3 perldebguts - Guts of Perl debugging
7 This is not L<perldebug>, which tells you how to use
8 the debugger. This manpage describes low-level details concerning
9 the debugger's internals, which range from difficult to impossible
10 to understand for anyone who isn't incredibly intimate with Perl's guts.
13 =head1 Debugger Internals
15 Perl has special debugging hooks at compile-time and run-time used
16 to create debugging environments. These hooks are not to be confused
17 with the I<perl -Dxxx> command described in L<perlrun>, which is
18 usable only if a special Perl is built per the instructions in the
19 F<INSTALL> podpage in the Perl source tree.
21 For example, whenever you call Perl's built-in C<caller> function
22 from the package C<DB>, the arguments that the corresponding stack
23 frame was called with are copied to the C<@DB::args> array. These
24 mechanisms are enabled by calling Perl with the B<-d> switch.
25 Specifically, the following additional features are enabled
32 Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require
33 'perl5db.pl'}> if not present) before the first line of your program.
37 Each array C<@{"_<$filename"}> holds the lines of $filename for a
38 file compiled by Perl. The same is also true for C<eval>ed strings
39 that contain subroutines, or which are currently being executed.
40 The $filename for C<eval>ed strings looks like C<(eval 34)>.
42 Values in this array are magical in numeric context: they compare
43 equal to zero only if the line is not breakable.
47 Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
48 by line number. Individual entries (as opposed to the whole hash)
49 are settable. Perl only cares about Boolean true here, although
50 the values used by F<perl5db.pl> have the form
51 C<"$break_condition\0$action">.
53 The same holds for evaluated strings that contain subroutines, or
54 which are currently being executed. The $filename for C<eval>ed strings
55 looks like C<(eval 34)>.
59 Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is
60 also the case for evaluated strings that contain subroutines, or
61 which are currently being executed. The $filename for C<eval>ed
62 strings looks like C<(eval 34)>.
66 After each C<require>d file is compiled, but before it is executed,
67 C<DB::postponed(*{"_<$filename"})> is called if the subroutine
68 C<DB::postponed> exists. Here, the $filename is the expanded name of
69 the C<require>d file, as found in the values of %INC.
73 After each subroutine C<subname> is compiled, the existence of
74 C<$DB::postponed{subname}> is checked. If this key exists,
75 C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine
80 A hash C<%DB::sub> is maintained, whose keys are subroutine names
81 and whose values have the form C<filename:startline-endline>.
82 C<filename> has the form C<(eval 34)> for subroutines defined inside
87 When the execution of your program reaches a point that can hold a
88 breakpoint, the C<DB::DB()> subroutine is called if any of the variables
89 C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables
90 are not C<local>izable. This feature is disabled when executing
91 inside C<DB::DB()>, including functions called from it
92 unless C<< $^D & (1<<30) >> is true.
96 When execution of the program reaches a subroutine call, a call to
97 C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> set to identify
98 the called subroutine. (This doesn't happen if the calling subroutine
99 was compiled in the C<DB> package.) C<$DB::sub> normally holds the name
100 of the called subroutine, if it has a name by which it can be looked up.
101 Failing that, C<$DB::sub> will hold a reference to the called subroutine.
102 Either way, the C<&DB::sub> subroutine can use C<$DB::sub> as a reference
103 by which to call the called subroutine, which it will normally want to do.
105 X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub>
106 is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling
107 back to C<&DB::sub>(I<args>).
111 When execution of the program uses C<goto> to enter a non-XS subroutine
112 and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto> is made, with
113 C<$DB::sub> set to identify the subroutine being entered. The call to
114 C<&DB::goto> does not replace the C<goto>; the requested subroutine will
115 still be entered once C<&DB::goto> has returned. C<$DB::sub> normally
116 holds the name of the subroutine being entered, if it has one. Failing
117 that, C<$DB::sub> will hold a reference to the subroutine being entered.
