3 # regen_perly.pl, DAPM 12-Feb-04
5 # Copyright (c) 2004, 2005 Larry Wall
7 # Given an input file perly.y, run bison on it and produce
8 # the following output files:
10 # perly.h standard bison header file with minor doctoring of
11 # #line directives plus adding a #ifdef PERL_CORE
13 # perly.tab the parser table C definitions extracted from the bison output
14 # plus an extra table generated by this script.
16 # perly.act the action case statements extracted from the bison output
18 # Note that perly.c is *not* regenerated - this is now a static file which
19 # is not dependent on perly.y any more.
21 # If a filename of the form foo.y is given on the command line, then
22 # this is used instead as the basename for all the files mentioned
25 # Note that temporary files of the form perlytmp.h and perlytmp.c are
26 # created and then deleted during this process
28 # Note also that this script is intended to be run on a UNIX system;
29 # it may work elsewhere but no specific attempt has been made to make it
32 sub usage { die "usage: $0 [ -b bison_executable ] [ file.y ]\n" }
37 BEGIN { require 'regen/regen_lib.pl'; }
41 if (@ARGV >= 2 and $ARGV[0] eq '-b') {
46 my $y_file = shift || 'perly.y';
48 usage unless @ARGV==0 && $y_file =~ /\.y$/;
50 (my $h_file = $y_file) =~ s/\.y$/.h/;
51 (my $act_file = $y_file) =~ s/\.y$/.act/;
52 (my $tab_file = $y_file) =~ s/\.y$/.tab/;
53 (my $tmpc_file = $y_file) =~ s/\.y$/tmp.c/;
54 (my $tmph_file = $y_file) =~ s/\.y$/tmp.h/;
55 my $new_h_file = "$h_file-new";
56 my $new_act_file = "$act_file-new";
57 my $new_tab_file = "$tab_file-new";
59 # the yytranslate[] table generated by bison is ASCII/EBCDIC sensitive
61 die "$0: must be run on an ASCII system\n" unless ord 'A' == 65;
63 # check for correct version number. The constraints are:
64 # * must be >= 1.24 to avoid licensing issues.
65 # * it must generate the yystos[] table. Version 1.28 doesn't generate
67 # * Must produce output which is extractable by the regexes below
68 # * Must produce the right values.
69 # These last two contstraints may well be met by earlier versions, but
70 # I simply haven't tested them yet. If it works for you, then modify
71 # the test below to allow that version too. DAPM Feb 04.
73 my $version = `$bison -V`;
74 unless ($version) { die <<EOF; }
75 Could not find a version of bison in your path. Please install bison.
78 unless ($version =~ /\b(1\.875[a-z]?|2\.[0134])\b/) { die <<EOF; }
80 You have the wrong version of bison in your path; currently 1.875
81 2.0, 2.1, 2.3 or 2.4 is required. Try installing
82 http://ftp.gnu.org/gnu/bison/bison-2.4.1.tar.gz
83 or similar. Your bison identifies itself as:
88 # creates $tmpc_file and $tmph_file
89 my_system("$bison -d -o $tmpc_file $y_file");
91 open CTMPFILE, $tmpc_file or die "Can't open $tmpc_file: $!\n";
93 { local $/; $clines = <CTMPFILE>; }
94 die "failed to read $tmpc_file: length mismatch\n"
95 unless length $clines == -s $tmpc_file;
98 my ($actlines, $tablines) = extract($clines);
100 $tablines .= make_type_tab($y_file, $tablines);
102 chmod 0644, $new_act_file;
103 open ACTFILE, ">$new_act_file" or die "can't open $new_act_file: $!\n";
104 print ACTFILE $actlines;
106 chmod 0444, $new_act_file;
108 chmod 0644, $new_tab_file;
109 open TABFILE, ">$new_tab_file" or die "can't open $new_tab_file: $!\n";
110 print TABFILE $tablines;
112 chmod 0444, $new_tab_file;
116 # Wrap PERL_CORE round the symbol definitions. Also, the
117 # C<#line 30 "perly.y"> confuses the Win32 resource compiler and the
118 # C<#line 188 "perlytmp.h"> gets picked up by make depend, so remove them.
