X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/572909b5db7143c7b62df384508fdba80be8b0da..1e9798d5699ca67718fd3b0d6baa9abcf27e6cd3:/src/reader.c?ds=sidebyside diff --git a/src/reader.c b/src/reader.c index ae7e1d64..e5aca8c4 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,414 +1,448 @@ /* Input parser for bison - Copyright (C) 1984, 1986, 1989 Free Software Foundation, Inc. + Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000 + Free Software Foundation, Inc. -This file is part of Bison, the GNU Compiler Compiler. + This file is part of Bison, the GNU Compiler Compiler. -Bison 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, or (at your option) -any later version. + Bison 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, or (at your option) + any later version. -Bison 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. + Bison 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 Bison; see the file COPYING. If not, write to -the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ + You should have received a copy of the GNU General Public License + along with Bison; see the file COPYING. If not, write to + the Free Software Foundation, Inc., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ -/* read in the grammar specification and record it in the format described in gram.h. - All guards are copied into the fguard file and all actions into faction, - in each case forming the body of a C function (yyguard or yyaction) - which contains a switch statement to decide which guard or action to execute. - -The entry point is reader(). */ - -#include -#include #include "system.h" +#include "getargs.h" #include "files.h" -#include "new.h" +#include "xalloc.h" #include "symtab.h" #include "lex.h" #include "gram.h" -#include "machine.h" - -#define LTYPESTR "\n#ifndef YYLTYPE\ntypedef\n struct yyltype\n\ - {\n int timestamp;\n int first_line;\n int first_column;\ -\n int last_line;\n int last_column;\n char *text;\n }\n\ - yyltype;\n\n#define YYLTYPE yyltype\n#endif\n\n" +#include "complain.h" +#include "output.h" +#include "reader.h" +#include "conflicts.h" +#include "quote.h" /* Number of slots allocated (but not necessarily used yet) in `rline' */ -int rline_allocated; - -extern char *program_name; -extern int definesflag; -extern int nolinesflag; -extern bucket *symval; -extern int numval; -extern int failure; -extern int expected_conflicts; -extern char *token_buffer; - -extern void init_lex(); -extern void tabinit(); -extern void output_headers(); -extern void output_trailers(); -extern void free_symtab(); -extern void open_extra_files(); -extern void fatal(); -extern void fatals(); -extern void unlex(); -extern void done(); - -extern int skip_white_space(); -extern int parse_percent_token(); -extern int lex(); - -void read_declarations(); -void copy_definition(); -void parse_token_decl(); -void parse_start_decl(); -void parse_type_decl(); -void parse_assoc_decl(); -void parse_union_decl(); -void parse_expect_decl(); -void copy_action(); -void readgram(); -void record_rule_line(); -void packsymbols(); -void output_token_defines(); -void packgram(); -int read_signed_integer(); -int get_type(); - -typedef - struct symbol_list - { - struct symbol_list *next; - bucket *sym; - bucket *ruleprec; - } - symbol_list; - +static int rline_allocated; +typedef struct symbol_list +{ + struct symbol_list *next; + bucket *sym; + bucket *ruleprec; +} +symbol_list; int lineno; -symbol_list *grammar; -int start_flag; -bucket *startval; char **tags; +short *user_toknums; +static symbol_list *grammar; +static int start_flag; +static bucket *startval; /* Nonzero if components of semantic values are used, implying they must be unions. */ static int value_components_used; -static int typed; /* nonzero if %union has been seen. */ +/* Nonzero if %union has been seen. */ +static int typed; -static int lastprec; /* incremented for each %left, %right or %nonassoc seen */ +/* Incremented for each %left, %right or %nonassoc seen */ +static int lastprec; -static int gensym_count; /* incremented for each generated symbol */ +/* Incremented for each generated symbol */ +static int gensym_count; static bucket *errtoken; +static bucket *undeftoken; + -/* Nonzero if any action or guard uses the @n construct. */ -static int yylsp_needed; - -extern char *version_string; +/*===================\ +| Low level lexing. | +\===================*/ -void -reader() +static void +skip_to_char (int target) { - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ + int c; + if (target == '\n') + complain (_(" Skipping to next \\n")); + else + complain (_(" Skipping to next %c"), target); -#if 0 - translations = 0; /* initially assume token number translation not needed. */ -#endif - /* Nowadays translations is always set to 1, - since we give `error' a user-token-number - to satisfy the Posix demand for YYERRCODE==256. */ - translations = 1; + do + c = skip_white_space (); + while (c != target && c != EOF); + if (c != EOF) + ungetc (c, finput); +} - nsyms = 1; - nvars = 0; - nrules = 0; - nitems = 0; - rline_allocated = 10; - rline = NEW2(rline_allocated, short); - typed = 0; - lastprec = 0; +/*---------------------------------------------------------. +| Read a signed integer from STREAM and return its value. | +`---------------------------------------------------------*/ - gensym_count = 0; +static inline int +read_signed_integer (FILE *stream) +{ + int c = getc (stream); + int sign = 1; + int n = 0; - semantic_parser = 0; - pure_parser = 0; - yylsp_needed = 0; + if (c == '-') + { + c = getc (stream); + sign = -1; + } - grammar = NULL; + while (isdigit (c)) + { + n = 10 * n + (c - '0'); + c = getc (stream); + } - init_lex(); - lineno = 1; + ungetc (c, stream); + + return sign * n; +} + +/*--------------------------------------------------------------. +| Get the data type (alternative in the union) of the value for | +| symbol N in rule RULE. | +`--------------------------------------------------------------*/ - /* initialize the symbol table. */ - tabinit(); - /* construct the error token */ - errtoken = getsym("error"); - errtoken->class = STOKEN; - errtoken->user_token_number = 256; /* Value specified by posix. */ - /* construct a token that represents all undefined literal tokens. */ - /* it is always token number 2. */ - getsym("$undefined.")->class = STOKEN; - /* Read the declaration section. Copy %{ ... %} groups to ftable and fdefines file. - Also notice any %token, %left, etc. found there. */ - fprintf(ftable, "\n/* A Bison parser, made from %s", infile); - fprintf(ftable, " with Bison version %s */\n\n", version_string); - fprintf(ftable, "#define YYBISON 1 /* Identify Bison output. */\n\n"); - read_declarations(); - /* output the definition of YYLTYPE into the fattrs and fdefines files. */ - /* fattrs winds up in the .tab.c file, before bison.simple. */ - fprintf(fattrs, LTYPESTR); - /* start writing the guard and action files, if they are needed. */ - output_headers(); - /* read in the grammar, build grammar in list form. write out guards and actions. */ - readgram(); - /* Now we know whether we need the line-number stack. - If we do, write its type into the .tab.h file. */ - if (yylsp_needed) +static char * +get_type_name (int n, symbol_list * rule) +{ + int i; + symbol_list *rp; + + if (n < 0) { - if (fdefines) - fprintf(fdefines, LTYPESTR); + complain (_("invalid $ value")); + return NULL; } - /* write closing delimiters for actions and guards. */ - output_trailers(); - if (yylsp_needed) - fprintf(ftable, "#define YYLSP_NEEDED\n\n"); - /* assign the symbols their symbol numbers. - Write #defines for the token symbols into fdefines if requested. */ - packsymbols(); - /* convert the grammar into the format described in gram.h. */ - packgram(); - /* free the symbol table data structure - since symbols are now all referred to by symbol number. */ - free_symtab(); -} + rp = rule; + i = 0; + while (i < n) + { + rp = rp->next; + if (rp == NULL || rp->sym == NULL) + { + complain (_("invalid $ value")); + return NULL; + } + i++; + } -/* read from finput until %% is seen. Discard the %%. -Handle any % declarations, -and copy the contents of any %{ ... %} groups to fattrs. */ + return rp->sym->type_name; +} + +/*-------------------------------------------------------------------. +| Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of | +| the string (either ' or "). | +`-------------------------------------------------------------------*/ -void -read_declarations () +static