118 Unlike when C<&DB::sub> is called, it is not guaranteed that C<$DB::sub>
119 can be used as a reference to operate on the subroutine being entered.
123 Note that if C<&DB::sub> needs external data for it to work, no
124 subroutine call is possible without it. As an example, the standard
125 debugger's C<&DB::sub> depends on the C<$DB::deep> variable
126 (it defines how many levels of recursion deep into the debugger you can go
127 before a mandatory break). If C<$DB::deep> is not defined, subroutine
128 calls are not possible, even though C<&DB::sub> exists.
130 =head2 Writing Your Own Debugger
132 =head3 Environment Variables
134 The C<PERL5DB> environment variable can be used to define a debugger.
135 For example, the minimal "working" debugger (it actually doesn't do anything)
136 consists of one line:
140 It can easily be defined like this:
142 $ PERL5DB="sub DB::DB {}" perl -d your-script
144 Another brief debugger, slightly more useful, can be created
147 sub DB::DB {print ++$i; scalar <STDIN>}
149 This debugger prints a number which increments for each statement
150 encountered and waits for you to hit a newline before continuing
151 to the next statement.
153 The following debugger is actually useful:
158 sub sub {print ++$i, " $sub\n"; &$sub}
161 It prints the sequence number of each subroutine call and the name of the
162 called subroutine. Note that C<&DB::sub> is being compiled into the
163 package C<DB> through the use of the C<package> directive.
165 When it starts, the debugger reads your rc file (F<./.perldb> or
166 F<~/.perldb> under Unix), which can set important options.
167 (A subroutine (C<&afterinit>) can be defined here as well; it is executed
168 after the debugger completes its own initialization.)
170 After the rc file is read, the debugger reads the PERLDB_OPTS
171 environment variable and uses it to set debugger options. The
172 contents of this variable are treated as if they were the argument
173 of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">).
175 =head3 Debugger Internal Variables
177 In addition to the file and subroutine-related variables mentioned above,
178 the debugger also maintains various magical internal variables.
184 C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
185 holds the lines of the currently-selected file (compiled by Perl), either
186 explicitly chosen with the debugger's C<f> command, or implicitly by flow
189 Values in this array are magical in numeric context: they compare
190 equal to zero only if the line is not breakable.
194 C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which
195 contains breakpoints and actions keyed by line number in
196 the currently-selected file, either explicitly chosen with the
197 debugger's C<f> command, or implicitly by flow of execution.
199 As previously noted, individual entries (as opposed to the whole hash)
200 are settable. Perl only cares about Boolean true here, although
201 the values used by F<perl5db.pl> have the form
202 C<"$break_condition\0$action">.
206 =head3 Debugger Customization Functions
208 Some functions are provided to simplify customization.
214 See L<perldebug/"Configurable Options"> for a description of options parsed by
215 C<DB::parse_options(string)>.
219 C<DB::dump_trace(skip[,count])> skips the specified number of frames
220 and returns a list containing information about the calling frames (all
221 of them, if C<count> is missing). Each entry is reference to a hash
222 with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
223 name, or info about C<eval>), C<args> (C<undef> or a reference to
224 an array), C<file>, and C<line>.
228 C<DB::print_trace(FH, skip[, count[, short]])> prints
229 formatted info about caller frames. The last two functions may be
230 convenient as arguments to C<< < >>, C<< << >> commands.
234 Note that any variables and functions that are not documented in
235 this manpages (or in L<perldebug>) are considered for internal
236 use only, and as such are subject to change without notice.
238 =head1 Frame Listing Output Examples
240 The C<frame> option can be used to control the output of frame
241 information. For example, contrast this expression trace:
244 Stack dump during die enabled outside of evals.
246 Loading DB routines from perl5db.pl patch level 0.94
247 Emacs support available.
249 Enter h or 'h h' for help.
256 DB<3> t print foo() * bar()
257 main::((eval 172):3): print foo() + bar();
258 main::foo((eval 168):2):
259 main::bar((eval 170):2):
262 with this one, once the C<o>ption C<frame=2> has been set:
266 DB<5> t print foo() * bar()
276 By way of demonstration, we present below a laborious listing
277 resulting from setting your C<PERLDB_OPTS> environment variable to
278 the value C<f=n N>, and running I<perl -d -V> from the command line.
