has already be dealt with,
     skip it silently, else add it to the list. */
  {
    typedef struct processed_file
    {
      char *filename;
      struct processed_file *next;
    } processed_file;
    static processed_file *pf_head = NULL;
    register processed_file *fnp;

    for (fnp = pf_head; fnp != NULL; fnp = fnp->next)
      if (streq (uncompressed_name, fnp->filename))
	goto exit;
    fnp = pf_head;
    pf_head = xnew (1, struct processed_file);
    pf_head->filename = savestr (uncompressed_name);
    pf_head->next = fnp;
  }

  if (stat (real_name, &stat_buf) != 0)
    {
      /* Reset real_name and try with a different name. */
      real_name = NULL;
      if (compressed_name != NULL) /* try with the given suffix */
	{
	  if (stat (uncompressed_name, &stat_buf) == 0)
	    real_name = uncompressed_name;
	}
      else			/* try all possible suffixes */
	{
	  for (compr = compressors; compr->suffix != NULL; compr++)
	    {
	      compressed_name = concat (file, ".", compr->suffix);
	      if (stat (compressed_name, &stat_buf) != 0)
		{
		  if (MSDOS)
		    {
		      char *suf = compressed_name + strlen (file);
		      size_t suflen = strlen (compr->suffix) + 1;
		      for ( ; suf[1]; suf++, suflen--)
			{
			  memmove (suf, suf + 1, suflen);
			  if (stat (compressed_name, &stat_buf) == 0)
			    {
			      real_name = compressed_name;
			      break;
			    }
			}
		      if (real_name != NULL)
			break;
		    } /* MSDOS */
		  free (compressed_name);
		  compressed_name = NULL;
		}
	      else
		{
		  real_name = compressed_name;
		  break;
		}
	    }
	}
      if (real_name == NULL)
	{
	  perror (file);
	  goto exit;
	}
    } /* try with a different name */

  if (!S_ISREG (stat_buf.st_mode))
    {
      error ("skipping %s: it is not a regular file.", real_name);
      goto exit;
    }
  if (real_name == compressed_name)
    {
      char *cmd = concat (compr->command, " ", real_name);
      inf = popen (cmd, "r");
      free (cmd);
    }
  else
    inf = fopen (real_name, "r");
  if (inf == NULL)
    {
      perror (real_name);
      goto exit;
    }

  find_entries (uncompressed_name, inf);

  if (real_name == compressed_name)
    pclose (inf);
  else
    fclose (inf);

  if (!CTAGS)
    {
      char *filename;

      if (filename_is_absolute (uncompressed_name))
	{
	  /* file is an absolute file name.  Canonicalise it. */
	  filename = absolute_filename (uncompressed_name, cwd);
	}
      else
	{
	  /* file is a file name relative to cwd.  Make it relative
	     to the directory of the tags file. */
	  filename = relative_filename (uncompressed_name, tagfiledir);
	}
      fprintf (tagf, "\f\n%s,%d\n", filename, total_size_of_entries (head));
      free (filename);
      put_entries (head);
      free_tree (head);
      head = NULL;
    }

 exit:
  if (compressed_name) free(compressed_name);
  if (uncompressed_name) free(uncompressed_name);
  return;
}

/*
 * This routine sets up the boolean pseudo-functions which work
 * by setting boolean flags dependent upon the corresponding character.
 * Every char which is NOT in that string is not a white char.  Therefore,
 * all of the array "_wht" is set to FALSE, and then the elements
 * subscripted by the chars in "white" are set to TRUE.  Thus "_wht"
 * of a char is TRUE if it is the string "white", else FALSE.
 */
static void
init ()
{
  register char *sp;
  register int i;

  for (i = 0; i < CHARS; i++)
    iswhite(i) = notinname(i) = begtoken(i) = intoken(i) = endtoken(i) = FALSE;
  for (sp = white; *sp != '\0'; sp++) iswhite (*sp) = TRUE;
  for (sp = nonam; *sp != '\0'; sp++) notinname (*sp) = TRUE;
  notinname('\0') = notinname('\n');
  for (sp = begtk; *sp != '\0'; sp++) begtoken (*sp) = TRUE;
  begtoken('\0') = begtoken('\n');
  for (sp = midtk; *sp != '\0'; sp++) intoken (*sp) = TRUE;
  intoken('\0') = intoken('\n');
  for (sp = endtk; *sp != '\0'; sp++) endtoken (*sp) = TRUE;
  endtoken('\0') = endtoken('\n');
}

/*
 * This routine opens the specified file and calls the function
 * which finds the function and type definitions.
 */
node *last_node = NULL;

static void
find_entries (file, inf)
     char *file;
     FILE *inf;
{
  char *cp;
  language *lang;
  node *old_last_node;

  /* Memory leakage here: the string pointed by curfile is
     never released, because curfile is copied into np->file
     for each node, to be used in CTAGS mode.  The amount of
     memory leaked here is the sum of the lengths of the
     file names. */
  curfile = savestr (file);

  /* If user specified a language, use it. */
  lang = forced_lang;
  if (lang != NULL && lang->function != NULL)
    {
      curlang = lang;
      lang->function (inf);
      return;
    }

  /* Try to guess the language given the file name. */
  lang = get_language_from_suffix (file);
  if (lang != NULL && lang->function != NULL)
    {
      curlang = lang;
      lang->function (inf);
      return;
    }

  /* Look for sharp-bang as the first two characters. */
  if (readline_internal (&lb, inf) > 0
      && lb.len >= 2
      && lb.buffer[0] == '#'
      && lb.buffer[1] == '!')
    {
      char *lp;

      /* Set lp to point at the first char after the last slash in the
         line or, if no slashes, at the first nonblank.  Then set cp to
	 the first successive blank and terminate the string. */
      lp = etags_strrchr (lb.buffer+2, '/');
      if (lp != NULL)
	lp += 1;
      else
	lp = skip_spaces (lb.buffer + 2);
      cp = skip_non_spaces (lp);
      *cp = '\0';

      if (strlen (lp) > 0)
	{
	  lang = get_language_from_interpreter (lp);
	  if (lang != NULL && lang->function != NULL)
	    {
	      curlang = lang;
	      lang->function (inf);
	      return;
	    }
	}
    }
  /* We rewind here, even if inf may be a pipe.  We fail if the
     length of the first line is longer than the pipe block size,
     which is unlikely. */
  rewind (inf);

  /* Try Fortran. */
  old_last_node = last_node;
  curlang = get_language_from_name ("fortran");
  Fortran_functions (inf);

  /* No Fortran entries found.  Try C. */
  if (old_last_node == last_node)
    {
      /* We do not tag if rewind fails.
	 Only the file name will be recorded in the tags file. */
      rewind (inf);
      curlang = get_language_from_name (cplusplus ? "c++" : "c");
      default_C_entries (inf);
    }
  return;
}

/* Record a tag. */
static void
pfnote (name, is_func, linestart, linelen, lno, cno)
     char *name;		/* tag name, or NULL if unnamed */
     bool is_func;		/* tag is a function */
     char *linestart;		/* start of the line where tag is */
     int linelen;		/* length of the line where tag is */
     int lno;			/* line number */
     long cno;			/* character number */
{
  register node *np;

  if (CTAGS && name == NULL)
    return;

  np = xnew (1, node);

  /* If ctags mode, change name "main" to M<thisfilename>. */
  if (CTAGS && !cxref_style && streq (name, "main"))
    {
      register char *fp = etags_strrchr (curfile, '/');
      np->name = concat ("M", fp == NULL ? curfile : fp + 1, "");
      fp = etags_strrchr (np->name, '.');
      if (fp != NULL && fp[1] != '\0' && fp[2] == '\0')
	fp[0] = '\0';
    }
  else
    np->name = name;
  np->been_warned = FALSE;
  np->file = curfile;
  np->is_func = is_func;
  np->lno = lno;
  /* Our char numbers are 0-base, because of C language tradition?
     ctags compatibility?  old versions compatibility?   I don't know.
     Anyway, since emacs's are 1-base we expect etags.el to take care
     of the difference.  If we wanted to have 1-based numbers, we would
     uncomment the +1 below. */
  np->cno = cno /* + 1 */ ;
  np->left = np->right = NULL;
  if (CTAGS && !cxref_style)
    {
      if (strlen (linestart) < 50)
	np->pat = concat (linestart, "$", "");
      else
	np->pat = savenstr (linestart, 50);
    }
  else
    np->pat = savenstr (linestart, linelen);

  add_node (np, &head);
}

/* Date: Wed, 22 Jan 1997 02:56:31 -0500 [last amended 18 Sep 1997]
 * From: Sam Kendall <[e-mail]>
 * Subject: Proposal for firming up the TAGS format specification
 * To: [e-mail]
 *
 * pfnote should emit the optimized form [unnamed tag] only if:
 *  1. name does not contain any of the characters " \t\r\n(),;";
 *  2. linestart contains name as either a rightmost, or rightmost but
 *     one character, substring;
 *  3. the character, if any, immediately before name in linestart must
 *     be one of the characters " \t(),;";
 *  4. the character, if any, immediately after name in linestart must
 *     also be one of the characters " \t(),;".
 *
 * The real implementation uses the notinname() macro, which recognises
 * characters slightly different form " \t\r\n(),;".  See the variable
 * `nonam'.
 */
#define traditional_tag_style TRUE
static void
new_pfnote (name, namelen, is_func, linestart, linelen, lno, cno)
     char *name;		/* tag name, or NULL if unnamed */
     int namelen;		/* tag length */
     bool is_func;		/* tag is a function */
     char *linestart;		/* start of the line where tag is */
     int linelen;		/* length of the line where tag is */
     int lno;			/* line number */
     long cno;			/* character number */
{
  register char *cp;
  bool named;

  named = TRUE;
  if (!CTAGS)
    {
      for (cp = name; !notinname (*cp); cp++)
	continue;
      if (*cp == '\0')				/* rule #1 */
	{
	  cp = linestart + linelen - namelen;
	  if (notinname (linestart[linelen-1]))
	    cp -= 1;				/* rule #4 */
	  if (cp >= linestart			/* rule #2 */
	      && (cp == linestart
		  || notinname (cp[-1]))	/* rule #3 */
	      && strneq (name, cp, namelen))	/* rule #2 */
	    named = FALSE;	/* use unnamed tag */
	}
    }

  if (named)
    name = savenstr (name, namelen);
  else
    name = NULL;
  pfnote (name, is_func, linestart, linelen, lno, cno);
}

/*
 * free_tree ()
 *	recurse on left children, iterate on right children.
 */
static void
free_tree (np)
     register node *np;
{
  while (np)
    {
      register node *node_right = np->right;
      free_tree (np->left);
      if (np->name != NULL)
	free (np->name);
      free (np->pat);
      free (np);
      np = node_right;
    }
}

/*
 * add_node ()
 *	Adds a node to the tree of nodes.  In etags mode, we don't keep
 *	it sorted; we just keep a linear list.  In ctags mode, maintain
 *	an ordered tree, with no attempt at balancing.
 *
 *	add_node is the only function allowed to add nodes, so it can
 *	maintain state.
 */
static void
add_node (np, cur_node_p)
     node *np, **cur_node_p;
{
  register int dif;
  register node *cur_node = *cur_node_p;

  if (cur_node == NULL)
    {
      *cur_node_p = np;
      last_node = np;
      return;
    }

  if (!CTAGS)
    {
      /* Etags Mode */
      if (last_node == NULL)
	fatal ("internal error in add_node", (char *)NULL);
      last_node->right = np;
      last_node = np;
    }
  else
    {
      /* Ctags Mode */
      dif = strcmp (np->name, cur_node->name);

      /*
       * If this tag name matches an existing one, then
       * do not add the node, but maybe print a warning.
       */
      if (!dif)
	{
	  if (streq (np->file, cur_node->file))
	    {
	      if (!no_warnings)
		{
		  fprintf (stderr, "Duplicate entry in file %s, line %d: %s\n",
			   np->file, lineno, np->name);
		  fprintf (stderr, "Second entry ignored\n");
		}
	    }
	  else if (!cur_node->been_warned && !no_warnings)
	    {
	      fprintf
		(stderr,
		 "Duplicate entry in files %s and %s: %s (Warning only)\n",
		 np->file, cur_node->file, np->name);
	      cur_node->been_warned = TRUE;
	    }
	  return;
	}

      /* Actually add the node */
      add_node (np, dif < 0 ? &cur_node->left : &cur_node->right);
    }
}

static void
put_entries (np)
     register node *np;
{
  register char *sp;

  if (np == NULL)
    return;

  /* Output subentries that precede this one */
  put_entries (np->left);

  /* Output this entry */

  if (!CTAGS)
    {
      if (np->name != NULL)
	fprintf (tagf, "%s\177%s\001%d,%ld\n",
		 np->pat, np->name, np->lno, np->cno);
      else
	fprintf (tagf, "%s\177%d,%ld\n",
		 np->pat, np->lno, np->cno);
    }
  else
    {
      if (np->name == NULL)
	error ("internal error: NULL name in ctags mode.", (char *)NULL);

      if (cxref_style)
	{
	  if (vgrind_style)
	    fprintf (stdout, "%s %s %d\n",
		     np->name, np->file, (np->lno + 63) / 64);
	  else
	    fprintf (stdout, "%-16s %3d %-16s %s\n",
		     np->name, np->lno, np->file, np->pat);
	}
      else
	{
	  fprintf (tagf, "%s\t%s\t", np->name, np->file);

	  if (np->is_func)
	    {			/* a function */
	      putc (searchar, tagf);
	      putc ('^', tagf);

	      for (sp = np->pat; *sp; sp++)
		{
		  if (*sp == '\\' || *sp == searchar)
		    putc ('\\', tagf);
		  putc (*sp, tagf);
		}
	      putc (searchar, tagf);
	    }
	  else
	    {			/* a typedef; text pattern inadequate */
	      fprintf (tagf, "%d", np->lno);
	    }
	  putc ('\n', tagf);
	}
    }

