X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/4dc58e7c764829fac17f2affe3c8b8ec8decaa87..131e2fef683836d45262423bbdf004a014335aba:/src/reader.c diff --git a/src/reader.c b/src/reader.c index 5e585e1b..cdacc071 100644 --- a/src/reader.c +++ b/src/reader.c @@ -1,5 +1,6 @@ /* Input parser for bison - Copyright (C) 1984, 86, 89, 92, 98, 2000 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. @@ -19,469 +20,423 @@ 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 "system.h" +#include "getargs.h" #include "files.h" -#include "alloc.h" +#include "xalloc.h" #include "symtab.h" #include "lex.h" #include "gram.h" -#include "machine.h" - -#define LTYPESTR "\ -\n\ -#ifndef YYLTYPE\n\ -typedef\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 int noparserflag; -extern int rawtoknumflag; -extern bucket *symval; -extern int numval; -extern int expected_conflicts; -extern char *token_buffer; -extern int maxtoken; - -extern void init_lex PARAMS((void)); -extern char *grow_token_buffer PARAMS((char *)); -extern void tabinit PARAMS((void)); -extern void output_headers PARAMS((void)); -extern void output_trailers PARAMS((void)); -extern void free_symtab PARAMS((void)); -extern void open_extra_files PARAMS((void)); -extern char *int_to_string PARAMS((int)); -extern char *printable_version PARAMS((int)); -extern void fatal PARAMS((char *)); -extern void fatals PARAMS((char *, char *)); -extern void warn PARAMS((char *)); -extern void warni PARAMS((char *, int)); -extern void warns PARAMS((char *, char *)); -extern void warnss PARAMS((char *, char *, char *)); -extern void warnsss PARAMS((char *, char *, char *, char *)); -extern void unlex PARAMS((int)); -extern void done PARAMS((int)); - -extern int skip_white_space PARAMS((void)); -extern int parse_percent_token PARAMS((void)); -extern int lex PARAMS((void)); - -typedef - struct symbol_list - { - struct symbol_list *next; - bucket *sym; - bucket *ruleprec; - } - symbol_list; - - -void reader PARAMS((void)); -void reader_output_yylsp PARAMS((FILE *)); -void read_declarations PARAMS((void)); -void copy_definition PARAMS((void)); -void parse_token_decl PARAMS((int, int)); -void parse_start_decl PARAMS((void)); -void parse_type_decl PARAMS((void)); -void parse_assoc_decl PARAMS((int)); -void parse_union_decl PARAMS((void)); -void parse_expect_decl PARAMS((void)); -char *get_type_name PARAMS((int, symbol_list *)); -void copy_guard PARAMS((symbol_list *, int)); -void parse_thong_decl PARAMS((void)); -void copy_action PARAMS((symbol_list *, int)); -bucket *gensym PARAMS((void)); -void readgram PARAMS((void)); -void record_rule_line PARAMS((void)); -void packsymbols PARAMS((void)); -void output_token_defines PARAMS((FILE *)); -void packgram PARAMS((void)); -int read_signed_integer PARAMS((FILE *)); +static int rline_allocated; -#if 0 -static int get_type PARAMS((void)); -#endif +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; -int *user_toknums; +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; - +/*===================\ +| Low level lexing. | +\===================*/ static void skip_to_char (int target) { int c; if (target == '\n') - warn(_(" Skipping to next \\n")); + complain (_(" Skipping to next \\n")); else - warni(_(" Skipping to next %c"), target); + complain (_(" Skipping to next %c"), target); do - c = skip_white_space(); + c = skip_white_space (); while (c != target && c != EOF); if (c != EOF) - ungetc(c, finput); + ungetc (c, finput); } -void -reader (void) +/*---------------------------------------------------------. +| Read a signed integer from STREAM and return its value. | +`---------------------------------------------------------*/ + +static inline int +read_signed_integer (FILE *stream) { - start_flag = 0; - startval = NULL; /* start symbol not specified yet. */ + int c = getc (stream); + int sign = 1; + int n = 0; -#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; + if (c == '-') + { + c = getc (stream); + sign = -1; + } - nsyms = 1; - nvars = 0; - nrules = 0; - nitems = 0; - rline_allocated = 10; - rline = NEW2(rline_allocated, short); + while (isdigit (c)) + { + n = 10 * n + (c - '0'); + c = getc (stream); + } - typed = 0; - lastprec = 0; + ungetc (c, stream); - gensym_count = 0; + return sign * n; +} + +/*--------------------------------------------------------------. +| Get the data type (alternative in the union) of the value for | +| symbol N in rule RULE. | +`--------------------------------------------------------------*/ - semantic_parser = 0; - pure_parser = 0; - yylsp_needed = 0; +static char * +get_type_name (int n, symbol_list * rule) +{ + int i; + symbol_list *rp; - grammar = NULL; + if (n < 0) + { + complain (_("invalid $ value")); + return NULL; + } - init_lex(); - lineno = 1; + rp = rule; + i = 0; - /* 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. */ - undeftoken = getsym("$undefined."); - undeftoken->class = STOKEN; - undeftoken->user_token_number = 2; - /* Read the declaration section. Copy %{ ... %} groups to ftable and fdefines file. - Also notice any %token, %left, etc. found there. */ - if (noparserflag) - fprintf(ftable, "\n/* Bison-generated parse tables, made from %s\n", - infile); - else - fprintf(ftable, "\n/* A Bison parser, made from %s\n", infile); - fprintf(ftable, " by %s */\n\n", VERSION_STRING); - fprintf(ftable, "#define YYBISON 1 /* Identify Bison output. */\n\n"); - 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 (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(); + while (i < n) + { + rp = rp->next; + if (rp == NULL || rp->sym == NULL) + { + complain (_("invalid $ value")); + return NULL; + } + i++; + } + + return rp->sym->type_name; } + +/*-------------------------------------------------------------------. +| Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of | +| the string (either ' or "). | +`-------------------------------------------------------------------*/ -void -reader_output_yylsp (FILE *f) +static inline void +copy_string (FILE *fin, FILE *fout, int match) { - if (yylsp_needed) - fprintf(f, LTYPESTR); + int c; + + putc (match, fout); + c = getc (fin); + + while (c != match) + { + if (c == EOF) + fatal (_("unterminated string at end of file")); + if (c == '\n') + { + complain (_("unterminated string")); + ungetc (c, fin); + c = match; /* invent terminator */ + continue; + } + + putc (c, fout); + + if (c == '\\') + { + c = getc (fin); + if (c == EOF) + fatal (_("unterminated string at end of file")); + putc (c, fout); + if (c == '\n') + lineno++; + } + + c = getc (fin); + } + + putc (c, fout); } -/* Read from finput until `%%' is seen. Discard the `%%'. Handle any - `%' declarations, and copy the contents of any `%{ ... %}' groups - to fattrs. */ -void -read_declarations (void) +/*---------------------------------------------------------------. +| Dump the comment from IN to OUT1 and OUT2. C is either `*' or | +| `/', depending upon the type of comments used. OUT2 might be | +| NULL. | +`---------------------------------------------------------------*/ + +static inline void +copy_comment2 (FILE *in, FILE *out1, FILE *out2, int c) { - register int c; - register int tok; + int cplus_comment; + int ended; - for (;;) - { - c = skip_white_space(); + cplus_comment = (c == '/'); + putc (c, out1); + if (out2) + putc (c, out2); + c = getc (in); - if (c == '%') + ended = 0; + while (!ended) + { + if (!cplus_comment && c == '*') { - tok = parse_percent_token(); + while (c == '*') + { + putc (c, out1); + if (out2) + putc (c, out2); + c = getc (in); + } - switch (tok) + if (c == '/') { - case TWO_PERCENTS: - return; + 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 (in); + } + else if (c == EOF) + fatal (_("unterminated comment")); + else + { + putc (c, out1); + if (out2) + putc (c, out2); + c = getc (in); + } + } +} - case PERCENT_LEFT_CURLY: - copy_definition(); - break; - case TOKEN: - parse_token_decl (STOKEN, SNTERM); - break; +/*------------------------------------------------------------. +| Dump the comment from FIN to FOUT. C is either `*' or `/', | +| depending upon the type of comments used. | +`------------------------------------------------------------*/ - case NTERM: - parse_token_decl (SNTERM, STOKEN); - break; +static inline void +copy_comment (FILE *fin, FILE *fout, int c) +{ + copy_comment2 (fin, fout, NULL, c); +} - case TYPE: - parse_type_decl(); - break; - case START: - parse_start_decl(); - break; +/*-----------------------------------------------------------------. +| 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 `$0' with | +| respect to the top of the stack. | +`-----------------------------------------------------------------*/ - case UNION: - parse_union_decl(); - break; +static inline void +copy_at (FILE *fin, FILE *fout, int stack_offset) +{ + int c; - case EXPECT: - parse_expect_decl(); - break; - case THONG: - parse_thong_decl(); - break; - case LEFT: - parse_assoc_decl(LEFT_ASSOC); - break; + c = getc (fin); + if (c == '$') + { + fprintf (fout, "yyloc"); + locations_flag = 1; + } + else if (isdigit (c) || c == '-') + { + int n; - case RIGHT: - parse_assoc_decl(RIGHT_ASSOC); - break; + ungetc (c, fin); + n = read_signed_integer (fin); - case NONASSOC: - parse_assoc_decl(NON_ASSOC); - break; + fprintf (fout, "yylsp[%d]", n - stack_offset); + locations_flag = 1; + } + else + { + char buf[] = "@c"; + buf[1] = c; + complain (_("%s is invalid"), quote (buf)); + } +} - case SEMANTIC_PARSER: - if (semantic_parser == 0) - { - semantic_parser = 1; - open_extra_files(); - } - break; - case PURE_PARSER: - pure_parser = 1; - break; +/*-------------------------------------------------------------------. +| 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 `$0' with respect to the top of the stack. | +`-------------------------------------------------------------------*/ - case NOOP: - break; +static inline void +copy_dollar (FILE *fin, FILE *fout, + symbol_list *rule, int stack_offset) +{ + int c = getc (fin); + char *type_name = NULL; - default: - warns(_("unrecognized: %s"), token_buffer); - skip_to_char('%'); - } - } - else if (c == EOF) - fatal(_("no input grammar")); - else + /* Get the typename if explicit. */ + if (c == '<') + { + char *cp = token_buffer; + + while ((c = getc (fin)) != '>' && c > 0) { - warns (_("unknown character: %s"), printable_version(c)); - skip_to_char('%'); + if (cp == token_buffer + maxtoken) + cp = grow_token_buffer (cp); + + *cp++ = c; } - } -} + *cp = 0; + type_name = token_buffer; + value_components_used = 1; + c = getc (fin); + } -/* copy the contents of a %{ ... %} into the definitions file. -The %{ has already been read. Return after reading the %}. */ + if (c == '$') + { + 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 `%}'. | +`-------------------------------------------------------------------*/ -void +static void copy_definition (void) { - 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; + int c; + /* -1 while reading a character if prev char was %. */ + int after_percent; - if (!nolinesflag) - fprintf(fattrs, "#line %d \"%s\"\n", lineno, infile); + 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) - fatal(_("unterminated string at end of file")); - if (c == '\n') - { - warn(_("unterminated string")); - ungetc(c, finput); - c = match; - continue; - } - - putc(c, fattrs); - - if (c == '\\') - { - c = getc(finput); - if (c == EOF) - fatal(_("unterminated string at end of file")); - 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); + 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, c); 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; @@ -490,18 +445,18 @@ copy_definition (void) } +/*-------------------------------------------------------------------. +| 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 (int what_is, int what_is_not) +static void +parse_token_decl (symbol_class what_is, symbol_class what_is_not) { - register int token = 0; - register char *typename = 0; - register struct bucket *symbol = NULL; /* pts to symbol being defined */ - int k; + int token = 0; + char *typename = 0; + struct bucket *symbol = NULL; /* pts to symbol being defined */ for (;;) { @@ -510,9 +465,9 @@ parse_token_decl (int what_is, int what_is_not) if (tmp_char == '%') return; if (tmp_char == EOF) - fatals ("Premature EOF after %s", token_buffer); + fatal (_("Premature EOF after %s"), token_buffer); - token = lex(); + token = lex (); if (token == COMMA) { symbol = NULL; @@ -520,26 +475,31 @@ parse_token_decl (int what_is, int what_is_not) } 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) + 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; - symval->class = STOKEN; - symval->type_name = typename; - symval->user_token_number = symbol->user_token_number; - symbol->user_token_number = SALIAS; - - symval->alias = symbol; - symbol->alias = symval; - symbol = NULL; - - nsyms--; /* symbol and symval combined are only one symbol */ + symbol = NULL; } else if (token == IDENTIFIER) { @@ -547,121 +507,46 @@ parse_token_decl (int what_is, int what_is_not) symbol = symval; if (symbol->class == what_is_not) - warns(_("symbol %s redefined"), symbol->tag); + complain (_("symbol %s redefined"), symbol->tag); symbol->class = what_is; - if (what_is == SNTERM && oldclass != SNTERM) + if (what_is == nterm_sym && oldclass != nterm_sym) symbol->value = nvars++; if (typename) { if (symbol->type_name == NULL) symbol->type_name = typename; - else if (strcmp(typename, symbol->type_name) != 0) - warns(_("type redeclaration for %s"), symbol->tag); + else if (strcmp (typename, symbol->type_name) != 0) + complain (_("type redeclaration for %s"), symbol->tag); } } else if (symbol && token == NUMBER) - { + { symbol->user_token_number = numval; translations = 1; - } + } else { - warnss(_("`%s' is invalid in %s"), - token_buffer, - (what_is == STOKEN) ? "%token" : "%nterm"); - skip_to_char('%'); - } - } - -} - -/* 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 -*/ - -void -parse_thong_decl (void) -{ - register int token; - register struct bucket *symbol; - register char *typename = 0; - int k, usrtoknum; - - translations = 