inline void +copy_string (FILE *fin, FILE *fout, int match) { - register int c; - register int tok; + int c; - for (;;) + putc (match, fout); + c = getc (fin); + + while (c != match) { - c = skip_white_space(); + if (c == EOF) + fatal (_("unterminated string at end of file")); + if (c == '\n') + { + complain (_("unterminated string")); + ungetc (c, fin); + c = match; /* invent terminator */ + continue; + } - if (c == '%') + putc (c, fout); + + if (c == '\\') { - tok = parse_percent_token(); + c = getc (fin); + if (c == EOF) + fatal (_("unterminated string at end of file")); + putc (c, fout); + if (c == '\n') + lineno++; + } - switch (tok) - { - case TWO_PERCENTS: - return; + c = getc (fin); + } - case PERCENT_LEFT_CURLY: - copy_definition(); - break; + putc (c, fout); +} - case TOKEN: - parse_token_decl (STOKEN, SNTERM); - break; - - case NTERM: - parse_token_decl (SNTERM, STOKEN); - break; - - case TYPE: - parse_type_decl(); - break; - - case START: - parse_start_decl(); - break; - - case UNION: - parse_union_decl(); - break; - - case EXPECT: - parse_expect_decl(); - break; - - case LEFT: - parse_assoc_decl(LEFT_ASSOC); - break; - case RIGHT: - parse_assoc_decl(RIGHT_ASSOC); - break; +/*----------------------------------------------------------------. +| Dump the wannabee comment from IN to OUT1 and OUT2. In fact we | +| just saw a `/', which might or might not be a comment. In any | +| case, copy what we saw. | +| | +| OUT2 might be NULL. | +`----------------------------------------------------------------*/ - case NONASSOC: - parse_assoc_decl(NON_ASSOC); - break; +static inline void +copy_comment2 (FILE *fin, FILE *out1, FILE *out2) +{ + int cplus_comment; + int ended; + int c; - case SEMANTIC_PARSER: - if (semantic_parser == 0) - { - semantic_parser = 1; - open_extra_files(); - } - break; + /* We read a `/', output it. */ + putc ('/', out1); + if (out2) + putc ('/', out2); - case PURE_PARSER: - pure_parser = 1; - break; + switch ((c = getc (fin))) + { + case '/': + cplus_comment = 1; + break; + case '*': + cplus_comment = 0; + break; + default: + ungetc (c, fin); + return; + } - default: - fatal("junk after `%%' in definition section"); + putc (c, out1); + if (out2) + putc (c, out2); + c = getc (fin); + + ended = 0; + while (!ended) + { + if (!cplus_comment && c == '*') + { + while (c == '*') + { + putc (c, out1); + if (out2) + putc (c, out2); + c = getc (fin); + } + + if (c == '/') + { + putc (c, out1); + if (out2) + putc (c, out2); + ended = 1; } } + else if (c == '\n') + { + lineno++; + putc (c, out1); + if (out2) + putc (c, out2); + if (cplus_comment) + ended = 1; + else + c = getc (fin); + } else if (c == EOF) - fatal("no input grammar"); - else if (c >= 040 && c <= 0177) - fatals ("unknown character `%c' in declaration section", c); + fatal (_("unterminated comment")); else - fatals ("unknown character with code 0x%x in declaration section", c); + { + putc (c, out1); + if (out2) + putc (c, out2); + c = getc (fin); + } } } -/* copy the contents of a %{ ... %} into the definitions file. -The %{ has already been read. Return after reading the %}. */ +/*-------------------------------------------------------------------. +| Dump the comment (actually the current string starting with a `/') | +| from FIN to FOUT. | +`-------------------------------------------------------------------*/ -void -copy_definition () +static inline void +copy_comment (FILE *fin, FILE *fout) { - register int c; - register int match; - register int ended; - register int after_percent; /* -1 while reading a character if prev char was % */ - int cplus_comment; + copy_comment2 (fin, fout, NULL); +} - if (!nolinesflag) - fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile); + +/*-----------------------------------------------------------------. +| FIN is pointing to a location (i.e., a `@'). Output to FOUT a | +| reference to this location. STACK_OFFSET is the number of values | +| in the current rule so far, which says where to find `$0' with | +| respect to the top of the stack. | +`-----------------------------------------------------------------*/ + +static inline void +copy_at (FILE *fin, FILE *fout, int stack_offset) +{ + int c; + + c = getc (fin); + if (c == '$') + { + fprintf (fout, "yyloc"); + locations_flag = 1; + } + else if (isdigit (c) || c == '-') + { + int n; + + ungetc (c, fin); + n = read_signed_integer (fin); + + fprintf (fout, "yylsp[%d]", n - stack_offset); + locations_flag = 1; + } + else + { + char buf[] = "@c"; + buf[1] = c; + complain (_("%s is invalid"), quote (buf)); + } +} + + +/*-------------------------------------------------------------------. +| FIN is pointing to a wannabee semantic value (i.e., a `$'). | +| | +| Possible inputs: $[]($|integer) | +| | +| Output to FOUT a reference to this semantic value. STACK_OFFSET is | +| the number of values in the current rule so far, which says where | +| to find `$0' with respect to the top of the stack. | +`-------------------------------------------------------------------*/ + +static inline void +copy_dollar (FILE *fin, FILE *fout, + symbol_list *rule, int stack_offset) +{ + int c = getc (fin); + char *type_name = NULL; + + /* Get the type name if explicit. */ + if (c == '<') + { + read_type_name (fin); + type_name = token_buffer; + value_components_used = 1; + c = getc (fin); + } + + if (c == '$') + { + fprintf (fout, "yyval"); + if (!type_name) + type_name = get_type_name (0, rule); + if (type_name) + fprintf (fout, ".%s", type_name); + if (!type_name && typed) + complain (_("$$ of `%s' has no declared type"), + rule->sym->tag); + } + else if (isdigit (c) || c == '-') + { + int n; + ungetc (c, fin); + n = read_signed_integer (fin); + + if (!type_name && n > 0) + type_name = get_type_name (n, rule); + + fprintf (fout, "yyvsp[%d]", n - stack_offset); + if (type_name) + fprintf (fout, ".%s", type_name); + if (!type_name && typed) + complain (_("$%d of `%s' has no declared type"), + n, rule->sym->tag); + } + else + { + char buf[] = "$c"; + buf[1] = c; + complain (_("%s is invalid"), quote (buf)); + } +} + +/*-------------------------------------------------------------------. +| Copy the contents of a `%{ ... %}' into the definitions file. The | +| `%{' has already been read. Return after reading the `%}'. | +`-------------------------------------------------------------------*/ + +static void +copy_definition (void) +{ + int c; + /* -1 while reading a character if prev char was %. */ + int after_percent; + + if (!no_lines_flag) + fprintf (fattrs, "#line %d \"%s\"\n", lineno, infile); after_percent = 0; - c = getc(finput); + c = getc (finput); for (;;) { switch (c) { case '\n': - putc(c, fattrs); + putc (c, fattrs); lineno++; break; case '%': - after_percent = -1; + after_percent = -1; break; - + case '\'': case '"': - match = c; - putc(c, fattrs); - c = getc(finput); - - while (c != match) - { - if (c == EOF || c == '\n') - fatal("unterminated string"); - - putc(c, fattrs); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal("unterminated string"); - putc(c, fattrs); - if (c == '\n') - lineno++; - } - - c = getc(finput); - } - - putc(c, fattrs); + copy_string (finput, fattrs, c); break; case '/': - putc(c, fattrs); - c = getc(finput); - if (c != '*' && c != '/') - continue; - - cplus_comment = (c == '/'); - putc(c, fattrs); - c = getc(finput); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc(c, fattrs); - c = getc(finput); - } - - if (c == '/') - { - putc(c, fattrs); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc(c, fattrs); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal("unterminated comment in `%{' definition"); - else - { - putc(c, fattrs); - c = getc(finput); - } - } - + copy_comment (finput, fattrs); break; case EOF: - fatal("unterminated `%{' definition"); + fatal ("%s", _("unterminated `%{' definition")); default: - putc(c, fattrs); + putc (c, fattrs); } - c = getc(finput); + c = getc (finput); if (after_percent) { if (c == '}') return; - putc('%', fattrs); + putc ('%', fattrs); } after_percent = 0; @@ -417,121 +451,151 @@ copy_definition () } +/*-------------------------------------------------------------------. +| Parse what comes after %token or %nterm. For %token, WHAT_IS is | +| token_sym and WHAT_IS_NOT is nterm_sym. For %nterm, the arguments | +| are reversed. | +`-------------------------------------------------------------------*/ -/* parse what comes after %token or %nterm. -For %token, what_is is STOKEN and what_is_not is SNTERM. -For %nterm, the arguments are