  /* Output subentries that follow this one */
  put_entries (np->right);
}

/* Length of a number's decimal representation. */
static int
number_len (num)
     long num;
{
  int len = 1;
  while ((num /= 10) > 0)
    len += 1;
  return len;
}

/*
 * Return total number of characters that put_entries will output for
 * the nodes in the subtree of the specified node.  Works only if
 * we are not ctags, but called only in that case.  This count
 * is irrelevant with the new tags.el, but is still supplied for
 * backward compatibility.
 */
static int
total_size_of_entries (np)
     register node *np;
{
  register int total;

  if (np == NULL)
    return 0;

  for (total = 0; np != NULL; np = np->right)
    {
      /* Count left subentries. */
      total += total_size_of_entries (np->left);

      /* Count this entry */
      total += strlen (np->pat) + 1;
      total += number_len ((long) np->lno) + 1 + number_len (np->cno) + 1;
      if (np->name != NULL)
	total += 1 + strlen (np->name);	/* \001name */
    }

  return total;
}

/*
 * The C symbol tables.
 */
enum sym_type
{
  st_none,
  st_C_objprot, st_C_objimpl, st_C_objend,
  st_C_gnumacro,
  st_C_ignore,
  st_C_javastruct,
  st_C_operator,
  st_C_struct, st_C_extern, st_C_enum, st_C_define, st_C_typedef, st_C_typespec
};

static unsigned int hash P_((const char *, unsigned int));
static struct C_stab_entry * in_word_set P_((const char *, unsigned int));
static enum sym_type C_symtype P_((char *, int, int));

/* Feed stuff between (but not including) %[ and %] lines to:
      gperf -c -k 1,3 -o -p -r -t
%[
struct C_stab_entry { char *name; int c_ext; enum sym_type type; }
%%
if,		0,	st_C_ignore
for,		0,	st_C_ignore
while,		0,	st_C_ignore
switch,		0,	st_C_ignore
return,		0,	st_C_ignore
@interface,	0,	st_C_objprot
@protocol,	0,	st_C_objprot
@implementation,0,	st_C_objimpl
@end,		0,	st_C_objend
import,		C_JAVA,	st_C_ignore
package,	C_JAVA,	st_C_ignore
friend,		C_PLPL,	st_C_ignore
extends,  	C_JAVA,	st_C_javastruct
implements,  	C_JAVA,	st_C_javastruct
interface,	C_JAVA, st_C_struct
class,  	C_PLPL,	st_C_struct
namespace,	C_PLPL,	st_C_struct
domain, 	C_STAR,	st_C_struct
union,  	0,	st_C_struct
struct, 	0,	st_C_struct
extern,  	0,	st_C_extern
enum,    	0,	st_C_enum
typedef, 	0,	st_C_typedef
define,  	0,	st_C_define
operator,	C_PLPL, st_C_operator
bool,		C_PLPL,	st_C_typespec
long,    	0,	st_C_typespec
short,   	0,	st_C_typespec
int,     	0,	st_C_typespec
char,    	0,	st_C_typespec
float,   	0,	st_C_typespec
double,  	0,	st_C_typespec
signed,  	0,	st_C_typespec
unsigned,	0,	st_C_typespec
auto,    	0,	st_C_typespec
void,    	0,	st_C_typespec
static,  	0,	st_C_typespec
const,   	0,	st_C_typespec
volatile,	0,	st_C_typespec
explicit,	C_PLPL,	st_C_typespec
mutable,	C_PLPL,	st_C_typespec
typename,	C_PLPL,	st_C_typespec
# DEFUN used in emacs, the next three used in glibc (SYSCALL only for mach).
DEFUN,		0,	st_C_gnumacro
SYSCALL,	0,	st_C_gnumacro
ENTRY,		0,	st_C_gnumacro
PSEUDO,		0,	st_C_gnumacro
# These are defined inside C functions, so currently they are not met.
# EXFUN used in glibc, DEFVAR_* in emacs.
#EXFUN,		0,	st_C_gnumacro
#DEFVAR_,	0,	st_C_gnumacro
%]
and replace lines between %< and %> with its output. */
/*%<*/
/* C code produced by gperf version 2.7.1 (19981006 egcs) */
/* Command-line: gperf -c -k 1,3 -o -p -r -t  */
struct C_stab_entry { char *name; int c_ext; enum sym_type type; };

#define TOTAL_KEYWORDS 46
#define MIN_WORD_LENGTH 2
#define MAX_WORD_LENGTH 15
#define MIN_HASH_VALUE 13
#define MAX_HASH_VALUE 123
/* maximum key range = 111, duplicates = 0 */

#ifdef __GNUC__
__inline
#endif
static unsigned int
hash (str, len)
     register const char *str;
     register unsigned int len;
{
  static unsigned char asso_values[] =
    {
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124,   3, 124, 124, 124,  43,   6,
       11, 124, 124, 124, 124, 124, 124, 124, 124, 124,
       11, 124, 124,  58,   7, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124,  57,   7,  42,
        4,  14,  52,   0, 124,  53, 124, 124,  29,  11,
        6,  35,  32, 124,  29,  34,  59,  58,  51,  24,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124, 124, 124, 124, 124,
      124, 124, 124, 124, 124, 124
    };
  register int hval = len;

  switch (hval)
    {
      default:
      case 3:
        hval += asso_values[(unsigned char)str[2]];
      case 2:
      case 1:
        hval += asso_values[(unsigned char)str[0]];
        break;
    }
  return hval;
}

#ifdef __GNUC__
__inline
#endif
static struct C_stab_entry *
in_word_set (str, len)
     register const char *str;
     register unsigned int len;
{
  static struct C_stab_entry wordlist[] =
    {
      {""}, {""}, {""}, {""}, {""}, {""}, {""}, {""}, {""},
      {""}, {""}, {""}, {""},
      {"@end",		0,	st_C_objend},
      {""}, {""}, {""}, {""},
      {"ENTRY",		0,	st_C_gnumacro},
      {"@interface",	0,	st_C_objprot},
      {""},
      {"domain", 	C_STAR,	st_C_struct},
      {""},
      {"PSEUDO",		0,	st_C_gnumacro},
      {""}, {""},
      {"namespace",	C_PLPL,	st_C_struct},
      {""}, {""},
      {"@implementation",0,	st_C_objimpl},
      {""}, {""}, {""}, {""}, {""}, {""}, {""}, {""}, {""},
      {"long",    	0,	st_C_typespec},
      {"signed",  	0,	st_C_typespec},
      {"@protocol",	0,	st_C_objprot},
      {""}, {""}, {""}, {""},
      {"bool",		C_PLPL,	st_C_typespec},
      {""}, {""}, {""}, {""}, {""}, {""},
      {"const",   	0,	st_C_typespec},
      {"explicit",	C_PLPL,	st_C_typespec},
      {"if",		0,	st_C_ignore},
      {""},
      {"operator",	C_PLPL, st_C_operator},
      {""},
      {"DEFUN",		0,	st_C_gnumacro},
      {""}, {""},
      {"define",  	0,	st_C_define},
      {""}, {""}, {""}, {""}, {""},
      {"double",  	0,	st_C_typespec},
      {"struct", 	0,	st_C_struct},
      {""}, {""}, {""}, {""},
      {"short",   	0,	st_C_typespec},
      {""},
      {"enum",    	0,	st_C_enum},
      {"mutable",	C_PLPL,	st_C_typespec},
      {""},
      {"extern",  	0,	st_C_extern},
      {"extends",  	C_JAVA,	st_C_javastruct},
      {"package",	C_JAVA,	st_C_ignore},
      {"while",		0,	st_C_ignore},
      {""},
      {"for",		0,	st_C_ignore},
      {""}, {""}, {""},
      {"volatile",	0,	st_C_typespec},
      {""}, {""},
      {"import",		C_JAVA,	st_C_ignore},
      {"float",   	0,	st_C_typespec},
      {"switch",		0,	st_C_ignore},
      {"return",		0,	st_C_ignore},
      {"implements",  	C_JAVA,	st_C_javastruct},
      {""},
      {"static",  	0,	st_C_typespec},
      {"typedef", 	0,	st_C_typedef},
      {"typename",	C_PLPL,	st_C_typespec},
      {"unsigned",	0,	st_C_typespec},
      {""}, {""},
      {"char",    	0,	st_C_typespec},
      {"class",  	C_PLPL,	st_C_struct},
      {""}, {""}, {""},
      {"void",    	0,	st_C_typespec},
      {""}, {""},
      {"friend",		C_PLPL,	st_C_ignore},
      {""}, {""}, {""},
      {"int",     	0,	st_C_typespec},
      {"union",  	0,	st_C_struct},
      {""}, {""}, {""},
      {"auto",    	0,	st_C_typespec},
      {"interface",	C_JAVA, st_C_struct},
      {""},
      {"SYSCALL",	0,	st_C_gnumacro}
    };

  if (len <= MAX_WORD_LENGTH && len >= MIN_WORD_LENGTH)
    {
      register int key = hash (str, len);

      if (key <= MAX_HASH_VALUE && key >= 0)
        {
          register const char *s = wordlist[key].name;

          if (*str == *s && !strncmp (str + 1, s + 1, len - 1))
            return &wordlist[key];
        }
    }
  return 0;
}
/*%>*/

static enum sym_type
C_symtype (str, len, c_ext)
     char *str;
     int len;
     int c_ext;
{
  register struct C_stab_entry *se = in_word_set (str, len);

  if (se == NULL || (se->c_ext && !(c_ext & se->c_ext)))
    return st_none;
  return se->type;
}

 /*
  * C functions and variables are recognized using a simple
  * finite automaton.  fvdef is its state variable.
  */
enum
{
  fvnone,			/* nothing seen */
  foperator,			/* func: operator keyword seen (cplpl) */
  fvnameseen,			/* function or variable name seen */
  fstartlist,			/* func: just after open parenthesis */
  finlist,			/* func: in parameter list */
  flistseen,			/* func: after parameter list */
  fignore,			/* func: before open brace */
  vignore			/* var-like: ignore until ';' */
} fvdef;

bool fvextern;			/* func or var: extern keyword seen; */

 /*
  * typedefs are recognized using a simple finite automaton.
  * typdef is its state variable.
  */
enum
{
  tnone,			/* nothing seen */
  tkeyseen,			/* typedef keyword seen */
  ttypeseen,			/* defined type seen */
  tinbody,			/* inside typedef body */
  tend,				/* just before typedef tag */
  tignore			/* junk after typedef tag */
} typdef;


 /*
  * struct-like structures (enum, struct and union) are recognized
  * using another simple finite automaton.  `structdef' is its state
  * variable.
  */
enum
{
  snone,			/* nothing seen yet */
  skeyseen,			/* struct-like keyword seen */
  stagseen,			/* struct-like tag seen */
  scolonseen,			/* colon seen after struct-like tag */
  sinbody			/* in struct body: recognize member func defs*/
} structdef;

/*
 * When structdef is stagseen, scolonseen, or sinbody, structtag is the
 * struct tag, and structtype is the type of the preceding struct-like
 * keyword.
 */
char *structtag = "<uninited>";
enum sym_type structtype;

/*
 * When objdef is different from onone, objtag is the name of the class.
 */
char *objtag = "<uninited>";

/*
 * Yet another little state machine to deal with preprocessor lines.
 */
enum
{
  dnone,			/* nothing seen */
  dsharpseen,			/* '#' seen as first char on line */
  ddefineseen,			/* '#' and 'define' seen */
  dignorerest			/* ignore rest of line */
} definedef;

/*
 * State machine for Objective C protocols and implementations.
 * Tom R.Hageman <[e-mail]>
 */
enum
{
  onone,			/* nothing seen */
  oprotocol,			/* @interface or @protocol seen */
  oimplementation,		/* @implementations seen */
  otagseen,			/* class name seen */
  oparenseen,			/* parenthesis before category seen */
  ocatseen,			/* category name seen */
  oinbody,			/* in @implementation body */
  omethodsign,			/* in @implementation body, after +/- */
  omethodtag,			/* after method name */
  omethodcolon,			/* after method colon */
  omethodparm,			/* after method parameter */
  oignore			/* wait for @end */
} objdef;


/*
 * Use this structure to keep info about the token read, and how it
 * should be tagged.  Used by the make_C_tag function to build a tag.
 */
typedef struct
{
  bool valid;
  char *str;
  bool named;
  int linelen;
  int lineno;
  long linepos;
  char *buffer;
} token;

token tok;			/* latest token read */

/*
 * Set this to TRUE, and the next token considered is called a function.
 * Used only for GNU emacs's function-defining macros.
 */
bool next_token_is_func;

/*
 * TRUE in the rules part of a yacc file, FALSE outside (parse as C).
 */
bool yacc_rules;

/*
 * methodlen is the length of the method name stored in token_name.
 */
int methodlen;

static bool consider_token P_((char *, int, int, int, int, int, bool *));
static void make_C_tag P_((bool));

/*
 * consider_token ()
 *	checks to see if the current token is at the start of a
 *	function or variable, or corresponds to a typedef, or
 * 	is a struct/union/enum tag, or #define, or an enum constant.
 *
 *	*IS_FUNC gets TRUE iff the token is a function or #define macro
 *	with args.  C_EXT is which language we are looking at.
 *
 * Globals
 *	fvdef			IN OUT
 *	structdef		IN OUT
 *	definedef		IN OUT
 *	typdef			IN OUT
 *	objdef			IN OUT
 *	next_token_is_func	IN OUT
 */

static bool
consider_token (str, len, c, c_ext, cblev, parlev, is_func_or_var)
     register char *str;	/* IN: token pointer */
     register int len;		/* IN: token length */
     register int c;		/* IN: first char after the token */
     int c_ext;			/* IN: C extensions mask */
     int cblev;			/* IN: curly brace level */
     int parlev;		/* IN: parenthesis level */
     bool *is_func_or_var;	/* OUT: function or variable found */
{
  enum sym_type toktype = C_symtype (str, len, c_ext);