1; - token = lex(); /* fetch typename or first token */ - if (token == TYPENAME) { - k = strlen(token_buffer); - typename = NEW2(k + 1, char); - strcpy(typename, token_buffer); - value_components_used = 1; - token = lex(); /* fetch first token */ - } - - /* process first token */ - - if (token != IDENTIFIER) - { - warns(_("unrecognized item %s, expected an identifier"), - token_buffer); - skip_to_char('%'); - return; - } - symval->class = STOKEN; - 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; - - /* process literal string token */ - - if (token != IDENTIFIER || *symval->tag != '\"') - { - warns(_("expected string constant instead of %s"), - token_buffer); - skip_to_char('%'); - return; - } - symval->class = STOKEN; - symval->type_name = typename; - symval->user_token_number = usrtoknum; - - symval->alias = symbol; - symbol->alias = symval; + complain (_("`%s' is invalid in %s"), + token_buffer, (what_is == token_sym) ? "%token" : "%nterm"); + skip_to_char ('%'); + } + } - nsyms--; /* symbol and symval combined are only one symbol */ } -/* parse what comes after %start */ +/*------------------------------. +| Parse what comes after %start | +`------------------------------*/ -void +static void parse_start_decl (void) { if (start_flag) - warn(_("multiple %start declarations")); - if (lex() != IDENTIFIER) - warn(_("invalid %start declaration")); + complain (_("multiple %s declarations"), "%start"); + if (lex () != IDENTIFIER) + complain (_("invalid %s declaration"), "%start"); else { start_flag = 1; @@ -669,38 +554,36 @@ parse_start_decl (void) } } +/*-----------------------------------------------------------. +| 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 +static void parse_type_decl (void) { - register int k; - register char *name; + char *name; - if (lex() != TYPENAME) + if (lex () != TYPENAME) { - warn(_("%type declaration has no ")); - skip_to_char('%'); + complain ("%s", _("%type declaration has no ")); + skip_to_char ('%'); return; } - k = strlen(token_buffer); - name = NEW2(k + 1, char); - strcpy(name, token_buffer); + name = xstrdup (token_buffer); for (;;) { - register int t; + int t; int tmp_char = ungetc (skip_white_space (), finput); if (tmp_char == '%') return; if (tmp_char == EOF) - fatals ("Premature EOF after %s", token_buffer); + fatal (_("Premature EOF after %s"), token_buffer); - t = lex(); + t = lex (); switch (t) { @@ -712,51 +595,50 @@ parse_type_decl (void) case IDENTIFIER: if (symval->type_name == NULL) symval->type_name = name; - else if (strcmp(name, symval->type_name) != 0) - warns(_("type redeclaration for %s"), symval->tag); + else if (strcmp (name, symval->type_name) != 0) + complain (_("type redeclaration for %s"), symval->tag); break; default: - warns(_("invalid %%type declaration due to item: `%s'"), token_buffer); - skip_to_char('%'); + 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 (int assoc) +static void +parse_assoc_decl (associativity assoc) { - register int k; - register char *name = NULL; - register int prev = 0; + char *name = NULL; + int prev = 0; - lastprec++; /* Assign a new precedence level, never 0. */ + lastprec++; /* Assign a new precedence level, never 0. */ for (;;) { - register int t; + int t; int tmp_char = ungetc (skip_white_space (), finput); if (tmp_char == '%') return; if (tmp_char == EOF) - fatals ("Premature EOF after %s", token_buffer); + fatal (_("Premature EOF after %s"), token_buffer); - 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: @@ -764,41 +646,42 @@ parse_assoc_decl (int assoc) case IDENTIFIER: if (symval->prec != 0) - warns(_("redefining precedence of %s"), symval->tag); + complain (_("redefining precedence of %s"), symval->tag); symval->prec = lastprec; symval->assoc = assoc; - if (symval->class == SNTERM) - warns(_("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 if (strcmp(name, symval->type_name) != 0) - warns(_("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 - { - warns(_("invalid text (%s) - number should be after identifier"), - token_buffer); - skip_to_char('%'); - } + } + else + { + complain (_ + ("invalid text (%s) - number should be after identifier"), +token_buffer); + skip_to_char ('%'); + } break; case SEMICOLON: return; default: - warns(_("unexpected item: %s"), token_buffer); - skip_to_char('%'); + complain (_("unexpected item: %s"), token_buffer); + skip_to_char ('%'); } prev = t; @@ -808,42 +691,39 @@ parse_assoc_decl (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 +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) - warn(_("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) { @@ -852,50 +732,10 @@ parse_union_decl (void) break; case '/': - c = getc(finput); + 