+my $version = Unicode::UCD::UnicodeVersion();
+my ($major, $dot, $dotdot) = $version =~ / (.*?) \. (.*?) (?: \. (.*) )? $ /x;
+$dotdot = 0 unless defined $dotdot;
+
+print $out_fh <<END;
+#define UNICODE_MAJOR_VERSION $major
+#define UNICODE_DOT_VERSION $dot
+#define UNICODE_DOT_DOT_VERSION $dotdot
+
+END
+
+# The data are at __DATA__ in this file.
+
+my @data = <DATA>;
+
+foreach my $charset (get_supported_code_pages()) {
+ print $out_fh "\n" . get_conditional_compile_line_start($charset);
+
+ my @a2n = @{get_a2n($charset)};
+
+ for ( @data ) {
+ chomp;
+
+ # Convert any '#' comments to /* ... */; empty lines and comments are
+ # output as blank lines
+ if ($_ =~ m/ ^ \s* (?: \# ( .* ) )? $ /x) {
+ my $comment_body = $1 // "";
+ if ($comment_body ne "") {
+ print $out_fh "/* $comment_body */\n";
+ }
+ else {
+ print $out_fh "\n";
+ }
+ next;
+ }
+
+ unless ($_ =~ m/ ^ ( [^\ ]* ) # Name or code point token
+ (?: [\ ]+ ( [^ ]* ) )? # optional flag
+ (?: [\ ]+ ( .* ) )? # name if unnamed; flag is required
+ /x)
+ {
+ die "Unexpected syntax at line $.: $_\n";
+ }
+
+ my $name_or_cp = $1;
+ my $flag = $2;
+ my $desired_name = $3;
+
+ my $name;
+ my $cp;
+ my $U_cp; # code point in Unicode (not-native) terms
+
+ if ($name_or_cp =~ /^U\+(.*)/) {
+ $U_cp = hex $1;
+ $name = charnames::viacode($name_or_cp);
+ if (! defined $name) {
+ next if $flag =~ /skip_if_undef/;
+ die "Unknown code point '$name_or_cp' at line $.: $_\n" unless $desired_name;
+ $name = "";
+ }
+ }
+ else {
+ $name = $name_or_cp;
+ die "Unknown name '$name' at line $.: $_\n" unless defined $name;
+ $U_cp = charnames::vianame($name =~ s/_/ /gr);
+ }
+
+ $cp = ($U_cp < 256)
+ ? $a2n[$U_cp]
+ : $U_cp;
+
+ $name = $desired_name if $name eq "" && $desired_name;
+ $name =~ s/[- ]/_/g; # The macro name can have no blanks nor dashes
+
+ my $str;
+ my $suffix;
+ if (defined $flag && $flag eq 'native') {
+ die "Are you sure you want to run this on an above-Latin1 code point?" if $cp > 0xff;
+ $suffix = '_NATIVE';
+ $str = sprintf "0x%02X", $cp; # Is a numeric constant
+ }
+ else {
+ $str = join "", map { sprintf "\\x%02X", ord $_ } split //, cp_2_utfbytes($U_cp, $charset);
+
+ $suffix = '_UTF8';
+ if (! defined $flag || $flag =~ /^ string (_skip_if_undef)? $/x) {
+ $str = "\"$str\""; # Will be a string constant
+ } elsif ($flag eq 'tail') {
+ $str =~ s/\\x..//; # Remove the first byte
+ $suffix .= '_TAIL';
+ $str = "\"$str\""; # Will be a string constant
+ }
+ elsif ($flag eq 'first') {
+ $str =~ s/ \\x ( .. ) .* /$1/x; # Get the two nibbles of the 1st byte
+ $suffix .= '_FIRST_BYTE';
+ $str = "0x$str"; # Is a numeric constant
+ }
+ else {
+ die "Unknown flag at line $.: $_\n";
+ }
+ }
+ printf $out_fh "# define %s%s %s /* U+%04X */\n", $name, $suffix, $str, $U_cp;
+ }
+
+ my $max_PRINT_A = 0;
+ for my $i (0x20 .. 0x7E) {
+ $max_PRINT_A = $a2n[$i] if $a2n[$i] > $max_PRINT_A;
+ }
+ printf $out_fh "# define MAX_PRINT_A_FOR_USE_ONLY_BY_REGCOMP_DOT_C 0x%02X /* The max code point that isPRINT_A */\n", $max_PRINT_A;
+
+ print $out_fh "\n" . get_conditional_compile_line_end();
+
+}
+
+use Unicode::UCD 'prop_invlist';
+
+my $count = 0;
+my @other_invlist = prop_invlist("Other");
+for (my $i = 0; $i < @other_invlist; $i += 2) {
+ $count += ((defined $other_invlist[$i+1])
+ ? $other_invlist[$i+1]
+ : 0x110000)
+ - $other_invlist[$i];
+}
+printf $out_fh "\n/* The number of code points not matching \\pC */\n"
+ . "#define NON_OTHER_COUNT_FOR_USE_ONLY_BY_REGCOMP_DOT_C %d\n",
+ 0x110000 - $count;
+
+# If this release has both the CWCM and CWCF properties, find the highest code
+# point which changes under any case change. We can use this to short-circuit
+# code
+my @cwcm = prop_invlist('CWCM');
+if (@cwcm) {
+ my @cwcf = prop_invlist('CWCF');
+ if (@cwcf) {
+ my $max = ($cwcm[-1] < $cwcf[-1])
+ ? $cwcf[-1]
+ : $cwcm[-1];
+ printf $out_fh "\n/* The highest code point that has any type of case change */\n"
+ . "#define HIGHEST_CASE_CHANGING_CP_FOR_USE_ONLY_BY_UTF8_DOT_C 0x%X\n",
+ $max - 1;
+ }
+}
+
+print $out_fh "\n#endif /* H_UNICODE_CONSTANTS */\n";
+
+read_only_bottom_close_and_rename($out_fh);
+
+# DATA FORMAT
+#
+# A blank line is output as-is.
+# Comments (lines whose first non-blank is a '#') are converted to C-style,
+# though empty comments are converted to blank lines. Otherwise, each line
+# represents one #define, and begins with either a Unicode character name with
+# the blanks and dashes in it squeezed out or replaced by underscores; or it
+# may be a hexadecimal Unicode code point of the form U+xxxx. In the latter