  /*
   * Advance the definedef state machine.
   */
  switch (definedef)
    {
    case dnone:
      /* We're not on a preprocessor line. */
      break;
    case dsharpseen:
      if (toktype == st_C_define)
	{
	  definedef = ddefineseen;
	}
      else
	{
	  definedef = dignorerest;
	}
      return FALSE;
    case ddefineseen:
      /*
       * Make a tag for any macro, unless it is a constant
       * and constantypedefs is FALSE.
       */
      definedef = dignorerest;
      *is_func_or_var = (c == '(');
      if (!*is_func_or_var && !constantypedefs)
	return FALSE;
      else
	return TRUE;
    case dignorerest:
      return FALSE;
    default:
      error ("internal error: definedef value.", (char *)NULL);
    }

  /*
   * Now typedefs
   */
  switch (typdef)
    {
    case tnone:
      if (toktype == st_C_typedef)
	{
	  if (typedefs)
	    typdef = tkeyseen;
	  fvextern = FALSE;
	  fvdef = fvnone;
	  return FALSE;
	}
      break;
    case tkeyseen:
      switch (toktype)
	{
	case st_none:
	case st_C_typespec:
	case st_C_struct:
	case st_C_enum:
	  typdef = ttypeseen;
	  break;
	}
      /* Do not return here, so the structdef stuff has a chance. */
      break;
    case tend:
      switch (toktype)
	{
	case st_C_typespec:
	case st_C_struct:
	case st_C_enum:
	  return FALSE;
	}
      return TRUE;
    }

  /*
   * This structdef business is currently only invoked when cblev==0.
   * It should be recursively invoked whatever the curly brace level,
   * and a stack of states kept, to allow for definitions of structs
   * within structs.
   *
   * This structdef business is NOT invoked when we are ctags and the
   * file is plain C.  This is because a struct tag may have the same
   * name as another tag, and this loses with ctags.
   */
  switch (toktype)
    {
    case st_C_javastruct:
      if (structdef == stagseen)
        structdef = scolonseen;
      return FALSE;
    case st_C_struct:
    case st_C_enum:
      if (typdef == tkeyseen
	  || (typedefs_and_cplusplus && cblev == 0 && structdef == snone))
	{
	  structdef = skeyseen;
	  structtype = toktype;
	}
      return FALSE;
    }

  if (structdef == skeyseen)
    {
      /* Save the tag for struct/union/class, for functions and variables
	 that may be defined inside. */
      if (structtype == st_C_struct)
	structtag = savenstr (str, len);
      else
	structtag = "<enum>";
      structdef = stagseen;
      return TRUE;
    }

  if (typdef != tnone)
    definedef = dnone;

  /* Detect GNU macros.

     Writers of emacs code are recommended to put the
     first two args of a DEFUN on the same line.

      The DEFUN macro, used in emacs C source code, has a first arg
     that is a string (the lisp function name), and a second arg that
     is a C function name.  Since etags skips strings, the second arg
     is tagged.  This is unfortunate, as it would be better to tag the
     first arg.  The simplest way to deal with this problem would be
     to name the tag with a name built from the function name, by
     removing the initial 'F' character and substituting '-' for '_'.
     Anyway, this assumes that the conventions of naming lisp
     functions will never change.  Currently, this method is not
     implemented. */
  if (definedef == dnone && toktype == st_C_gnumacro)
    {
      next_token_is_func = TRUE;
      return FALSE;
    }
  if (next_token_is_func)
    {
      next_token_is_func = FALSE;
      fvdef = fignore;
      *is_func_or_var = TRUE;
      return TRUE;
    }

  /* Detect Objective C constructs. */
  switch (objdef)
    {
    case onone:
      switch (toktype)
	{
	case st_C_objprot:
	  objdef = oprotocol;
	  return FALSE;
	case st_C_objimpl:
	  objdef = oimplementation;
	  return FALSE;
	}
      break;
    case oimplementation:
      /* Save the class tag for functions or variables defined inside. */
      objtag = savenstr (str, len);
      objdef = oinbody;
      return FALSE;
    case oprotocol:
      /* Save the class tag for categories. */
      objtag = savenstr (str, len);
      objdef = otagseen;
      *is_func_or_var = TRUE;
      return TRUE;
    case oparenseen:
      objdef = ocatseen;
      *is_func_or_var = TRUE;
      return TRUE;
    case oinbody:
      break;
    case omethodsign:
      if (parlev == 0)
	{
	  objdef = omethodtag;
	  methodlen = len;
	  grow_linebuffer (&token_name, methodlen + 1);
	  strncpy (token_name.buffer, str, len);
	  token_name.buffer[methodlen] = '\0';
	  token_name.len = methodlen;
	  return TRUE;
	}
      return FALSE;
    case omethodcolon:
      if (parlev == 0)
	objdef = omethodparm;
      return FALSE;
    case omethodparm:
      if (parlev == 0)
	{
	  objdef = omethodtag;
	  methodlen += len;
	  grow_linebuffer (&token_name, methodlen + 1);
	  strncat (token_name.buffer, str, len);
	  token_name.len = methodlen;
	  return TRUE;
	}
      return FALSE;
    case oignore:
      if (toktype == st_C_objend)
	{
	  /* Memory leakage here: the string pointed by objtag is
	     never released, because many tests would be needed to
	     avoid breaking on incorrect input code.  The amount of
	     memory leaked here is the sum of the lengths of the
	     class tags.
	  free (objtag); */
	  objdef = onone;
	}
      return FALSE;
    }

  /* A function, variable or enum constant? */
  switch (toktype)
    {
    case st_C_extern:
      fvextern = TRUE;
      /* FALLTHRU */
    case st_C_typespec:
      if (fvdef != finlist && fvdef != fignore && fvdef != vignore)
	fvdef = fvnone;		/* should be useless */
      return FALSE;
    case st_C_ignore:
      fvextern = FALSE;
      fvdef = vignore;
      return FALSE;
    case st_C_operator:
      fvdef = foperator;
      *is_func_or_var = TRUE;
      return TRUE;
    case st_none:
      if ((c_ext & C_PLPL) && strneq (str+len-10, "::operator", 10))
	{
	  fvdef = foperator;
	  *is_func_or_var = TRUE;
	  return TRUE;
	}
      if (constantypedefs && structdef == sinbody && structtype == st_C_enum)
	return TRUE;
      if (fvdef == fvnone)
	{
	  fvdef = fvnameseen;	/* function or variable */
	  *is_func_or_var = TRUE;
	  return TRUE;
	}
      break;
    }

  return FALSE;
}

/*
 * C_entries ()
 *	This routine finds functions, variables, typedefs,
 * 	#define's, enum constants and struct/union/enum definitions in
 * 	C syntax and adds them to the list.
 */
#define current_lb_is_new (newndx == curndx)
#define switch_line_buffers() (curndx = 1 - curndx)

#define curlb (lbs[curndx].lb)
#define othlb (lbs[1-curndx].lb)
#define newlb (lbs[newndx].lb)
#define curlinepos (lbs[curndx].linepos)
#define othlinepos (lbs[1-curndx].linepos)
#define newlinepos (lbs[newndx].linepos)

#define CNL_SAVE_DEFINEDEF()						\
do {									\
  curlinepos = charno;							\
  lineno++;								\
  linecharno = charno;							\
  charno += readline (&curlb, inf);					\
  lp = curlb.buffer;							\
  quotednl = FALSE;							\
  newndx = curndx;							\
} while (0)

#define CNL()								\
do {									\
  CNL_SAVE_DEFINEDEF();							\
  if (savetok.valid)							\
    {									\
      tok = savetok;							\
      savetok.valid = FALSE;						\
    }									\
  definedef = dnone;							\
} while (0)


static void
make_C_tag (isfun)
     bool isfun;
{
  /* This function should never be called when tok.valid is FALSE, but
     we must protect against invalid input or internal errors. */
  if (tok.valid)
    {
      if (traditional_tag_style)
	{
	  /* This was the original code.  Now we call new_pfnote instead,
	     which uses the new method for naming tags (see new_pfnote). */
	  char *name = NULL;

	  if (CTAGS || tok.named)
	    name = savestr (token_name.buffer);
	  pfnote (name, isfun,
		  tok.buffer, tok.linelen, tok.lineno, tok.linepos);
	}
      else
	new_pfnote (token_name.buffer, token_name.len, isfun,
		    tok.buffer, tok.linelen, tok.lineno, tok.linepos);
      tok.valid = FALSE;
    }
  else if (DEBUG)
    abort ();
}


static void
C_entries (c_ext, inf)
     int c_ext;			/* extension of C */
     FILE *inf;			/* input file */
{
  register char c;		/* latest char read; '\0' for end of line */
  register char *lp;		/* pointer one beyond the character `c' */
  int curndx, newndx;		/* indices for current and new lb */
  register int tokoff;		/* offset in line of start of current token */
  register int toklen;		/* length of current token */
  char *qualifier;		/* string used to qualify names */
  int qlen;			/* length of qualifier */
  int cblev;			/* current curly brace level */
  int parlev;			/* current parenthesis level */
  bool incomm, inquote, inchar, quotednl, midtoken;
  bool purec, cplpl, cjava;
  token savetok;		/* token saved during preprocessor handling */


  tokoff = toklen = 0;		/* keep compiler quiet */
  curndx = newndx = 0;
  lineno = 0;
  charno = 0;
  lp = curlb.buffer;
  *lp = 0;

  fvdef = fvnone; fvextern = FALSE; typdef = tnone;
  structdef = snone; definedef = dnone; objdef = onone;
  next_token_is_func = yacc_rules = FALSE;
  midtoken = inquote = inchar = incomm = quotednl = FALSE;
  tok.valid = savetok.valid = FALSE;
  cblev = 0;
  parlev = 0;
  purec = !(c_ext & ~YACC);	/* no extensions (apart from possibly yacc) */
  cplpl = (c_ext & C_PLPL) == C_PLPL;
  cjava = (c_ext & C_JAVA) == C_JAVA;
  if (cjava)
    { qualifier = "."; qlen = 1; }
  else
    { qualifier = "::"; qlen = 2; }

  while (!feof (inf))
    {
      c = *lp++;
      if (c == '\\')
	{
	  /* If we're at the end of the line, the next character is a
	     '\0'; don't skip it, because it's the thing that tells us
	     to read the next line.  */
	  if (*lp == '\0')
	    {
	      quotednl = TRUE;
	      continue;
	    }
	  lp++;
	  c = ' ';
	}
      else if (incomm)
	{
	  switch (c)
	    {
	    case '*':
	      if (*lp == '/')
		{
		  c = *lp++;
		  incomm = FALSE;
		}
	      break;
	    case '\0':
	      /* Newlines inside comments do not end macro definitions in
		 traditional cpp. */
	      CNL_SAVE_DEFINEDEF ();
	      break;
	    }
	  continue;
	}
      else if (inquote)
	{
	  switch (c)
	    {
	    case '"':
	      inquote = FALSE;
	      break;
	    case '\0':
	      /* Newlines inside strings do not end macro definitions
		 in traditional cpp, even though compilers don't
		 usually accept them. */
	      CNL_SAVE_DEFINEDEF ();
	      break;
	    }
	  continue;
	}
      else if (inchar)
	{
	  switch (c)
	    {
	    case '\0':
	      /* Hmmm, something went wrong. */
	      CNL ();
	      /* FALLTHRU */
	    case '\'':
	      inchar = FALSE;
	      break;
	    }
	  continue;
	}
      else
	switch (c)
	  {
	  case '"':
	    inquote = TRUE;
	    if (fvdef != finlist && fvdef != fignore && fvdef !=vignore)
	      {
		fvextern = FALSE;
		fvdef = fvnone;
	      }
	    continue;
	  case '\'':
	    inchar = TRUE;
	    if (fvdef != finlist && fvdef != fignore && fvdef !=vignore)
	      {
		fvextern = FALSE;
		fvdef = fvnone;
	      }
	    continue;
	  case '/':
	    if (*lp == '*')
	      {
		lp++;
		incomm = TRUE;
		continue;
	      }
	    else if (/* cplpl && */ *lp == '/')
	      {
		c = '\0';
		break;
	      }
	    else
	      break;
	  case '%':
	    if ((c_ext & YACC) && *lp == '%')
	      {
		/* entering or exiting rules section in yacc file */
		lp++;
		definedef = dnone; fvdef = fvnone; fvextern = FALSE;
		typdef = tnone; structdef = snone;
		next_token_is_func = FALSE;
		midtoken = inquote = inchar = incomm = quotednl = FALSE;
		cblev = 0;
		yacc_rules = !yacc_rules;
		continue;
 	      }
	    else
	      break;
	  case '#':
	    if (definedef == dnone)
	      {
		char *cp;
		bool cpptoken = TRUE;

