3 # Regenerate (overwriting only if changed):
10 # from information stored in this file. pod/perlguts.pod is not completely
11 # regenerated. Only the magic table is replaced; the other parts remain
14 # Accepts the standard regen_lib -q and -v args.
16 # This script is normally invoked from regen.pl.
18 # Its output files contain:
21 # included by dump.c - the textual representation for each magic type.
24 # { PERL_MAGIC_foo, "foo(f)" }
30 # processed by generate_uudmap.c into mg_data.h which eventually
31 # populates PL_magic_vtables[].
35 # { 'f', "want_vtbl_foo | FLAGS", "description for comments" }
37 # triplets. FLAGS can be:
38 # PERL_MAGIC_READONLY_ACCEPTABLE
39 # ok set this type of magic on an SvREADONLY() SV
40 # PERL_MAGIC_VALUE_MAGIC
41 # this is value magic (pos, taint etc)
42 # rather than container magic (%ENV, $1 etc)
46 # This contains five kinds of entries:
48 # #define PERL_MAGIC_arylen '#'
51 # enum { /* pass one of these to get_vtbl */
56 # PL_magic_vtable_names[] = {
61 # PL_magic_vtables[] = {
62 # /* per-magic sets of vtable function pointers */
63 # { get, set, len, clear, free, copy, dup, local },
66 # define PL_vtbl_arylen PL_magic_vtables[want_vtbl_arylen]
72 # updates the list of magic types between
73 # =for mg_vtable.pl begin
75 # =for mg_vtable.pl end
82 # Get function prototypes
83 require './regen/regen_lib.pl';
86 # =====================================================================
88 # START OF CONFIGURATION DATA
93 # This hash is mainly concerned with populating all the other stuff
94 # ancillary to the vtable.
96 # The key is the name, e.g. 'regdata' for PERL_MAGIC_regdata
98 # The keys of the value hash are:
100 # the magic's char identifier
103 # a description which appears in code comments in generated files
105 # readonly_acceptable
106 # If true, set PERL_MAGIC_READONLY_ACCEPTABLE flag;
107 # SvREADONLY() svs are allowed to have this magic added to them
109 # unknown_to_sv_magic
110 # if true, this isn't one of the standard magic types which
111 # Perl_sv_magic() knows how to deal with
114 # If true, set PERL_MAGIC_VALUE_MAGIC flag;
115 # this kind of magic is value (pos, taint etc) rather than
116 # container magic (%ENV, $1 etc)
119 # name of the vtable lookup enum, e.g. 'foo' creates want_vtbl_foo
123 sv => { char => "\0", vtable => 'sv', readonly_acceptable => 1,
124 desc => 'Special scalar variable' },
125 # overload, or type "A" magic, used to be here. Hence overloaded is
126 # often called AMAGIC internally, even though it does not use "A"
128 overload_table => { char => 'c', vtable => 'ovrld',
129 desc => 'Holds overload table (AMT) on stash' },
130 bm => { char => 'B', vtable => 'regexp', value_magic => 1,
131 readonly_acceptable => 1,
132 desc => 'Boyer-Moore (fast string search)' },
133 regdata => { char => 'D', vtable => 'regdata',
134 desc => "Regex match position data\n(\@+ and \@- vars)" },
135 regdatum => { char => 'd', vtable => 'regdatum',
136 desc => 'Regex match position data element' },
137 env => { char => 'E', vtable => 'env', desc => '%ENV hash' },
138 envelem => { char => 'e', vtable => 'envelem',
139 desc => '%ENV hash element' },
140 fm => { char => 'f', vtable => 'regexp', value_magic => 1,
141 readonly_acceptable => 1, desc => "Formline ('compiled' format)" },
142 regex_global => { char => 'g', vtable => 'mglob', value_magic => 1,
143 readonly_acceptable => 1, desc => 'm//g target' },
144 hints => { char => 'H', vtable => 'hints', desc => '%^H hash' },
145 hintselem => { char => 'h', vtable => 'hintselem',
146 desc => '%^H hash element' },
147 isa => { char => 'I', vtable => 'isa', desc => '@ISA array' },
148 isaelem => { char => 'i', vtable => 'isaelem',
149 desc => '@ISA array element' },
150 nkeys => { char => 'k', vtable => 'nkeys', value_magic => 1,
151 desc => 'scalar(keys()) lvalue' },
