X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/0d6508efab13d47fb08e9b49a90579abfb3a6d20..e83d80b8ed8587fd05437d33f2011403f55f1f9d:/src/lex.c diff --git a/src/lex.c b/src/lex.c index 371fbed9..78b8a8f3 100644 --- a/src/lex.c +++ b/src/lex.c @@ -1,5 +1,5 @@ /* Token-reader for Bison's input parser, - Copyright 1984, 1986, 1989, 1992, 2000 Free Software Foundation, Inc. + Copyright 1984, 1986, 1989, 1992, 2000, 2001 Free Software Foundation, Inc. This file is part of Bison, the GNU Compiler Compiler. @@ -21,43 +21,37 @@ #include "system.h" #include "getargs.h" #include "files.h" -#include "getopt.h" /* for optarg */ #include "symtab.h" +#include "options.h" #include "lex.h" -#include "xalloc.h" #include "complain.h" #include "gram.h" #include "quote.h" /* Buffer for storing the current token. */ -char *token_buffer; +static struct obstack token_obstack; +const char *token_buffer = NULL; -/* Allocated size of token_buffer, not including space for terminator. */ -static int maxtoken; - -bucket *symval; +bucket *symval = NULL; int numval; -static int unlexed; /* these two describe a token to be reread */ -static bucket *unlexed_symval; /* by the next call to lex */ - +/* A token to be reread, see unlex and lex. */ +static token_t unlexed = tok_undef; +static bucket *unlexed_symval = NULL; +static const char *unlexed_token_buffer = NULL; void -init_lex (void) +lex_init (void) { - maxtoken = 100; - token_buffer = XCALLOC (char, maxtoken + 1); - unlexed = -1; + obstack_init (&token_obstack); + unlexed = tok_undef; } -static char * -grow_token_buffer (char *p) +void +lex_free (void) { - int offset = p - token_buffer; - maxtoken *= 2; - token_buffer = XREALLOC (token_buffer, char, maxtoken + 1); - return token_buffer + offset; + obstack_free (&token_obstack, NULL); } @@ -148,16 +142,22 @@ xgetc (FILE *f) /*------------------------------------------------------------------. | Read one literal character from finput. Process \ escapes. | -| Append the normalized string version of the char to *PP. Assign | +| Append the normalized string version of the char to OUT. Assign | | the character code to *PCODE. Return 1 unless the character is an | -| unescaped `term' or \n report error for \n | +| unescaped `term' or \n report error for \n. | `------------------------------------------------------------------*/ -static int -literalchar (char **pp, int *pcode, char term) +/* FIXME: We could directly work in the obstack, but that would make + it more difficult to move to quotearg some day. So for the time + being, I prefer have literalchar behave like quotearg, and change + my mind later if I was wrong. */ + +int +literalchar (struct obstack *out, int *pcode, char term) { int c; - char *p; + char buf[4096]; + char *cp; int code; int wasquote = 0; @@ -241,85 +241,89 @@ literalchar (char **pp, int *pcode, char term) } else { - char buf [] = "c"; - buf[0] = c; + char badchar [] = "c"; + badchar[0] = c; complain (_("unknown escape sequence: `\\' followed by `%s'"), - quote (buf)); + quote (badchar)); code = '?'; } } /* has \ */ - /* now fill token_buffer with the canonical name for this character - as a literal token. Do not use what the user typed, - so that `\012' and `\n' can be interchangeable. */ + /* now fill BUF with the canonical name for this character as a + literal token. Do not use what the user typed, so that `\012' + and `\n' can be interchangeable. */ - p = *pp; + cp = buf; if (code == term && wasquote) - *p++ = code; + *cp++ = code; else if (code == '\\') { - *p++ = '\\'; - *p++ = '\\'; + *cp++ = '\\'; + *cp++ = '\\'; } else if (code == '\'') { - *p++ = '\\'; - *p++ = '\''; + *cp++ = '\\'; + *cp++ = '\''; } else if (code == '\"') { - *p++ = '\\'; - *p++ = '\"'; + *cp++ = '\\'; + *cp++ = '\"'; } else if (code >= 040 && code < 0177) - *p++ = code; + *cp++ = code; else if (code == '\t') { - *p++ = '\\'; - *p++ = 't'; + *cp++ = '\\'; + *cp++ = 't'; } else if (code == '\n') { - *p++ = '\\'; - *p++ = 'n'; + *cp++ = '\\'; + *cp++ = 'n'; } else if (code == '\r') { - *p++ = '\\'; - *p++ = 'r'; + *cp++ = '\\'; + *cp++ = 'r'; } else if (code == '\v') { - *p++ = '\\'; - *p++ = 'v'; + *cp++ = '\\'; + *cp++ = 'v'; } else if (code == '\b') { - *p++ = '\\'; - *p++ = 'b'; + *cp++ = '\\'; + *cp++ = 'b'; } else if (code == '\f') { - *p++ = '\\'; - *p++ = 'f'; + *cp++ = '\\'; + *cp++ = 'f'; } else { - *p++ = '\\'; - *p++ = code / 0100 + '0'; - *p++ = ((code / 010) & 07) + '0'; - *p++ = (code & 07) + '0'; + *cp++ = '\\'; + *cp++ = code / 0100 + '0'; + *cp++ = ((code / 010) & 07) + '0'; + *cp++ = (code & 07) + '0'; } - *pp = p; + *cp = '\0'; + + if (out) + obstack_sgrow (out, buf); *pcode = code; return !wasquote; } void -unlex (int token) +unlex (token_t token) { unlexed = token; + unlexed_token_buffer = token_buffer; unlexed_symval = symval; } @@ -331,7 +335,6 @@ unlex (int token) void read_type_name (FILE *fin) { - char *p = token_buffer; int c = getc (fin); while (c != '>') @@ -345,40 +348,38 @@ read_type_name (FILE *fin) break; } - if (p == token_buffer + maxtoken) - p = grow_token_buffer (p); - - *p++ = c; + obstack_1grow (&token_obstack, c); c = getc (fin); } - *p = 0; + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); } -int +token_t lex (void) { int c; - char *p; - if (unlexed >= 0) + /* Just to make sure. */ + token_buffer = NULL; + + if (unlexed != tok_undef) { + token_t res = unlexed; symval = unlexed_symval; - c = unlexed; - unlexed = -1; - return c; + token_buffer = unlexed_token_buffer; + unlexed = tok_undef; + return res; } c = skip_white_space (); - /* for error messages (token buffer always valid) */ - *token_buffer = c; - token_buffer[1] = 0; switch (c) { case EOF: - strcpy (token_buffer, "EOF"); - return ENDFILE; + token_buffer = "EOF"; + return tok_eof; case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G': case 'H': case 'I': case 'J': @@ -394,267 +395,280 @@ lex (void) case 'z': case '.': case '_': - p = token_buffer; while (isalnum (c) || c == '_' || c == '.') { - if (p == token_buffer + maxtoken) - p = grow_token_buffer (p); - - *p++ = c; + obstack_1grow (&token_obstack, c); c = getc (finput); } - - *p = 0; + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); ungetc (c, finput); symval = getsym (token_buffer); - return IDENTIFIER; + return tok_identifier; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': { numval = 0; - p = token_buffer; while (isdigit (c)) { - if (p == token_buffer + maxtoken) - p = grow_token_buffer (p); - - *p++ = c; + obstack_1grow (&token_obstack, c); numval = numval * 10 + c - '0'; c = getc (finput); } - *p = 0; + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); ungetc (c, finput); - return NUMBER; + return tok_number; } case '\'': /* parse the literal token and compute character code in code */ - translations = -1; { - int code, discode; - char discard[10], *dp; + int code; - p = token_buffer; - *p++ = '\''; - literalchar (&p, &code, '\''); + obstack_1grow (&token_obstack, '\''); + literalchar (&token_obstack, &code, '\''); c = getc (finput); if (c != '\'') { + int discode; complain (_("use \"...