		/* Look back on this line.  If all blanks, or nonblanks
		   followed by an end of comment, this is a preprocessor
		   token. */
		for (cp = newlb.buffer; cp < lp-1; cp++)
		  if (!iswhite (*cp))
		    {
		      if (*cp == '*' && *(cp+1) == '/')
			{
			  cp++;
			  cpptoken = TRUE;
			}
		      else
			cpptoken = FALSE;
		    }
		if (cpptoken)
		  definedef = dsharpseen;
	      } /* if (definedef == dnone) */

	    continue;
	  } /* switch (c) */


      /* Consider token only if some complicated conditions are satisfied. */
      if ((definedef != dnone
	   || (cblev == 0 && structdef != scolonseen)
	   || (cblev == 1 && cplpl && structdef == sinbody)
	   || (structdef == sinbody && purec))
	  && typdef != tignore
	  && definedef != dignorerest
	  && fvdef != finlist)
	{
	  if (midtoken)
	    {
	      if (endtoken (c))
		{
		  bool funorvar = FALSE;

		  if (c == ':' && cplpl && *lp == ':' && begtoken (lp[1]))
		    {
		      /*
		       * This handles :: in the middle, but not at the
		       * beginning of an identifier.  Also, space-separated
		       * :: is not recognised.
		       */
		      lp += 2;
		      toklen += 2;
		      c = lp[-1];
		      goto intok;
		    }
		  else
		    {
		      if (yacc_rules
			  || consider_token (newlb.buffer + tokoff, toklen, c,
					     c_ext, cblev, parlev, &funorvar))
			{
			  if (fvdef == foperator)
			    {
			      char *oldlp = lp;
			      lp = skip_spaces (lp-1);
			      if (*lp != '\0')
				lp += 1;
			      while (*lp != '\0'
				     && !iswhite (*lp) && *lp != '(')
				lp += 1;
			      c = *lp++;
			      toklen += lp - oldlp;
			    }
			  tok.named = FALSE;
			  if (!purec
			      && funorvar
			      && definedef == dnone
			      && structdef == sinbody)
			    /* function or var defined in C++ class body */
			    {
			      int len = strlen (structtag) + qlen + toklen;
			      grow_linebuffer (&token_name, len + 1);
			      strcpy (token_name.buffer, structtag);
			      strcat (token_name.buffer, qualifier);
			      strncat (token_name.buffer,
				       newlb.buffer + tokoff, toklen);
			      token_name.len = len;
			      tok.named = TRUE;
			    }
			  else if (objdef == ocatseen)
			    /* Objective C category */
			    {
			      int len = strlen (objtag) + 2 + toklen;
			      grow_linebuffer (&token_name, len + 1);
			      strcpy (token_name.buffer, objtag);
			      strcat (token_name.buffer, "(");
			      strncat (token_name.buffer,
				       newlb.buffer + tokoff, toklen);
			      strcat (token_name.buffer, ")");
			      token_name.len = len;
			      tok.named = TRUE;
			    }
			  else if (objdef == omethodtag
				   || objdef == omethodparm)
			    /* Objective C method */
			    {
			      tok.named = TRUE;
			    }
			  else
			    {
			      grow_linebuffer (&token_name, toklen + 1);
			      strncpy (token_name.buffer,
				       newlb.buffer + tokoff, toklen);
			      token_name.buffer[toklen] = '\0';
			      token_name.len = toklen;
			      /* Name macros and members. */
			      tok.named = (structdef == stagseen
					   || typdef == ttypeseen
					   || typdef == tend
					   || (funorvar
					       && definedef == dignorerest)
					   || (funorvar
					       && definedef == dnone
					       && structdef == sinbody));
			    }
			  tok.lineno = lineno;
			  tok.linelen = tokoff + toklen + 1;
			  tok.buffer = newlb.buffer;
			  tok.linepos = newlinepos;
			  tok.valid = TRUE;

			  if (definedef == dnone
			      && (fvdef == fvnameseen
				  || fvdef == foperator
				  || structdef == stagseen
				  || typdef == tend
				  || objdef != onone))
			    {
			      if (current_lb_is_new)
				switch_line_buffers ();
			    }
			  else
			    make_C_tag (funorvar);
			}
		      midtoken = FALSE;
		    }
		} /* if (endtoken (c)) */
	      else if (intoken (c))
		intok:
		{
		  toklen++;
		  continue;
		}
	    } /* if (midtoken) */
	  else if (begtoken (c))
	    {
	      switch (definedef)
		{
		case dnone:
		  switch (fvdef)
		    {
		    case fstartlist:
		      fvdef = finlist;
		      continue;
		    case flistseen:
		      make_C_tag (TRUE); /* a function */
		      fvdef = fignore;
		      break;
		    case fvnameseen:
		      fvdef = fvnone;
		      break;
		    }
		  if (structdef == stagseen && !cjava)
		    structdef = snone;
		  break;
		case dsharpseen:
		  savetok = tok;
		}
	      if (!yacc_rules || lp == newlb.buffer + 1)
		{
		  tokoff = lp - 1 - newlb.buffer;
		  toklen = 1;
		  midtoken = TRUE;
		}
	      continue;
	    } /* if (begtoken) */
	} /* if must look at token */


      /* Detect end of line, colon, comma, semicolon and various braces
	 after having handled a token.*/
      switch (c)
	{
	case ':':
	  if (definedef != dnone)
	    break;
	  switch (objdef)
	    {
	    case  otagseen:
	      objdef = oignore;
	      make_C_tag (TRUE); /* an Objective C class */
	      break;
	    case omethodtag:
	    case omethodparm:
	      objdef = omethodcolon;
	      methodlen += 1;
	      grow_linebuffer (&token_name, methodlen + 1);
	      strcat (token_name.buffer, ":");
	      token_name.len = methodlen;
	      break;
	    }
	  if (structdef == stagseen)
	    structdef = scolonseen;
	  else
	    switch (fvdef)
	      {
	      case fvnameseen:
		if (yacc_rules)
		  {
		    make_C_tag (FALSE); /* a yacc function */
		    fvdef = fignore;
		  }
		break;
	      case fstartlist:
		fvextern = FALSE;
		fvdef = fvnone;
		break;
	      }
	  break;
	case ';':
	  if (definedef != dnone)
	    break;
	  if (cblev == 0)
	    switch (typdef)
	      {
	      case tend:
		make_C_tag (FALSE); /* a typedef */
		/* FALLTHRU */
	      default:
		typdef = tnone;
	      }
	  switch (fvdef)
	    {
	    case fignore:
	      break;
	    case fvnameseen:
	      if ((members && cblev == 1)
		  || (globals && cblev == 0 && (!fvextern || declarations)))
		make_C_tag (FALSE); /* a variable */
	      fvextern = FALSE;
	      fvdef = fvnone;
	      tok.valid = FALSE;
	      break;
	    case flistseen:
	      if (declarations && (cblev == 0 || cblev == 1))
		make_C_tag (TRUE); /* a function declaration */
	      /* FALLTHRU */
	    default:
	      fvextern = FALSE;
	      fvdef = fvnone;
	      /* The following instruction invalidates the token.
		 Probably the token should be invalidated in all
		 other cases  where some state machine is reset. */
	      tok.valid = FALSE;
	    }
	  if (structdef == stagseen)
	    structdef = snone;
	  break;
	case ',':
	  if (definedef != dnone)
	    break;
	  switch (objdef)
	    {
	    case omethodtag:
	    case omethodparm:
	      make_C_tag (TRUE); /* an Objective C method */
	      objdef = oinbody;
	      break;
	    }
	  switch (fvdef)
	    {
	    case foperator:
	    case finlist:
	    case fignore:
	    case vignore:
	      break;
	    case fvnameseen:
	      if ((members && cblev == 1)
		  || (globals && cblev == 0 && (!fvextern || declarations)))
		make_C_tag (FALSE); /* a variable */
	      break;
	    default:
	      fvdef = fvnone;
	    }
	  if (structdef == stagseen)
	    structdef = snone;
	  break;
	case '[':
	  if (definedef != dnone)
	    break;
	  if (cblev == 0 && typdef == tend)
	    {
	      typdef = tignore;
	      make_C_tag (FALSE);	/* a typedef */
	      break;
	    }
	  switch (fvdef)
	    {
	    case foperator:
	    case finlist:
	    case fignore:
	    case vignore:
	      break;
	    case fvnameseen:
	      if ((members && cblev == 1)
		  || (globals && cblev == 0 && (!fvextern || declarations)))
		make_C_tag (FALSE); /* a variable */
	      /* FALLTHRU */
	    default:
	      fvdef = fvnone;
	    }
	  if (structdef == stagseen)
	    structdef = snone;
	  break;
	case '(':
	  if (definedef != dnone)
	    break;
	  if (objdef == otagseen && parlev == 0)
	    objdef = oparenseen;
	  switch (fvdef)
	    {
	    case fvnameseen:
	      if (typdef == ttypeseen
		  && tok.valid
		  && *lp != '*'
		  && structdef != sinbody)
		{
		  /* This handles constructs like:
		     typedef void OperatorFun (int fun); */
		  make_C_tag (FALSE);
		  typdef = tignore;
		}
	      /* FALLTHRU */
	    case foperator:
	      fvdef = fstartlist;
	      break;
	    case flistseen:
	      fvdef = finlist;
	      break;
	    }
	  parlev++;
	  break;
	case ')':
	  if (definedef != dnone)
	    break;
	  if (objdef == ocatseen && parlev == 1)
	    {
	      make_C_tag (TRUE); /* an Objective C category */
	      objdef = oignore;
	    }
	  if (--parlev == 0)
	    {
	      switch (fvdef)
		{
		case fstartlist:
		case finlist:
		  fvdef = flistseen;
		  break;
		}
	      if (cblev == 0 && (typdef == tend))
		{
		  typdef = tignore;
		  make_C_tag (FALSE); /* a typedef */
		}
	    }
	  else if (parlev < 0)	/* can happen due to ill-conceived #if's. */
	    parlev = 0;
	  break;
	case '{':
	  if (definedef != dnone)
	    break;
	  if (typdef == ttypeseen)
	    typdef = tinbody;
	  switch (structdef)
	    {
	    case skeyseen:	/* unnamed struct */
	      structdef = sinbody;
	      structtag = "_anonymous_";
	      break;
	    case stagseen:
	    case scolonseen:	/* named struct */
	      structdef = sinbody;
	      make_C_tag (FALSE);	/* a struct */
	      break;
	    }
	  switch (fvdef)
	    {
	    case flistseen:
	      make_C_tag (TRUE); /* a function */
	      /* FALLTHRU */
	    case fignore:
	      fvdef = fvnone;
	      break;
	    case fvnone:
	      switch (objdef)
		{
		case otagseen:
		  make_C_tag (TRUE); /* an Objective C class */
		  objdef = oignore;
		  break;
		case omethodtag:
		case omethodparm:
		  make_C_tag (TRUE); /* an Objective C method */
		  objdef = oinbody;
		  break;
		default:
		  /* Neutralize `extern "C" {' grot. */
		  if (cblev == 0 && structdef == snone && typdef == tnone)
		    cblev = -1;
		}
	    }
	  cblev++;
	  break;
	case '*':
	  if (definedef != dnone)
	    break;
	  if (fvdef == fstartlist)
	    fvdef = fvnone;	/* avoid tagging `foo' in `foo (*bar()) ()' */
	  break;
	case '}':
	  if (definedef != dnone)
	    break;
	  if (!noindentypedefs && lp == newlb.buffer + 1)
	    {
	      cblev = 0;	/* reset curly brace level if first column */
	      parlev = 0;	/* also reset paren level, just in case... */
	    }
	  else if (cblev > 0)
	    cblev--;
	  if (cblev == 0)
	    {
	      if (typdef == tinbody)
		typdef = tend;
	      /* Memory leakage here: the string pointed by structtag is
	         never released, because I fear to miss something and
	         break things while freeing the area.  The amount of
	         memory leaked here is the sum of the lengths of the
	         struct tags.
	      if (structdef == sinbody)
		free (structtag); */

	      structdef = snone;
	      structtag = "<error>";
	    }
	  break;
	case '=':
	  if (definedef != dnone)
	    break;
	  switch (fvdef)
	    {
	    case foperator:
	    case finlist:
	    case fignore:
	    case vignore:
	      break;
	    case fvnameseen:
	      if ((members && cblev == 1)
		  || (globals && cblev == 0 && (!fvextern || declarations)))
		make_C_tag (FALSE); /* a variable */
	      /* FALLTHRU */
	    default:
	      fvdef = vignore;
	    }
	  break;
	case '+':
	case '-':
	  if (objdef == oinbody && cblev == 0)
	    {
	      objdef = omethodsign;
	      break;
	    }
	  /* FALLTHRU */
	case '#': case '~': case '&': case '%': case '/': case '|':
	case '^': case '!': case '<': case '>': case '.': case '?': case ']':
	  if (definedef != dnone)
	    break;
	  /* These surely cannot follow a function tag in C. */
	  switch (fvdef)
	    {
	    case foperator:
	    case finlist:
	    case fignore:
	    case vignore:
	      break;
	    default:
	      fvdef = fvnone;
	    }
	  break;
	case '\0':
	  if (objdef == otagseen)
	    {
	      make_C_tag (TRUE); /* an Objective C class */
	      objdef = oignore;
	    }
	  /* If a macro spans multiple lines don't reset its state. */
	  if (quotednl)
	    CNL_SAVE_DEFINEDEF ();
	  else
	    CNL ();
	  break;
	} /* switch (c) */

    } /* while not eof */
}

/*
 * Process either a C++ file or a C file depending on the setting
 * of a global flag.
 */
static void
default_C_entries (inf)
     FILE *inf;
{
  C_entries (cplusplus ? C_PLPL : 0, inf);
}

/* Always do plain ANSI C. */
static void
plain_C_entries (inf)
     FILE *inf;
{
  C_entries (0, inf);
}

/* Always do C++. */
static void
Cplusplus_entries (inf)
     FILE *inf;
{
  C_entries (C_PLPL, inf);
}

/* Always do Java. */
static void
Cjava_entries (inf)
     FILE *inf;
{
  C_entries (C_JAVA, inf);
}

/* Always do C*. */
static void
Cstar_entries (inf)
     FILE *inf;
{
  C_entries (C_STAR, inf);
}

/* Always do Yacc. */
static void
Yacc_entries (inf)
     FILE *inf;
{
  C_entries (YACC, inf);
}

