xref: /onnv-gate/usr/src/cmd/perl/5.8.4/distrib/pod/perldebguts.pod (revision 0:68f95e015346)
1=head1 NAME
2
3perldebguts - Guts of Perl debugging
4
5=head1 DESCRIPTION
6
7This is not the perldebug(1) manpage, which tells you how to use
8the debugger.  This manpage describes low-level details concerning
9the debugger's internals, which range from difficult to impossible
10to understand for anyone who isn't incredibly intimate with Perl's guts.
11Caveat lector.
12
13=head1 Debugger Internals
14
15Perl has special debugging hooks at compile-time and run-time used
16to create debugging environments.  These hooks are not to be confused
17with the I<perl -Dxxx> command described in L<perlrun>, which is
18usable only if a special Perl is built per the instructions in the
19F<INSTALL> podpage in the Perl source tree.
20
21For example, whenever you call Perl's built-in C<caller> function
22from the package C<DB>, the arguments that the corresponding stack
23frame was called with are copied to the C<@DB::args> array.  These
24mechanisms are enabled by calling Perl with the B<-d> switch.
25Specifically, the following additional features are enabled
26(cf. L<perlvar/$^P>):
27
28=over 4
29
30=item *
31
32Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require
33'perl5db.pl'}> if not present) before the first line of your program.
34
35=item *
36
37Each array C<@{"_<$filename"}> holds the lines of $filename for a
38file compiled by Perl.  The same is also true for C<eval>ed strings
39that contain subroutines, or which are currently being executed.
40The $filename for C<eval>ed strings looks like C<(eval 34)>.
41Code assertions in regexes look like C<(re_eval 19)>.
42
43Values in this array are magical in numeric context: they compare
44equal to zero only if the line is not breakable.
45
46=item *
47
48Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
49by line number.  Individual entries (as opposed to the whole hash)
50are settable.  Perl only cares about Boolean true here, although
51the values used by F<perl5db.pl> have the form
52C<"$break_condition\0$action">.
53
54The same holds for evaluated strings that contain subroutines, or
55which are currently being executed.  The $filename for C<eval>ed strings
56looks like C<(eval 34)> or  C<(re_eval 19)>.
57
58=item *
59
60Each scalar C<${"_<$filename"}> contains C<"_<$filename">.  This is
61also the case for evaluated strings that contain subroutines, or
62which are currently being executed.  The $filename for C<eval>ed
63strings looks like C<(eval 34)> or C<(re_eval 19)>.
64
65=item *
66
67After each C<require>d file is compiled, but before it is executed,
68C<DB::postponed(*{"_<$filename"})> is called if the subroutine
69C<DB::postponed> exists.  Here, the $filename is the expanded name of
70the C<require>d file, as found in the values of %INC.
71
72=item *
73
74After each subroutine C<subname> is compiled, the existence of
75C<$DB::postponed{subname}> is checked.  If this key exists,
76C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine
77also exists.
78
79=item *
80
81A hash C<%DB::sub> is maintained, whose keys are subroutine names
82and whose values have the form C<filename:startline-endline>.
83C<filename> has the form C<(eval 34)> for subroutines defined inside
84C<eval>s, or C<(re_eval 19)> for those within regex code assertions.
85
86=item *
87
88When the execution of your program reaches a point that can hold a
89breakpoint, the C<DB::DB()> subroutine is called if any of the variables
90C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true.  These variables
91are not C<local>izable.  This feature is disabled when executing
92inside C<DB::DB()>, including functions called from it
93unless C<< $^D & (1<<30) >> is true.
94
95=item *
96
97When execution of the program reaches a subroutine call, a call to
98C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> holding the
99name of the called subroutine. (This doesn't happen if the subroutine
100was compiled in the C<DB> package.)
101
102=back
103
104Note that if C<&DB::sub> needs external data for it to work, no
105subroutine call is possible without it. As an example, the standard
106debugger's C<&DB::sub> depends on the C<$DB::deep> variable
107(it defines how many levels of recursion deep into the debugger you can go
108before a mandatory break).  If C<$DB::deep> is not defined, subroutine
109calls are not possible, even though C<&DB::sub> exists.