120 open TMPH_FILE, $tmph_file or die "Can't open $tmph_file: $!\n";
121 chmod 0644, $new_h_file;
122 open H_FILE, ">$new_h_file" or die "Can't open $new_h_file: $!\n";
123 my $endcore_done = 0;
124 # Token macros need to be generated manually on bison 2.4
125 my $gather_tokens = ($version =~ /\b2\.4\b/ ? undef : 0);
127 while (<TMPH_FILE>) {
128 print H_FILE "#ifdef PERL_CORE\n" if $. == 1;
129 if (!$endcore_done and /YYSTYPE_IS_DECLARED/) {
130 print H_FILE "#endif /* PERL_CORE */\n";
133 next if /^#line \d+ ".*"/;
134 if (not defined $gather_tokens) {
135 $gather_tokens = 1 if /^\s* enum \s* yytokentype \s* \{/x;
137 elsif ($gather_tokens) {
138 if (/^\# \s* endif/x) { # The #endif just after the end of the token enum
140 $_ .= "\n/* Tokens. */\n$tokens";
143 my ($tok, $val) = /(\w+) \s* = \s* (\d+)/x;
144 $tokens .= "#define $tok $val\n" if $tok;
151 chmod 0444, $new_h_file;
154 rename_if_different($new_h_file, $h_file);
155 rename_if_different($new_tab_file, $tab_file);
156 rename_if_different($new_act_file, $act_file);
168 ^/* YYFINAL[^\n]+\n #optional comment
170 \# \s* define \s* YYFINAL # first #define
171 .*? # other defines + most tables
172 yystos\[\]\s*= # start of last table
174 }\s*; # end of last table
176 or die "Can't extract tables from $tmpc_file\n";
181 switch \s* \( \s* \w+ \s* \) \s* { \s*
185 (?: \s* /\* .*? \*/ \s* )* # optional C-comments
187 \#line [^\n]+"\Q$y_file\E"
192 (?: \s* /\* .*? \*/ \s* )* # optional C-comments
197 \#line[^\n]+\.simple"
202 or die "Can't extract actions from $tmpc_file\n";
205 # Remove extraneous comments from bison 2.4
206 $actlines =~ s!\s* /\* \s* Line \s* \d+ \s* of \s* yacc\.c \s* \*/!!gx;
208 # C<#line 188 "perlytmp.c"> gets picked up by make depend, so remove them.
209 $actlines =~ s/^#line \d+ "\Q$tmpc_file\E".*$//gm;
211 # convert yyvsp[nnn] into ps[nnn].val
213 $actlines =~ s/yyvsp\[(.*?)\]/ps[$1].val/g
214 or die "Can't convert value stack name\n";
216 return $actlines. "\n", $tablines. "\n";
219 # Generate a table, yy_type_tab[], that specifies for each token, what
220 # type of value it holds.
222 # Read the .y file and extract a list of all the token names and
223 # non-terminal names; then scan the string $tablines for the table yytname,
224 # which gives the token index of each token/non-terminal; then use this to
225 # create yy_type_tab.
227 # ie given (in perly.y),
234 # and (in $tablines),
236 # yytname[] = { "A" "B", "C", "D", "E" };
240 # typedef enum { toketype_ival, toketype_opval, toketype_pval } toketypes;
242 # static const toketypes yy_type_tab[]
243 # = { toketype_opval, toketype_ival, toketype_pval,
244 # toketype_opval, toketype_ival }
246 # where "E" has the default type. The default type is determined
247 # by the __DEFAULT__ comment next to the appropriate union member in
251 my ($y_file, $tablines) = @_;
255 open my $fh, '<', $y_file or die "Can't open $y_file: $!\n";
258 warn "$y_file:$.: dangerous assignment to $1: $_";
262 m{(\w+) \s* ; \s* /\* \s* __DEFAULT__}x
263 or die "$y_file: can't parse __DEFAULT__ line: $_";
264 die "$y_file: duplicate __DEFAULT__ line: $_"
265 if defined $default_token;
270 next unless /^%(token|type)/;
271 s/^%(token|type)\s+<(\w+)>\s+//
272 or die "$y_file: unparseable token/type line: $_";
273 $tokens{$_} = $2 for (split ' ', $_);
276 die "$y_file: no __DEFAULT__ token defined\n" unless $default_token;
277 $types{$default_token} = 1;
279 $tablines =~ /^\Qstatic const char *const yytname[] =\E\n
284 or die "Can't extract yytname[] from table string\n";
286 $fields =~ s{"([^"]+)"}
288 (defined $tokens{$1} ? $tokens{$1} : $default_token)
290 $fields =~ s/, \s* 0 \s* $//x
291 or die "make_type_tab: couldn't delete trailing ',0'\n";
294 "\ntypedef enum {\n\t"
295 . join(", ", map "toketype_$_", sort keys %types)
296 . "\n} toketypes;\n\n"
297 . "/* type of each token/terminal */\n"
298 . "static const toketypes yy_type_tab[] =\n{\n"
307 die "failed to execute command '@_': $!\n";
310 die sprintf "command '@_' died with signal %d\n",
314 die sprintf "command '@_' exited with value %d\n", $? >> 8;