/* A useful macro. */
#define LOOP_ON_INPUT_LINES(file_pointer, line_buffer, char_pointer)	\
  for (lineno = charno = 0;	/* loop initialization */		\
       !feof (file_pointer)	/* loop test */				\
       && (lineno++,		/* instructions at start of loop */	\
	   linecharno = charno,						\
	   charno += readline (&line_buffer, file_pointer),		\
	   char_pointer = lb.buffer,					\
	   TRUE);							\
      )


/*
 * Read a file, but do no processing.  This is used to do regexp
 * matching on files that have no language defined.
 */
static void
just_read_file (inf)
     FILE *inf;
{
  register char *dummy;

  LOOP_ON_INPUT_LINES (inf, lb, dummy)
    continue;
}

/* Fortran parsing */

static bool tail P_((char *));
static void takeprec P_((void));
static void getit P_((FILE *));

static bool
tail (cp)
     char *cp;
{
  register int len = 0;

  while (*cp != '\0' && lowcase (*cp) == lowcase (dbp[len]))
    cp++, len++;
  if (*cp == '\0' && !intoken (dbp[len]))
    {
      dbp += len;
      return TRUE;
    }
  return FALSE;
}

static void
takeprec ()
{
  dbp = skip_spaces (dbp);
  if (*dbp != '*')
    return;
  dbp++;
  dbp = skip_spaces (dbp);
  if (strneq (dbp, "(*)", 3))
    {
      dbp += 3;
      return;
    }
  if (!isdigit (*dbp))
    {
      --dbp;			/* force failure */
      return;
    }
  do
    dbp++;
  while (isdigit (*dbp));
}

static void
getit (inf)
     FILE *inf;
{
  register char *cp;

  dbp = skip_spaces (dbp);
  if (*dbp == '\0')
    {
      lineno++;
      linecharno = charno;
      charno += readline (&lb, inf);
      dbp = lb.buffer;
      if (dbp[5] != '&')
	return;
      dbp += 6;
      dbp = skip_spaces (dbp);
    }
  if (!isalpha (*dbp) && *dbp != '_' && *dbp != '$')
    return;
  for (cp = dbp + 1; *cp != '\0' && intoken (*cp); cp++)
    continue;
  pfnote (savenstr (dbp, cp-dbp), TRUE,
	  lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
}


static void
Fortran_functions (inf)
     FILE *inf;
{
  LOOP_ON_INPUT_LINES (inf, lb, dbp)
    {
      if (*dbp == '%')
	dbp++;			/* Ratfor escape to fortran */
      dbp = skip_spaces (dbp);
      if (*dbp == '\0')
	continue;
      switch (lowcase (*dbp))
	{
	case 'i':
	  if (tail ("integer"))
	    takeprec ();
	  break;
	case 'r':
	  if (tail ("real"))
	    takeprec ();
	  break;
	case 'l':
	  if (tail ("logical"))
	    takeprec ();
	  break;
	case 'c':
	  if (tail ("complex") || tail ("character"))
	    takeprec ();
	  break;
	case 'd':
	  if (tail ("double"))
	    {
	      dbp = skip_spaces (dbp);
	      if (*dbp == '\0')
		continue;
	      if (tail ("precision"))
		break;
	      continue;
	    }
	  break;
	}
      dbp = skip_spaces (dbp);
      if (*dbp == '\0')
	continue;
      switch (lowcase (*dbp))
	{
	case 'f':
	  if (tail ("function"))
	    getit (inf);
	  continue;
	case 's':
	  if (tail ("subroutine"))
	    getit (inf);
	  continue;
	case 'e':
	  if (tail ("entry"))
	    getit (inf);
	  continue;
	case 'b':
	  if (tail ("blockdata") || tail ("block data"))
	    {
	      dbp = skip_spaces (dbp);
	      if (*dbp == '\0')	/* assume un-named */
		pfnote (savestr ("blockdata"), TRUE,
			lb.buffer, dbp - lb.buffer, lineno, linecharno);
	      else
		getit (inf);	/* look for name */
	    }
	  continue;
	}
    }
}

/*
 * Philippe Waroquiers <[e-mail]>, 1998-04-24
 * Ada parsing
 */

static void adagetit P_((FILE *, char *));

/* Once we are positioned after an "interesting" keyword, let's get
   the real tag value necessary. */
static void
adagetit (inf, name_qualifier)
     FILE *inf;
     char *name_qualifier;
{
  register char *cp;
  char *name;
  char c;

  while (!feof (inf))
    {
      dbp = skip_spaces (dbp);
      if (*dbp == '\0'
	  || (dbp[0] == '-' && dbp[1] == '-'))
	{
	  lineno++;
	  linecharno = charno;
	  charno += readline (&lb, inf);
	  dbp = lb.buffer;
	}
      switch (*dbp)
        {
        case 'b':
        case 'B':
          if (tail ("body"))
            {
              /* Skipping body of   procedure body   or   package body or ....
		 resetting qualifier to body instead of spec. */
              name_qualifier = "/b";
              continue;
            }
          break;
        case 't':
        case 'T':
          /* Skipping type of   task type   or   protected type ... */
          if (tail ("type"))
            continue;
          break;
        }
      if (*dbp == '"')
	{
	  dbp += 1;
	  for (cp = dbp; *cp != '\0' && *cp != '"'; cp++)
	    continue;
	}
      else
	{
	  dbp = skip_spaces (dbp);
	  for (cp = dbp;
	       (*cp != '\0'
		&& (isalpha (*cp) || isdigit (*cp) || *cp == '_' || *cp == '.'));
	       cp++)
	    continue;
	  if (cp == dbp)
	    return;
	}
      c = *cp;
      *cp = '\0';
      name = concat (dbp, name_qualifier, "");
      *cp = c;
      pfnote (name, TRUE, lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
      if (c == '"')
	dbp = cp + 1;
      return;
    }
}

static void
Ada_funcs (inf)
     FILE *inf;
{
  bool inquote = FALSE;

  LOOP_ON_INPUT_LINES (inf, lb, dbp)
    {
      while (*dbp != '\0')
	{
	  /* Skip a string i.e. "abcd". */
	  if (inquote || (*dbp == '"'))
	    {
	      dbp = etags_strchr ((inquote) ? dbp : dbp+1, '"');
	      if (dbp != NULL)
		{
		  inquote = FALSE;
		  dbp += 1;
		  continue;	/* advance char */
		}
	      else
		{
		  inquote = TRUE;
		  break;	/* advance line */
		}
	    }

	  /* Skip comments. */
	  if (dbp[0] == '-' && dbp[1] == '-')
	    break;		/* advance line */

	  /* Skip character enclosed in single quote i.e. 'a'
	     and skip single quote starting an attribute i.e. 'Image. */
	  if (*dbp == '\'')
	    {
	      dbp++ ;
	      if (*dbp != '\0')
		dbp++;
	      continue;
	    }

	  /* Search for beginning of a token.  */
	  if (!begtoken (*dbp))
	    {
	      dbp++;
	      continue;		/* advance char */
	    }

	  /* We are at the beginning of a token. */
	  switch (*dbp)
	    {
	    case 'f':
	    case 'F':
	      if (!packages_only && tail ("function"))
		adagetit (inf, "/f");
	      else
		break;		/* from switch */
	      continue;		/* advance char */
	    case 'p':
	    case 'P':
	      if (!packages_only && tail ("procedure"))
		adagetit (inf, "/p");
	      else if (tail ("package"))
		adagetit (inf, "/s");
	      else if (tail ("protected")) /* protected type */
		adagetit (inf, "/t");
	      else
		break;		/* from switch */
	      continue;		/* advance char */
	    case 't':
	    case 'T':
	      if (!packages_only && tail ("task"))
		adagetit (inf, "/k");
	      else if (typedefs && !packages_only && tail ("type"))
		{
		  adagetit (inf, "/t");
		  while (*dbp != '\0')
		    dbp += 1;
		}
	      else
		break;		/* from switch */
	      continue;		/* advance char */
	    }

	  /* Look for the end of the token. */
	  while (!endtoken (*dbp))
	    dbp++;

	} /* advance char */
    } /* advance line */
}

/*
 * Bob Weiner, Motorola Inc., 4/3/94
 * Unix and microcontroller assembly tag handling
 * look for '^[a-zA-Z_.$][a-zA_Z0-9_.$]*[: ^I^J]'
 */
static void
Asm_labels (inf)
     FILE *inf;
{
  register char *cp;

  LOOP_ON_INPUT_LINES (inf, lb, cp)
    {
      /* If first char is alphabetic or one of [_.$], test for colon
	 following identifier. */
      if (isalpha (*cp) || *cp == '_' || *cp == '.' || *cp == '$')
 	{
 	  /* Read past label. */
	  cp++;
 	  while (isalnum (*cp) || *cp == '_' || *cp == '.' || *cp == '$')
 	    cp++;
 	  if (*cp == ':' || iswhite (*cp))
 	    {
 	      /* Found end of label, so copy it and add it to the table. */
 	      pfnote (savenstr(lb.buffer, cp-lb.buffer), TRUE,
		      lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
 	    }
 	}
    }
}

/*
 * Perl support by Bart Robinson <[e-mail]>
 *              enhanced by Michael Ernst <[e-mail]>
 * Perl sub names: look for /^sub[ \t\n]+[^ \t\n{]+/
 * Perl variable names: /^(my|local).../
 */
static void
Perl_functions (inf)
     FILE *inf;
{
  register char *cp;

  LOOP_ON_INPUT_LINES (inf, lb, cp)
    {
      if (*cp++ == 's'
	  && *cp++ == 'u'
	  && *cp++ == 'b' && iswhite (*cp++))
	{
	  cp = skip_spaces (cp);
 	  if (*cp != '\0')
 	    {
	      char *sp = cp;
 	      while (*cp != '\0'
		     && !iswhite (*cp) && *cp != '{' && *cp != '(')
		cp++;
	      pfnote (savenstr (sp, cp-sp), TRUE,
 		      lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
 	    }
 	}
       else if (globals		/* only if tagging global vars is enabled */
		&& ((cp = lb.buffer,
		     *cp++ == 'm'
		     && *cp++ == 'y')
		    || (cp = lb.buffer,
			*cp++ == 'l'
			&& *cp++ == 'o'
			&& *cp++ == 'c'
			&& *cp++ == 'a'
			&& *cp++ == 'l'))
		&& (*cp == '(' || iswhite (*cp)))
 	{
 	  /* After "my" or "local", but before any following paren or space. */
 	  char *varname = NULL;

 	  cp = skip_spaces (cp);
 	  if (*cp == '$' || *cp == '@' || *cp == '%')
 	    {
 	      char* varstart = ++cp;
 	      while (isalnum (*cp) || *cp == '_')
 		cp++;
 	      varname = savenstr (varstart, cp-varstart);
 	    }
 	  else
 	    {
 	      /* Should be examining a variable list at this point;
 		 could insist on seeing an open parenthesis. */
 	      while (*cp != '\0' && *cp != ';' && *cp != '=' &&  *cp != ')')
 		cp++;
 	    }

 	  /* Perhaps I should back cp up one character, so the TAGS table
 	     doesn't mention (and so depend upon) the following char. */
 	  pfnote ((CTAGS) ? savenstr (lb.buffer, cp-lb.buffer) : varname,
 		  FALSE, lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
	}
    }
}

/*
 * Python support by Eric S. Raymond <[e-mail]>
 * Look for /^def[ \t\n]+[^ \t\n(:]+/ or /^class[ \t\n]+[^ \t\n(:]+/
 */
static void
Python_functions (inf)
     FILE *inf;
{
  register char *cp;

  LOOP_ON_INPUT_LINES (inf, lb, cp)
    {
      if (*cp++ == 'd'
	  && *cp++ == 'e'
	  && *cp++ == 'f' && iswhite (*cp++))
	{
	  cp = skip_spaces (cp);
	  while (*cp != '\0' && !iswhite (*cp) && *cp != '(' && *cp != ':')
	    cp++;
	  pfnote (NULL, TRUE,
		  lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
	}

      cp = lb.buffer;
      if (*cp++ == 'c'
	  && *cp++ == 'l'
	  && *cp++ == 'a'
	  && *cp++ == 's'
	  && *cp++ == 's' && iswhite (*cp++))
	{
	  cp = skip_spaces (cp);
	  while (*cp != '\0' && !iswhite (*cp) && *cp != '(' && *cp != ':')
	    cp++;
	  pfnote (NULL, TRUE,
		  lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
	}
    }
}

/* Idea by Corny de Souza
 * Cobol tag functions
 * We could look for anything that could be a paragraph name.
 * i.e. anything that starts in column 8 is one word and ends in a full stop.
 */
static void
Cobol_paragraphs (inf)
     FILE *inf;
{
  register char *bp, *ep;

  LOOP_ON_INPUT_LINES (inf, lb, bp)
    {
      if (lb.len < 9)
	continue;
      bp += 8;

      /* If eoln, compiler option or comment ignore whole line. */
      if (bp[-1] != ' ' || !isalnum (bp[0]))
        continue;

      for (ep = bp; isalnum (*ep) || *ep == '-'; ep++)
	continue;
      if (*ep++ == '.')
	pfnote (savenstr (bp, ep-bp), TRUE,
		lb.buffer, ep - lb.buffer + 1, lineno, linecharno);
    }
}