\" for multi-character literal tokens")); while (1) - { - dp = discard; - if (!literalchar (&dp, &discode, '\'')) - break; - } + if (!literalchar (0, &discode, '\'')) + break; } - *p++ = '\''; - *p = 0; + obstack_1grow (&token_obstack, '\''); + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); symval = getsym (token_buffer); symval->class = token_sym; - if (!symval->user_token_number) + if (symval->user_token_number == SUNDEF) symval->user_token_number = code; - return IDENTIFIER; + return tok_identifier; } case '\"': /* parse the literal string token and treat as an identifier */ - translations = -1; { int code; /* ignored here */ - p = token_buffer; - *p++ = '\"'; + + obstack_1grow (&token_obstack, '\"'); /* Read up to and including ". */ - while (literalchar (&p, &code, '\"')) - { - if (p >= token_buffer + maxtoken - 4) - p = grow_token_buffer (p); - } - *p = 0; + while (literalchar (&token_obstack, &code, '\"')) + /* nothing */; + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); symval = getsym (token_buffer); symval->class = token_sym; - return IDENTIFIER; + return tok_identifier; } case ',': - return COMMA; + token_buffer = ","; + return tok_comma; case ':': - return COLON; + token_buffer = ":"; + return tok_colon; case ';': - return SEMICOLON; + token_buffer = ";"; + return tok_semicolon; case '|': - return BAR; + token_buffer = "|"; + return tok_bar; case '{': - return LEFT_CURLY; + token_buffer = "{"; + return tok_left_curly; case '=': + obstack_1grow (&token_obstack, c); do { c = getc (finput); + obstack_1grow (&token_obstack, c); if (c == '\n') lineno++; } while (c == ' ' || c == '\n' || c == '\t'); + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); if (c == '{') { - strcpy (token_buffer, "={"); - return LEFT_CURLY; + return tok_left_curly; } else { ungetc (c, finput); - return ILLEGAL; + return tok_illegal; } case '<': read_type_name (finput); - return TYPENAME; + return tok_typename; case '%': return parse_percent_token (); default: - return ILLEGAL; + obstack_1grow (&token_obstack, c); + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); + return tok_illegal; } } -/* the following table dictates the action taken for the various % - directives. A set_flag value causes the named flag to be set. A - retval action returns the code. */ -struct percent_table_struct -{ - const char *name; - void *set_flag; - int retval; -}; +/* This function is a strcmp, which doesn't differentiate `-' and `_' + chars. */ -struct percent_table_struct percent_table[] = +static int +option_strcmp (const char *left, const char *right) { - { "token", NULL, TOKEN }, - { "term", NULL, TOKEN }, - { "nterm", NULL, NTERM }, - { "type", NULL, TYPE }, - { "guard", NULL, GUARD }, - { "union", NULL, UNION }, - { "expect", NULL, EXPECT }, - { "thong", NULL, THONG }, - { "start", NULL, START }, - { "left", NULL, LEFT }, - { "right", NULL, RIGHT }, - { "nonassoc", NULL, NONASSOC }, - { "binary", NULL, NONASSOC }, - { "prec", NULL, PREC }, - { "locations", &locations_flag, NOOP }, /* -l */ - { "no_lines", &no_lines_flag, NOOP }, /* -l */ - { "raw", &raw_flag, NOOP }, /* -r */ - { "token_table", &token_table_flag, NOOP }, /* -k */ - { "yacc", &yacc_flag, NOOP }, /* -y */ - { "fixed_output_files",&yacc_flag, NOOP }, /* -y */ - { "defines", &defines_flag, NOOP }, /* -d */ - { "no_parser", &no_parser_flag, NOOP }, /* -n */ -#if 0 - /* For the time being, this is not enabled yet, while it's possible - though, since we use obstacks. The only risk is with semantic - parsers which will output an `include' of an output file: be sure - that the naem included is indeed the name of the output file. */ - { "output_file", &spec_outfile, SETOPT }, /* -o */ - { "file_prefix", &spec_file_prefix, SETOPT }, /* -b */ - { "name_prefix", &spec_name_prefix, SETOPT }, /* -p */ -#endif - { "verbose", &verbose_flag, NOOP }, /* -v */ - { "debug", &debug_flag, NOOP }, /* -t */ - { "semantic_parser", &semantic_parser, NOOP }, - { "pure_parser", &pure_parser, NOOP }, -/* {"help", , NOOP}, *//* -h */ -/* {"version", , NOOP}, *//* -V */ - { NULL, NULL, ILLEGAL} -}; + const