110
111=head2 Writing Your Own Debugger
112
113=head3 Environment Variables
114
115The C<PERL5DB> environment variable can be used to define a debugger.
116For example, the minimal "working" debugger (it actually doesn't do anything)
117consists of one line:
118
119  sub DB::DB {}
120
121It can easily be defined like this:
122
123  $ PERL5DB="sub DB::DB {}" perl -d your-script
124
125Another brief debugger, slightly more useful, can be created
126with only the line:
127
128  sub DB::DB {print ++$i; scalar <STDIN>}
129
130This debugger prints a number which increments for each statement
131encountered and waits for you to hit a newline before continuing
132to the next statement.
133
134The following debugger is actually useful:
135
136  {
137    package DB;
138    sub DB  {}
139    sub sub {print ++$i, " $sub\n"; &$sub}
140  }
141
142It prints the sequence number of each subroutine call and the name of the
143called subroutine.  Note that C<&DB::sub> is being compiled into the
144package C<DB> through the use of the C<package> directive.
145
146When it starts, the debugger reads your rc file (F<./.perldb> or
147F<~/.perldb> under Unix), which can set important options.
148(A subroutine (C<&afterinit>) can be defined here as well; it is executed
149after the debugger completes its own initialization.)
150
151After the rc file is read, the debugger reads the PERLDB_OPTS
152environment variable and uses it to set debugger options. The
153contents of this variable are treated as if they were the argument
154of an C<o ...> debugger command (q.v. in L<perldebug/Options>).
155
156=head3 Debugger internal variables
157In addition to the file and subroutine-related variables mentioned above,
158the debugger also maintains various magical internal variables.
159
160=over 4
161
162=item *
163
164C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
165holds the lines of the currently-selected file (compiled by Perl), either
166explicitly chosen with the debugger's C<f> command, or implicitly by flow
167of execution.
168
169Values in this array are magical in numeric context: they compare
170equal to zero only if the line is not breakable.
171
172=item *
173
174C<%DB::dbline>, is an alias for C<%{"::_<current_file"}>, which
175contains breakpoints and actions keyed by line number in
176the currently-selected file, either explicitly chosen with the
177debugger's C<f> command, or implicitly by flow of execution.
178
179As previously noted, individual entries (as opposed to the whole hash)
180are settable.  Perl only cares about Boolean true here, although
181the values used by F<perl5db.pl> have the form
182C<"$break_condition\0$action">.
183
184=back
185
186=head3 Debugger customization functions
187
188Some functions are provided to simplify customization.
189
190=over 4
191
192=item *
193
194See L<perldebug/"Options"> for description of options parsed by
195C<DB::parse_options(string)> parses debugger options; see
196L<pperldebug/Options> for a description of options recognized.
197
198=item *
199
200C<DB::dump_trace(skip[,count])> skips the specified number of frames
201and returns a list containing information about the calling frames (all
202of them, if C<count> is missing).  Each entry is reference to a hash
203with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
204name, or info about C<eval>), C<args> (C<undef> or a reference to
205an array), C<file>, and C<line>.
206
207=item *
208
209C<DB::print_trace(FH, skip[, count[, short]])> prints
210formatted info about caller frames.  The last two functions may be
211convenient as arguments to C<< < >>, C<< << >> commands.
212
213=back
214
215Note that any variables and functions that are not documented in
216this manpages (or in L<perldebug>) are considered for internal
217use only, and as such are subject to change without notice.
218
219=head1 Frame Listing Output Examples
220
221The C<frame> option can be used to control the output of frame
222information.  For example, contrast this expression trace:
223
224 $ perl -de 42
225 Stack dump during die enabled outside of evals.
226
227 Loading DB routines from perl5db.pl patch level 0.94
228 Emacs support available.
229
230 Enter h or `h h' for help.
231
232 main::(-e:1):   0
233   DB<1> sub foo { 14 }
234
235   DB<2> sub bar { 3 }
236
237   DB<3> t print foo() * bar()
238 main::((eval 172):3):   print foo() + bar();
239 main::foo((eval 168):2):
240 main::bar((eval 170):2):
241 42
242
243with this one, once the C<o>ption C<frame=2> has been set:
244
245   DB<4> o f=2
246                frame = '2'
247   DB<5> t print foo() * bar()
248 3:      foo() * bar()
249 entering main::foo
250  2:     sub foo { 14 };
251 exited main::foo
252 entering main::bar
253  2:     sub bar { 3 };
254 exited main::bar
255 42
256
257By way of demonstration, we present below a laborious listing
258resulting from setting your C<PERLDB_OPTS> environment variable to
259the value C<f=n N>, and running I<perl -d -V> from the command line.