/* Added by Mosur Mohan, 4/22/88 */
/* Pascal parsing                */

/*
 *  Locates tags for procedures & functions.  Doesn't do any type- or
 *  var-definitions.  It does look for the keyword "extern" or
 *  "forward" immediately following the procedure statement; if found,
 *  the tag is skipped.
 */
static void
Pascal_functions (inf)
     FILE *inf;
{
  linebuffer tline;		/* mostly copied from C_entries */
  long save_lcno;
  int save_lineno, save_len;
  char c, *cp, *namebuf;

  bool				/* each of these flags is TRUE iff: */
    incomment,			/* point is inside a comment */
    inquote,			/* point is inside '..' string */
    get_tagname,		/* point is after PROCEDURE/FUNCTION
				   keyword, so next item = potential tag */
    found_tag,			/* point is after a potential tag */
    inparms,			/* point is within parameter-list */
    verify_tag;			/* point has passed the parm-list, so the
				   next token will determine whether this
				   is a FORWARD/EXTERN to be ignored, or
				   whether it is a real tag */

  save_lcno = save_lineno = save_len = 0; /* keep compiler quiet */
  namebuf = NULL;		/* keep compiler quiet */
  lineno = 0;
  charno = 0;
  dbp = lb.buffer;
  *dbp = '\0';
  initbuffer (&tline);

  incomment = inquote = FALSE;
  found_tag = FALSE;		/* have a proc name; check if extern */
  get_tagname = FALSE;		/* have found "procedure" keyword    */
  inparms = FALSE;		/* found '(' after "proc"            */
  verify_tag = FALSE;		/* check if "extern" is ahead        */


  while (!feof (inf))		/* long main loop to get next char */
    {
      c = *dbp++;
      if (c == '\0')		/* if end of line */
	{
	  lineno++;
	  linecharno = charno;
	  charno += readline (&lb, inf);
	  dbp = lb.buffer;
	  if (*dbp == '\0')
	    continue;
	  if (!((found_tag && verify_tag)
		|| get_tagname))
	    c = *dbp++;		/* only if don't need *dbp pointing
				   to the beginning of the name of
				   the procedure or function */
	}
      if (incomment)
	{
	  if (c == '}')		/* within { } comments */
	    incomment = FALSE;
	  else if (c == '*' && *dbp == ')') /* within (* *) comments */
	    {
	      dbp++;
	      incomment = FALSE;
	    }
	  continue;
	}
      else if (inquote)
	{
	  if (c == '\'')
	    inquote = FALSE;
	  continue;
	}
      else
	switch (c)
	  {
	  case '\'':
	    inquote = TRUE;	/* found first quote */
	    continue;
	  case '{':		/* found open { comment */
	    incomment = TRUE;
	    continue;
	  case '(':
	    if (*dbp == '*')	/* found open (* comment */
	      {
		incomment = TRUE;
		dbp++;
	      }
	    else if (found_tag)	/* found '(' after tag, i.e., parm-list */
	      inparms = TRUE;
	    continue;
	  case ')':		/* end of parms list */
	    if (inparms)
	      inparms = FALSE;
	    continue;
	  case ';':
	    if (found_tag && !inparms) /* end of proc or fn stmt */
	      {
		verify_tag = TRUE;
		break;
	      }
	    continue;
	  }
      if (found_tag && verify_tag && (*dbp != ' '))
	{
	  /* check if this is an "extern" declaration */
	  if (*dbp == '\0')
	    continue;
	  if (lowcase (*dbp == 'e'))
	    {
	      if (tail ("extern"))	/* superfluous, really! */
		{
		  found_tag = FALSE;
		  verify_tag = FALSE;
		}
	    }
	  else if (lowcase (*dbp) == 'f')
	    {
	      if (tail ("forward"))	/*  check for forward reference */
		{
		  found_tag = FALSE;
		  verify_tag = FALSE;
		}
	    }
	  if (found_tag && verify_tag) /* not external proc, so make tag */
	    {
	      found_tag = FALSE;
	      verify_tag = FALSE;
	      pfnote (namebuf, TRUE,
		      tline.buffer, save_len, save_lineno, save_lcno);
	      continue;
	    }
	}
      if (get_tagname)		/* grab name of proc or fn */
	{
	  if (*dbp == '\0')
	    continue;

	  /* save all values for later tagging */
	  grow_linebuffer (&tline, lb.len + 1);
	  strcpy (tline.buffer, lb.buffer);
	  save_lineno = lineno;
	  save_lcno = linecharno;

	  /* grab block name */
	  for (cp = dbp + 1; *cp != '\0' && !endtoken (*cp); cp++)
	    continue;
	  namebuf = savenstr (dbp, cp-dbp);
	  dbp = cp;		/* set dbp to e-o-token */
	  save_len = dbp - lb.buffer + 1;
	  get_tagname = FALSE;
	  found_tag = TRUE;
	  continue;

	  /* and proceed to check for "extern" */
	}
      else if (!incomment && !inquote && !found_tag)
	{
	  /* check for proc/fn keywords */
	  switch (lowcase (c))
	    {
	    case 'p':
	      if (tail ("rocedure"))	/* c = 'p', dbp has advanced */
		get_tagname = TRUE;
	      continue;
	    case 'f':
	      if (tail ("unction"))
		get_tagname = TRUE;
	      continue;
	    }
	}
    }				/* while not eof */

  free (tline.buffer);
}

/*
 * lisp tag functions
 *  look for (def or (DEF, quote or QUOTE
 */

static int L_isdef P_((char *));
static int L_isquote P_((char *));
static void L_getit P_((void));

static int
L_isdef (strp)
     register char *strp;
{
  return ((strp[1] == 'd' || strp[1] == 'D')
	  && (strp[2] == 'e' || strp[2] == 'E')
	  && (strp[3] == 'f' || strp[3] == 'F'));
}

static int
L_isquote (strp)
     register char *strp;
{
  return ((*++strp == 'q' || *strp == 'Q')
	  && (*++strp == 'u' || *strp == 'U')
	  && (*++strp == 'o' || *strp == 'O')
	  && (*++strp == 't' || *strp == 'T')
	  && (*++strp == 'e' || *strp == 'E')
	  && iswhite (*++strp));
}

static void
L_getit ()
{
  register char *cp;

  if (*dbp == '\'')		/* Skip prefix quote */
    dbp++;
  else if (*dbp == '(')
  {
    if (L_isquote (dbp))
      dbp += 7;			/* Skip "(quote " */
    else
      dbp += 1;			/* Skip "(" before name in (defstruct (foo)) */
    dbp = skip_spaces (dbp);
  }

  for (cp = dbp /*+1*/;
       *cp != '\0' && *cp != '(' && !iswhite(*cp) && *cp != ')';
       cp++)
    continue;
  if (cp == dbp)
    return;

  pfnote (savenstr (dbp, cp-dbp), TRUE,
	  lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
}

static void
Lisp_functions (inf)
     FILE *inf;
{
  LOOP_ON_INPUT_LINES (inf, lb, dbp)
    {
      if (dbp[0] == '(')
	{
	  if (L_isdef (dbp))
	    {
	      dbp = skip_non_spaces (dbp);
	      dbp = skip_spaces (dbp);
	      L_getit ();
	    }
	  else
	    {
	      /* Check for (foo::defmumble name-defined ... */
	      do
		dbp++;
	      while (*dbp != '\0' && !iswhite (*dbp)
		     && *dbp != ':' && *dbp != '(' && *dbp != ')');
	      if (*dbp == ':')
		{
		  do
		    dbp++;
		  while (*dbp == ':');

		  if (L_isdef (dbp - 1))
		    {
		      dbp = skip_non_spaces (dbp);
		      dbp = skip_spaces (dbp);
		      L_getit ();
		    }
		}
	    }
	}
    }
}

/*
 * Postscript tag functions
 * Just look for lines where the first character is '/'
 * Richard Mlynarik <[e-mail]>
 * Also look at "defineps" for PSWrap
 * suggested by Masatake YAMATO <[e-mail]>
 */
static void
Postscript_functions (inf)
     FILE *inf;
{
  register char *bp, *ep;

  LOOP_ON_INPUT_LINES (inf, lb, bp)
    {
      if (bp[0] == '/')
	{
	  for (ep = bp+1;
	       *ep != '\0' && *ep != ' ' && *ep != '{';
	       ep++)
	    continue;
	  pfnote (savenstr (bp, ep-bp), TRUE,
		  lb.buffer, ep - lb.buffer + 1, lineno, linecharno);
	}
      else if (strneq (bp, "defineps", 8))
	{
	  bp = skip_non_spaces (bp);
 	  bp = skip_spaces (bp);
	  get_tag (bp);
	}
    }
}


/*
 * Scheme tag functions
 * look for (def... xyzzy
 * look for (def... (xyzzy
 * look for (def ... ((...(xyzzy ....
 * look for (set! xyzzy
 */

static void
Scheme_functions (inf)
     FILE *inf;
{
  register char *bp;

  LOOP_ON_INPUT_LINES (inf, lb, bp)
    {
      if (bp[0] == '('
	  && (bp[1] == 'D' || bp[1] == 'd')
	  && (bp[2] == 'E' || bp[2] == 'e')
	  && (bp[3] == 'F' || bp[3] == 'f'))
	{
	  bp = skip_non_spaces (bp);
	  /* Skip over open parens and white space */
	  while (iswhite (*bp) || *bp == '(')
	    bp++;
	  get_tag (bp);
	}
      if (bp[0] == '('
	  && (bp[1] == 'S' || bp[1] == 's')
	  && (bp[2] == 'E' || bp[2] == 'e')
	  && (bp[3] == 'T' || bp[3] == 't')
	  && (bp[4] == '!' || bp[4] == '!')
	  && (iswhite (bp[5])))
	{
	  bp = skip_non_spaces (bp);
	  bp = skip_spaces (bp);
	  get_tag (bp);
	}
    }
}

/* Find tags in TeX and LaTeX input files.  */

/* TEX_toktab is a table of TeX control sequences that define tags.
   Each TEX_tabent records one such control sequence.
   CONVERT THIS TO USE THE Stab TYPE!! */
struct TEX_tabent
{
  char *name;
  int len;
};

struct TEX_tabent *TEX_toktab = NULL;	/* Table with tag tokens */

/* Default set of control sequences to put into TEX_toktab.
   The value of environment var TEXTAGS is prepended to this.  */

char *TEX_defenv = "\
:chapter:section:subsection:subsubsection:eqno:label:ref:cite:bibitem\
:part:appendix:entry:index";

static void TEX_mode P_((FILE *));
static struct TEX_tabent *TEX_decode_env P_((char *, char *));
static int TEX_Token P_((char *));

char TEX_esc = '\\';
char TEX_opgrp = '{';
char TEX_clgrp = '}';

/*
 * TeX/LaTeX scanning loop.
 */
static void
TeX_functions (inf)
     FILE *inf;
{
  char *cp, *lasthit;
  register int i;

  /* Select either \ or ! as escape character.  */
  TEX_mode (inf);

  /* Initialize token table once from environment. */
  if (!TEX_toktab)
    TEX_toktab = TEX_decode_env ("TEXTAGS", TEX_defenv);

  LOOP_ON_INPUT_LINES (inf, lb, cp)
    {
      lasthit = cp;
      /* Look at each esc in line. */
      while ((cp = etags_strchr (cp, TEX_esc)) != NULL)
	{
	  if (*++cp == '\0')
	    break;
	  linecharno += cp - lasthit;
	  lasthit = cp;
	  i = TEX_Token (lasthit);
	  if (i >= 0)
	    {
	      /* We seem to include the TeX command in the tag name.
	      register char *p;
	      for (p = lasthit + TEX_toktab[i].len;
		   *p != '\0' && *p != TEX_clgrp;
		   p++)
		continue; */
	      pfnote (/*savenstr (lasthit, p-lasthit)*/ (char *)NULL, TRUE,
		      lb.buffer, lb.len, lineno, linecharno);
	      break;		/* We only tag a line once */
	    }
	}
    }
}

#define TEX_LESC '\\'
#define TEX_SESC '!'
#define TEX_cmt  '%'

/* Figure out whether TeX's escapechar is '\\' or '!' and set grouping
   chars accordingly. */
static void
TEX_mode (inf)
     FILE *inf;
{
  int c;

  while ((c = getc (inf)) != EOF)
    {
      /* Skip to next line if we hit the TeX comment char. */
      if (c == TEX_cmt)
	while (c != '\n')
	  c = getc (inf);
      else if (c == TEX_LESC || c == TEX_SESC )
	break;
    }

  if (c == TEX_LESC)
    {
      TEX_esc = TEX_LESC;
      TEX_opgrp = '{';
      TEX_clgrp = '}';
    }
  else
    {
      TEX_esc = TEX_SESC;
      TEX_opgrp = '<';
      TEX_clgrp = '>';
    }
  /* If the input file is compressed, inf is a pipe, and rewind may fail.
     No attempt is made to correct the situation. */
  rewind (inf);
}

/* Read environment and prepend it to the default string.
   Build token table. */
static struct TEX_tabent *
TEX_decode_env (evarname, defenv)
     char *evarname;
     char *defenv;
{
  register char *env, *p;

  struct TEX_tabent *tab;
  int size, i;

  /* Append default string to environment. */
  env = getenv (evarname);
  if (!env)
    env = defenv;
  else
    {
      char *oldenv = env;
      env = concat (oldenv, defenv, "");
    }

  /* Allocate a token table */
  for (size = 1, p = env; p;)
    if ((p = etags_strchr (p, ':')) && *++p != '\0')
      size++;
  /* Add 1 to leave room for null terminator.  */
  tab = xnew (size + 1, struct TEX_tabent);

  /* Unpack environment string into token table. Be careful about */
  /* zero-length strings (leading ':', "::" and trailing ':') */
  for (i = 0; *env;)
    {
      p = etags_strchr (env, ':');
      if (!p)			/* End of environment string. */
	p = env + strlen (env);
      if (p - env > 0)
	{			/* Only non-zero strings. */
	  tab[i].name = savenstr (env, p - env);
	  tab[i].len = strlen (tab[i].name);
	  i++;
	}
      if (*p)
	env = p + 1;
      else
	{
	  tab[i].name = NULL;	/* Mark end of table. */
	  tab[i].len = 0;
	  break;
	}
    }
  return tab;
}

/* If the text at CP matches one of the tag-defining TeX command names,
   return the pointer to the first occurrence of that command in TEX_toktab.
   Otherwise return -1.
   Keep the capital `T' in `token' for dumb truncating compilers
   (this distinguishes it from `TEX_toktab' */
static int
TEX_Token (cp)
     char *cp;
{
  int i;

  for (i = 0; TEX_toktab[i].len > 0; i++)
    if (strneq (TEX_toktab[i].name, cp, TEX_toktab[i].len))
      return i;
  return -1;
}