279 Examples using various values of C<n> are shown to give you a feel
280 for the difference between settings. Long though it may be, this
281 is not a complete listing, but only excerpts.
288 entering Config::BEGIN
289 Package lib/Exporter.pm.
291 Package lib/Config.pm.
292 entering Config::TIEHASH
293 entering Exporter::import
294 entering Exporter::export
295 entering Config::myconfig
296 entering Config::FETCH
297 entering Config::FETCH
298 entering Config::FETCH
299 entering Config::FETCH
304 entering Config::BEGIN
305 Package lib/Exporter.pm.
308 Package lib/Config.pm.
309 entering Config::TIEHASH
310 exited Config::TIEHASH
311 entering Exporter::import
312 entering Exporter::export
313 exited Exporter::export
314 exited Exporter::import
316 entering Config::myconfig
317 entering Config::FETCH
319 entering Config::FETCH
321 entering Config::FETCH
325 in $=main::BEGIN() from /dev/null:0
326 in $=Config::BEGIN() from lib/Config.pm:2
327 Package lib/Exporter.pm.
329 Package lib/Config.pm.
330 in $=Config::TIEHASH('Config') from lib/Config.pm:644
331 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
332 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
333 in @=Config::myconfig() from /dev/null:0
334 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
335 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
336 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
337 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
338 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
339 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
343 in $=main::BEGIN() from /dev/null:0
344 in $=Config::BEGIN() from lib/Config.pm:2
345 Package lib/Exporter.pm.
347 out $=Config::BEGIN() from lib/Config.pm:0
348 Package lib/Config.pm.
349 in $=Config::TIEHASH('Config') from lib/Config.pm:644
350 out $=Config::TIEHASH('Config') from lib/Config.pm:644
351 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
352 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
353 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
354 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
355 out $=main::BEGIN() from /dev/null:0
356 in @=Config::myconfig() from /dev/null:0
357 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
358 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
359 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
360 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
361 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
362 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
363 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
367 in $=main::BEGIN() from /dev/null:0
368 in $=Config::BEGIN() from lib/Config.pm:2
369 Package lib/Exporter.pm.
371 out $=Config::BEGIN() from lib/Config.pm:0
372 Package lib/Config.pm.
373 in $=Config::TIEHASH('Config') from lib/Config.pm:644
374 out $=Config::TIEHASH('Config') from lib/Config.pm:644
375 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
376 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
377 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
378 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
379 out $=main::BEGIN() from /dev/null:0
380 in @=Config::myconfig() from /dev/null:0
381 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
382 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
383 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
384 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
388 in $=CODE(0x15eca4)() from /dev/null:0
389 in $=CODE(0x182528)() from lib/Config.pm:2
390 Package lib/Exporter.pm.
391 out $=CODE(0x182528)() from lib/Config.pm:0
392 scalar context return from CODE(0x182528): undef
393 Package lib/Config.pm.
394 in $=Config::TIEHASH('Config') from lib/Config.pm:628
395 out $=Config::TIEHASH('Config') from lib/Config.pm:628
396 scalar context return from Config::TIEHASH: empty hash
397 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
398 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
399 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
400 scalar context return from Exporter::export: ''
401 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
402 scalar context return from Exporter::import: ''
406 In all cases shown above, the line indentation shows the call tree.
407 If bit 2 of C<frame> is set, a line is printed on exit from a
408 subroutine as well. If bit 4 is set, the arguments are printed
409 along with the caller info. If bit 8 is set, the arguments are
410 printed even if they are tied or references. If bit 16 is set, the
411 return value is printed, too.
413 When a package is compiled, a line like this
417 is printed with proper indentation.