152 dbfile => { char => 'L',
153 desc => 'Debugger %_<filename' },
154 dbline => { char => 'l', vtable => 'dbline',
155 desc => 'Debugger %_<filename element' },
156 shared => { char => 'N', desc => 'Shared between threads',
157 unknown_to_sv_magic => 1 },
158 shared_scalar => { char => 'n', desc => 'Shared between threads',
159 unknown_to_sv_magic => 1 },
160 collxfrm => { char => 'o', vtable => 'collxfrm', value_magic => 1,
161 desc => 'Locale transformation' },
162 tied => { char => 'P', vtable => 'pack',
163 value_magic => 1, # treat as value, so 'local @tied' isn't tied
164 desc => 'Tied array or hash' },
165 tiedelem => { char => 'p', vtable => 'packelem',
166 desc => 'Tied array or hash element' },
167 tiedscalar => { char => 'q', vtable => 'packelem',
168 desc => 'Tied scalar or handle' },
169 qr => { char => 'r', vtable => 'regexp', value_magic => 1,
170 readonly_acceptable => 1, desc => 'Precompiled qr// regex' },
171 sig => { char => 'S', vtable => 'sig',
172 desc => '%SIG hash' },
173 sigelem => { char => 's', vtable => 'sigelem',
174 desc => '%SIG hash element' },
175 taint => { char => 't', vtable => 'taint', value_magic => 1,
176 desc => 'Taintedness' },
177 uvar => { char => 'U', vtable => 'uvar',
178 desc => 'Available for use by extensions' },
179 uvar_elem => { char => 'u', desc => 'Reserved for use by extensions',
180 unknown_to_sv_magic => 1 },
181 vec => { char => 'v', vtable => 'vec', value_magic => 1,
182 desc => 'vec() lvalue' },
183 vstring => { char => 'V', value_magic => 1,
184 desc => 'SV was vstring literal' },
185 utf8 => { char => 'w', vtable => 'utf8', value_magic => 1,
186 desc => 'Cached UTF-8 information' },
187 substr => { char => 'x', vtable => 'substr', value_magic => 1,
188 desc => 'substr() lvalue' },
189 defelem => { char => 'y', vtable => 'defelem', value_magic => 1,
190 desc => "Shadow \"foreach\" iterator variable /\nsmart parameter vivification" },
191 nonelem => { char => 'Y', vtable => 'nonelem', value_magic => 1,
192 desc => "Array element that does not exist" },
193 arylen => { char => '#', vtable => 'arylen', value_magic => 1,
194 desc => 'Array length ($#ary)' },
195 pos => { char => '.', vtable => 'pos', value_magic => 1,
196 desc => 'pos() lvalue' },
197 backref => { char => '<', vtable => 'backref', value_magic => 1,
198 readonly_acceptable => 1, desc => 'For weak ref data' },
199 symtab => { char => ':', value_magic => 1,
200 desc => 'Extra data for symbol tables' },
201 rhash => { char => '%', value_magic => 1,
202 desc => 'Extra data for restricted hashes' },
203 arylen_p => { char => '@', value_magic => 1,
204 desc => 'To move arylen out of XPVAV' },
205 ext => { char => '~', desc => 'Available for use by extensions',
206 readonly_acceptable => 1 },
207 checkcall => { char => ']', value_magic => 1, vtable => 'checkcall',
208 desc => 'Inlining/mutation of call to this CV'},
209 debugvar => { char => '*', desc => '$DB::single, signal, trace vars',
210 vtable => 'debugvar' },
211 lvref => { char => '\\', vtable => 'lvref',
212 desc => "Lvalue reference constructor" },
218 # This hash is mainly concerned with populating the vtable.
219 # (despite the name it has nothing to do with signals!)
221 # These have a subtly different "namespace" from the magic types.
223 # The key is the name, e.g. 'regdata' for PERL_MAGIC_regdata
224 # The keys of the value hash are:
227 # for each entry in the anon array, add
228 # add "#define want_vtbl_$_ want_vtbl_$name"
231 # prefix the vtable with the specified entry (e.g. '#ifdef FOO')
232 # and suffix it with '#else { 0, 0, 0, 0, 0, 0, 0, 0 } #endif'
235 # special-case cast a 'get' function whose signature expects
236 # a pointer to constant magic, so that it can be added to a vtable
237 # which expects pointers to functions without the 'const'.