260Examples use various values of C<n> are shown to give you a feel
261for the difference between settings.  Long those it may be, this
262is not a complete listing, but only excerpts.
263
264=over 4
265
266=item 1
267
268  entering main::BEGIN
269   entering Config::BEGIN
270    Package lib/Exporter.pm.
271    Package lib/Carp.pm.
272   Package lib/Config.pm.
273   entering Config::TIEHASH
274   entering Exporter::import
275    entering Exporter::export
276  entering Config::myconfig
277   entering Config::FETCH
278   entering Config::FETCH
279   entering Config::FETCH
280   entering Config::FETCH
281
282=item 2
283
284  entering main::BEGIN
285   entering Config::BEGIN
286    Package lib/Exporter.pm.
287    Package lib/Carp.pm.
288   exited Config::BEGIN
289   Package lib/Config.pm.
290   entering Config::TIEHASH
291   exited Config::TIEHASH
292   entering Exporter::import
293    entering Exporter::export
294    exited Exporter::export
295   exited Exporter::import
296  exited main::BEGIN
297  entering Config::myconfig
298   entering Config::FETCH
299   exited Config::FETCH
300   entering Config::FETCH
301   exited Config::FETCH
302   entering Config::FETCH
303
304=item 4
305
306  in  $=main::BEGIN() from /dev/null:0
307   in  $=Config::BEGIN() from lib/Config.pm:2
308    Package lib/Exporter.pm.
309    Package lib/Carp.pm.
310   Package lib/Config.pm.
311   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
312   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
313    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
314  in  @=Config::myconfig() from /dev/null:0
315   in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
316   in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
317   in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
318   in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
319   in  $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
320   in  $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
321
322=item 6
323
324  in  $=main::BEGIN() from /dev/null:0
325   in  $=Config::BEGIN() from lib/Config.pm:2
326    Package lib/Exporter.pm.
327    Package lib/Carp.pm.
328   out $=Config::BEGIN() from lib/Config.pm:0
329   Package lib/Config.pm.
330   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
331   out $=Config::TIEHASH('Config') from lib/Config.pm:644
332   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
333    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
334    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
335   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
336  out $=main::BEGIN() from /dev/null:0
337  in  @=Config::myconfig() from /dev/null:0
338   in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
339   out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
340   in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
341   out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
342   in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
343   out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
344   in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
345
346=item 14
347
348  in  $=main::BEGIN() from /dev/null:0
349   in  $=Config::BEGIN() from lib/Config.pm:2
350    Package lib/Exporter.pm.
351    Package lib/Carp.pm.
352   out $=Config::BEGIN() from lib/Config.pm:0
353   Package lib/Config.pm.
354   in  $=Config::TIEHASH('Config') from lib/Config.pm:644
355   out $=Config::TIEHASH('Config') from lib/Config.pm:644
356   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
357    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
358    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
359   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
360  out $=main::BEGIN() from /dev/null:0
361  in  @=Config::myconfig() from /dev/null:0
362   in  $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
363   out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
364   in  $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
365   out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
366
367=item 30
368
369  in  $=CODE(0x15eca4)() from /dev/null:0
370   in  $=CODE(0x182528)() from lib/Config.pm:2
371    Package lib/Exporter.pm.
372   out $=CODE(0x182528)() from lib/Config.pm:0
373   scalar context return from CODE(0x182528): undef
374   Package lib/Config.pm.
375   in  $=Config::TIEHASH('Config') from lib/Config.pm:628
376   out $=Config::TIEHASH('Config') from lib/Config.pm:628
377   scalar context return from Config::TIEHASH:   empty hash
378   in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
379    in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
380    out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
381    scalar context return from Exporter::export: ''
382   out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
383   scalar context return from Exporter::import: ''
384
385=back
386
387In all cases shown above, the line indentation shows the call tree.
