1#!/usr/bin/perl -w 2# 3# Regenerate (overwriting only if changed): 4# 5# pod/perldebguts.pod 6# regnodes.h 7# 8# from information stored in 9# 10# regcomp.sym 11# regexp.h 12# 13# pod/perldebguts.pod is not completely regenerated. Only the table of 14# regexp nodes is replaced; other parts remain unchanged. 15# 16# Accepts the standard regen_lib -q and -v args. 17# 18# This script is normally invoked from regen.pl. 19 20BEGIN { 21 # Get function prototypes 22 require 'regen/regen_lib.pl'; 23} 24use strict; 25 26open DESC, 'regcomp.sym'; 27 28my $ind = 0; 29my (@name,@rest,@type,@code,@args,@flags,@longj,@cmnt); 30my ($longest_name_length,$desc,$lastregop) = 0; 31while (<DESC>) { 32 # Special pod comments 33 if (/^#\* ?/) { $cmnt[$ind] .= "# $'"; } 34 # Truly blank lines possibly surrounding pod comments 35 elsif (/^\s*$/) { $cmnt[$ind] .= "\n" } 36 37 next if /^(?:#|\s*$)/; 38 chomp; # No \z in 5.004 39 s/\s*$//; 40 if (/^-+\s*$/) { 41 $lastregop= $ind; 42 next; 43 } 44 unless ($lastregop) { 45 ($name[$ind], $desc, $rest[$ind]) = /^(\S+)\s+([^\t]+?)\s*;\s*(.*)/; 46 ($type[$ind], $code[$ind], $args[$ind], $flags[$ind], $longj[$ind]) 47 = split /[,\s]\s*/, $desc; 48 $longest_name_length = length $name[$ind] 49 if length $name[$ind] > $longest_name_length; 50 ++$ind; 51 } else { 52 my ($type,@lists)=split /\s+/, $_; 53 die "No list? $type" if !@lists; 54 foreach my $list (@lists) { 55 my ($names,$special)=split /:/, $list , 2; 56 $special ||= ""; 57 foreach my $name (split /,/,$names) { 58 my $real= $name eq 'resume' 59 ? "resume_$type" 60 : "${type}_$name"; 61 my @suffix; 62 if (!$special) { 63 @suffix=(""); 64 } elsif ($special=~/\d/) { 65 @suffix=(1..$special); 66 } elsif ($special eq 'FAIL') { 67 @suffix=("","_fail"); 68 } else { 69 die "unknown :type ':$special'"; 70 } 71 foreach my $suffix (@suffix) { 72 $name[$ind]="$real$suffix"; 73 $type[$ind]=$type; 74 $rest[$ind]="state for $type"; 75 ++$ind; 76 } 77 } 78 } 79 80 } 81} 82# use fixed width to keep the diffs between regcomp.pl recompiles 83# as small as possible. 84my ($width,$rwidth,$twidth)=(22,12,9); 85$lastregop ||= $ind; 86my $tot = $ind; 87close DESC; 88die "Too many regexp/state opcodes! Maximum is 256, but there are $lastregop in file!" 89 if $lastregop>256; 90 91sub process_flags { 92 my ($flag, $varname, $comment) = @_; 93 $comment = '' unless defined $comment; 94 95 $ind = 0; 96 my @selected; 97 my $bitmap = ''; 98 do { 99 my $set = $flags[$ind] && $flags[$ind] eq $flag ? 1 : 0; 100 # Whilst I could do this with vec, I'd prefer to do longhand the arithmetic 101 # ops in the C code. 102 my $current = do { 103 local $^W; 104 ord do { 105 substr $bitmap, ($ind >> 3); 106 } 107 }; 108 substr($bitmap, ($ind >> 3), 1) = chr($current | ($set << ($ind & 7))); 109 110 push @selected, $name[$ind] if $set; 111 } while (++$ind < $lastregop); 112 my $out_string = join ', ', @selected, 0; 113 $out_string =~ s/(.