reversed. */ - -void -parse_token_decl (what_is, what_is_not) - int what_is, what_is_not; +static void +parse_token_decl (symbol_class what_is, symbol_class what_is_not) { -/* register int start_lineno; JF */ - register int token = 0; - register int prev; - register char *typename = 0; - int k; + int token = 0; + char *typename = 0; -/* start_lineno = lineno; JF */ + /* The symbol being defined. */ + struct bucket *symbol = NULL; + /* After `%token' and `%nterm', any number of symbols maybe be + defined. */ for (;;) { - if(ungetc(skip_white_space(), finput) == '%') - return; - -/* if (lineno != start_lineno) - return; JF */ + int tmp_char = ungetc (skip_white_space (), finput); - /* we have not passed a newline, so the token now starting is in this declaration */ - prev = token; + /* `%' (for instance from `%token', or from `%%' etc.) is the + only valid means to end this declaration. */ + if (tmp_char == '%') + return; + if (tmp_char == EOF) + fatal (_("Premature EOF after %s"), token_buffer); - token = lex(); + token = lex (); if (token == COMMA) - continue; + { + symbol = NULL; + continue; + } if (token == TYPENAME) { - k = strlen(token_buffer); - typename = NEW2(k + 1, char); - strcpy(typename, token_buffer); + typename = xstrdup (token_buffer); value_components_used = 1; + symbol = NULL; + } + else if (token == IDENTIFIER && *symval->tag == '\"' && symbol) + { + if (symval->alias) + warn (_("symbol `%s' used more than once as a literal string"), + symval->tag); + else if (symbol->alias) + warn (_("symbol `%s' given more than one literal string"), + symbol->tag); + else + { + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = symbol->user_token_number; + symbol->user_token_number = SALIAS; + symval->alias = symbol; + symbol->alias = symval; + /* symbol and symval combined are only one symbol */ + nsyms--; + } + translations = 1; + symbol = NULL; } else if (token == IDENTIFIER) { int oldclass = symval->class; + symbol = symval; - if (symval->class == what_is_not) - fatals("symbol %s redefined", symval->tag); - symval->class = what_is; - if (what_is == SNTERM && oldclass != SNTERM) - symval->value = nvars++; + if (symbol->class == what_is_not) + complain (_("symbol %s redefined"), symbol->tag); + symbol->class = what_is; + if (what_is == nterm_sym && oldclass != nterm_sym) + symbol->value = nvars++; if (typename) { - if (symval->type_name == NULL) - symval->type_name = typename; - else - fatals("type redeclaration for %s", symval->tag); + if (symbol->type_name == NULL) + symbol->type_name = typename; + else if (strcmp (typename, symbol->type_name) != 0) + complain (_("type redeclaration for %s"), symbol->tag); } } - else if (prev == IDENTIFIER && token == NUMBER) - { - symval->user_token_number = numval; + else if (symbol && token == NUMBER) + { + symbol->user_token_number = numval; translations = 1; - } + } else - fatal("invalid text in %token or %nterm declaration"); + { + complain (_("`%s' is invalid in %s"), + token_buffer, (what_is == token_sym) ? "%token" : "%nterm"); + skip_to_char ('%'); + } } } +/*------------------------------. +| Parse what comes after %start | +`------------------------------*/ -/* parse what comes after %start */ - -void -parse_start_decl () +static void +parse_start_decl (void) { if (start_flag) - fatal("multiple %start declarations"); - start_flag = 1; - if (lex() != IDENTIFIER) - fatal("invalid %start declaration"); - startval = symval; + complain (_("multiple %s declarations"), "%start"); + if (lex () != IDENTIFIER) + complain (_("invalid %s declaration"), "%start"); + else + { + start_flag = 1; + startval = symval; + } } +/*-----------------------------------------------------------. +| read in a %type declaration and record its information for | +| get_type_name to access | +`-----------------------------------------------------------*/ - -/* read in a %type declaration and record its information for get_type_name to access */ - -void -parse_type_decl () +static void +parse_type_decl (void) { - register int k; - register char *name; -/* register int start_lineno; JF */ - - if (lex() != TYPENAME) - fatal("ill-formed %type declaration"); + char *name; - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + if (lex () != TYPENAME) + { + complain ("%s", _("%type declaration has no ")); + skip_to_char ('%'); + return; + } -/* start_lineno = lineno; */ + name = xstrdup (token_buffer); for (;;) { - register int t; + int t; + int tmp_char = ungetc (skip_white_space (), finput); - if(ungetc(skip_white_space(), finput) == '%') + if (tmp_char == '%') return; + if (tmp_char == EOF) + fatal (_("Premature EOF after %s"), token_buffer); -/* if (lineno != start_lineno) - return; JF */ - - /* we have not passed a newline, so the token now starting is in this declaration */ - - t = lex(); + t = lex (); switch (t) { @@ -543,56 +607,50 @@ parse_type_decl () case IDENTIFIER: if (symval->type_name == NULL) symval->type_name = name; - else - fatals("type redeclaration for %s", symval->tag); + else if (strcmp (name, symval->type_name) != 0) + complain (_("type redeclaration for %s"), symval->tag); break; default: - fatal("invalid %type declaration"); + complain (_("invalid %%type declaration due to item: %s"), + token_buffer); + skip_to_char ('%'); } } } -/* read in a %left, %right or %nonassoc declaration and record its information. */ -/* assoc is either LEFT_ASSOC, RIGHT_ASSOC or NON_ASSOC. */ +/*----------------------------------------------------------------. +| Read in a %left, %right or %nonassoc declaration and record its | +| information. | +`----------------------------------------------------------------*/ -void -parse_assoc_decl (assoc) -int assoc; +static void +parse_assoc_decl (associativity assoc) { - register int k; - register char *name = NULL; -/* register int start_lineno; JF */ - register int prev = 0; /* JF added = 0 to keep lint happy */ - - lastprec++; /* Assign a new precedence level, never 0. */ + char *name = NULL; + int prev = 0; -/* start_lineno = lineno; */ + lastprec++; /* Assign a new precedence level, never 0. */ for (;;) { - register int t; + int t; + int tmp_char = ungetc (skip_white_space (), finput); - if(ungetc(skip_white_space(), finput) == '%') + if (tmp_char == '%') return; + if (tmp_char == EOF) + fatal (_("Premature EOF after %s"), token_buffer); - /* if (lineno != start_lineno) - return; JF */ - - /* we have not passed a newline, so the token now starting is in this declaration */ - - t = lex(); + t = lex (); switch (t) { - case TYPENAME: - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + name = xstrdup (token_buffer); break; case COMMA: @@ -600,36 +658,42 @@ int assoc; case IDENTIFIER: if (symval->prec != 0) - fatals("redefining precedence of %s", symval->tag); + complain (_("redefining precedence of %s"), symval->tag); symval->prec = lastprec; symval->assoc = assoc; - if (symval->class == SNTERM) - fatals("symbol %s redefined", symval->tag); - symval->class = STOKEN; + if (symval->class == nterm_sym) + complain (_("symbol %s redefined"), symval->tag); + symval->class = token_sym; if (name) - { /* record the type, if one is specified */ + { /* record the type, if one is specified */ if (symval->type_name == NULL) symval->type_name = name; - else - fatals("type redeclaration for %s", symval->tag); + else if (strcmp (name, symval->type_name) != 0) + complain (_("type redeclaration for %s"), symval->tag); } break; case NUMBER: if (prev == IDENTIFIER) - { + { symval->user_token_number = numval; translations = 1; - } - else - fatal("invalid text in association declaration"); + } + else + { + complain (_ + ("invalid text (%s) - number should be after identifier"), +token_buffer); + skip_to_char ('%'); + } break; case SEMICOLON: return; default: - fatal("malformatted association declaration"); + complain (_("unexpected item: %s"), token_buffer); + skip_to_char ('%'); } prev = t; @@ -639,42 +703,39 @@ int assoc; -/* copy the union declaration into fattrs (and fdefines), - where it is made into the - definition of YYSTYPE, the type of elements of the parser value stack. */ +/*-------------------------------------------------------------------. +| Copy the union declaration into fattrs (and fdefines), where it is | +| made into the definition of YYSTYPE, the type of elements of the | +| parser value stack. | +`-------------------------------------------------------------------*/ -void -parse_union_decl() +static void +parse_union_decl (void) { - register int c; - register int count; - register int