388If bit 2 of C<frame> is set, a line is printed on exit from a
389subroutine as well.  If bit 4 is set, the arguments are printed
390along with the caller info.  If bit 8 is set, the arguments are
391printed even if they are tied or references.  If bit 16 is set, the
392return value is printed, too.
393
394When a package is compiled, a line like this
395
396    Package lib/Carp.pm.
397
398is printed with proper indentation.
399
400=head1 Debugging regular expressions
401
402There are two ways to enable debugging output for regular expressions.
403
404If your perl is compiled with C<-DDEBUGGING>, you may use the
405B<-Dr> flag on the command line.
406
407Otherwise, one can C<use re 'debug'>, which has effects at
408compile time and run time.  It is not lexically scoped.
409
410=head2 Compile-time output
411
412The debugging output at compile time looks like this:
413
414  Compiling REx `[bc]d(ef*g)+h[ij]k$'
415  size 45 Got 364 bytes for offset annotations.
416  first at 1
417  rarest char g at 0
418  rarest char d at 0
419     1: ANYOF[bc](12)
420    12: EXACT <d>(14)
421    14: CURLYX[0] {1,32767}(28)
422    16:   OPEN1(18)
423    18:     EXACT <e>(20)
424    20:     STAR(23)
425    21:       EXACT <f>(0)
426    23:     EXACT <g>(25)
427    25:   CLOSE1(27)
428    27:   WHILEM[1/1](0)
429    28: NOTHING(29)
430    29: EXACT <h>(31)
431    31: ANYOF[ij](42)
432    42: EXACT <k>(44)
433    44: EOL(45)
434    45: END(0)
435  anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
436        stclass `ANYOF[bc]' minlen 7
437  Offsets: [45]
438  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
439  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
440  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
441  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
442  Omitting $` $& $' support.
443
444The first line shows the pre-compiled form of the regex.  The second
445shows the size of the compiled form (in arbitrary units, usually
4464-byte words) and the total number of bytes allocated for the
447offset/length table, usually 4+C<size>*8.  The next line shows the
448label I<id> of the first node that does a match.
449
450The
451
452  anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
453        stclass `ANYOF[bc]' minlen 7
454
455line (split into two lines above) contains optimizer
456information.  In the example shown, the optimizer found that the match
457should contain a substring C<de> at offset 1, plus substring C<gh>
458at some offset between 3 and infinity.  Moreover, when checking for
459these substrings (to abandon impossible matches quickly), Perl will check
460for the substring C<gh> before checking for the substring C<de>.  The
461optimizer may also use the knowledge that the match starts (at the
462C<first> I<id>) with a character class, and no string
463shorter than 7 characters can possibly match.
464
465The fields of interest which may appear in this line are
466
467=over 4
468
469=item C<anchored> I<STRING> C<at> I<POS>
470
471=item C<floating> I<STRING> C<at> I<POS1..POS2>
472
473See above.
474
475=item C<matching floating/anchored>
476
477Which substring to check first.
478
479=item C<minlen>
480
481The minimal length of the match.
482
483=item C<stclass> I<TYPE>
484
485Type of first matching node.
486
487=item C<noscan>
488
489Don't scan for the found substrings.
490
491=item C<isall>
492
493Means that the optimizer information is all that the regular
494expression contains, and thus one does not need to enter the regex engine at
495all.
496
497=item C<GPOS>
498
499Set if the pattern contains C<\G>.
500
501=item C<plus>
502
503Set if the pattern starts with a repeated char (as in C<x+y>).
504
505=item C<implicit>
506
507Set if the pattern starts with C<.*>.
508
509=item C<with eval>
510
511Set if the pattern contain eval-groups, such as C<(?{ code })> and
512C<(??{ code })>.
513
514=item C<anchored(TYPE)>
515
516If the pattern may match only at a handful of places, (with C<TYPE>
517being C<BOL>, C<MBOL>, or C<GPOS>.  See the table below.
518
519=back
520
521If a substring is known to match at end-of-line only, it may be
522followed by C<$>, as in C<floating `k'$>.
523
524The optimizer-specific information is used to avoid entering (a slow) regex
525engine on strings that will not definitely match.  If the C<isall> flag
526is set, a call to the regex engine may be avoided even when the optimizer
527found an appropriate place for the match.