419 =head1 Debugging Regular Expressions
421 There are two ways to enable debugging output for regular expressions.
423 If your perl is compiled with C<-DDEBUGGING>, you may use the
424 B<-Dr> flag on the command line, and C<-Drv> for more verbose
427 Otherwise, one can C<use re 'debug'>, which has effects at both
428 compile time and run time. Since Perl 5.9.5, this pragma is lexically
431 =head2 Compile-time Output
433 The debugging output at compile time looks like this:
435 Compiling REx '[bc]d(ef*g)+h[ij]k$'
436 size 45 Got 364 bytes for offset annotations.
442 14: CURLYX[0] {1,32767}(28)
456 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
457 stclass 'ANYOF[bc]' minlen 7
459 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
460 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
461 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
462 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
463 Omitting $` $& $' support.
465 The first line shows the pre-compiled form of the regex. The second
466 shows the size of the compiled form (in arbitrary units, usually
467 4-byte words) and the total number of bytes allocated for the
468 offset/length table, usually 4+C<size>*8. The next line shows the
469 label I<id> of the first node that does a match.
473 anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
474 stclass 'ANYOF[bc]' minlen 7
476 line (split into two lines above) contains optimizer
477 information. In the example shown, the optimizer found that the match
478 should contain a substring C<de> at offset 1, plus substring C<gh>
479 at some offset between 3 and infinity. Moreover, when checking for
480 these substrings (to abandon impossible matches quickly), Perl will check
481 for the substring C<gh> before checking for the substring C<de>. The
482 optimizer may also use the knowledge that the match starts (at the
483 C<first> I<id>) with a character class, and no string
484 shorter than 7 characters can possibly match.
486 The fields of interest which may appear in this line are
490 =item C<anchored> I<STRING> C<at> I<POS>
492 =item C<floating> I<STRING> C<at> I<POS1..POS2>
496 =item C<matching floating/anchored>
498 Which substring to check first.
502 The minimal length of the match.
504 =item C<stclass> I<TYPE>
506 Type of first matching node.
510 Don't scan for the found substrings.
514 Means that the optimizer information is all that the regular
515 expression contains, and thus one does not need to enter the regex engine at
520 Set if the pattern contains C<\G>.
524 Set if the pattern starts with a repeated char (as in C<x+y>).
528 Set if the pattern starts with C<.*>.
532 Set if the pattern contain eval-groups, such as C<(?{ code })> and
535 =item C<anchored(TYPE)>
537 If the pattern may match only at a handful of places, with C<TYPE>
538 being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below.
542 If a substring is known to match at end-of-line only, it may be
543 followed by C<$>, as in C<floating 'k'$>.
545 The optimizer-specific information is used to avoid entering (a slow) regex
546 engine on strings that will not definitely match. If the C<isall> flag
547 is set, a call to the regex engine may be avoided even when the optimizer
548 found an appropriate place for the match.
550 Above the optimizer section is the list of I<nodes> of the compiled
551 form of the regex. Each line has format
553 C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
555 =head2 Types of Nodes
557 Here are the current possible types, with short descriptions:
560 This table is generated by regen/regcomp.pl. Any changes made here
563 =for regcomp.pl begin
565 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
569 END no End of program.
570 SUCCEED no Return from a subroutine, basically.
572 # Line Start Anchors:
573 SBOL no Match "" at beginning of line: /^/, /\A/
574 MBOL no Same, assuming multiline: /^/m
577 SEOL no Match "" at end of line: /$/
578 MEOL no Same, assuming multiline: /$/m
579 EOS no Match "" at end of string: /\z/
581 # Match Start Anchors:
582 GPOS no Matches where last m//g left off.
584 # Word Boundary Opcodes:
585 BOUND no Like BOUNDA for non-utf8, otherwise match
586 "" between any Unicode \w\W or \W\w
587 BOUNDL no Like BOUND/BOUNDU, but \w and \W are
588 defined by current locale
589 BOUNDU no Match "" at any boundary of a given type
591 BOUNDA no Match "" at any boundary between \w\W or
592 \W\w, where \w is [_a-zA-Z0-9]
593 NBOUND no Like NBOUNDA for non-utf8, otherwise match
594 "" between any Unicode \w\w or \W\W
595 NBOUNDL no Like NBOUND/NBOUNDU, but \w and \W are
596 defined by current locale
597 NBOUNDU no Match "" at any non-boundary of a given
598 type using using Unicode rules
599 NBOUNDA no Match "" betweeen any \w\w or \W\W, where
602 # [Special] alternatives:
603 REG_ANY no Match any one character (except newline).
604 SANY no Match any one character.
605 ANYOF sv Match character in (or not in) this class,
606 charclass single char match only
607 ANYOFD sv Like ANYOF, but /d is in effect
609 ANYOFL sv Like ANYOF, but /l is in effect
611 ANYOFPOSIXL sv Like ANYOFL, but matches [[:posix:]]
614 ANYOFM byte 1 Like ANYOF, but matches an invariant byte
615 as determined by the mask and arg
616 NANYOFM byte 1 complement of ANYOFM
618 # POSIX Character Classes:
619 POSIXD none Some [[:class:]] under /d; the FLAGS field
621 POSIXL none Some [[:class:]] under /l; the FLAGS field
623 POSIXU none Some [[:class:]] under /u; the FLAGS field
625 POSIXA none Some [[:class:]] under /a; the FLAGS field
627 NPOSIXD none complement of POSIXD, [[:^class:]]
628 NPOSIXL none complement of POSIXL, [[:^class:]]
629 NPOSIXU none complement of POSIXU, [[:^class:]]
630 NPOSIXA none complement of POSIXA, [[:^class:]]
632 ASCII none [[:ascii:]]
633 NASCII none [[:^ascii:]]
635 CLUMP no Match any extended grapheme cluster
640 # BRANCH The set of branches constituting a single choice are
641 # hooked together with their "next" pointers, since
642 # precedence prevents anything being concatenated to
643 # any individual branch. The "next" pointer of the last
644 # BRANCH in a choice points to the thing following the
645 # whole choice. This is also where the final "next"
646 # pointer of each individual branch points; each branch
647 # starts with the operand node of a BRANCH node.
649 BRANCH node Match this alternative, or the next...
653 EXACT str Match this string (preceded by length).
654 EXACTL str Like EXACT, but /l is in effect (used so
655 locale-related warnings can be checked
657 EXACTF str Match this non-UTF-8 string (not guaranteed
658 to be folded) using /id rules (w/len).
659 EXACTFL str Match this string (not guaranteed to be
660 folded) using /il rules (w/len).
661 EXACTFU str Match this string (folded iff in UTF-8,
662 length in folding doesn't change if not in
663 UTF-8) using /iu rules (w/len).
664 EXACTFAA str Match this string (not guaranteed to be
665 folded) using /iaa rules (w/len).
667 EXACTFU_SS str Match this string (folded iff in UTF-8,
668 length in folding may change even if not in
669 UTF-8) using /iu rules (w/len).
670 EXACTFLU8 str Rare circumstances: like EXACTFU, but is
671 under /l, UTF-8, folded, and everything in
673 EXACTFAA_NO_TRIE str Match this string (which is not trie-able;
674 not guaranteed to be folded) using /iaa
677 EXACT_ONLY8 str Like EXACT, but no strings that aren't in
679 EXACTFU_ONLY8 str Like EXACTFU, but no strings that aren't in
684 NOTHING no Match empty string.
685 # A variant of above which delimits a group, thus stops optimizations
686 TAIL no Match empty string. Can jump here from
691 # STAR,PLUS '?', and complex '*' and '+', are implemented as
692 # circular BRANCH structures. Simple cases
693 # (one character per match) are implemented with STAR
694 # and PLUS for speed and to minimize recursive plunges.