in_comment; - int cplus_comment; + int c; + int count = 0; if (typed) - fatal("multiple %union declarations"); + complain (_("multiple %s declarations"), "%union"); typed = 1; - if (!nolinesflag) - fprintf(fattrs, "\n#line %d \"%s\"\n", lineno, infile); + if (!no_lines_flag) + fprintf (fattrs, "\n#line %d \"%s\"\n", lineno, infile); else - fprintf(fattrs, "\n"); + fprintf (fattrs, "\n"); - fprintf(fattrs, "typedef union"); + fprintf (fattrs, "typedef union"); if (fdefines) - fprintf(fdefines, "typedef union"); + fprintf (fdefines, "typedef union"); - count = 0; - in_comment = 0; - - c = getc(finput); + c = getc (finput); while (c != EOF) { - putc(c, fattrs); + putc (c, fattrs); if (fdefines) - putc(c, fdefines); + putc (c, fdefines); switch (c) { @@ -683,635 +744,558 @@ parse_union_decl() break; case '/': - c = getc(finput); - if (c != '*' && c != '/') - ungetc(c, finput); - else - { - putc(c, fattrs); - if (fdefines) - putc(c, fdefines); - cplus_comment = (c == '/'); - in_comment = 1; - c = getc(finput); - while (in_comment) - { - putc(c, fattrs); - if (fdefines) - putc(c, fdefines); - - if (c == '\n') - { - lineno++; - if (cplus_comment) - { - in_comment = 0; - break; - } - } - if (c == EOF) - fatal("unterminated comment"); - - if (!cplus_comment && c == '*') - { - c = getc(finput); - if (c == '/') - { - putc('/', fattrs); - if (fdefines) - putc('/', fdefines); - in_comment = 0; - } - } - else - c = getc(finput); - } - } + copy_comment2 (finput, fattrs, fdefines); break; - case '{': count++; break; case '}': if (count == 0) - fatal ("unmatched close-brace (`}')"); + complain (_("unmatched %s"), "`}'"); count--; - if (count == 0) + if (count <= 0) { - fprintf(fattrs, " YYSTYPE;\n"); + fprintf (fattrs, " YYSTYPE;\n"); if (fdefines) - fprintf(fdefines, " YYSTYPE;\n"); + fprintf (fdefines, " YYSTYPE;\n"); /* JF don't choke on trailing semi */ - c=skip_white_space(); - if(c!=';') ungetc(c,finput); + c = skip_white_space (); + if (c != ';') + ungetc (c, finput); return; } } - c = getc(finput); + c = getc (finput); } } -/* parse the declaration %expect N which says to expect N - shift-reduce conflicts. */ -void -parse_expect_decl() +/*-------------------------------------------------------. +| Parse the declaration %expect N which says to expect N | +| shift-reduce conflicts. | +`-------------------------------------------------------*/ + +static void +parse_expect_decl (void) { - register int c; - register int count; - char buffer[20]; + int c = skip_white_space (); + ungetc (c, finput); - c = getc(finput); - while (c == ' ' || c == '\t') - c = getc(finput); + if (!isdigit (c)) + complain (_("argument of %%expect is not an integer")); + else + expected_conflicts = read_signed_integer (finput); +} - count = 0; - while (c >= '0' && c <= '9') + +/*-------------------------------------------------------------------. +| Parse what comes after %thong. the full syntax is | +| | +| %thong token number literal | +| | +| the or number may be omitted. The number specifies the | +| user_token_number. | +| | +| Two symbols are entered in the table, one for the token symbol and | +| one for the literal. Both are given the , if any, from the | +| declaration. The ->user_token_number of the first is SALIAS and | +| the ->user_token_number of the second is set to the number, if | +| any, from the declaration. The two symbols are linked via | +| pointers in their ->alias fields. | +| | +| During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, | +| only the literal string is retained it is the literal string that | +| is output to yytname | +`-------------------------------------------------------------------*/ + +static void +parse_thong_decl (void) +{ + int token; + struct bucket *symbol; + char *typename = 0; + int usrtoknum; + + translations = 1; + token = lex (); /* fetch typename or first token */ + if (token == TYPENAME) { - if (count < 20) - buffer[count++] = c; - c = getc(finput); + typename = xstrdup (token_buffer); + value_components_used = 1; + token = lex (); /* fetch first token */ } - buffer[count] = 0; - ungetc (c, finput); + /* process first token */ + + if (token != IDENTIFIER) + { + complain (_("unrecognized item %s, expected an identifier"), + token_buffer); + skip_to_char ('%'); + return; + } + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = SALIAS; + symbol = symval; + + token = lex (); /* get number or literal string */ + + if (token == NUMBER) + { + usrtoknum = numval; + token = lex (); /* okay, did number, now get literal */ + } + else + usrtoknum = 0; - expected_conflicts = atoi (buffer); + /* process literal string token */ + + if (token != IDENTIFIER || *symval->tag != '\"') + { + complain (_("expected string constant instead of %s"), token_buffer); + skip_to_char ('%'); + return; + } + symval->class = token_sym; + symval->type_name = typename; + symval->user_token_number = usrtoknum; + + symval->alias = symbol; + symbol->alias = symval; + + /* symbol and symval combined are only one symbol. */ + nsyms--; } -/* that's all of parsing the declaration section */ - -/* Get the data type (alternative in the union) of the value for symbol n in rule rule. */ -char * -get_type_name(n, rule) -int n; -symbol_list *rule; +/*----------------------------------------------------------------. +| Read from finput until `%%' is seen. Discard the `%%'. Handle | +| any `%' declarations, and copy the contents of any `%{ ... %}' | +| groups to fattrs. | +`----------------------------------------------------------------*/ + +static void +read_declarations (void) { - static char *msg = "invalid $ value"; + int c; + int tok; - register int i; - register symbol_list *rp; + for (;;) + { + c = skip_white_space (); - if (n < 0) - fatal(msg); + if (c == '%') + { + tok = parse_percent_token (); - rp = rule; - i = 0; + switch (tok) + { + case TWO_PERCENTS: + return; - while (i < n) - { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - fatal(msg); - i++; - } + case PERCENT_LEFT_CURLY: + copy_definition (); + break; + + case TOKEN: + parse_token_decl (token_sym, nterm_sym); + break; + + case NTERM: + parse_token_decl (nterm_sym, token_sym); + break; + + case TYPE: + parse_type_decl (); + break; + + case START: + parse_start_decl (); + break; + + case UNION: + parse_union_decl (); + break; + + case EXPECT: + parse_expect_decl (); + break; + case THONG: + parse_thong_decl (); + break; + + case LEFT: + parse_assoc_decl (left_assoc); + break; + + case RIGHT: + parse_assoc_decl (right_assoc); + break; + + case NONASSOC: + parse_assoc_decl (non_assoc); + break; + + case SEMANTIC_PARSER: + if (semantic_parser == 0) + { + semantic_parser = 1; + open_extra_files (); + } + break; + + case PURE_PARSER: + pure_parser = 1; + break; - return (rp->sym->type_name); + case NOOP: + break; + + default: + complain (_("unrecognized: %s"), token_buffer); + skip_to_char ('%'); + } + } + else if (c == EOF) + fatal (_("no input grammar")); + else + { + char buf[] = "c"; + buf[0] = c; + complain (_("unknown character: %s"), quote (buf)); + skip_to_char ('%'); + } + } } + +/*-------------------------------------------------------------------. +| Assuming that a `{' has just been seen, copy everything up to the | +| matching `}' into the actions file. STACK_OFFSET is the number of | +| values in the current rule so far, which says where to find `$0' | +| with respect to the top of the stack. | +`-------------------------------------------------------------------*/ + +static void +copy_action (symbol_list *rule, int stack_offset) +{ + int c; + int count; + /* offset is always 0 if parser has already popped the stack pointer */ + if (semantic_parser) + stack_offset = 0; + fprintf (faction, "\ncase %d:\n", nrules); + if (!no_lines_flag) + fprintf (faction, "#line %d \"%s\"\n", lineno, infile); + putc ('{', faction); -/* after %guard is seen in the input file, -copy the actual guard into the guards file. -If the guard is followed by an action, copy that into the actions file. -stack_offset is the number of values in the current rule so far, -which says where to find $0 with respect to the top of the stack, -for the simple parser in which the stack is not popped until after the guard is run. */ + count = 1; + c = getc (finput); -void -copy_guard(rule, stack_offset) -symbol_list *rule; -int stack_offset; + while (count > 0) + { + while (c != '}') + { + switch (c) + { + case '\n': + putc (c, faction); + lineno++; + break; + + case '{': + putc (c, faction); + count++; + break; + + case '\'': + case '"': + copy_string (finput, faction, c); + break; + + case '/': + copy_comment (finput, faction); + break; + + case '$': + copy_dollar (finput, faction, rule, stack_offset); + break; + + case '@': + copy_at (finput, faction, stack_offset); + break; + + case EOF: + fatal (_("unmatched %s"), "`{'"); + + default: + putc (c, faction); + } + + c = getc (finput); + } + + /* above loop exits when c is '}' */ + + if (--count) + { + putc (c, faction); + c = getc (finput); + } + } + + fprintf (faction, ";\n break;}"); +} + +/*-------------------------------------------------------------------. +| After `%guard' is seen in the input file, copy the actual guard | +| into the guards file. If the guard is followed by an action, copy | +| that into the actions file. STACK_OFFSET is the number of values | +| in the current rule so far, which says where to find `$0' with | +| respect to the top of the stack, for the simple parser in which | +| the stack is not popped until after the guard is run. | +`-------------------------------------------------------------------*/ + +static void +copy_guard (symbol_list *rule, int stack_offset) { - register int c; - register int n; - register int count; - register int match; - register int ended; - register char *type_name; + int c; + int count; int brace_flag = 0; - int cplus_comment; /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) stack_offset = 0; + if (semantic_parser) + stack_offset = 0; - fprintf(fguard, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf(fguard, "#line %d \"%s\"\n", lineno, infile); - putc('{', fguard); + fprintf (fguard, "\ncase %d:\n", nrules); + if (!no_lines_flag) + fprintf (fguard, "#line %d \"%s\"\n", lineno, infile); + putc ('{', fguard); count = 0; - c = getc(finput); + c = getc (finput); while (brace_flag ? (count > 0) : (c != ';')) { switch (c) { case '\n': - putc(c, fguard); + putc (c, fguard); lineno++; break; case '{': - putc(c, fguard); + putc (c, fguard); brace_flag = 1; count++; break; case '}': - putc(c, fguard); + putc (c, fguard); if (count > 0) count--; else - fatal("unmatched right brace ('}')"); - break; - - case '\'': - case '"': - match = c; - putc(c, fguard); - c = getc(finput); - - while (c != match) - { - if (c == EOF || c == '\n') - fatal("unterminated string"); - - putc(c, fguard); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal("unterminated string"); - putc(c, fguard); - if (c == '\n') - lineno++; - } - - c = getc(finput); - } - - putc(c, fguard); - break; - - case '/': - putc(c, fguard); - c = getc(finput); - if (c != '*' && c != '/') - continue; - - cplus_comment = (c == '/'); - putc(c, fguard); - c = getc(finput); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc(c, fguard); - c = getc(finput); - } - - if (c == '/') - { - putc(c, fguard); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc(c, fguard); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal("unterminated comment"); - else - { - putc(c, fguard); - c = getc(finput); - } - } - - break; - - case '$': - c = getc(finput); - type_name = NULL; - - if (c == '<') - { - register char *cp = token_buffer; - - while ((c = getc(finput)) != '>' && c > 0) - *cp++ = c; - *cp = 0; - type_name = token_buffer; - - c = getc(finput); - } - - if (c == '$') - { - fprintf(fguard, "yyval"); - if (!type_name) type_name = rule->sym->type_name; - if (type_name) - fprintf(fguard, ".%s", type_name); - if(!type_name && typed) /* JF */ - fprintf(stderr,"%s:%d: warning: $$ of '%s' has no declared type.\n",infile,lineno,rule->sym->tag); - } - - else if (isdigit(c) || c == '-') { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - - if (!type_name && n > 0) - type_name = get_type_name(n, rule); - - fprintf(fguard, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf(fguard, ".%s", type_name); - if(!type_name && typed) /* JF */ - fprintf(stderr,"%s:%d: warning: $%d of '%s' has no declared type.\n",infile,lineno,n,rule->sym->tag); - continue; + complain (_("unmatched %s"), "`}'"); + c = getc (finput); /* skip it */ } - else - fatals("$%c is invalid",c); /* JF changed style */ + break; + case '\'': + case '"': + copy_string (finput, fguard, c); break; - case '@': - c = getc(finput); - if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - } - else - fatals("@%c is invalid",c); /* JF changed style */ + case '/': + copy_comment (finput, fguard); + break; - fprintf(fguard, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; + case '$': + copy_dollar (finput, fguard, rule, stack_offset); + break; - continue; + case '@': + copy_at (finput, fguard, stack_offset); + break; case EOF: - fatal("unterminated %guard clause"); + fatal ("%s", _("unterminated %guard clause")); default: - putc(c, fguard); + putc (c, fguard); } if (c != '}' || count != 0) - c = getc(finput); + c = getc (finput); } - c = skip_white_space(); + c = skip_white_space (); - fprintf(fguard, ";\n break;}"); + fprintf (fguard, ";\n break;}"); if (c == '{') - copy_action(rule, stack_offset); + copy_action (rule, stack_offset); else if (c == '=') { - c = getc(finput); + c = getc (finput); /* why not skip_white_space -wjh */ if (c == '{') - copy_action(rule, stack_offset); + copy_action (rule, stack_offset); } else - ungetc(c, finput); + ungetc (c, finput); } + - - -/* Assuming that a { has just been seen, copy everything up to the matching } -into the actions file. -stack_offset is the number of values in the current rule so far, -which says where to find $0 with respect to the top of the stack. */ - -void -copy_action(rule, stack_offset) -symbol_list *rule; -int stack_offset; +static void +record_rule_line (void) { - register int c; - register int n; - register int count; - register int match; - register int ended; - register char *type_name; - int cplus_comment; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) stack_offset = 0; - - fprintf(faction, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf(faction, "#line %d \"%s\"\n", lineno, infile); - putc('{', faction); - - count = 1; - c = getc(finput); + /* Record each rule's source line number in rline table. */ - while (count > 0) + if (nrules >= rline_allocated) { - while (c != '}') - { - switch (c) - { - case '\n': - putc(c, faction); - lineno++; - break; - - case '{': - putc(c, faction); - count++; - break; - - case '\'': - case '"': - match = c; - putc(c, faction); - c = getc(finput); - - while (c != match) - { - if (c == EOF || c == '\n') - fatal("unterminated string"); - - putc(c, faction); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal("unterminated string"); - putc(c, faction); - if (c == '\n') - lineno++; - } - - c = getc(finput); - } - - putc(c, faction); - break; - - case '/': - putc(c, faction); - c = getc(finput); - if (c != '*' && c != '/') - continue; - - cplus_comment = (c == '/'); - putc(c, faction); - c = getc(finput); - - ended = 0; - while (!ended) - { - if (!cplus_comment && c == '*') - { - while (c == '*') - { - putc(c, faction); - c = getc(finput); - } - - if (c == '/') - { - putc(c, faction); - ended = 1; - } - } - else if (c == '\n') - { - lineno++; - putc(c, faction); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal("unterminated comment"); - else - { - putc(c, faction); - c = getc(finput); - } - } + rline_allocated = nrules * 2; + rline = XREALLOC (rline, short, rline_allocated); + } + rline[nrules] = lineno; +} - break; - case '$': - c = getc(finput); - type_name = NULL; +/*-------------------------------------------------------------------. +| Generate a dummy symbol, a nonterminal, whose name cannot conflict | +| with the user's names. | +`-------------------------------------------------------------------*/ - if (c == '<') - { - register char *cp = token_buffer; +static bucket * +gensym (void) +{ + bucket *sym; - while ((c = getc(finput)) != '>' && c > 0) - *cp++ = c; - *cp = 0; - type_name = token_buffer; - value_components_used = 1; + sprintf (token_buffer, "@%d", ++gensym_count); + sym = getsym (token_buffer); + sym->class = nterm_sym; + sym->value = nvars++; + return sym; +} - c = getc(finput); - } - if (c == '$') - { - fprintf(faction, "yyval"); - if (!type_name) type_name = get_type_name(0, rule); - if (type_name) - fprintf(faction, ".%s", type_name); - if(!type_name && typed) /* JF */ - fprintf(stderr,"%s:%d: warning: $$ of '%s' has no declared type.\n",infile,lineno,rule->sym->tag); - } - else if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - - if (!type_name && n > 0) - type_name = get_type_name(n, rule); - - fprintf(faction, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf(faction, ".