/*
 * Prolog support (rewritten) by Anders Lindgren, Mar. 96
 *
 * Assumes that the predicate starts at column 0.
 * Only the first clause of a predicate is added.
 */
static int prolog_pred P_((char *, char *));
static void prolog_skip_comment P_((linebuffer *, FILE *));
static int prolog_atom P_((char *, int));

static void
Prolog_functions (inf)
     FILE *inf;
{
  char *cp, *last;
  int len;
  int allocated;

  allocated = 0;
  len = 0;
  last = NULL;

  LOOP_ON_INPUT_LINES (inf, lb, cp)
    {
      if (cp[0] == '\0')	/* Empty line */
	continue;
      else if (iswhite (cp[0])) /* Not a predicate */
	continue;
      else if (cp[0] == '/' && cp[1] == '*')	/* comment. */
	prolog_skip_comment (&lb, inf);
      else if ((len = prolog_pred (cp, last)) > 0)
	{
	  /* Predicate.  Store the function name so that we only
	     generate a tag for the first clause.  */
	  if (last == NULL)
	    last = xnew(len + 1, char);
	  else if (len + 1 > allocated)
	    last = xrnew (last, len + 1, char);
	  allocated = len + 1;
	  strncpy (last, cp, len);
	  last[len] = '\0';
	}
    }
}


static void
prolog_skip_comment (plb, inf)
     linebuffer *plb;
     FILE *inf;
{
  char *cp;

  do
    {
      for (cp = plb->buffer; *cp != '\0'; cp++)
	if (cp[0] == '*' && cp[1] == '/')
	  return;
      lineno++;
      linecharno += readline (plb, inf);
    }
  while (!feof(inf));
}

/*
 * A predicate definition is added if it matches:
 *     <beginning of line><Prolog Atom><whitespace>(
 *
 * It is added to the tags database if it doesn't match the
 * name of the previous clause header.
 *
 * Return the size of the name of the predicate, or 0 if no header
 * was found.
 */
static int
prolog_pred (s, last)
     char *s;
     char *last;		/* Name of last clause. */
{
  int pos;
  int len;

  pos = prolog_atom (s, 0);
  if (pos < 1)
    return 0;

  len = pos;
  pos = skip_spaces (s + pos) - s;

  if ((s[pos] == '(') || (s[pos] == '.'))
    {
      if (s[pos] == '(')
	pos++;

      /* Save only the first clause. */
      if (last == NULL
	  || len != (int)strlen (last)
	  || !strneq (s, last, len))
	{
	  pfnote (savenstr (s, len), TRUE, s, pos, lineno, linecharno);
	  return len;
	}
    }
  return 0;
}

/*
 * Consume a Prolog atom.
 * Return the number of bytes consumed, or -1 if there was an error.
 *
 * A prolog atom, in this context, could be one of:
 * - An alphanumeric sequence, starting with a lower case letter.
 * - A quoted arbitrary string. Single quotes can escape themselves.
 *   Backslash quotes everything.
 */
static int
prolog_atom (s, pos)
     char *s;
     int pos;
{
  int origpos;

  origpos = pos;

  if (islower(s[pos]) || (s[pos] == '_'))
    {
      /* The atom is unquoted. */
      pos++;
      while (isalnum(s[pos]) || (s[pos] == '_'))
	{
	  pos++;
	}
      return pos - origpos;
    }
  else if (s[pos] == '\'')
    {
      pos++;

      while (1)
	{
	  if (s[pos] == '\'')
	    {
	      pos++;
	      if (s[pos] != '\'')
		break;
	      pos++;		/* A double quote */
	    }
	  else if (s[pos] == '\0')
	    /* Multiline quoted atoms are ignored. */
	    return -1;
	  else if (s[pos] == '\\')
	    {
	      if (s[pos+1] == '\0')
		return -1;
	      pos += 2;
	    }
	  else
	    pos++;
	}
      return pos - origpos;
    }
  else
    return -1;
}

/*
 * Support for Erlang  --  Anders Lindgren, Feb 1996.
 *
 * Generates tags for functions, defines, and records.
 *
 * Assumes that Erlang functions start at column 0.
 */
static int erlang_func P_((char *, char *));
static void erlang_attribute P_((char *));
static int erlang_atom P_((char *, int));

static void
Erlang_functions (inf)
     FILE *inf;
{
  char *cp, *last;
  int len;
  int allocated;

  allocated = 0;
  len = 0;
  last = NULL;

  LOOP_ON_INPUT_LINES (inf, lb, cp)
    {
      if (cp[0] == '\0')	/* Empty line */
	continue;
      else if (iswhite (cp[0])) /* Not function nor attribute */
	continue;
      else if (cp[0] == '%')	/* comment */
	continue;
      else if (cp[0] == '"')	/* Sometimes, strings start in column one */
	continue;
      else if (cp[0] == '-') 	/* attribute, e.g. "-define" */
	{
	  erlang_attribute (cp);
	  last = NULL;
	}
      else if ((len = erlang_func (cp, last)) > 0)
	{
	  /*
	   * Function.  Store the function name so that we only
	   * generates a tag for the first clause.
	   */
	  if (last == NULL)
	    last = xnew (len + 1, char);
	  else if (len + 1 > allocated)
	    last = xrnew (last, len + 1, char);
	  allocated = len + 1;
	  strncpy (last, cp, len);
	  last[len] = '\0';
	}
    }
}


/*
 * A function definition is added if it matches:
 *     <beginning of line><Erlang Atom><whitespace>(
 *
 * It is added to the tags database if it doesn't match the
 * name of the previous clause header.
 *
 * Return the size of the name of the function, or 0 if no function
 * was found.
 */
static int
erlang_func (s, last)
     char *s;
     char *last;		/* Name of last clause. */
{
  int pos;
  int len;

  pos = erlang_atom (s, 0);
  if (pos < 1)
    return 0;

  len = pos;
  pos = skip_spaces (s + pos) - s;

  /* Save only the first clause. */
  if (s[pos++] == '('
      && (last == NULL
	  || len != (int)strlen (last)
	  || !strneq (s, last, len)))
	{
	  pfnote (savenstr (s, len), TRUE, s, pos, lineno, linecharno);
	  return len;
	}

  return 0;
}


/*
 * Handle attributes.  Currently, tags are generated for defines
 * and records.
 *
 * They are on the form:
 * -define(foo, bar).
 * -define(Foo(M, N), M+N).
 * -record(graph, {vtab = notable, cyclic = true}).
 */
static void
erlang_attribute (s)
     char *s;
{
  int pos;
  int len;

  if (strneq (s, "-define", 7) || strneq (s, "-record", 7))
    {
      pos = skip_spaces (s + 7) - s;
      if (s[pos++] == '(')
	{
	  pos = skip_spaces (s + pos) - s;
	  len = erlang_atom (s, pos);
	  if (len != 0)
	    pfnote (savenstr (& s[pos], len), TRUE,
		    s, pos + len, lineno, linecharno);
	}
    }
  return;
}


/*
 * Consume an Erlang atom (or variable).
 * Return the number of bytes consumed, or -1 if there was an error.
 */
static int
erlang_atom (s, pos)
     char *s;
     int pos;
{
  int origpos;

  origpos = pos;

  if (isalpha (s[pos]) || s[pos] == '_')
    {
      /* The atom is unquoted. */
      pos++;
      while (isalnum (s[pos]) || s[pos] == '_')
	pos++;
      return pos - origpos;
    }
  else if (s[pos] == '\'')
    {
      pos++;

      while (1)
	{
	  if (s[pos] == '\'')
	    {
	      pos++;
	      break;
	    }
	  else if (s[pos] == '\0')
	    /* Multiline quoted atoms are ignored. */
	    return -1;
	  else if (s[pos] == '\\')
	    {
	      if (s[pos+1] == '\0')
		return -1;
	      pos += 2;
	    }
	  else
	    pos++;
	}
      return pos - origpos;
    }
  else
    return -1;
}

#ifdef ETAGS_REGEXPS

static char *scan_separators P_((char *));
static void analyse_regex P_((char *, bool));
static void add_regex P_((char *, bool, language *));
static char *substitute P_((char *, char *, struct re_registers *));

/* Take a string like "/blah/" and turn it into "blah", making sure
   that the first and last characters are the same, and handling
   quoted separator characters.  Actually, stops on the occurrence of
   an unquoted separator.  Also turns "\t" into a Tab character.
   Returns pointer to terminating separator.  Works in place.  Null
   terminates name string. */
static char *
scan_separators (name)
     char *name;
{
  char sep = name[0];
  char *copyto = name;
  bool quoted = FALSE;

  for (++name; *name != '\0'; ++name)
    {
      if (quoted)
	{
	  if (*name == 't')
	    *copyto++ = '\t';
	  else if (*name == sep)
	    *copyto++ = sep;
	  else
	    {
	      /* Something else is quoted, so preserve the quote. */
	      *copyto++ = '\\';
	      *copyto++ = *name;
	    }
	  quoted = FALSE;
	}
      else if (*name == '\\')
	quoted = TRUE;
      else if (*name == sep)
	break;
      else
	*copyto++ = *name;
    }

  /* Terminate copied string. */
  *copyto = '\0';
  return name;
}

/* Look at the argument of --regex or --no-regex and do the right
   thing.  Same for each line of a regexp file. */
static void
analyse_regex (regex_arg, ignore_case)
     char *regex_arg;
     bool ignore_case;
{
  if (regex_arg == NULL)
    free_patterns ();		/* --no-regex: remove existing regexps */

  /* A real --regexp option or a line in a regexp file. */
  switch (regex_arg[0])
    {
      /* Comments in regexp file or null arg to --regex. */
    case '\0':
    case ' ':
    case '\t':
      break;

      /* Read a regex file.  This is recursive and may result in a
	 loop, which will stop when the file descriptors are exhausted. */
    case '@':
      {
	FILE *regexfp;
	linebuffer regexbuf;
	char *regexfile = regex_arg + 1;

	/* regexfile is a file containing regexps, one per line. */
	regexfp = fopen (regexfile, "r");
	if (regexfp == NULL)
	  {
	    pfatal (regexfile);
	    return;
	  }
	initbuffer (&regexbuf);
	while (readline_internal (&regexbuf, regexfp) > 0)
	  analyse_regex (regexbuf.buffer, ignore_case);
	free (regexbuf.buffer);
	fclose (regexfp);
      }
      break;

      /* Regexp to be used for a specific language only. */
    case '{':
      {
	language *lang;
	char *lang_name = regex_arg + 1;
	char *cp;

	for (cp = lang_name; *cp != '}'; cp++)
	  if (*cp == '\0')
	    {
	      error ("unterminated language name in regex: %s", regex_arg);
	      return;
	    }
	*cp = '\0';
	lang = get_language_from_name (lang_name);
	if (lang == NULL)
	  return;
	add_regex (cp + 1, ignore_case, lang);
      }
      break;

      /* Regexp to be used for any language. */
    default:
      add_regex (regex_arg, ignore_case, NULL);
      break;
    }
}

/* Turn a name, which is an ed-style (but Emacs syntax) regular
   expression, into a real regular expression by compiling it. */
static void
add_regex (regexp_pattern, ignore_case, lang)
     char *regexp_pattern;
     bool ignore_case;
     language *lang;
{
  char *name;
  const char *err;
  struct re_pattern_buffer *patbuf;
  pattern *pp;


  if (regexp_pattern[strlen(regexp_pattern)-1] != regexp_pattern[0])
    {
      error ("%s: unterminated regexp", regexp_pattern);
      return;
    }
  name = scan_separators (regexp_pattern);
  if (regexp_pattern[0] == '\0')
    {
      error ("null regexp", (char *)NULL);
      return;
    }
  (void) scan_separators (name);

  patbuf = xnew (1, struct re_pattern_buffer);
  /* Translation table to fold case if appropriate. */
  patbuf->translate = (ignore_case) ? lc_trans : NULL;
  patbuf->fastmap = NULL;
  patbuf->buffer = NULL;
  patbuf->allocated = 0;

  err = re_compile_pattern (regexp_pattern, strlen (regexp_pattern), patbuf);
  if (err != NULL)
    {
      error ("%s while compiling pattern", err);
      return;
    }

  pp = p_head;
  p_head = xnew (1, pattern);
  p_head->regex = savestr (regexp_pattern);
  p_head->p_next = pp;
  p_head->language = lang;
  p_head->pattern = patbuf;
  p_head->name_pattern = savestr (name);
  p_head->error_signaled = FALSE;
}

/*
 * Do the substitutions indicated by the regular expression and
 * arguments.
 */
static char *
substitute (in, out, regs)
     char *in, *out;
     struct re_registers *regs;
{
  char *result, *t;
  int size, dig, diglen;

  result = NULL;
  size = strlen (out);

  /* Pass 1: figure out how much to allocate by finding all \N strings. */
  if (out[size - 1] == '\\')
    fatal ("pattern error in \"%s\"", out);
  for (t = etags_strchr (out, '\\');
       t != NULL;
       t = etags_strchr (t + 2, '\\'))
    if (isdigit (t[1]))
      {
	dig = t[1] - '0';
	diglen = regs->end[dig] - regs->start[dig];
	size += diglen - 2;
      }
    else
      size -= 1;