696 STAR node Match this (simple) thing 0 or more times.
697 PLUS node Match this (simple) thing 1 or more times.
699 CURLY sv 2 Match this simple thing {n,m} times.
700 CURLYN no 2 Capture next-after-this simple thing
701 CURLYM no 2 Capture this medium-complex thing {n,m}
703 CURLYX sv 2 Match this complex thing {n,m} times.
705 # This terminator creates a loop structure for CURLYX
706 WHILEM no Do curly processing and see if rest
711 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
712 OPEN num 1 Mark this point in input as start of #n.
713 CLOSE num 1 Close corresponding OPEN of #n.
714 SROPEN none Same as OPEN, but for script run
715 SRCLOSE none Close preceding SROPEN
717 REF num 1 Match some already matched string
718 REFF num 1 Match already matched string, folded using
719 native charset rules for non-utf8
720 REFFL num 1 Match already matched string, folded in
722 REFFU num 1 Match already matched string, folded using
723 unicode rules for non-utf8
724 REFFA num 1 Match already matched string, folded using
725 unicode rules for non-utf8, no mixing
728 # Named references. Code in regcomp.c assumes that these all are after
729 # the numbered references
730 NREF no-sv 1 Match some already matched string
731 NREFF no-sv 1 Match already matched string, folded using
732 native charset rules for non-utf8
733 NREFFL no-sv 1 Match already matched string, folded in
735 NREFFU num 1 Match already matched string, folded using
736 unicode rules for non-utf8
737 NREFFA num 1 Match already matched string, folded using
738 unicode rules for non-utf8, no mixing
741 # Support for long RE
742 LONGJMP off 1 1 Jump far away.
743 BRANCHJ off 1 1 BRANCH with long offset.
745 # Special Case Regops
746 IFMATCH off 1 1 Succeeds if the following matches.
747 UNLESSM off 1 1 Fails if the following matches.
748 SUSPEND off 1 1 "Independent" sub-RE.
749 IFTHEN off 1 1 Switch, should be preceded by switcher.
750 GROUPP num 1 Whether the group matched.
754 EVAL evl/flags Execute some Perl code.
759 MINMOD no Next operator is not greedy.
760 LOGICAL no Next opcode should set the flag only.
762 # This is not used yet
763 RENUM off 1 1 Group with independently numbered parens.
767 # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
768 # have inline charclass data (ascii only), the 'C' store it in the
771 TRIE trie 1 Match many EXACT(F[ALU]?)? at once.
773 TRIEC trie Same as TRIE, but with embedded charclass
776 AHOCORASICK trie 1 Aho Corasick stclass. flags==type
777 AHOCORASICKC trie Same as AHOCORASICK, but with embedded
778 charclass charclass data
781 GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
783 # Special conditionals
784 NGROUPP no-sv 1 Whether the group matched.
785 INSUBP num 1 Whether we are in a specific recurse.
786 DEFINEP none 1 Never execute directly.
789 ENDLIKE none Used only for the type field of verbs
790 OPFAIL no-sv 1 Same as (?!), but with verb arg
791 ACCEPT no-sv/num Accepts the current matched string, with
794 # Verbs With Arguments
795 VERB no-sv 1 Used only for the type field of verbs
796 PRUNE no-sv 1 Pattern fails at this startpoint if no-
797 backtracking through this
798 MARKPOINT no-sv 1 Push the current location for rollback by
800 SKIP no-sv 1 On failure skip forward (to the mark)
802 COMMIT no-sv 1 Pattern fails outright if backtracking
804 CUTGROUP no-sv 1 On failure go to the next alternation in
807 # Control what to keep in $&.
808 KEEPS no $& begins here.
810 # New charclass like patterns
811 LNBREAK none generic newline pattern
815 # This is not really a node, but an optimized away piece of a "long"
816 # node. To simplify debugging output, we mark it as if it were a node
817 OPTIMIZED off Placeholder for dump.
819 # Special opcode with the property that no opcode in a compiled program
820 # will ever be of this type. Thus it can be used as a flag value that
821 # no other opcode has been seen. END is used similarly, in that an END
822 # node cant be optimized. So END implies "unoptimizable" and PSEUDO
823 # mean "not seen anything to optimize yet".