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 at end of file")); - - if (!cplus_comment && c == '*') - { - c = getc(finput); - if (c == '/') - { - putc('/', fattrs); - if (fdefines) - putc('/', fdefines); - in_comment = 0; - } - } - else - c = getc(finput); - } - } + continue; + copy_comment2 (finput, fattrs, fdefines, c); break; @@ -905,234 +745,362 @@ parse_union_decl (void) case '}': if (count == 0) - warn (_("unmatched close-brace (`}')")); + complain (_("unmatched %s"), "`}'"); count--; 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 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]; - - c = getc(finput); - while (c == ' ' || c == '\t') - c = getc(finput); - - count = 0; - while (c >= '0' && c <= '9') - { - if (count < 20) - buffer[count++] = c; - c = getc(finput); - } - buffer[count] = 0; - + int c = skip_white_space (); ungetc (c, finput); - if (count <= 0 || count > 10) - warn(_("argument of %expect is not an integer")); - expected_conflicts = atoi (buffer); + if (!isdigit (c)) + complain (_("argument of %%expect is not an integer")); + else + expected_conflicts = read_signed_integer (finput); } -/* 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 (int n, symbol_list *rule) +/*-------------------------------------------------------------------. +| 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) { - static char *msg = N_("invalid $ value"); + int token; + struct bucket *symbol; + char *typename = 0; + int usrtoknum; + + translations = 1; + token = lex (); /* fetch typename or first token */ + if (token == TYPENAME) + { + typename = xstrdup (token_buffer); + value_components_used = 1; + token = lex (); /* fetch first token */ + } - register int i; - register symbol_list *rp; + /* process first token */ - if (n < 0) + if (token != IDENTIFIER) { - warn(_(msg)); - return NULL; + 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; - rp = rule; - i = 0; + token = lex (); /* get number or literal string */ - while (i < n) + if (token == NUMBER) { - rp = rp->next; - if (rp == NULL || rp->sym == NULL) - { - warn(_(msg)); - return NULL; - } - i++; + usrtoknum = numval; + token = lex (); /* okay, did number, now get literal */ + } + else + usrtoknum = 0; + + /* 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; - return (rp->sym->type_name); + /* symbol and symval combined are only one symbol. */ + nsyms--; } -/* Dump the string from FINPUT to FOUTPUT. MATCH is the delimiter of - the string (either ' or "). */ +/*----------------------------------------------------------------. +| Read from finput until `%%' is seen. Discard the `%%'. Handle | +| any `%' declarations, and copy the contents of any `%{ ... %}' | +| groups to fattrs. | +`----------------------------------------------------------------*/ -static inline void -copy_string (FILE *finput, FILE *foutput, int match) +static void +read_declarations (void) { int c; + int tok; - putc (match, foutput); - c = getc (finput); - - while (c != match) + for (;;) { - if (c == EOF) - fatal(_("unterminated string at end of file")); - if (c == '\n') + c = skip_white_space (); + + if (c == '%') { - warn (_("unterminated string")); - ungetc (c, finput); - c = match; /* invent terminator */ - continue; - } + tok = parse_percent_token (); + + switch (tok) + { + case TWO_PERCENTS: + return; - putc(c, foutput); + case PERCENT_LEFT_CURLY: + copy_definition (); + break; - if (c == '\\') + 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; + + case NOOP: + break; + + default: + complain (_("unrecognized: %s"), token_buffer); + skip_to_char ('%'); + } + } + else if (c == EOF) + fatal (_("no input grammar")); + else { - c = getc (finput); - if (c == EOF) - fatal (_("unterminated string")); - putc (c, foutput); - if (c == '\n') - lineno++; + char buf[] = "c"; + buf[0] = c; + complain (_("unknown character: %s"), quote (buf)); + skip_to_char ('%'); } - - c = getc(finput); } - - putc(c, foutput); } + +/*-------------------------------------------------------------------. +| 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; -/* Dump the comment from FINPUT to FOUTPUT. C is either `*' or `/', - depending upon the type of comments used. */ + /* offset is always 0 if parser has already popped the stack pointer */ + if (semantic_parser) + stack_offset = 0; -static inline void -copy_comment (FILE *finput, FILE *foutput, int c) -{ - int cplus_comment; - register int match; - register int ended; + fprintf (faction, "\ncase %d:\n", nrules); + if (!no_lines_flag) + fprintf (faction, "#line %d \"%s\"\n", lineno, infile); + putc ('{', faction); - cplus_comment = (c == '/'); - putc (c, foutput); + count = 1; c = getc (finput); - ended = 0; - while (!ended) + while (count > 