528
529Above the optimizer section is the list of I<nodes> of the compiled
530form of the regex.  Each line has format
531
532C<   >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
533
534=head2 Types of nodes
535
536Here are the possible types, with short descriptions:
537
538    # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
539
540    # Exit points
541    END		no	End of program.
542    SUCCEED	no	Return from a subroutine, basically.
543
544    # Anchors:
545    BOL		no	Match "" at beginning of line.
546    MBOL	no	Same, assuming multiline.
547    SBOL	no	Same, assuming singleline.
548    EOS		no	Match "" at end of string.
549    EOL		no	Match "" at end of line.
550    MEOL	no	Same, assuming multiline.
551    SEOL	no	Same, assuming singleline.
552    BOUND	no	Match "" at any word boundary
553    BOUNDL	no	Match "" at any word boundary
554    NBOUND	no	Match "" at any word non-boundary
555    NBOUNDL	no	Match "" at any word non-boundary
556    GPOS	no	Matches where last m//g left off.
557
558    # [Special] alternatives
559    ANY		no	Match any one character (except newline).
560    SANY	no	Match any one character.
561    ANYOF	sv	Match character in (or not in) this class.
562    ALNUM	no	Match any alphanumeric character
563    ALNUML	no	Match any alphanumeric char in locale
564    NALNUM	no	Match any non-alphanumeric character
565    NALNUML	no	Match any non-alphanumeric char in locale
566    SPACE	no	Match any whitespace character
567    SPACEL	no	Match any whitespace char in locale
568    NSPACE	no	Match any non-whitespace character
569    NSPACEL	no	Match any non-whitespace char in locale
570    DIGIT	no	Match any numeric character
571    NDIGIT	no	Match any non-numeric character
572
573    # BRANCH	The set of branches constituting a single choice are hooked
574    #		together with their "next" pointers, since precedence prevents
575    #		anything being concatenated to any individual branch.  The
576    #		"next" pointer of the last BRANCH in a choice points to the
577    #		thing following the whole choice.  This is also where the
578    #		final "next" pointer of each individual branch points; each
579    #		branch starts with the operand node of a BRANCH node.
580    #
581    BRANCH	node	Match this alternative, or the next...
582
583    # BACK	Normal "next" pointers all implicitly point forward; BACK
584    #		exists to make loop structures possible.
585    # not used
586    BACK	no	Match "", "next" ptr points backward.
587
588    # Literals
589    EXACT	sv	Match this string (preceded by length).
590    EXACTF	sv	Match this string, folded (prec. by length).
591    EXACTFL	sv	Match this string, folded in locale (w/len).
592
593    # Do nothing
594    NOTHING	no	Match empty string.
595    # A variant of above which delimits a group, thus stops optimizations
596    TAIL	no	Match empty string. Can jump here from outside.
597
598    # STAR,PLUS	'?', and complex '*' and '+', are implemented as circular
599    #		BRANCH structures using BACK.  Simple cases (one character
600    #		per match) are implemented with STAR and PLUS for speed
601    #		and to minimize recursive plunges.
602    #
603    STAR	node	Match this (simple) thing 0 or more times.
604    PLUS	node	Match this (simple) thing 1 or more times.
605
606    CURLY	sv 2	Match this simple thing {n,m} times.
607    CURLYN	no 2	Match next-after-this simple thing
608    #			{n,m} times, set parens.
609    CURLYM	no 2	Match this medium-complex thing {n,m} times.
610    CURLYX	sv 2	Match this complex thing {n,m} times.
611
612    # This terminator creates a loop structure for CURLYX
613    WHILEM	no	Do curly processing and see if rest matches.
614
615    # OPEN,CLOSE,GROUPP	...are numbered at compile time.