{1,70},) /$1\n /g; 114 115 my $out_mask = join ', ', map {sprintf "0x%02X", ord $_} split '', $bitmap; 116 117 return $comment . <<"EOP"; 118#define REGNODE_\U$varname\E(node) (PL_${varname}_bitmask[(node) >> 3] & (1 << ((node) & 7))) 119 120#ifndef DOINIT 121EXTCONST U8 PL_${varname}\[] __attribute__deprecated__; 122#else 123EXTCONST U8 PL_${varname}\[] __attribute__deprecated__ = { 124 $out_string 125}; 126#endif /* DOINIT */ 127 128#ifndef DOINIT 129EXTCONST U8 PL_${varname}_bitmask[]; 130#else 131EXTCONST U8 PL_${varname}_bitmask[] = { 132 $out_mask 133}; 134#endif /* DOINIT */ 135EOP 136} 137 138my $out = open_new('regnodes.h', '>', 139 { by => 'regen/regcomp.pl', from => 'regcomp.sym' }); 140printf $out <<EOP, 141/* Regops and State definitions */ 142 143#define %*s\t%d 144#define %*s\t%d 145 146EOP 147 -$width, REGNODE_MAX => $lastregop - 1, 148 -$width, REGMATCH_STATE_MAX => $tot - 1 149; 150 151 152for ($ind=0; $ind < $lastregop ; ++$ind) { 153 printf $out "#define\t%*s\t%d\t/* %#04x %s */\n", 154 -$width, $name[$ind], $ind, $ind, $rest[$ind]; 155} 156print $out "\t/* ------------ States ------------- */\n"; 157for ( ; $ind < $tot ; $ind++) { 158 printf $out "#define\t%*s\t(REGNODE_MAX + %d)\t/* %s */\n", 159 -$width, $name[$ind], $ind - $lastregop + 1, $rest[$ind]; 160} 161 162print $out <<EOP; 163 164/* PL_regkind[] What type of regop or state is this. */ 165 166#ifndef DOINIT 167EXTCONST U8 PL_regkind[]; 168#else 169EXTCONST U8 PL_regkind[] = { 170EOP 171 172$ind = 0; 173do { 174 printf $out "\t%*s\t/* %*s */\n", 175 -1-$twidth, "$type[$ind],", -$width, $name[$ind]; 176 print $out "\t/* ------------ States ------------- */\n" 177 if $ind + 1 == $lastregop and $lastregop != $tot; 178} while (++$ind < $tot); 179 180print $out <<EOP; 181}; 182#endif 183 184/* regarglen[] - How large is the argument part of the node (in regnodes) */ 185 186#ifdef REG_COMP_C 187static const U8 regarglen[] = { 188EOP 189 190$ind = 0; 191do { 192 my $size = 0; 193 $size = "EXTRA_SIZE(struct regnode_$args[$ind])" if $args[$ind]; 194 195 printf $out "\t%*s\t/* %*s */\n", 196 -37, "$size,",-$rwidth,$name[$ind]; 197} while (++$ind < $lastregop); 198 199print $out <<EOP; 200}; 201 202/* reg_off_by_arg[] - Which argument holds the offset to the next node */ 203 204static const char reg_off_by_arg[] = { 205EOP 206 207$ind = 0; 208do { 209 my $size = $longj[$ind] || 0; 210 211 printf $out "\t%d,\t/* %*s */\n", 212 $size, -$rwidth, $name[$ind] 213} while (++$ind < $lastregop); 214 215print $out <<EOP; 216}; 217 218#endif /* REG_COMP_C */ 219 220/* reg_name[] - Opcode/state names in string form, for debugging */ 221 222#ifndef DOINIT 223EXTCONST char * PL_reg_name[]; 224#else 225EXTCONST char * const PL_reg_name[] = { 226EOP 227 228$ind = 0; 229my $ofs = 0; 230my $sym = ""; 231do { 232 my $size = $longj[$ind] || 0; 233 234 