0) { - if (!cplus_comment && c == '*') + while (c != '}') { - while (c == '*') + switch (c) { - putc(c, foutput); - c = getc(finput); - } + case '\n': + putc (c, faction); + lineno++; + break; - if (c == '/') - { - putc(c, foutput); - ended = 1; + case '{': + putc (c, faction); + count++; + break; + + case '\'': + case '"': + copy_string (finput, faction, c); + break; + + case '/': + putc (c, faction); + c = getc (finput); + if (c != '*' && c != '/') + continue; + copy_comment (finput, faction, c); + 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); } - else if (c == '\n') - { - lineno++; - putc (c, foutput); - if (cplus_comment) - ended = 1; - else - c = getc(finput); - } - else if (c == EOF) - fatal (_("unterminated comment")); - else + + /* above loop exits when c is '}' */ + + if (--count) { - putc (c, foutput); + putc (c, faction); c = getc (finput); } } -} -/* 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. */ + 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. | +`-------------------------------------------------------------------*/ -void +static void copy_guard (symbol_list *rule, int stack_offset) { - register int c; - register int n; - register int count; - register char *type_name; + int c; + int count; int brace_flag = 0; /* 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, "\ncase %d:\n", nrules); + if (!no_lines_flag) fprintf (fguard, "#line %d \"%s\"\n", lineno, infile); - putc('{', fguard); + 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 { - warn(_("unmatched right brace (`}')")); - c = getc(finput); /* skip it */ + complain (_("unmatched %s"), "`}'"); + c = getc (finput); /* skip it */ } - break; + break; case '\'': case '"': @@ -1145,313 +1113,163 @@ copy_guard (symbol_list *rule, int stack_offset) if (c != '*' && c != '/') continue; copy_comment (finput, fguard, c); - break; - - case '$': - c = getc(finput); - type_name = NULL; - - if (c == '<') - { - register char *cp = token_buffer; - - while ((c = getc(finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer(cp); - - *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) - warns(_("$$ of `%s' has no declared type"), 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); + break; - fprintf(fguard, "yyvsp[%d]", n - stack_offset); - if (type_name) - fprintf(fguard, ".%s", type_name); - if (!type_name && typed) - warnss (_("$%s of `%s' has no declared type"), - int_to_string(n), rule->sym->tag); - continue; - } - else - warns(_("$%s is invalid"), printable_version(c)); + case '$': + copy_dollar (finput, fguard, rule, stack_offset); break; case '@': - c = getc (finput); - if (c == '$') - { - fprintf (fguard, "yyloc"); - yylsp_needed = 1; - } - else if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer (finput); - c = getc (finput); - fprintf (fguard, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; - continue; - } - else - { - warns (_("@%s is invalid"), printable_version (c)); - n = 1; - } + copy_at (finput, fguard, stack_offset); break; case EOF: - fatal (_("unterminated %%guard clause")); + fatal ("%s", _("unterminated %guard clause")); default: 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); /* why not skip_white_space -wjh */ + 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 `$0' - with respect to the top of the stack. */ - -void -copy_action (symbol_list *rule, int stack_offset) +static void +record_rule_line (void) { - register int c; - register int n; - register int count; - register char *type_name; - - /* offset is always 0 if parser has already popped the stack pointer */ - if (semantic_parser) - stack_offset = 0; - - fprintf (faction, "\ncase %d:\n", nrules); - if (!nolinesflag) - fprintf (faction, "#line %d \"%s\"\n", lineno, infile); - putc ('{', faction); - - count = 1; - c = getc(finput); + /* Record each rule's source line number in rline table. */ - while (count > 0) + if (nrules >= rline_allocated) { - while (c != '}') - { - switch (c) - { - case '\n': - putc(c, faction); - lineno++; - break; + rline_allocated = nrules * 2; + rline = XREALLOC (rline, short, rline_allocated); + } + rline[nrules] = lineno; +} - case '{': - putc(c, faction); - count++; - break; - case '\'': - case '"': - copy_string (finput, faction, c); - break; +/*-------------------------------------------------------------------. +| Generate a dummy symbol, a nonterminal, whose name cannot conflict | +| with the user's names. | +`-------------------------------------------------------------------*/ - case '/': - putc(c, faction); - c = getc(finput); - if (c != '*' && c != '/') - continue; - copy_comment (finput, faction, c); - break; +static bucket * +gensym (void) +{ + bucket *sym; - case '$': - c = getc(finput); - type_name = NULL; + sprintf (token_buffer, "@%d", ++gensym_count); + sym = getsym (token_buffer); + sym->class = nterm_sym; + sym->value = nvars++; + return sym; +} - if (c == '<') - { - register char *cp = token_buffer; +#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 | +`------------------------------------------------------------------*/ - while ((c = getc(finput)) != '>' && c > 0) - { - if (cp == token_buffer + maxtoken) - cp = grow_token_buffer(cp); +static int +get_type (void) +{ + int k; + int t; + char *name; - *cp++ = c; - } - *cp = 0; - type_name = token_buffer; - value_components_used = 1; + t = lex (); - 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) - warns(_("$$ of `%s' has no declared type"), - 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) - warnss(_("$%s of `%s' has no declared type"), - int_to_string(n), rule->sym->tag); - continue; - } - else - warns(_("$%s is invalid"), printable_version(c)); + if (t != TYPENAME) + { + complain (_("invalid %s declaration"), "%type"); + return t; + } - break; + name = xstrdup (token_buffer); - case '@': - c = getc (finput); - if (c == '$') - { - fprintf (faction, "yyloc"); - yylsp_needed = 1; - } - else if (isdigit(c) || c == '-') - { - ungetc (c, finput); - n = read_signed_integer (finput); - c = getc (finput); - fprintf (faction, "yylsp[%d]", n - stack_offset); - yylsp_needed = 1; - continue; - } - else - { - warns (_("@%s is invalid"), printable_version (c)); - n = 1; - } - break; + 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 (void) -{ - 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. */ +#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. | +`------------------------------------------------------------------*/ -void +static void readgram (void) { - register int t; - register bucket *lhs = NULL; - 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; @@ -1465,17 +1283,17 @@ readgram (void) start_flag = 1; } - t = lex(); + t = lex (); if (t != COLON) { - warn(_("ill-formed rule: initial symbol not followed by colon")); - unlex(t); + complain (_("ill-formed rule: initial symbol not followed by colon")); + unlex (t); } } if (nrules == 0 && t == BAR) { - warn(_("grammar starts with vertical bar")); + complain (_("grammar starts with vertical bar")); lhs = symval; /* BOGUS: use a random symval */ } /* start a new rule and record its lhs. */ @@ -1485,7 +1303,7 @@ readgram (void) record_rule_line (); - p = NEW(symbol_list); + p = XCALLOC (symbol_list, 1); p->sym = lhs; crule1 = p1; @@ -1499,61 +1317,63 @@ readgram (void) /* 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) - warns(_("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(); + t = lex (); crule->ruleprec = symval; - t = lex(); + 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 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 @@ -1561,93 +1381,93 @@ readgram (void) 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. */ 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 */ + } /* 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) { - warn(_("two @prec's in a row")); - t = lex(); + complain (_("two @prec's in a row")); + t = lex (); crule->ruleprec = symval; - t = lex(); + t = lex (); } if (t == GUARD) { - if (! semantic_parser) - warn(_("%%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) { - /* This case never occurs -wjh */ - if (actionflag) - warn(_("two actions at end of one rule")); - copy_action(crule, rulelength); - actionflag = 1; + /* 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(); + t = lex (); } - /* If $$ is being set in default way, warn if any type + /* If $$ is being set in default way, report if any type mismatch. */ else if (!xactions - && first_rhs - && lhs->type_name != first_rhs->type_name) + && first_rhs && lhs->type_name != first_rhs->type_name) { if (lhs->type_name == 0 || first_rhs->type_name == 0 - || strcmp(lhs->type_name,first_rhs->type_name)) - warnss(_("type clash (`%s' `%s') on default action"), - lhs->type_name ? lhs->type_name : "", - first_rhs->type_name ? first_rhs->type_name : ""); + || strcmp (lhs->type_name, first_rhs->type_name)) + complain (_("type clash (`%s' `%s') on default action"), + lhs->type_name ? lhs->type_name : "", + first_rhs->type_name ? first_rhs->type_name : ""); } /* Warn if there is no default for $$ but we need one. */ else if (!xactions && !first_rhs && lhs->type_name != 0) - warn(_("empty rule for typed nonterminal, and no action")); + 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. */ + /* 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 % @@ -1655,169 +1475,158 @@ readgram (void) */ 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 { - warns(_("invalid