824 PSEUDO off Pseudo opcode for internal use.
828 =for unprinted-credits
829 Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
831 Following the optimizer information is a dump of the offset/length
832 table, here split across several lines:
835 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
836 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
837 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
838 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
840 The first line here indicates that the offset/length table contains 45
841 entries. Each entry is a pair of integers, denoted by C<offset[length]>.
842 Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
843 entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:>
844 (the C<1: ANYOF[bc]>) begins at character position 1 in the
845 pre-compiled form of the regex, and has a length of 4 characters.
846 C<5[1]> in position 12
847 indicates that the node labeled C<12:>
848 (the C<< 12: EXACT <d> >>) begins at character position 5 in the
849 pre-compiled form of the regex, and has a length of 1 character.
850 C<12[1]> in position 14
851 indicates that the node labeled C<14:>
852 (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
853 pre-compiled form of the regex, and has a length of 1 character---that
854 is, it corresponds to the C<+> symbol in the precompiled regex.
856 C<0[0]> items indicate that there is no corresponding node.
858 =head2 Run-time Output
860 First of all, when doing a match, one may get no run-time output even
861 if debugging is enabled. This means that the regex engine was never
862 entered and that all of the job was therefore done by the optimizer.
864 If the regex engine was entered, the output may look like this:
866 Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
867 Setting an EVAL scope, savestack=3
868 2 <ab> <cdefg__gh_> | 1: ANYOF
869 3 <abc> <defg__gh_> | 11: EXACT <d>
870 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
871 4 <abcd> <efg__gh_> | 26: WHILEM
872 0 out of 1..32767 cc=effff31c
873 4 <abcd> <efg__gh_> | 15: OPEN1
874 4 <abcd> <efg__gh_> | 17: EXACT <e>
875 5 <abcde> <fg__gh_> | 19: STAR
876 EXACT <f> can match 1 times out of 32767...
877 Setting an EVAL scope, savestack=3
878 6 <bcdef> <g__gh__> | 22: EXACT <g>
879 7 <bcdefg> <__gh__> | 24: CLOSE1
880 7 <bcdefg> <__gh__> | 26: WHILEM
881 1 out of 1..32767 cc=effff31c
882 Setting an EVAL scope, savestack=12
883 7 <bcdefg> <__gh__> | 15: OPEN1
884 7 <bcdefg> <__gh__> | 17: EXACT <e>
885 restoring \1 to 4(4)..7
886 failed, try continuation...
887 7 <bcdefg> <__gh__> | 27: NOTHING
888 7 <bcdefg> <__gh__> | 28: EXACT <h>
892 The most significant information in the output is about the particular I<node>
893 of the compiled regex that is currently being tested against the target string.
894 The format of these lines is
896 C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
898 The I<TYPE> info is indented with respect to the backtracking level.
899 Other incidental information appears interspersed within.
901 =head1 Debugging Perl Memory Usage
903 Perl is a profligate wastrel when it comes to memory use. There
904 is a saying that to estimate memory usage of Perl, assume a reasonable
905 algorithm for memory allocation, multiply that estimate by 10, and
906 while you still may miss the mark, at least you won't be quite so
907 astonished. This is not absolutely true, but may provide a good
908 grasp of what happens.
910 Assume that an integer cannot take less than 20 bytes of memory, a
911 float cannot take less than 24 bytes, a string cannot take less
912 than 32 bytes (all these examples assume 32-bit architectures, the
913 result are quite a bit worse on 64-bit architectures). If a variable
914 is accessed in two of three different ways (which require an integer,
915 a float, or a string), the memory footprint may increase yet another
916 20 bytes. A sloppy malloc(3) implementation can inflate these
917 numbers dramatically.