printf $out "\t%*s\t/* $sym%#04x */\n", 235 -3-$width,qq("$name[$ind]",), $ind - $ofs; 236 if ($ind + 1 == $lastregop and $lastregop != $tot) { 237 print $out "\t/* ------------ States ------------- */\n"; 238 $ofs = $lastregop - 1; 239 $sym = 'REGNODE_MAX +'; 240 } 241 242} while (++$ind < $tot); 243 244print $out <<EOP; 245}; 246#endif /* DOINIT */ 247 248EOP 249 250{ 251print $out <<EOP; 252/* PL_reg_extflags_name[] - Opcode/state names in string form, for debugging */ 253 254#ifndef DOINIT 255EXTCONST char * PL_reg_extflags_name[]; 256#else 257EXTCONST char * const PL_reg_extflags_name[] = { 258EOP 259 260my %rxfv; 261my %definitions; # Remember what the symbol definitions are 262my $val = 0; 263my %reverse; 264my $REG_EXTFLAGS_NAME_SIZE = 0; 265foreach my $file ("op_reg_common.h", "regexp.h") { 266 open FH,"<$file" or die "Can't read $file: $!"; 267 while (<FH>) { 268 269 # optional leading '_'. Return symbol in $1, and strip it from 270 # rest of line 271 if (s/^ \# \s* define \s+ ( _? RXf_ \w+ ) \s+ //xi) { 272 chomp; 273 my $define = $1; 274 my $orig= $_; 275 s{ /\* .*? \*/ }{ }x; # Replace comments by a blank 276 277 # Replace any prior defined symbols by their values 278 foreach my $key (keys %definitions) { 279 s/\b$key\b/$definitions{$key}/g; 280 } 281 282 # Remove the U suffix from unsigned int literals 283 s/\b([0-9]+)U\b/$1/g; 284 285 my $newval = eval $_; # Get numeric definition 286 287 $definitions{$define} = $newval; 288 289 next unless $_ =~ /<</; # Bit defines use left shift 290 if($val & $newval) { 291 my @names=($define, $reverse{$newval}); 292 s/PMf_// for @names; 293 if ($names[0] ne $names[1]) { 294 die sprintf "ERROR: both $define and $reverse{$newval} use 0x%08X (%s:%s)", $newval, $orig, $_; 295 } 296 next; 297 } 298 $val|=$newval; 299 $rxfv{$define}= $newval; 300 $reverse{$newval} = $define; 301 } 302 } 303} 304my %vrxf=reverse %rxfv; 305printf $out "\t/* Bits in extflags defined: %s */\n", unpack 'B*', pack 'N', $val; 306my %multibits; 307for (0..31) { 308 my $power_of_2 = 2**$_; 309 my $n=$vrxf{$power_of_2}; 310 my $extra = ""; 311 if (! $n) { 312 313 # Here, there was no name that matched exactly the bit. It could be 314 # either that it is unused, or the name matches multiple bits. 315 if (! ($val & $power_of_2)) { 316 $n = "UNUSED_BIT_$_"; 317 } 318 else { 319 320 # Here, must be because it matches multiple bits. Look through 321 # all possibilities until find one that matches this one. Use 322 # that name, and all the bits it matches 323 foreach my $name (keys %rxfv) { 324 if ($rxfv{$name} & $power_of_2) { 325 $n = $name . ( $multibits{$name}++ ); 326 $extra= sprintf qq{ : "%s" - 0x%08x}, $name, $rxfv{$name} 327 if $power_of_2 != $rxfv{$name}; 328 last; 329 } 330 } 331 } 332 } 333 s/\bRXf_(PMf_)?