247 # For each specified method, add a vtable function pointer
248 # of the form "Perl_magic_$sig{foo}{get}" etc
252 'sv' => {get => 'get', set => 'set'},
253 'env' => {set => 'set_all_env', clear => 'clear_all_env'},
254 'envelem' => {set => 'setenv', clear => 'clearenv'},
255 'sig' => { set => 'setsigall' },
256 'sigelem' => {get => 'getsig', set => 'setsig', clear => 'clearsig',
257 cond => '#ifndef PERL_MICRO'},
258 'pack' => {len => 'sizepack', clear => 'wipepack'},
259 'packelem' => {get => 'getpack', set => 'setpack', clear => 'clearpack'},
260 'dbline' => {set => 'setdbline'},
261 'isa' => {set => 'setisa', clear => 'clearisa'},
262 'isaelem' => {set => 'setisa'},
263 'arylen' => {get => 'getarylen', set => 'setarylen', const => 1},
264 'arylen_p' => {clear => 'cleararylen_p', free => 'freearylen_p'},
265 'mglob' => {set => 'setmglob',
266 free => 'freemglob' },
267 'nkeys' => {get => 'getnkeys', set => 'setnkeys'},
268 'taint' => {get => 'gettaint', set => 'settaint'},
269 'substr' => {get => 'getsubstr', set => 'setsubstr'},
270 'vec' => {get => 'getvec', set => 'setvec'},
271 'pos' => {get => 'getpos', set => 'setpos'},
272 'uvar' => {get => 'getuvar', set => 'setuvar'},
273 'defelem' => {get => 'getdefelem', set => 'setdefelem'},
274 'nonelem' => {set => 'setnonelem'},
275 'regexp' => {set => 'setregexp', alias => [qw(bm fm)]},
276 'regdata' => {len => 'regdata_cnt'},
277 'regdatum' => {get => 'regdatum_get', set => 'regdatum_set'},
278 'backref' => {free => 'killbackrefs'},
279 'ovrld' => {free => 'freeovrld'},
280 'utf8' => {set => 'setutf8',
281 free => 'freeutf8' },
282 'collxfrm' => {set => 'setcollxfrm',
283 free => 'freecollxfrm',
284 cond => '#ifdef USE_LOCALE_COLLATE'},
285 'hintselem' => {set => 'sethint', clear => 'clearhint'},
286 'hints' => {clear => 'clearhints'},
287 'checkcall' => {copy => 'copycallchecker'},
288 'debugvar' => { set => 'setdebugvar', get => 'getdebugvar' },
289 'lvref' => {set => 'setlvref'},
293 # END OF CONFIGURATION DATA
295 # =====================================================================
299 my ($vt, $raw, $names) = map {
301 { by => 'regen/mg_vtable.pl', file => $_, style => '*' });
302 } 'mg_vtable.h', 'mg_raw.h', 'mg_names.inc';
303 my $guts = open_new("pod/perlguts.pod", ">");
306 /* These constants should be used in preference to raw characters
307 * when using magic. Note that some perl guts still assume
308 * certain character properties of these constants, namely that
309 * isUPPER() and toLOWER() may do useful mappings.
314 # Of course, it would be *much* easier if we could output this table directly
315 # here and now. However, for our sins, we try to support EBCDIC, which wouldn't
316 # be *so* bad, except that there are (at least) 3 EBCDIC charset variants, and
317 # they don't agree on the code point for '~'. Which we use. Great.
318 # So we have to get the local build runtime to sort our table in character order
319 # (And of course, just to be helpful, in POSIX BC '~' is \xFF, so we can't even
320 # simplify the C code by assuming that the last element of the array is
328 $longest = length $_ if length $_ > $longest;
331 my $longest_p1 = $longest + 1;
334 while (my ($name, $data) = each %mg) {
335 my $byte = $data->{char};
336 if ($byte =~ /[[:print:]]/) {
337 $data->{r_char} = $byte; # readable char
338 ($data->{c_char} = $byte) =~ s/([\\"])/\\$1/g; # for C strings
341 $data->{c_char} = $data->{r_char} = '\\'.ord $byte;
343 $mg_order{(uc $byte) . $byte} = $name;
348 foreach (sort keys %mg_order) {
349 my $name = $mg_order{$_};
351 my $data = $mg{$name};
352 my $i = ord $data->{char};
354 # add entry to mg_raw.h
356 unless ($data->{unknown_to_sv_magic}) {
357 my $value = $data->{vtable}
358 ? "want_vtbl_$data->{vtable}" : 'magic_vtable_max';
359 $value .= ' | PERL_MAGIC_READONLY_ACCEPTABLE'
360 if $data->{readonly_acceptable};
361 $value .= ' | PERL_MAGIC_VALUE_MAGIC' if $data->{value_magic};
