X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/bb92250c78de62973527486051533f6dec9afeee..41f83caf1cbad0e79f7fcd3e35386a523e463784:/ChangeLog diff --git a/ChangeLog b/ChangeLog index 426b5c55..b96950b0 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,154 @@ +2002-12-08 Paul Eggert + + * data/glr.c, data/lalr1.cc, data/yacc.c: Do not use __file__, + since this causes problems when __file__ contains character + sequences like "@" that are treated specially by src/scan-skel.l. + Instead, just use the file's basename. This fixes the bug + reported by Martin Mokrejs in + . + +2002-12-06 Paul Eggert + + Add support for rules that do not have trailing semicolons, as + POSIX requires. Improve the quality of locations in Bison + diagnostics. + + * src/location.c: Include . + (empty_location): Now const. + (location_print): New function. Follow the recommendation of the + GNU Coding Standards for locations that span file boundaries. + * src/location.h: Do not include ; no longer needed. + (boundary): New type. + (location_t): Use it. This allows locations to span file boundaries. + All member uses changed: file -> start.file or end.file (as needed), + first_line -> start.line, first_column -> start.column, + last_line -> end.line, last_column -> end.column. + (equal_boundaries): New function. + (LOCATION_RESET, LOCATION_STEP): Remove. + (LOCATION_PRINT): Remove. All callers changed to use location_print. + (empty_location): Now const. + (location_print): New decl. + * src/parse-gram.y (lloc_default): New function, which handles + empty locations more accurately. + (YYLLOC_DEFAULT): Use it. + (%token COLON): Remove. + (%token ID_COLON): New token. + (rules): Use it. + (declarations, rules): Remove trailing semicolon. + (declaration, rules_or_grammar_declaration): + Allow empty (";") declaration. + (symbol_def): Remove empty actions; no longer needed. + (rules_or_grammar_declaration): Remove trailing semicolon. + (semi_colon.opt): Remove. + * src/reader.h: Include location.h. + (scanner_cursor): New decl. + * src/reduce.c (nonterminals_reduce): Use warn_at rather than + rolling our own. + * src/scan-gram.l (YY_USER_INIT): Initialize scanner_cursor instead + of *loc. + (STEP): Remove. No longer needed, now that adjust_location does + the work. All uses removed. + (scanner_cursor): New var. + (adjust_location): Renamed from extend_location. It now sets + *loc and adjusts the scanner cursor. All uses changed. + Don't bother testing for CR. + (handle_syncline): Remove location arg; now updates scanner cursor. + All callers changed. + (unexpected_end_of_file): Now accepts start boundary of token or + comment, not location. All callers changed. Update scanner cursor, + not the location. + (SC_AFTER_IDENTIFIER): New state. + (context_state): Renamed from c_context. All uses changed. + (id_loc, code_start, token_start): New local vars. + (): New initial context. Move all + processing of Yacc white space and equivalents here. + ({id}): Save id_loc. Begin state SC_AFTER_IDENTIFIER + instead of returning ID immediately, since we need to search for + a subsequent colon. + ("'", "\""): Save token_start. + ("%{", "{", "%%"): Save code_start. + (): New state, looking for a colon. + (, , ): + BEGIN context_state at end, not INITIAL. + ("\"", "'", + "}", "%}", <>): + Return correct token start. + (): Save start boundary when + the start of a character, string or multiline comment is found. + * tests/conflicts.at (S/R in initial, Defaulted Conflicted + Reduction): Adjust reported locations to match the more-precise + results now expected. + * tests/input.at (Invalid $n, Invalid @n, Type Clashes): Likewise. + * tests/reduce.at (Useless Rules, Reduced Automaton, + Underivable Rules): Likewise. + * tests/regression.at (Invalid inputs): No longer `expecting ";" + or "|"' now that so many other tokens are allowed by the new grammar. + + * src/complain.h (current_file): Remove duplicate decl; + current_file is now owned by files.h. + * src/complain.c, src/scan-gram.l: Include files.h. + +2002-12-06 Paul Eggert + + * data/glr.c (yy_reduce_print): Don't assume that yyrline[yyrule] + promotes to int; it might be unsigned int. + * data/yacc.c (yy_reduce_print): Likewise. + + * doc/bison.texinfo (Table of Symbols): YYERROR_VERBOSE should + be #defined in the prologue, not in the Bison declarations. + This fixes Debian Bug 102878, reported by Shaul Karl. + +2002-12-02 Paul Eggert + + * configure.ac (AC_REPLACE_FUNCS): Add strtoul. + * lib/strtoul.c: New file, from gnulib. + This fixes a porting bug reported by Peter Klein in + . + +2002-11-30 Paul Eggert + + * src/scan-gram.l (no_cr_read, extend_location): Move to epilogue, + and put only a forward declaration in the prologue. This is for + consistency with the other scanner helper functions. + + Type clashes now generate warnings, not errors, since it + appears that POSIX may allow some grammars with type clashes. + * src/reader.c (grammar_current_rule_check): Warn about + type clashes instead of complaining. + * tests/input.at (Type Clashes): Expect warnings, not complaints. + + Add Yacc library, since POSIX requires it. + * doc/bison.texinfo (Yacc Library): New node. Regenerate top menu. + * lib/Makefile.am (lib_LIBRARIES, liby_a_SOURCES): New macros. + * lib/main.c, lib/yyerror.c: New files. + + gram_error can be static; it need not be extern. + * src/reader.h (gram_error): Remove decl. + * src/parse-gram.y (gram_error): Now static. Add static decl. + (print_token_value): Omit parameter names from forward decl, + for consistency. + +2002-11-29 Paul Eggert + + * doc/bison.texinfo: Emphasize that yylex and yyerror must + be declared before being used. E.g., one should typically + declare them in the prologue. Use GNU coding style in examples. + Put "const" consistently after the type it modifies. Mention + that C99 supports "inline". Mention that yyerror traditionally + returns "int". + + %parse-param and %lex-param now take just one argument, the + declaration; the argument name is deduced from the declaration. + + * doc/bison.texinfo (Parser Function, Pure Calling, Error + Reporting, Table of Symbols): Document this. + * src/parse-gram.y (add_param): New function. + (COMMA): Remove. + (declaration): Implement new rule for %parse-param and %lex-param. + * src/scan-gram.l: "," now elicits a warning, rather than being + a token; this is more compatible with byacc. + * tests/calc.at (Simple LALR Calculator): Adopt new convention. + 2002-11-27 Paul Eggert Rename identifiers to avoid real and potential collisions.