preproc: add support for builtin include modules (%use)

Add a builtin equivalent to the %include directive called %use.
%use includes a standard macro file compiled into the binary; these
come from the macros/ directory in the source code.

The idea here is to be able to provide optional macro packages with
the distribution, without adding complex host filesystem dependencies.

1 files changed, 100 insertions, 22 deletions

diff --git a/macros.pl b/macros.pl
index 7bba2b1..c22185b 100755
--- a/macros.pl
+++ b/macros.pl
@@ -1,4 +1,4 @@
-#!/usr/bin/perl -w
+#!/usr/bin/perl
 #
 # macros.pl   produce macros.c from standard.mac
 #
@@ -7,42 +7,58 @@
 # redistributable under the license given in the file "LICENSE"
 # distributed in the NASM archive.
 
-use strict;
+require 'phash.ph';
 
 my $fname;
 my $line = 0;
 my $index      = 0;
-my $tasm_count;
+my $tasm_count = 0;
 
-undef $tasm_count;
+open(OUT,">macros.c") or die "unable to open macros.c\n";
 
-open(OUTPUT,">macros.c") or die "unable to open macros.c\n";
+print OUT "/*\n";
+print OUT " * Do not edit - this file auto-generated by macros.pl from:\n";
+print OUT " * ", join(' ', @ARGV), "\n";
+print OUT " */\n";
+print OUT "\n";
+print OUT "#include \"tables.h\"\n";
+print OUT "#include \"nasmlib.h\"\n";
+print OUT "#include \"hashtbl.h\"\n";
+print OUT "\n";
+print OUT "const char * const nasm_stdmac[] = {";
 
-print OUTPUT "/*\n";
-print OUTPUT " * Do not edit - this file auto-generated by macros.pl from:\n";
-print OUTPUT " * ", join(' ', @ARGV), "\n";
-print OUTPUT " */\n";
-print OUTPUT "\n";
-print OUTPUT "#include \"tables.h\"\n";
-print OUTPUT "\n";
-print OUTPUT "const char * const nasm_stdmac[] = {";
+my $nmodule = 0;
+my @module_list   = ();
+my %module_number = ();
+my %module_index  = ();
+my $module;
 
 foreach $fname ( @ARGV ) {
     open(INPUT,$fname) or die "unable to open $fname\n";
-    print OUTPUT "\n    /* From $fname */\n";
+    print OUT "\n    /* From $fname */\n";
     while (<INPUT>) {
 	$line++;
 	chomp;
 	if (m/^\s*\*END\*TASM\*MACROS\*\s*$/) {
 	    $tasm_count = $index;
-	    print OUTPUT "    /* End of TASM macros */\n";
+	    print OUT "    /* End of TASM macros */\n";
+	} elsif (m/^USE:\s*(\S+)\s*$/) {
+	    $module = $1;
+	    if (defined($module_number{$module})) {
+		die "$0: $fname: duplicate module: $module\n";
+	    }
+	    printf OUT "        /* %4d */ NULL,\n", $index++;
+	    print OUT "    /* %use $module */\n";
+	    push(@module_list, $module);
+	    $module_number{$module} = $nmodule++;
+	    $module_index{$module}  = $index;
 	} elsif (m/^\s*((\s*([^\"\';\s]+|\"[^\"]*\"|\'[^\']*\'))*)\s*(;.*)?$/) {
 	    $_ = $1;
+	    s/(\s)\s+/$1/g;
 	    s/\\/\\\\/g;
 	    s/"/\\"/g;
 	    if (length > 0) {
-		print OUTPUT "        \"$_\",\n";
-		$index++;
+		printf OUT "        /* %4d */ \"%s\",\n", $index++, $_;
 	    }
 	} else {
 	    die "$fname:$line:  error unterminated quote";
@@ -50,8 +66,70 @@ foreach $fname ( @ARGV ) {
     }
     close(INPUT);
 }
-print OUTPUT "\n    NULL\n};\n\n";
-$tasm_count = $index unless ( defined($tasm_count) );
-print OUTPUT "const char * const * nasm_stdmac_after_tasm = ",
-    "&nasm_stdmac[$tasm_count];\n";
-close(OUTPUT);
+printf OUT "        /* %4d */ NULL\n};\n\n", $index++;
+print OUT "const char * const * const nasm_stdmac_after_tasm = ",
+    "&nasm_stdmac[$tasm_count];\n\n";
+
+my @hashinfo = gen_perfect_hash(\%module_number);
+if (!@hashinfo) {
+    die "$0: no hash found\n";
+}
+# Paranoia...
+verify_hash_table(\%module_number, \@hashinfo);
+my ($n, $sv, $g) = @hashinfo;
+die if ($n & ($n-1));
+
+print OUT "const char * const *nasm_stdmac_find_module(const char *module)\n";
+print OUT "{\n";
+print OUT "    static const struct {\n";
+print OUT "         const char *module;\n";
+print OUT "         const char * const *macros;\n";
+print OUT "    } modules[$nmodule] = {\n";
+foreach $module (@module_list) {
+    printf OUT "        { \"%s\", nasm_stdmac+%d },\n",
+	$module, $module_index{$module};
+}
+print OUT "    };\n";
+
+# Put a large value in unused slots.  This makes it extremely unlikely
+# that any combination that involves unused slot will pass the range test.
+# This speeds up rejection of unrecognized tokens, i.e. identifiers.
+print OUT "#define UNUSED 16383\n";
+
+print OUT "    static const int16_t hash1[$n] = {\n";
+for ($i = 0; $i < $n; $i++) {
+    my $h = ${$g}[$i*2+0];
+    print OUT "        ", defined($h) ? $h : 'UNUSED', ",\n";
+}
+print OUT "    };\n";
+
+print OUT "    static const int16_t hash2[$n] = {\n";
+for ($i = 0; $i < $n; $i++) {
+    my $h = ${$g}[$i*2+1];
+    print OUT "        ", defined($h) ? $h : 'UNUSED', ",\n";
+}
+print OUT "    };\n";
+
+print OUT  "    uint32_t k1, k2;\n";
+print OUT  "    uint64_t crc;\n";
+# For correct overflow behavior, "ix" should be unsigned of the same
+# width as the hash arrays.
+print OUT  "    uint16_t ix;\n";
+print OUT  "\n";
+
+printf OUT "    crc = crc64i(UINT64_C(0x%08x%08x), module);\n",
+    $$sv[0], $$sv[1];
+print  OUT "    k1 = (uint32_t)crc;\n";
+print  OUT "    k2 = (uint32_t)(crc >> 32);\n";
+print  OUT "\n";
+printf OUT "    ix = hash1[k1 & 0x%x] + hash2[k2 & 0x%x];\n", $n-1, $n-1;
+printf OUT "    if (ix >= %d)\n", scalar(@module_list);
+print OUT  "        return NULL;\n";
+print OUT  "\n";
+print OUT  "    if (nasm_stricmp(modules[ix].module, module))\n";
+print OUT  "        return NULL;\n";
+print OUT  "\n";
+print OUT  "    return modules[ix].macros;\n";
+print OUT  "}\n";
+
+close(OUT);