616    OPEN	num 1	Mark this point in input as start of #n.
617    CLOSE	num 1	Analogous to OPEN.
618
619    REF		num 1	Match some already matched string
620    REFF	num 1	Match already matched string, folded
621    REFFL	num 1	Match already matched string, folded in loc.
622
623    # grouping assertions
624    IFMATCH	off 1 2	Succeeds if the following matches.
625    UNLESSM	off 1 2	Fails if the following matches.
626    SUSPEND	off 1 1	"Independent" sub-regex.
627    IFTHEN	off 1 1	Switch, should be preceded by switcher .
628    GROUPP	num 1	Whether the group matched.
629
630    # Support for long regex
631    LONGJMP	off 1 1	Jump far away.
632    BRANCHJ	off 1 1	BRANCH with long offset.
633
634    # The heavy worker
635    EVAL	evl 1	Execute some Perl code.
636
637    # Modifiers
638    MINMOD	no	Next operator is not greedy.
639    LOGICAL	no	Next opcode should set the flag only.
640
641    # This is not used yet
642    RENUM	off 1 1	Group with independently numbered parens.
643
644    # This is not really a node, but an optimized away piece of a "long" node.
645    # To simplify debugging output, we mark it as if it were a node
646    OPTIMIZED	off	Placeholder for dump.
647
648=for unprinted-credits
649Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
650
651Following the optimizer information is a dump of the offset/length
652table, here split across several lines:
653
654  Offsets: [45]
655  	1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
656  	0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
657  	11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
658  	0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
659
660The first line here indicates that the offset/length table contains 45
661entries.  Each entry is a pair of integers, denoted by C<offset[length]>.
662Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
663entry #12 is C<5[1]>.  C<1[4]> indicates that the node labeled C<1:>
664(the C<1: ANYOF[bc]>) begins at character position 1 in the
665pre-compiled form of the regex, and has a length of 4 characters.
666C<5[1]> in position 12
667indicates that the node labeled C<12:>
668(the C<< 12: EXACT <d> >>) begins at character position 5 in the
669pre-compiled form of the regex, and has a length of 1 character.
670C<12[1]> in position 14
671indicates that the node labeled C<14:>
672(the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
673pre-compiled form of the regex, and has a length of 1 character---that
674is, it corresponds to the C<+> symbol in the precompiled regex.
675
676C<0[0]> items indicate that there is no corresponding node.
677
678=head2 Run-time output
679
680First of all, when doing a match, one may get no run-time output even
681if debugging is enabled.  This means that the regex engine was never
682entered and that all of the job was therefore done by the optimizer.
683
684If the regex engine was entered, the output may look like this:
685
686  Matching `[bc]d(ef*g)+h[ij]k$' against `abcdefg__gh__'
687    Setting an EVAL scope, savestack=3
688     2 <ab> <cdefg__gh_>    |  1: ANYOF
689     3 <abc> <defg__gh_>    | 11: EXACT <d>
690     4 <abcd> <efg__gh_>    | 13: CURLYX {1,32767}
691     4 <abcd> <efg__gh_>    | 26:   WHILEM
692				0 out of 1..32767  cc=effff31c
693     4 <abcd> <efg__gh_>    | 15:     OPEN1
694     4 <abcd> <efg__gh_>    | 17:     EXACT <e>
695     5 <abcde> <fg__gh_>    | 19:     STAR
696			     EXACT <f> can match 1 times out of 32767...
697    Setting an EVAL scope, savestack=3
698     6 <bcdef> <g__gh__>    | 22:       EXACT <g>
699     7 <bcdefg> <__gh__>    | 24:       CLOSE1
700     7 <bcdefg> <__gh__>    | 26:       WHILEM
701				    1 out of 1..32767  cc=effff31c
702    Setting an EVAL scope, savestack=12
703     7 <bcdefg> <__gh__>    | 15:         OPEN1
704     7 <bcdefg> <__gh__>    | 17:         EXACT <e>
705       restoring \1 to 4(4)..7
706				    failed, try continuation...
707     7 <bcdefg> <__gh__>    | 27:         NOTHING
708     7 <bcdefg> <__gh__>    | 28:         EXACT <h>
709				    failed...
710				failed...
711
712The most significant information in the output is about the particular I<node>
713of the compiled regex that is currently being tested against the target string.
714The format of these lines is
715
716C<    >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>>   |I<ID>:  I<TYPE>
717
718The I<TYPE> info is indented with respect to the backtracking level.
719Other incidental information appears interspersed within.