%s", type_name); - if(!type_name && typed) /* JF */ - fprintf(stderr,"%s:%d: warning: $%d of '%s' has no declared type.\n",infile,lineno,n,rule->sym->tag); - continue; - } - else - fatals("$%c is invalid",c); /* JF changed format */ +#if 0 +/*------------------------------------------------------------------. +| read in a %type declaration and record its information for | +| get_type_name to access. This is unused. It is only called from | +| the #if 0 part of readgram | +`------------------------------------------------------------------*/ + +static int +get_type (void) +{ + int k; + int t; + char *name; - break; + t = lex (); - case '@': - c = getc(finput); - if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer(finput); - c = getc(finput); - } - else - fatal("invalid @-construct"); + if (t != TYPENAME) + { + complain (_("invalid %s declaration"), "%type"); + return t; + } - fprintf(faction, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; + name = xstrdup (token_buffer); - continue; + for (;;) + { + t = lex (); - case EOF: - fatal("unmatched '{'"); + switch (t) + { + case SEMICOLON: + return lex (); - default: - putc(c, faction); - } + case COMMA: + break; - c = getc(finput); - } + case IDENTIFIER: + if (symval->type_name == NULL) + symval->type_name = name; + else if (strcmp (name, symval->type_name) != 0) + complain (_("type redeclaration for %s"), symval->tag); - /* above loop exits when c is '}' */ + break; - if (--count) - { - putc(c, faction); - c = getc(finput); + default: + return t; } } - - fprintf(faction, ";\n break;}"); -} - - - -/* generate a dummy symbol, a nonterminal, -whose name cannot conflict with the user's names. */ - -bucket * -gensym() -{ - register bucket *sym; - - sprintf (token_buffer, "@%d", ++gensym_count); - sym = getsym(token_buffer); - sym->class = SNTERM; - sym->value = nvars++; - return (sym); } -/* Parse the input grammar into a one symbol_list structure. -Each rule is represented by a sequence of symbols: the left hand side -followed by the contents of the right hand side, followed by a null pointer -instead of a symbol to terminate the rule. -The next symbol is the lhs of the following rule. - -All guards and actions are copied out to the appropriate files, -labelled by the rule number they apply to. */ - -void -readgram() +#endif + +/*------------------------------------------------------------------. +| Parse the input grammar into a one symbol_list structure. Each | +| rule is represented by a sequence of symbols: the left hand side | +| followed by the contents of the right hand side, followed by a | +| null pointer instead of a symbol to terminate the rule. The next | +| symbol is the lhs of the following rule. | +| | +| All guards and actions are copied out to the appropriate files, | +| labelled by the rule number they apply to. | +`------------------------------------------------------------------*/ + +static void +readgram (void) { - register int t; - register bucket *lhs; - register symbol_list *p; - register symbol_list *p1; - register bucket *bp; + int t; + bucket *lhs = NULL; + symbol_list *p; + symbol_list *p1; + bucket *bp; - symbol_list *crule; /* points to first symbol_list of current rule. */ - /* its symbol is the lhs of the rule. */ - symbol_list *crule1; /* points to the symbol_list preceding crule. */ + /* Points to first symbol_list of current rule. its symbol is the + lhs of the rule. */ + symbol_list *crule; + /* Points to the symbol_list preceding crule. */ + symbol_list *crule1; p1 = NULL; - t = lex(); + t = lex (); while (t != TWO_PERCENTS && t != ENDFILE) { if (t == IDENTIFIER || t == BAR) { - register int actionflag = 0; - int rulelength = 0; /* number of symbols in rhs of this rule so far */ + int action_flag = 0; + /* Number of symbols in rhs of this rule so far */ + int rulelength = 0; int xactions = 0; /* JF for error checking */ bucket *first_rhs = 0; if (t == IDENTIFIER) { lhs = symval; - - t = lex(); + + if (!start_flag) + { + startval = lhs; + start_flag = 1; + } + + t = lex (); if (t != COLON) - fatal("ill-formed rule"); + { + complain (_("ill-formed rule: initial symbol not followed by colon")); + unlex (t); + } } - if (nrules == 0) + if (nrules == 0 && t == BAR) { - if (t == BAR) - fatal("grammar starts with vertical bar"); - - if (!start_flag) - startval = lhs; + complain (_("grammar starts with vertical bar")); + lhs = symval; /* BOGUS: use a random symval */ } - /* start a new rule and record its lhs. */ nrules++; @@ -1319,7 +1303,7 @@ readgram() record_rule_line (); - p = NEW(symbol_list); + p = XCALLOC (symbol_list, 1); p->sym = lhs; crule1 = p1; @@ -1333,55 +1317,63 @@ readgram() /* mark the rule's lhs as a nonterminal if not already so. */ - if (lhs->class == SUNKNOWN) + if (lhs->class == unknown_sym) { - lhs->class = SNTERM; + lhs->class = nterm_sym; lhs->value = nvars; nvars++; } - else if (lhs->class == STOKEN) - fatals("rule given for %s, which is a token", lhs->tag); + else if (lhs->class == token_sym) + complain (_("rule given for %s, which is a token"), lhs->tag); /* read the rhs of the rule. */ for (;;) { - t = lex(); + t = lex (); + if (t == PREC) + { + t = lex (); + crule->ruleprec = symval; + t = lex (); + } - if (! (t == IDENTIFIER || t == LEFT_CURLY)) break; + if (!(t == IDENTIFIER || t == LEFT_CURLY)) + break; /* If next token is an identifier, see if a colon follows it. - If one does, exit this rule now. */ + If one does, exit this rule now. */ if (t == IDENTIFIER) { - register bucket *ssave; - register int t1; + bucket *ssave; + int t1; ssave = symval; - t1 = lex(); - unlex(t1); + t1 = lex (); + unlex (t1); symval = ssave; - if (t1 == COLON) break; + if (t1 == COLON) + break; - if(!first_rhs) /* JF */ + if (!first_rhs) /* JF */ first_rhs = symval; /* Not followed by colon => process as part of this rule's rhs. */ } /* If we just passed an action, that action was in the middle - of a rule, so make a dummy rule to reduce it to a - non-terminal. */ - if (actionflag) + of a rule, so make a dummy rule to reduce it to a + non-terminal. */ + if (action_flag) { - register bucket *sdummy; + bucket *sdummy; - /* Since the action was written out with this rule's */ - /* number, we must write give the new rule this number */ - /* by inserting the new rule before it. */ + /* Since the action was written out with this rule's + number, we must give the new rule this number by + inserting the new rule before it. */ /* Make a dummy nonterminal, a gensym. */ - sdummy = gensym(); + sdummy = gensym (); /* Make a new rule, whose body is empty, before the current one, so that the action @@ -1389,254 +1381,307 @@ readgram() nrules++; nitems++; record_rule_line (); - p = NEW(symbol_list); + p = XCALLOC (symbol_list, 1); if (crule1) crule1->next = p; - else grammar = p; + else + grammar = p; p->sym = sdummy; - crule1 = NEW(symbol_list); + crule1 = XCALLOC (symbol_list, 1); p->next = crule1; crule1->next = crule; - /* insert the dummy generated by that rule into this rule. */ + /* Insert the dummy generated by that rule into this + rule. */ nitems++; - p = NEW(symbol_list); + p = XCALLOC (symbol_list, 1); p->sym = sdummy; p1->next = p; p1 = p; - actionflag = 0; + action_flag = 0; } if (t == IDENTIFIER) { nitems++; - p = NEW(symbol_list); + p = XCALLOC (symbol_list, 1); p->sym = symval; p1->next = p; p1 = p; } - else /* handle an action. */ + else /* handle an action. */ { - copy_action(crule, rulelength); - actionflag = 1; + copy_action (crule, rulelength); + action_flag = 1; xactions++; /* JF */ } rulelength++; - } + } /* end of read rhs of rule */ /* Put an empty link in the list to mark the end of this rule */ - p = NEW(symbol_list); + p = XCALLOC (symbol_list, 1); p1->next = p; p1 = p; if (t == PREC) { - t = lex(); + complain (_("two @prec's in a row")); + t = lex (); crule->ruleprec = symval; - t = lex(); + t = lex (); } if (t == GUARD) { - if (! semantic_parser) - fatal("%guard present but %semantic_parser not specified"); + if (!semantic_parser) + complain (_("%%guard present but %%semantic_parser not specified")); - copy_guard(crule, rulelength); - t = lex(); + copy_guard (crule, rulelength); + t = lex (); } else if (t == LEFT_CURLY) { - if (actionflag) fatal("two actions at end of one rule"); - copy_action(crule, rulelength); - t = lex(); + /* This case never occurs -wjh */ + if (action_flag) + complain (_("two actions at end of one rule")); + copy_action (crule, rulelength); + action_flag = 1; + xactions++; /* -wjh */ + t = lex (); } - /* If $$ is being set in default way, - warn if any type mismatch. */ - else if (!xactions && first_rhs && lhs->type_name != first_rhs->type_name) + /* If $$ is being set in default way, report if any type + mismatch. */ + else if (!xactions + && first_rhs && lhs->type_name != first_rhs->type_name) { - if (lhs->type_name == 0 || first_rhs->type_name == 0 - || strcmp(lhs->type_name,first_rhs->type_name)) - fprintf(stderr, "%s:%d: warning: type clash ('%s' '%s') on default action\n", - infile, - lineno, - lhs->type_name ? lhs->type_name : "", - first_rhs->type_name ? first_rhs->type_name : ""); + if (lhs->type_name == 0 + || first_rhs->type_name == 0 + || strcmp (lhs->type_name, first_rhs->type_name)) + complain (_("type clash (`%s' `%s') on default action"), + lhs->type_name ? lhs->type_name : "", + first_rhs->type_name ? first_rhs->type_name : ""); } /* Warn if there is no default for $$ but we need one. */ else if (!xactions && !first_rhs && lhs->type_name != 0) - fprintf(stderr, - "%s:%d: warning: empty rule for typed nonterminal, and no action\n", - infile, - lineno); + complain (_("empty rule for typed nonterminal, and no action")); if (t == SEMICOLON) - t = lex(); + t = lex (); } +#if 0 /* these things can appear as alternatives to rules. */ +/* NO, they cannot. + a) none of the documentation allows them + b) most of them scan forward until finding a next % + thus they may swallow lots of intervening rules +*/ else if (t == TOKEN) { - parse_token_decl(STOKEN, SNTERM); - t = lex(); + parse_token_decl (token_sym, nterm_sym); + t = lex (); } else if (t == NTERM) { - parse_token_decl(SNTERM, STOKEN); - t = lex(); + parse_token_decl (nterm_sym, token_sym); + t = lex (); } else if (t == TYPE) { - t = get_type(); + t = get_type (); } else if (t == UNION) { - parse_union_decl(); - t = lex(); + parse_union_decl (); + t = lex (); } else if (t == EXPECT) { - parse_expect_decl(); - t = lex(); + parse_expect_decl (); + t = lex (); } else if (t == START) { - parse_start_decl(); - t = lex(); + parse_start_decl (); + t = lex (); } +#endif + else - fatal("invalid input"); + { + complain (_("invalid input: %s"), token_buffer); + t = lex (); + } } + /* grammar has been read. Do some checking */ + if (nsyms > MAXSHORT) - fatals("too many symbols (tokens plus nonterminals); maximum %d", + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), MAXSHORT); if (nrules == 0) - fatal("no input grammar"); + fatal (_("no rules in the input grammar")); - if (typed == 0 /* JF put out same default YYSTYPE as YACC does */ + /* JF put out same default YYSTYPE as YACC does */ + if (typed == 0 && !value_components_used) { /* We used to use `unsigned long' as YYSTYPE on MSDOS, - but it seems better to be consistent. - Most programs should declare their own type anyway. */ - fprintf(fattrs, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); + but it seems better to be consistent. + Most programs should declare their own type anyway. */ + fprintf (fattrs, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); if (fdefines) - fprintf(fdefines, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); + fprintf (fdefines, "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n"); } /* Report any undefined symbols and consider them nonterminals. */ for (bp = firstsymbol; bp; bp = bp->next) - if (bp->class == SUNKNOWN) + if (bp->class == unknown_sym) { - fprintf(stderr, "symbol %s used, not defined as token, and no rules for it\n", - bp->tag); - failure = 1; - bp->class = SNTERM; + complain (_ + ("symbol %s is used, but is not defined as a token and has no rules"), + bp->tag); + bp->class = nterm_sym; bp->value = nvars++; } ntokens = nsyms - nvars; } + +/*--------------------------------------------------------------. +| For named tokens, but not literal ones, define the name. The | +| value is the user token number. | +`--------------------------------------------------------------*/ - -void -record_rule_line () -{ - /* Record each rule's source line number in rline table. */ - - if (nrules >= rline_allocated) - { - rline_allocated = nrules * 2; - rline = (short *) xrealloc (rline, - rline_allocated * sizeof (short)); - } - rline[nrules] = lineno; -} - - -/* read in a %type declaration and record its information for get_type_name to access */ - -int -get_type() +static void +output_token_defines (FILE *file) { - register int k; - register int t; - register char *name; - - t = lex(); - - if (t != TYPENAME) - fatal("ill-formed %type declaration"); - - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + bucket *bp; + char *cp, *symbol; + char c; - for (;;) + for (bp = firstsymbol; bp; bp = bp->next) { - t = lex(); + symbol = bp->tag; /* get symbol */ - switch (t) + if (bp->value >= ntokens) + continue; + if (bp->user_token_number == SALIAS) + continue; + if ('\'' == *symbol) + continue; /* skip literal character */ + if (bp == errtoken) + continue; /* skip error token */ + if ('\"' == *symbol) { - case SEMICOLON: - return (lex()); - - case COMMA: - break; - - case IDENTIFIER: - if (symval->type_name == NULL) - symval->type_name = name; + /* use literal string only if given a symbol with an alias */ + if (bp->alias) + symbol = bp->alias->tag; else - fatals("type redeclaration for %s", symval->tag); + continue; + } - break; + /* Don't #define nonliteral tokens whose names contain periods. */ + cp = symbol; + while ((c = *cp++) && c != '.'); + if (c != '\0') + continue; - default: - return (t); - } + fprintf (file, "#define\t%s\t%d\n", symbol, + ((translations && !raw_flag) + ? bp->user_token_number : bp->value)); + if (semantic_parser) + fprintf (file, "#define\tT%s\t%d\n", symbol, bp->value); } -} + putc ('\n', file); +} -/* assign symbol numbers, and write definition of token names into fdefines. -Set up vectors tags and sprec of names and precedences of symbols. */ +/*------------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors TAGS and SPREC of names and precedences | +| of symbols. | +`------------------------------------------------------------------*/ -void -packsymbols() +static void +packsymbols (void) { - register bucket *bp; - register int tokno = 1; - register int i; - register int last_user_token_number; + bucket *bp; + int tokno = 1; + int i; + int last_user_token_number; + static char DOLLAR[] = "$"; /* int lossage = 0; JF set but not used */ - tags = NEW2(nsyms + 1, char *); - tags[0] = "$"; + tags = XCALLOC (char *, nsyms + 1); + tags[0] = DOLLAR; + user_toknums = XCALLOC (short, nsyms + 1); + user_toknums[0] = 0; - sprec = NEW2(nsyms, short); - sassoc = NEW2(nsyms, short); + sprec = XCALLOC (short, nsyms); + sassoc = XCALLOC (short, nsyms); max_user_token_number = 256; last_user_token_number = 256; for (bp = firstsymbol; bp; bp = bp->next) { - if (bp->class == SNTERM) + if (bp->class == nterm_sym) { bp->value += ntokens; } - else + else if (bp->alias) + { + /* this symbol and its alias are a single token defn. + allocate a tokno, and assign to both check agreement of + ->prec and ->assoc fields and make both the same */ + if (bp->value == 0) + bp->value = bp->alias->value = tokno++; + + if (bp->prec != bp->alias->prec) + { + if (bp->prec != 0 && bp->alias->prec != 0 + && bp->user_token_number == SALIAS) + complain (_("conflicting precedences for %s and %s"), + bp->tag, bp->alias->tag); + if (bp->prec != 0) + bp->alias->prec = bp->prec; + else + bp->prec = bp->alias->prec; + } + + if (bp->assoc != bp->alias->assoc) + { + if (bp->assoc != 0 && bp->alias->assoc != 0 + && bp->user_token_number == SALIAS) + complain (_("conflicting assoc values for %s and %s"), + bp->tag, bp->alias->tag); + if (bp->assoc != 0) + bp->alias->assoc = bp->assoc; + else + bp->assoc = bp->alias->assoc; + } + + if (bp->user_token_number == SALIAS) + continue; /* do not do processing below for SALIASs */ + + } + else /* bp->class == token_sym */ + { + bp->value = tokno++; + } + + if (bp->class == token_sym) { if (translations && !