unsigned char *l, *r; + int c; + + assert (left); + assert (right); + l = (const unsigned char *)left; + r = (const unsigned char *)right; + while (((c = *l - *r++) == 0 && *l != '\0') + || ((*l == '-' || *l == '_') && (*r == '_' || *r == '-'))) + l++; + return c; +} /* Parse a token which starts with %. Assumes the % has already been read and discarded. */ -int +token_t parse_percent_token (void) { - int c; - char *p; - struct percent_table_struct *tx; + const struct option_table_struct *tx = NULL; + const char *arg = NULL; + /* Where the ARG was found in token_buffer. */ + size_t arg_offset = 0; - p = token_buffer; - c = getc (finput); - *p++ = '%'; - *p++ = c; /* for error msg */ - *p = 0; + int c = getc (finput); switch (c) { case '%': - return TWO_PERCENTS; + return tok_two_percents; case '{': - return PERCENT_LEFT_CURLY; + return tok_percent_left_curly; + /* FIXME: Who the heck are those 5 guys!?! `%<' = `%left'!!! + Let's ask for there removal. */ case '<': - return LEFT; + return tok_left; case '>': - return RIGHT; + return tok_right; case '2': - return NONASSOC; + return tok_nonassoc; case '0': - return TOKEN; + return tok_token; case '=': - return PREC; + return tok_prec; } + if (!isalpha (c)) - return ILLEGAL; + return tok_illegal; - p = token_buffer; - *p++ = '%'; + obstack_1grow (&token_obstack, '%'); while (isalpha (c) || c == '_' || c == '-') { - if (p == token_buffer + maxtoken) - p = grow_token_buffer (p); - - if (c == '-') - c = '_'; - *p++ = c; + if (c == '_') + c = '-'; + obstack_1grow (&token_obstack, c); c = getc (finput); } - ungetc (c, finput); + /* %DIRECTIVE="ARG". Separate into + TOKEN_BUFFER = `%DIRECTIVE\0ARG\0'. + This is a bit hackish, but once we move to a Bison parser, + things will be cleaned up. */ + if (c == '=') + { + /* End of the directive. We skip the `='. */ + obstack_1grow (&token_obstack, '\0'); + /* Fetch the ARG if present. */ + c = getc (finput); + if (c == '"') + { + int code; + arg_offset = obstack_object_size (&token_obstack); + /* Read up to and including `"'. Do not append the closing + `"' in the output: it's not part of the ARG. */ + while (literalchar (NULL, &code, '"')) + obstack_1grow (&token_obstack, code); + } + /* else: should be an error. */ + } + else + ungetc (c, finput); - *p = 0; + obstack_1grow (&token_obstack, '\0'); + token_buffer = obstack_finish (&token_obstack); + if (arg_offset) + arg = token_buffer + arg_offset; /* table lookup % directive */ - for (tx = percent_table; tx->name; tx++) - if (strcmp (token_buffer + 1, tx->name) == 0) + for (tx = option_table; tx->name; tx++) + if ((tx->access == opt_percent || tx->access == opt_both) + && option_strcmp (token_buffer + 1, tx->name) == 0) break; - if (tx->retval == SETOPT) - { - *((char **) (tx->set_flag)) = optarg; - return NOOP; - } - if (tx->set_flag) + if (arg && tx->ret_val != tok_stropt) + fatal (_("`%s' supports no argument: %s"), token_buffer, quote (arg)); + + + switch (tx->ret_val) { + case tok_stropt: + assert (tx->set_flag); + if (arg) + { + /* Keep only the first assignment: command line options have + already been processed, and we want them to have + precedence. Side effect: if this %-option is used + several times, only the first is honored. Bah. */ + if (!*((char **) (tx->set_flag))) + *((char **) (tx->set_flag)) = xstrdup (arg); + } + else + fatal (_("`%s' requires an argument"), token_buffer); + return tok_noop; + break; + + case tok_intopt: + assert (tx->set_flag); *((int *) (tx->set_flag)) = 1; - return NOOP; + return tok_noop; + break; + + case tok_obsolete: + fatal (_("`%s' is no longer supported"), token_buffer); + return tok_noop; + break; + + default: + return tx->ret_val; + break; } - return tx->retval; + abort (); }