// for $n, $extra; 334 printf $out qq(\t%-20s/* 0x%08x%s */\n), 335 qq("$n",),$power_of_2, $extra; 336 $REG_EXTFLAGS_NAME_SIZE++; 337} 338 339print $out <<EOP; 340}; 341#endif /* DOINIT */ 342 343EOP 344print $out <<EOQ 345#ifdef DEBUGGING 346# define REG_EXTFLAGS_NAME_SIZE $REG_EXTFLAGS_NAME_SIZE 347#endif 348 349EOQ 350} 351{ 352print $out <<EOP; 353/* PL_reg_intflags_name[] - Opcode/state names in string form, for debugging */ 354 355#ifndef DOINIT 356EXTCONST char * PL_reg_intflags_name[]; 357#else 358EXTCONST char * const PL_reg_intflags_name[] = { 359EOP 360 361my %rxfv; 362my %definitions; # Remember what the symbol definitions are 363my $val = 0; 364my %reverse; 365my $REG_INTFLAGS_NAME_SIZE = 0; 366foreach my $file ("regcomp.h") { 367 open my $fh, "<", $file or die "Can't read $file: $!"; 368 while (<$fh>) { 369 # optional leading '_'. Return symbol in $1, and strip it from 370 # rest of line 371 if (m/^ \# \s* define \s+ ( PREGf_ ( \w+ ) ) \s+ 0x([0-9a-f]+)(?:\s*\/\*(.*)\*\/)?/xi) { 372 chomp; 373 my $define = $1; 374 my $abbr= $2; 375 my $hex= $3; 376 my $comment= $4; 377 my $val= hex($hex); 378 $comment= $comment ? " - $comment" : ""; 379 380 printf $out qq(\t%-30s/* 0x%08x - %s%s */\n), qq("$abbr",), $val, $define, $comment; 381 $REG_INTFLAGS_NAME_SIZE++; 382 } 383 } 384} 385 386print $out <<EOP; 387}; 388#endif /* DOINIT */ 389 390EOP 391print $out <<EOQ; 392#ifdef DEBUGGING 393# define REG_INTFLAGS_NAME_SIZE $REG_INTFLAGS_NAME_SIZE 394#endif 395 396EOQ 397} 398 399print $out process_flags('V', 'varies', <<'EOC'); 400/* The following have no fixed length. U8 so we can do strchr() on it. */ 401EOC 402 403print $out process_flags('S', 'simple', <<'EOC'); 404 405/* The following always have a length of 1. U8 we can do strchr() on it. */ 406/* (Note that length 1 means "one character" under UTF8, not "one octet".) */ 407EOC 408 409read_only_bottom_close_and_rename($out); 410 411my $guts = open_new('pod/perldebguts.pod', '>'); 412 413my $code; 414my $name_fmt = '<' x ($longest_name_length-1); 415my $descr_fmt = '<' x (58-$longest_name_length); 416eval <<EOD; 417format GuTS = 418 ^*~~ 419 \$cmnt[\$_] 420 ^$name_fmt ^<<<<<<<<< ^$descr_fmt~~ 421 \$name[\$_], \$code, \$rest[\$_] 422. 423EOD 424 425select +(select($guts), do { 426 $~ = "GuTS"; 427 428 open my $oldguts, "pod/perldebguts.pod" 429 or die "$0 cannot open pod/perldebguts.pod for reading: $!"; 430 while(<$oldguts>) { 431 print; 432 last if /=for regcomp.pl begin/; 433 } 434 435 print <<'end'; 436 437 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION 438end 439 for (0..$lastregop-1) { 440 $code = "$code[$_] ".($args[$_]||""); 441 $code .= " $longj[$_]" if $longj[$_]; 442 if ($cmnt[$_] ||= "") { 443 # Trim multiple blanks 444 $cmnt[$_] =~ s/^\n\n+/\n/; $cmnt[$_] =~ s/\n\n+$/\n\n/ 445 } 446 write; 447 } 448 print "\n"; 449 450 while(<$oldguts>) { 451 last if /=for regcomp.pl end/; 452 } 453 do { print } while <$oldguts>; 454 455})[0]; 456 457close_and_rename($guts); 458