720
721=head1 Debugging Perl memory usage
722
723Perl is a profligate wastrel when it comes to memory use.  There
724is a saying that to estimate memory usage of Perl, assume a reasonable
725algorithm for memory allocation, multiply that estimate by 10, and
726while you still may miss the mark, at least you won't be quite so
727astonished.  This is not absolutely true, but may provide a good
728grasp of what happens.
729
730Assume that an integer cannot take less than 20 bytes of memory, a
731float cannot take less than 24 bytes, a string cannot take less
732than 32 bytes (all these examples assume 32-bit architectures, the
733result are quite a bit worse on 64-bit architectures).  If a variable
734is accessed in two of three different ways (which require an integer,
735a float, or a string), the memory footprint may increase yet another
73620 bytes.  A sloppy malloc(3) implementation can inflate these
737numbers dramatically.
738
739On the opposite end of the scale, a declaration like
740
741  sub foo;
742
743may take up to 500 bytes of memory, depending on which release of Perl
744you're running.
745
746Anecdotal estimates of source-to-compiled code bloat suggest an
747eightfold increase.  This means that the compiled form of reasonable
748(normally commented, properly indented etc.) code will take
749about eight times more space in memory than the code took
750on disk.
751
752The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
753(it was available only if Perl was built with C<-DDEBUGGING>).
754The switch was used to track Perl's memory allocations and possible
755memory leaks.  These days the use of malloc debugging tools like
756F<Purify> or F<valgrind> is suggested instead.
757
758One way to find out how much memory is being used by Perl data
759structures is to install the Devel::Size module from CPAN: it gives
760you the minimum number of bytes required to store a particular data
761structure.  Please be mindful of the difference between the size()
762and total_size().
763
764If Perl has been compiled using Perl's malloc you can analyze Perl
765memory usage by setting the $ENV{PERL_DEBUG_MSTATS}.
766
767=head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
768
769If your perl is using Perl's malloc() and was compiled with the
770necessary switches (this is the default), then it will print memory
771usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
772> 1 >>, and before termination of the program when C<<
773$ENV{PERL_DEBUG_MSTATS} >= 1 >>.  The report format is similar to
774the following example:
775
776  $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
777  Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
778     14216 free:   130   117    28     7     9   0   2     2   1 0 0
779		437    61    36     0     5
780     60924 used:   125   137   161    55     7   8   6    16   2 0 1
781		 74   109   304    84    20
782  Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
783  Memory allocation statistics after execution:   (buckets 4(4)..8188(8192)
784     30888 free:   245    78    85    13     6   2   1     3   2 0 1
785		315   162    39    42    11
786    175816 used:   265   176  1112   111    26  22  11    27   2 1 1
787		196   178  1066   798    39
788  Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
789
790It is possible to ask for such a statistic at arbitrary points in
791your execution using the mstat() function out of the standard
792Devel::Peek module.
793
794Here is some explanation of that format:
795
796=over 4
797
798=item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
799
800Perl's malloc() uses bucketed allocations.  Every request is rounded
801up to the closest bucket size available, and a bucket is taken from
802the pool of buckets of that size.
803
804The line above describes the limits of buckets currently in use.
805Each bucket has two sizes: memory footprint and the maximal size
806of user data that can fit into this bucket.  Suppose in the above
807example that the smallest bucket were size 4.  The biggest bucket
808would have usable size 8188, and the memory footprint would be 8192.
809
810In a Perl built for debugging, some buckets may have negative usable
811size.  This means that these buckets cannot (and will not) be used.
812For larger buckets, the memory footprint may be one page greater
813than a power of 2.  If so, case the corresponding power of two is
814printed in the C<APPROX> field above.
815
816=item Free/Used
817
818The 1 or 2 rows of numbers following that correspond to the number
819of buckets of each size between C<SMALLEST> and C<GREATEST>.  In
820the first row, the sizes (memory footprints) of buckets are powers
821of two--or possibly one page greater.  In the second row, if present,
822the memory footprints of the buckets are between the memory footprints
823of two buckets "above".
824
825For example, suppose under the previous example, the memory footprints
826were
827
828     free:    8     16    32    64    128  256 512 1024 2048 4096 8192
829	   4     12    24    48    80
830
831With non-C<DEBUGGING> perl, the buckets starting from C<128> have
832a 4-byte overhead, and thus an 8192-long bucket may take up to
8338188-byte allocations.