362 my $comment = "/* $name '$data->{r_char}' $data->{desc} */";
363 $comment =~ s/([\\"])/\\$1/g;
364 $comment =~ tr/\n/ /;
365 print $raw qq{ { '$data->{c_char}', "$value",\n "$comment" },\n};
368 # add #define PERL_MAGIC_foo entry to vt_table.h
370 my $comment = $data->{desc};
371 my $leader = ' ' x ($longest + 27);
372 $comment =~ s/\n/\n$leader/s;
373 printf $vt "#define PERL_MAGIC_%-${longest}s '%s' /* %s */\n",
374 $name, $data->{c_char}, $comment;
376 # add entry to mg_names.inc
378 my $char = $data->{r_char};
379 $char =~ s/([\\"])/\\$1/g;
380 printf $names qq[\t{ PERL_MAGIC_%-${longest_p1}s "%s(%s)" },\n],
381 "$name,", $name, $char;
383 # construct perlguts.pod entry
385 push @rows, [(sprintf "%-2s PERL_MAGIC_%s", $data->{r_char},$name),
386 $data->{vtable} ? "vtbl_$data->{vtable}" : '(none)',
390 # output @rows to perlguts.pod
392 select +(select($guts), do {
393 my @header = ('(old-style char and macro)', 'MGVTBL', 'Type of magic');
395 foreach my $row (@rows) {
397 $widths[$_] = length $row->[$_]
398 if length $row->[$_] > $widths[$_];
403 = sprintf "$indent%%-%ds%%-%ds%%s\n", $widths[0] + 1, $widths[1] + 1;
405 79 - 7 - (length $indent) - $widths[0] - $widths[1] - 2;
407 open my $oldguts, "<", "pod/perlguts.pod"
408 or die "$0 cannot open pod/perlguts.pod for reading: $!";
411 last if /^=for mg_vtable.pl begin/
414 print "\n", $indent . "mg_type\n";
415 printf $format, @header;
416 printf $format, map {'-' x length $_} @header;
418 my ($type, $vtbl, $desc) = @$_;
421 if (length $desc > $desc_wrap) {
422 # If it's too long, first split on '(', if there.
423 # [Which, if there, is always short enough, currently.
424 # Make this more robust if that changes]
425 ($desc, @cont) = split /(?=\()/, $desc;
427 ($desc, @cont) = $desc =~ /(.{1,$desc_wrap})(?: |\z)/g
430 printf $format, $type, $vtbl, $desc;
431 printf $format, '', '', $_ foreach @cont;
436 for my $magic (sort @names) {
439 print "=for apidoc AmnhU||PERL_MAGIC_$magic\n";
442 print "=for apidoc_item ||PERL_MAGIC_$magic\n";
448 last if /^=for mg_vtable.pl end/;
450 do { print } while <$oldguts>;
455 # Process %sig - everything goes to mg_vtable.h
457 my @names = sort keys %sig;
459 my $want = join ",\n ", (map {"want_vtbl_$_"} @names), 'magic_vtable_max';
460 my $names = join qq{",\n "}, @names;
464 enum { /* pass one of these to get_vtbl */
469 EXTCONST char * const PL_magic_vtable_names[magic_vtable_max] = {
473 EXTCONST char * const PL_magic_vtable_names[magic_vtable_max];
480 /* These all need to be 0, not NULL, as NULL can be (void*)0, which is a
481 * pointer to data, whereas we're assigning pointers to functions, which are
482 * not the same beast. ANSI doesn't allow the assignment from one to the other.
483 * (although most, but not all, compilers are prepared to do it)
498 EXT_MGVTBL PL_magic_vtables[magic_vtable_max] = {
504 while (my $name = shift @names) {
505 my $data = $sig{$name};
506 push @vtable_names, $name;
508 $data->{$_} ? "Perl_magic_$data->{$_}" : 0;
509 } qw(get set len clear free copy dup local);
511 $funcs[0] = "(int (*)(pTHX_ SV *, MAGIC *))" . $funcs[0] if $data->{const};
512 my $funcs = join ", ", @funcs;
514 # Because we can't have a , after the last {...}
515 my $comma = @names ? ',' : '';
517 print $vt "$data->{cond}\n" if $data->{cond};
518 print $vt " { $funcs }$comma\n";
519 print $vt <<"EOH" if $data->{cond};
521 { 0, 0, 0, 0, 0, 0, 0, 0 }$comma
524 foreach(@{$data->{alias}}) {
525 push @aliases, "#define want_vtbl_$_ want_vtbl_$name\n";
526 push @vtable_names, $_;
533 EXT_MGVTBL PL_magic_vtables[magic_vtable_max];
538 print $vt (sort @aliases), "\n";
540 print $vt "#define PL_vtbl_$_ PL_magic_vtables[want_vtbl_$_]\n"
541 foreach sort @vtable_names;
543 # 63, not 64, As we rely on the last possible value to mean "NULL vtable"
544 die "Too many vtable names" if @vtable_names > 63;
546 read_only_bottom_close_and_rename($_) foreach $vt, $raw, $names;
547 close_and_rename($guts);