  /* Allocate space and do the substitutions. */
  result = xnew (size + 1, char);

  for (t = result; *out != '\0'; out++)
    if (*out == '\\' && isdigit (*++out))
      {
	/* Using "dig2" satisfies my debugger.  Bleah. */
	dig = *out - '0';
	diglen = regs->end[dig] - regs->start[dig];
	strncpy (t, in + regs->start[dig], diglen);
	t += diglen;
      }
    else
      *t++ = *out;
  *t = '\0';

  if (DEBUG && (t > result + size || t - result != (int)strlen (result)))
    abort ();

  return result;
}

/* Deallocate all patterns. */
static void
free_patterns ()
{
  pattern *pp;
  while (p_head != NULL)
    {
      pp = p_head->p_next;
      free (p_head->regex);
      free (p_head->name_pattern);
      free (p_head);
      p_head = pp;
    }
  return;
}

static void
get_tag (bp)
     register char *bp;
{
  register char *cp;

  if (*bp == '\0')
    return;
  /* Go till you get to white space or a syntactic break */
  for (cp = bp + 1;
       *cp != '\0' && *cp != '(' && *cp != ')' && !iswhite (*cp);
       cp++)
    continue;
  pfnote (savenstr (bp, cp-bp), TRUE,
	  lb.buffer, cp - lb.buffer + 1, lineno, linecharno);
}

#endif /* ETAGS_REGEXPS */
/* Initialize a linebuffer for use */
static void
initbuffer (lbp)
     linebuffer *lbp;
{
  lbp->size = 200;
  lbp->buffer = xnew (200, char);
}

/*
 * Read a line of text from `stream' into `lbp', excluding the
 * newline or CR-NL, if any.  Return the number of characters read from
 * `stream', which is the length of the line including the newline.
 *
 * On DOS or Windows we do not count the CR character, if any, before the
 * NL, in the returned length; this mirrors the behavior of emacs on those
 * platforms (for text files, it translates CR-NL to NL as it reads in the
 * file).
 */
static long
readline_internal (lbp, stream)
     linebuffer *lbp;
     register FILE *stream;
{
  char *buffer = lbp->buffer;
  register char *p = lbp->buffer;
  register char *pend;
  int chars_deleted;

  pend = p + lbp->size;		/* Separate to avoid 386/IX compiler bug.  */

  while (1)
    {
      register int c = getc (stream);
      if (p == pend)
	{
	  /* We're at the end of linebuffer: expand it. */
	  lbp->size *= 2;
	  buffer = xrnew (buffer, lbp->size, char);
	  p += buffer - lbp->buffer;
	  pend = buffer + lbp->size;
	  lbp->buffer = buffer;
	}
      if (c == EOF)
	{
	  *p = '\0';
	  chars_deleted = 0;
	  break;
	}
      if (c == '\n')
	{
	  if (p > buffer && p[-1] == '\r')
	    {
	      p -= 1;
#ifdef DOS_NT
	     /* Assume CRLF->LF translation will be performed by Emacs
		when loading this file, so CRs won't appear in the buffer.
		It would be cleaner to compensate within Emacs;
		however, Emacs does not know how many CRs were deleted
		before any given point in the file.  */
	      chars_deleted = 1;
#else
	      chars_deleted = 2;
#endif
	    }
	  else
	    {
	      chars_deleted = 1;
	    }
	  *p = '\0';
	  break;
	}
      *p++ = c;
    }
  lbp->len = p - buffer;

  return lbp->len + chars_deleted;
}

/*
 * Like readline_internal, above, but in addition try to match the
 * input line against relevant regular expressions.
 */
static long
readline (lbp, stream)
     linebuffer *lbp;
     FILE *stream;
{
  /* Read new line. */
  long result = readline_internal (lbp, stream);
#ifdef ETAGS_REGEXPS
  int match;
  pattern *pp;

  /* Match against relevant patterns. */
  if (lbp->len > 0)
    for (pp = p_head; pp != NULL; pp = pp->p_next)
      {
	/* Only use generic regexps or those for the current language. */
	if (pp->language != NULL && pp->language != curlang)
	  continue;

	match = re_match (pp->pattern, lbp->buffer, lbp->len, 0, &pp->regs);
	switch (match)
	  {
	  case -2:
	    /* Some error. */
	    if (!pp->error_signaled)
	      {
		error ("error while matching \"%s\"", pp->regex);
		pp->error_signaled = TRUE;
	      }
	    break;
	  case -1:
	    /* No match. */
	    break;
	  default:
	    /* Match occurred.  Construct a tag. */
	    if (pp->name_pattern[0] != '\0')
	      {
		/* Make a named tag. */
		char *name = substitute (lbp->buffer,
					 pp->name_pattern, &pp->regs);
		if (name != NULL)
		  pfnote (name, TRUE, lbp->buffer, match, lineno, linecharno);
	      }
	    else
	      {
		/* Make an unnamed tag. */
		pfnote ((char *)NULL, TRUE,
			lbp->buffer, match, lineno, linecharno);
	      }
	    break;
	  }
      }
#endif /* ETAGS_REGEXPS */

  return result;
}

/*
 * Return a pointer to a space of size strlen(cp)+1 allocated
 * with xnew where the string CP has been copied.
 */
static char *
savestr (cp)
     char *cp;
{
  return savenstr (cp, strlen (cp));
}

/*
 * Return a pointer to a space of size LEN+1 allocated with xnew where
 * the string CP has been copied for at most the first LEN characters.
 */
static char *
savenstr (cp, len)
     char *cp;
     int len;
{
  register char *dp;

  dp = xnew (len + 1, char);
  strncpy (dp, cp, len);
  dp[len] = '\0';
  return dp;
}

/*
 * Return the ptr in sp at which the character c last
 * appears; NULL if not found
 *
 * Identical to POSIX strrchr, included for portability.
 */
static char *
etags_strrchr (sp, c)
     register const char *sp;
     register int c;
{
  register const char *r;

  r = NULL;
  do
    {
      if (*sp == c)
	r = sp;
  } while (*sp++);
  return (char *)r;
}


/*
 * Return the ptr in sp at which the character c first
 * appears; NULL if not found
 *
 * Identical to POSIX strchr, included for portability.
 */
static char *
etags_strchr (sp, c)
     register const char *sp;
     register int c;
{
  do
    {
      if (*sp == c)
	return (char *)sp;
    } while (*sp++);
  return NULL;
}

/* Skip spaces, return new pointer. */
static char *
skip_spaces (cp)
     char *cp;
{
  while (iswhite (*cp))
    cp++;
  return cp;
}

/* Skip non spaces, return new pointer. */
static char *
skip_non_spaces (cp)
     char *cp;
{
  while (*cp != '\0' && !iswhite (*cp))
    cp++;
  return cp;
}

/* Print error message and exit.  */
static void
fatal (s1, s2)
     char *s1, *s2;
{
  error (s1, s2);
  exit (BAD);
}

static void
pfatal (s1)
     char *s1;
{
  perror (s1);
  exit (BAD);
}

static void
suggest_asking_for_help ()
{
  fprintf (stderr, "\tTry `%s %s' for a complete list of options.\n",
	   progname,
#ifdef LONG_OPTIONS
	   "--help"
#else
	   "-h"
#endif
	   );
  exit (BAD);
}

/* Print error message.  `s1' is printf control string, `s2' is arg for it. */
static void
error (s1, s2)
     const char *s1, *s2;
{
  fprintf (stderr, "%s: ", progname);
  fprintf (stderr, s1, s2);
  fprintf (stderr, "\n");
}

/* Return a newly-allocated string whose contents
   concatenate those of s1, s2, s3.  */
static char *
concat (s1, s2, s3)
     char *s1, *s2, *s3;
{
  int len1 = strlen (s1), len2 = strlen (s2), len3 = strlen (s3);
  char *result = xnew (len1 + len2 + len3 + 1, char);

  strcpy (result, s1);
  strcpy (result + len1, s2);
  strcpy (result + len1 + len2, s3);
  result[len1 + len2 + len3] = '\0';

  return result;
}

/* Does the same work as the system V getcwd, but does not need to
   guess the buffer size in advance. */
static char *
etags_getcwd ()
{
#ifdef HAVE_GETCWD
  int bufsize = 200;
  char *path = xnew (bufsize, char);

  while (getcwd (path, bufsize) == NULL)
    {
      if (errno != ERANGE)
	pfatal ("getcwd");
      bufsize *= 2;
      free (path);
      path = xnew (bufsize, char);
    }

  canonicalize_filename (path);
  return path;

#else /* not HAVE_GETCWD */
#ifdef MSDOS
  char *p, path[MAXPATHLEN + 1]; /* Fixed size is safe on MSDOS.  */

  getwd (path);

  for (p = path; *p != '\0'; p++)
    if (*p == '\\')
      *p = '/';
    else
      *p = lowcase (*p);

  return strdup (path);
#else /* not MSDOS */
  linebuffer path;
  FILE *pipe;

  initbuffer (&path);
  pipe = (FILE *) popen ("pwd 2>/dev/null", "r");
  if (pipe == NULL || readline_internal (&path, pipe) == 0)
    pfatal ("pwd");
  pclose (pipe);

  return path.buffer;
#endif /* not MSDOS */
#endif /* not HAVE_GETCWD */
}

/* Return a newly allocated string containing the file name of FILE
   relative to the absolute directory DIR (which should end with a slash). */
static char *
relative_filename (file, dir)
     char *file, *dir;
{
  char *fp, *dp, *afn, *res;
  int i;

  /* Find the common root of file and dir (with a trailing slash). */
  afn = absolute_filename (file, cwd);
  fp = afn;
  dp = dir;
  while (*fp++ == *dp++)
    continue;
  fp--, dp--;			/* back to the first differing char */
#ifdef DOS_NT
  if (fp == afn && afn[0] != '/') /* cannot build a relative name */
    return afn;
#endif
  do				/* look at the equal chars until '/' */
    fp--, dp--;
  while (*fp != '/');

  /* Build a sequence of "../" strings for the resulting relative file name. */
  i = 0;
  while ((dp = etags_strchr (dp + 1, '/')) != NULL)
    i += 1;
  res = xnew (3*i + strlen (fp + 1) + 1, char);
  res[0] = '\0';
  while (i-- > 0)
    strcat (res, "../");

  /* Add the file name relative to the common root of file and dir. */
  strcat (res, fp + 1);
  free (afn);

  return res;
}

/* Return a newly allocated string containing the absolute file name
   of FILE given DIR (which should end with a slash). */
static char *
absolute_filename (file, dir)
     char *file, *dir;
{
  char *slashp, *cp, *res;

  if (filename_is_absolute (file))
    res = savestr (file);
#ifdef DOS_NT
  /* We don't support non-absolute file names with a drive
     letter, like `d:NAME' (it's too much hassle).  */
  else if (file[1] == ':')
    fatal ("%s: relative file names with drive letters not supported", file);
#endif
  else
    res = concat (dir, file, "");

  /* Delete the "/dirname/.." and "/." substrings. */
  slashp = etags_strchr (res, '/');
  while (slashp != NULL && slashp[0] != '\0')
    {
      if (slashp[1] == '.')
	{
	  if (slashp[2] == '.'
	      && (slashp[3] == '/' || slashp[3] == '\0'))
	    {
	      cp = slashp;
	      do
		cp--;
	      while (cp >= res && !filename_is_absolute (cp));
	      if (cp < res)
		cp = slashp;	/* the absolute name begins with "/.." */
#ifdef DOS_NT
	      /* Under MSDOS and NT we get `d:/NAME' as absolute
		 file name, so the luser could say `d:/../NAME'.
		 We silently treat this as `d:/NAME'.  */
	      else if (cp[0] != '/')
		cp = slashp;
#endif
	      strcpy (cp, slashp + 3);
	      slashp = cp;
	      continue;
	    }
	  else if (slashp[2] == '/' || slashp[2] == '\0')
	    {
	      strcpy (slashp, slashp + 2);
	      continue;
	    }
	}

      slashp = etags_strchr (slashp + 1, '/');
    }

  if (res[0] == '\0')
    return savestr ("/");
  else
    return res;
}

/* Return a newly allocated string containing the absolute
   file name of dir where FILE resides given DIR (which should
   end with a slash). */
static char *
absolute_dirname (file, dir)
     char *file, *dir;
{
  char *slashp, *res;
  char save;

  canonicalize_filename (file);
  slashp = etags_strrchr (file, '/');
  if (slashp == NULL)
    return savestr (dir);
  save = slashp[1];
  slashp[1] = '\0';
  res = absolute_filename (file, dir);
  slashp[1] = save;

  return res;
}

/* Whether the argument string is an absolute file name.  The argument
   string must have been canonicalized with canonicalize_filename. */
static bool
filename_is_absolute (fn)
     char *fn;
{
  return (fn[0] == '/'
#ifdef DOS_NT
	  || (isalpha(fn[0]) && fn[1] == ':' && fn[2] == '/')
#endif
	  );
}

/* Translate backslashes into slashes.  Works in place. */
static void
canonicalize_filename (fn)
     register char *fn;
{
#ifdef DOS_NT
  /* Canonicalize drive letter case.  */
  if (islower (fn[0]))
    fn[0] = toupper (fn[0]);
  /* Convert backslashes to slashes.  */
  for (; *fn != '\0'; fn++)
    if (*fn == '\\')
      *fn = '/';
#else
  /* No action. */
  fn = NULL;			/* shut up the compiler */
#endif
}

/* Increase the size of a linebuffer. */
static void
grow_linebuffer (lbp, toksize)
     linebuffer *lbp;
     int toksize;
{
  while (lbp->size < toksize)
    lbp->size *= 2;
  lbp->buffer = xrnew (lbp->buffer, lbp->size, char);
}

/* Like malloc but get fatal error if memory is exhausted.  */
static long *
xmalloc (size)
     unsigned int size;
{
  long *result = (long *) malloc (size);
  if (result == NULL)
    fatal ("virtual memory exhausted", (char *)NULL);
  return result;
}

static long *
xrealloc (ptr, size)
     char *ptr;
     unsigned int size;
{
  long *result =  (long *) realloc (ptr, size);
  if (result == NULL)
    fatal ("virtual memory exhausted", (char *)NULL);
  return result;
}