# from information stored in
#
# regcomp.sym
+# op_reg_common.h
# regexp.h
#
# pod/perldebguts.pod is not completely regenerated. Only the table of
# Accepts the standard regen_lib -q and -v args.
#
# This script is normally invoked from regen.pl.
+#
+# F<regcomp.sym> defines the opcodes and states used in the regex
+# engine, it also includes documentation on the opcodes. This script
+# parses those definitions out and turns them into typedefs, defines,
+# and data structures, and maybe even code which the regex engine can
+# use to operate.
+#
+# F<regexp.h> and op_reg_common.h contain defines C<RXf_xxx> and
+# C<PREGf_xxx> that are used in flags in our code. These defines are
+# parsed out and data structures are created to allow the debug mode of
+# the regex engine to show things such as which flags were set during
+# compilation. In some cases we transform the C code in the header files
+# into perl code which we execute to C<eval()> the contents. For instance
+# in a situation like this:
+#
+# #define RXf_X 0x1 /* the X mode */
+# #define RXf_Y 0x2 /* the Y mode */
+# #define RXf_Z (X|Y) /* the Z mode */
+#
+# this script might end up eval()ing something like C<0x1> and then
+# C<0x2> and then C<(0x1|0x2)> the results of which it then might use in
+# constructing a data structure, or pod in perldebguts, or a comment in
+# C<regnodes.h>. It also would separate out the "X", "Y", and "Z" and
+# use them, and would also use the data in the line comment if present.
+#
+# If you compile a regex under perl -Mre=Debug,ALL you can see much
+# of the content that this file generates and parses out of its input
+# files.
BEGIN {
# Get function prototypes
require './regen/regen_lib.pl';
+ require './regen/HeaderParser.pm';
}
use strict;
#ifndef DOINIT
EXTCONST struct regnode_meta PL_regnode_info[];
#else
-EXTCONST struct regnode_meta const PL_regnode_info[] = {
+EXTCONST struct regnode_meta PL_regnode_info[] = {
EOP
my @fields= qw(type arg_len arg_len_varies off_by_arg);
foreach my $node_idx (0..$#all) {
my $val= 0;
my %reverse;
my $REG_EXTFLAGS_NAME_SIZE= 0;
+ my $hp= HeaderParser->new();
foreach my $file ( "op_reg_common.h", "regexp.h" ) {
- open my $in_fh, "<", $file or die "Can't read '$file': $!";
- while (<$in_fh>) {
+ $hp->read_file($file);
+ foreach my $line_info (@{$hp->lines}) {
+ next unless $line_info->{type} eq "content"
+ and $line_info->{sub_type} eq "#define";
+ my $line= $line_info->{line};
+ $line=~s/\s*\\\n\s*/ /g;
# optional leading '_'. Return symbol in $1, and strip it from
# comment of line. Currently doesn't handle comments running onto
# next line
- if (s/^ \# \s* define \s+ ( _? RXf_ \w+ ) \s+ //xi) {
- chomp;
+ if ($line=~s/^ \# \s* define \s+ ( _? RXf_ \w+ ) \s+ //xi) {
+ chomp($line);
my $define= $1;
my $orig= $_;
- s{ /\* .*? \*/ }{ }x; # Replace comments by a blank
+ $line=~s{ /\* .*? \*/ }{ }x; # Replace comments by a blank
# Replace any prior defined symbols by their values
foreach my $key ( keys %definitions ) {
- s/\b$key\b/$definitions{$key}/g;
+ $line=~s/\b$key\b/$definitions{$key}/g;
}
# Remove the U suffix from unsigned int literals
- s/\b([0-9]+)U\b/$1/g;
+ $line=~s/\b([0-9]+)U\b/$1/g;
- my $newval= eval $_; # Get numeric definition
+ my $newval= eval $line; # Get numeric definition
$definitions{$define}= $newval;
- next unless $_ =~ /<</; # Bit defines use left shift
+ next unless $line =~ /<</; # Bit defines use left shift
if ( $val & $newval ) {
my @names= ( $define, $reverse{$newval} );
s/PMf_// for @names;
if ( $names[0] ne $names[1] ) {
die sprintf
"ERROR: both $define and $reverse{$newval} use 0x%08X (%s:%s)",
- $newval, $orig, $_;
+ $newval, $orig, $line;
}
next;
}
my $val= 0;
my %reverse;
my $REG_INTFLAGS_NAME_SIZE= 0;
+ my $hp= HeaderParser->new();
+ my $last_val = 0;
foreach my $file ("regcomp.h") {
- open my $fh, "<", $file or die "Can't read $file: $!";
- while (<$fh>) {
+ $hp->read_file($file);
+ my @bit_tuples;
+ foreach my $line_info (@{$hp->lines}) {
+ next unless $line_info->{type} eq "content"
+ and $line_info->{sub_type} eq "#define";
+ my $line= $line_info->{line};
+ $line=~s/\s*\\\n\s*/ /g;
# optional leading '_'. Return symbol in $1, and strip it from
# comment of line
if (
- m/^ \# \s* define \s+ ( PREGf_ ( \w+ ) ) \s+ 0x([0-9a-f]+)(?:\s*\/\*(.*)\*\/)?/xi
- )
- {
- chomp;
+ $line =~ m/^ \# \s* define \s+ ( PREGf_ ( \w+ ) ) \s+ 0x([0-9a-f]+)(?:\s*\/\*(.*)\*\/)?/xi
+ ){
+ chomp $line;
my $define= $1;
my $abbr= $2;
my $hex= $3;
my $comment= $4;
my $val= hex($hex);
+ my $bin= sprintf "%b", $val;
+ if ($bin=~/1.*?1/) { die "Not expecting multiple bits in PREGf" }
+ my $bit= length($bin) - 1 ;
$comment= $comment ? " - $comment" : "";
-
- printf $out qq(\t%-30s/* 0x%08x - %s%s */\n), qq("$abbr",),
- $val, $define, $comment;
- $REG_INTFLAGS_NAME_SIZE++;
+ if ($bit_tuples[$bit]) {
+ die "Duplicate PREGf bit '$bit': $define $val ($hex)";
+ }
+ $bit_tuples[$bit]= [ $bit, $val, $abbr, $define, $comment ];
+ }
+ }
+ foreach my $i (0..$#bit_tuples) {
+ my $bit_tuple= $bit_tuples[$i];
+ if (!$bit_tuple) {
+ $bit_tuple= [ $i, 1<<$i, "", "", "*UNUSED*" ];
}
+ my ($bit, $val, $abbr, $define, $comment)= @$bit_tuple;
+ printf $out qq(\t%-30s/* (1<<%2d) - 0x%08x - %s%s */\n),
+ qq("$abbr",), $bit, $val, $define, $comment;
}
+ $REG_INTFLAGS_NAME_SIZE=0+@bit_tuples;
}
print $out <<EOP;
exit(1);
}
my $out= open_new( 'regnodes.h', '>',
- { by => 'regen/regcomp.pl', from => 'regcomp.sym' } );
+ {
+ by => 'regen/regcomp.pl',
+ from => [ 'regcomp.sym', 'op_reg_common.h', 'regexp.h' ],
+ },
+);
print $out "#if $confine_to_core\n\n";
print_typedefs($out);
print_state_defs($out);