(bp->user_token_number)) bp->user_token_number = ++last_user_token_number; if (bp->user_token_number > max_user_token_number) max_user_token_number = bp->user_token_number; - bp->value = tokno++; } tags[bp->value] = bp->tag; + user_toknums[bp->value] = bp->user_token_number; sprec[bp->value] = bp->prec; sassoc[bp->value] = bp->assoc; @@ -1644,55 +1689,53 @@ packsymbols() if (translations) { - register int i; + int j; - token_translations = NEW2(max_user_token_number+1, short); + token_translations = XCALLOC (short, max_user_token_number + 1); - /* initialize all entries for literal tokens to 2, - the internal token number for $undefined., - which represents all invalid inputs. */ - for (i = 0; i <= max_user_token_number; i++) - token_translations[i] = 2; - } + /* initialize all entries for literal tokens to 2, the internal + token number for $undefined., which represents all invalid + inputs. */ + for (j = 0; j <= max_user_token_number; j++) + token_translations[j] = 2; - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->value >= ntokens) continue; - if (translations) + for (bp = firstsymbol; bp; bp = bp->next) { + if (bp->value >= ntokens) + continue; /* non-terminal */ + if (bp->user_token_number == SALIAS) + continue; if (token_translations[bp->user_token_number] != 2) - { - /* JF made this a call to fatals() */ - fatals( "tokens %s and %s both assigned number %d", - tags[token_translations[bp->user_token_number]], - bp->tag, - bp->user_token_number); - } + complain (_("tokens %s and %s both assigned number %d"), + tags[token_translations[bp->user_token_number]], + bp->tag, bp->user_token_number); token_translations[bp->user_token_number] = bp->value; } } error_token_number = errtoken->value; - output_token_defines(ftable); + if (!no_parser_flag) + output_token_defines (ftable); - if (startval->class == SUNKNOWN) - fatals("the start symbol %s is undefined", startval->tag); - else if (startval->class == STOKEN) - fatals("the start symbol %s is a token", startval->tag); + if (startval->class == unknown_sym) + fatal (_("the start symbol %s is undefined"), startval->tag); + else if (startval->class == token_sym) + fatal (_("the start symbol %s is a token"), startval->tag); start_symbol = startval->value; - if (definesflag) + if (defines_flag) { - output_token_defines(fdefines); + output_token_defines (fdefines); if (!pure_parser) { if (spec_name_prefix) - fprintf(fdefines, "\nextern YYSTYPE %slval;\n", spec_name_prefix); + fprintf (fdefines, "\nextern YYSTYPE %slval;\n", + spec_name_prefix); else - fprintf(fdefines, "\nextern YYSTYPE yylval;\n"); + fprintf (fdefines, "\nextern YYSTYPE yylval;\n"); } if (semantic_parser) @@ -1700,73 +1743,38 @@ packsymbols() { /* don't make these for dummy nonterminals made by gensym. */ if (*tags[i] != '@') - fprintf(fdefines, "#define\tNT%s\t%d\n", tags[i], i); + fprintf (fdefines, "#define\tNT%s\t%d\n", tags[i], i); } #if 0 /* `fdefines' is now a temporary file, so we need to copy its contents in `done', so we can't close it here. */ - fclose(fdefines); + fclose (fdefines); fdefines = NULL; #endif } } - - -void -output_token_defines(file) -FILE *file; -{ - bucket *bp; - - for (bp = firstsymbol; bp; bp = bp->next) - { - if (bp->value >= ntokens) continue; - - /* For named tokens, but not literal ones, define the name. */ - /* The value is the user token number. */ - - if ('\'' != *tags[bp->value] && bp != errtoken) - { - register char *cp = tags[bp->value]; - register char c; - - /* Don't #define nonliteral tokens whose names contain periods. */ - - while ((c = *cp++) && c != '.'); - if (!c) - { - fprintf(file, "#define\t%s\t%d\n", tags[bp->value], - (translations ? bp->user_token_number : bp->value)); - if (semantic_parser) - fprintf(file, "#define\tT%s\t%d\n", tags[bp->value], - bp->value); - } - } - } - - putc('\n', file); -} +/*---------------------------------------------------------------. +| Convert the rules into the representation using RRHS, RLHS and | +| RITEMS. | +`---------------------------------------------------------------*/ -/* convert the rules into the representation using rrhs, rlhs and ritems. */ - -void -packgram() +static void +packgram (void) { - register int itemno; - register int ruleno; - register symbol_list *p; -/* register bucket *bp; JF unused */ + int itemno; + int ruleno; + symbol_list *p; bucket *ruleprec; - ritem = NEW2(nitems + 1, short); - rlhs = NEW2(nrules, short) - 1; - rrhs = NEW2(nrules, short) - 1; - rprec = NEW2(nrules, short) - 1; - rprecsym = NEW2(nrules, short) - 1; - rassoc = NEW2(nrules, short) - 1; + ritem = XCALLOC (short, nitems + 1); + rlhs = XCALLOC (short, nrules) - 1; + rrhs = XCALLOC (short, nrules) - 1; + rprec = XCALLOC (short, nrules) - 1; + rprecsym = XCALLOC (short, nrules) - 1; + rassoc = XCALLOC (short, nrules) - 1; itemno = 0; ruleno = 1; @@ -1784,55 +1792,144 @@ packgram() ritem[itemno++] = p->sym->value; /* A rule gets by default the precedence and associativity of the last token in it. */ - if (p->sym->class == STOKEN) + if (p->sym->class == token_sym) { rprec[ruleno] = p->sym->prec; rassoc[ruleno] = p->sym->assoc; } - if (p) p = p->next; + if (p) + p = p->next; } /* If this rule has a %prec, - the specified symbol's precedence replaces the default. */ + the specified symbol's precedence replaces the default. */ if (ruleprec) { - rprec[ruleno] = ruleprec->prec; - rassoc[ruleno] = ruleprec->assoc; + rprec[ruleno] = ruleprec->prec; + rassoc[ruleno] = ruleprec->assoc; rprecsym[ruleno] = ruleprec->value; } ritem[itemno++] = -ruleno; ruleno++; - if (p) p = p->next; + if (p) + p = p->next; } ritem[itemno] = 0; } -/* Read a signed integer from STREAM and return its value. */ +/*-------------------------------------------------------------------. +| Read in the grammar specification and record it in the format | +| described in gram.h. All guards are copied into the FGUARD file | +| and all actions into FACTION, in each case forming the body of a C | +| function (YYGUARD or YYACTION) which contains a switch statement | +| to decide which guard or action to execute. | +`-------------------------------------------------------------------*/ -int -read_signed_integer (stream) - FILE *stream; +void +reader (void) { - register int c = getc(stream); - register int sign = 1; - register int n; + start_flag = 0; + startval = NULL; /* start symbol not specified yet. */ - if (c == '-') - { - c = getc(stream); - sign = -1; - } - n = 0; - while (isdigit(c)) - { - n = 10*n + (c - '0'); - c = getc(stream); - } +#if 0 + /* initially assume token number translation not needed. */ + translations = 0; +#endif + /* Nowadays translations is always set to 1, since we give `error' a + user-token-number to satisfy the Posix demand for YYERRCODE==256. + */ + translations = 1; + + nsyms = 1; + nvars = 0; + nrules = 0; + nitems = 0; + rline_allocated = 10; + rline = XCALLOC (short, rline_allocated); + + typed = 0; + lastprec = 0; + + gensym_count = 0; + + semantic_parser = 0; + pure_parser = 0; + + grammar = NULL; + + init_lex (); + lineno = 1; + + /* Initialize the symbol table. */ + tabinit (); + /* Construct the error token */ + errtoken = getsym ("error"); + errtoken->class = token_sym; + errtoken->user_token_number = 256; /* Value specified by POSIX. */ + /* Construct a token that represents all undefined literal tokens. + It is always token number 2. */ + undeftoken = getsym ("$undefined."); + undeftoken->class = token_sym; + undeftoken->user_token_number = 2; + + /* Read the declaration section. Copy %{ ... %} groups to FTABLE + and FDEFINES file. Also notice any %token, %left, etc. found + there. */ + putc ('\n', ftable); + fprintf (ftable, "\ +/* %s, made from %s\n\ + by GNU bison %s. */\n\ +\n", no_parser_flag ? "Bison-generated parse tables" : "A Bison parser", infile, VERSION); + + fputs ("#define YYBISON 1 /* Identify Bison output. */\n\n", ftable); + read_declarations (); + /* Start writing the guard and action files, if they are needed. */ + output_headers (); + /* Read in the grammar, build grammar in list form. Write out + guards and actions. */ + readgram (); + /* Now we know whether we need the line-number stack. If we do, + write its type into the .tab.h file. */ + if (fdefines) + reader_output_yylsp (fdefines); + /* Write closing delimiters for actions and guards. */ + output_trailers (); + if (locations_flag) + fputs ("#define YYLSP_NEEDED\n\n", ftable); + /* Assign the symbols their symbol numbers. Write #defines for the + token symbols into FDEFINES if requested. */ + packsymbols (); + /* Convert the grammar into the format described in gram.h. */ + packgram (); + /* Free the symbol table data structure since symbols are now all + referred to by symbol number. */ + free_symtab (); +} - ungetc(c, stream); - return n * sign; +void +reader_output_yylsp (FILE *f) +{ + if (locations_flag) + fputs ("\ +\n\ +#ifndef YYLTYPE\n\ +typedef struct yyltype\n\ +{\n\ + int timestamp;\n\ + int first_line;\n\ + int first_column;\ +\n\ + int last_line;\n\ + int last_column;\n\ + char *text;\n\ +} yyltype;\n\ +\n\ +# define YYLTYPE yyltype\n\ +#endif\n\ +\n", + f); }