input: %s"), token_buffer); - t = lex(); + 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 %s"), - int_to_string(MAXSHORT)); + fatal (_("too many symbols (tokens plus nonterminals); maximum %d"), + MAXSHORT); if (nrules == 0) - fatal(_("no rules in the 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) { - warns(_("symbol %s is used, but is not defined as a token and has no rules"), - bp->tag); - 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 (void) -{ - /* Record each rule's source line number in rline table. */ - - if (nrules >= rline_allocated) - { - rline_allocated = nrules * 2; - rline = (short *) xrealloc ((char *) rline, - rline_allocated * sizeof (short)); - } - rline[nrules] = lineno; -} - - -#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) +static void +output_token_defines (FILE *file) { - register int k; - register int t; - register char *name; - - t = lex(); - - if (t != TYPENAME) - { - warn(_("ill-formed %type declaration")); - return t; - } - - 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(); - - switch (t) + symbol = bp->tag; /* get symbol */ + + 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; - else if (strcmp(name, symval->type_name) != 0) - warns(_("type redeclaration for %s"), symval->tag); + /* use literal string only if given a symbol with an alias */ + if (bp->alias) + symbol = bp->alias->tag; + else + 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); } -#endif -/* 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 +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[] = "$"; /* int lossage = 0; JF set but not used */ - tags = NEW2(nsyms + 1, char *); - tags[0] = "$"; - user_toknums = NEW2(nsyms + 1, int); + 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; } @@ -1833,8 +1642,8 @@ packsymbols (void) { if (bp->prec != 0 && bp->alias->prec != 0 && bp->user_token_number == SALIAS) - warnss(_("conflicting precedences for %s and %s"), - bp->tag, bp->alias->tag); + complain (_("conflicting precedences for %s and %s"), + bp->tag, bp->alias->tag); if (bp->prec != 0) bp->alias->prec = bp->prec; else @@ -1843,26 +1652,26 @@ packsymbols (void) if (bp->assoc != bp->alias->assoc) { - if (bp->assoc != 0 && bp->alias->assoc != 0 - && bp->user_token_number == SALIAS) - warnss(_("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->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 */ + continue; /* do not do processing below for SALIASs */ - } - else /* bp->class == STOKEN */ + } + else /* bp->class == token_sym */ { bp->value = tokno++; } - if (bp->class == STOKEN) + if (bp->class == token_sym) { if (translations && !(bp->user_token_number)) bp->user_token_number = ++last_user_token_number; @@ -1879,53 +1688,53 @@ packsymbols (void) 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; + 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; /* non-terminal */ - if (bp->user_token_number == SALIAS) + { + if (bp->value >= ntokens) + continue; /* non-terminal */ + if (bp->user_token_number == SALIAS) continue; - if (token_translations[bp->user_token_number] != 2) - warnsss(_("tokens %s and %s both assigned number %s"), - tags[token_translations[bp->user_token_number]], - bp->tag, - int_to_string(bp->user_token_number)); - token_translations[bp->user_token_number] = bp->value; - } + if (token_translations[bp->user_token_number] != 2) + 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; - if (! noparserflag) - 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) @@ -1933,80 +1742,38 @@ packsymbols (void) { /* 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 } } -/* For named tokens, but not literal ones, define the name. The value - is the user token number. */ - -void -output_token_defines (FILE *file) -{ - bucket *bp; - register char *cp, *symbol; - register char c; - - for (bp = firstsymbol; bp; bp = bp->next) - { - symbol = bp->tag; /* get symbol */ - - 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) - { - /* use literal string only if given a symbol with an alias */ - if (bp->alias) - symbol = bp->alias->tag; - else - continue; - } - - /* Don't #define nonliteral tokens whose names contain periods. */ - cp = symbol; - while ((c = *cp++) && c != '.'); - if (c != '\0') continue; - - fprintf(file, "#define\t%s\t%d\n", symbol, - ((translations && ! rawtoknumflag) - ? bp->user_token_number - : bp->value)); - if (semantic_parser) - fprintf(file, "#define\tT%s\t%d\n", symbol, 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 +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; @@ -2024,54 +1791,144 @@ packgram (void) 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 (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); }