919 On the opposite end of the scale, a declaration like
923 may take up to 500 bytes of memory, depending on which release of Perl
926 Anecdotal estimates of source-to-compiled code bloat suggest an
927 eightfold increase. This means that the compiled form of reasonable
928 (normally commented, properly indented etc.) code will take
929 about eight times more space in memory than the code took
932 The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
933 (it was available only if Perl was built with C<-DDEBUGGING>).
934 The switch was used to track Perl's memory allocations and possible
935 memory leaks. These days the use of malloc debugging tools like
936 F<Purify> or F<valgrind> is suggested instead. See also
937 L<perlhacktips/PERL_MEM_LOG>.
939 One way to find out how much memory is being used by Perl data
940 structures is to install the Devel::Size module from CPAN: it gives
941 you the minimum number of bytes required to store a particular data
942 structure. Please be mindful of the difference between the size()
945 If Perl has been compiled using Perl's malloc you can analyze Perl
946 memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
948 =head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
950 If your perl is using Perl's malloc() and was compiled with the
951 necessary switches (this is the default), then it will print memory
952 usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
953 > 1 >>, and before termination of the program when C<<
954 $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to
955 the following example:
957 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
958 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
959 14216 free: 130 117 28 7 9 0 2 2 1 0 0
961 60924 used: 125 137 161 55 7 8 6 16 2 0 1
963 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
964 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
965 30888 free: 245 78 85 13 6 2 1 3 2 0 1
967 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
969 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
971 It is possible to ask for such a statistic at arbitrary points in
972 your execution using the mstat() function out of the standard
975 Here is some explanation of that format:
979 =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
981 Perl's malloc() uses bucketed allocations. Every request is rounded
982 up to the closest bucket size available, and a bucket is taken from
983 the pool of buckets of that size.
985 The line above describes the limits of buckets currently in use.
986 Each bucket has two sizes: memory footprint and the maximal size
987 of user data that can fit into this bucket. Suppose in the above
988 example that the smallest bucket were size 4. The biggest bucket
989 would have usable size 8188, and the memory footprint would be 8192.
991 In a Perl built for debugging, some buckets may have negative usable
992 size. This means that these buckets cannot (and will not) be used.
993 For larger buckets, the memory footprint may be one page greater
994 than a power of 2. If so, the corresponding power of two is
995 printed in the C<APPROX> field above.
999 The 1 or 2 rows of numbers following that correspond to the number
1000 of buckets of each size between C<SMALLEST> and C<GREATEST>. In
1001 the first row, the sizes (memory footprints) of buckets are powers
1002 of two--or possibly one page greater. In the second row, if present,
1003 the memory footprints of the buckets are between the memory footprints
1004 of two buckets "above".
1006 For example, suppose under the previous example, the memory footprints
1009 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
1012 With a non-C<DEBUGGING> perl, the buckets starting from C<128> have
1013 a 4-byte overhead, and thus an 8192-long bucket may take up to
1014 8188-byte allocations.
1016 =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
1018 The first two fields give the total amount of memory perl sbrk(2)ed
1019 (ess-broken? :-) and number of sbrk(2)s used. The third number is
1020 what perl thinks about continuity of returned chunks. So long as
1021 this number is positive, malloc() will assume that it is probable
1022 that sbrk(2) will provide continuous memory.
1024 Memory allocated by external libraries is not counted.
1028 The amount of sbrk(2)ed memory needed to keep buckets aligned.
1030 =item C<heads: 2192>
1032 Although memory overhead of bigger buckets is kept inside the bucket, for
1033 smaller buckets, it is kept in separate areas. This field gives the
1034 total size of these areas.
1038 malloc() may want to subdivide a bigger bucket into smaller buckets.
1039 If only a part of the deceased bucket is left unsubdivided, the rest
1040 is kept as an element of a linked list. This field gives the total
1041 size of these chunks.
1045 To minimize the number of sbrk(2)s, malloc() asks for more memory. This
1046 field gives the size of the yet unused part, which is sbrk(2)ed, but