834
835=item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
836
837The first two fields give the total amount of memory perl sbrk(2)ed
838(ess-broken? :-) and number of sbrk(2)s used.  The third number is
839what perl thinks about continuity of returned chunks.  So long as
840this number is positive, malloc() will assume that it is probable
841that sbrk(2) will provide continuous memory.
842
843Memory allocated by external libraries is not counted.
844
845=item C<pad: 0>
846
847The amount of sbrk(2)ed memory needed to keep buckets aligned.
848
849=item C<heads: 2192>
850
851Although memory overhead of bigger buckets is kept inside the bucket, for
852smaller buckets, it is kept in separate areas.  This field gives the
853total size of these areas.
854
855=item C<chain: 0>
856
857malloc() may want to subdivide a bigger bucket into smaller buckets.
858If only a part of the deceased bucket is left unsubdivided, the rest
859is kept as an element of a linked list.  This field gives the total
860size of these chunks.
861
862=item C<tail: 6144>
863
864To minimize the number of sbrk(2)s, malloc() asks for more memory.  This
865field gives the size of the yet unused part, which is sbrk(2)ed, but
866never touched.
867
868=back
869
870=head2 Example of using B<-DL> switch
871
872(Note that -DL is obsolete since circa 5.6.0, and even before that
873Perl needed to be compiled with -DDEBUGGING.)
874
875Below we show how to analyse memory usage by
876
877  do 'lib/auto/POSIX/autosplit.ix';
878
879The file in question contains a header and 146 lines similar to
880
881  sub getcwd;
882
883B<WARNING>: The discussion below supposes 32-bit architecture.  In
884newer releases of Perl, memory usage of the constructs discussed
885here is greatly improved, but the story discussed below is a real-life
886story.  This story is mercilessly terse, and assumes rather more than cursory
887knowledge of Perl internals.  Type space to continue, `q' to quit.
888(Actually, you just want to skip to the next section.)
889
890Here is the itemized list of Perl allocations performed during parsing
891of this file:
892
893 !!! "after" at test.pl line 3.
894    Id  subtot   4   8  12  16  20  24  28  32  36  40  48  56  64  72  80 80+
895  0 02   13752   .   .   .   . 294   .   .   .   .   .   .   .   .   .   .   4
896  0 54    5545   .   .   8 124  16   .   .   .   1   1   .   .   .   .   .   3
897  5 05      32   .   .   .   .   .   .   .   1   .   .   .   .   .   .   .   .
898  6 02    7152   .   .   .   .   .   .   .   .   .   . 149   .   .   .   .   .
899  7 02    3600   .   .   .   .   . 150   .   .   .   .   .   .   .   .   .   .
900  7 03      64   .  -1   .   1   .   .   2   .   .   .   .   .   .   .   .   .
901  7 04    7056   .   .   .   .   .   .   .   .   .   .   .   .   .   .   .   7
902  7 17   38404   .   .   .   .   .   .   .   1   .   . 442 149   .   . 147   .
903  9 03    2078  17 249  32   .   .   .   .   2   .   .   .   .   .   .   .   .
904
905
906To see this list, insert two C<warn('!...')> statements around the call:
907
908  warn('!');
909  do 'lib/auto/POSIX/autosplit.ix';
910  warn('!!! "after"');
911
912and run it with Perl's B<-DL> option.  The first warn() will print
913memory allocation info before parsing the file and will memorize
914the statistics at this point (we ignore what it prints).  The second
915warn() prints increments with respect to these memorized data.  This
916is the printout shown above.
917
918Different I<Id>s on the left correspond to different subsystems of
919the perl interpreter.  They are just the first argument given to
920the perl memory allocation API named New().  To find what C<9 03>
921means, just B<grep> the perl source for C<903>.  You'll find it in
922F<util.c>, function savepvn().  (I know, you wonder why we told you
923to B<grep> and then gave away the answer.  That's because grepping
924the source is good for the soul.)  This function is used to store
925a copy of an existing chunk of memory.  Using a C debugger, one can
926see that the function was called either directly from gv_init() or
927via sv_magic(), and that gv_init() is called from gv_fetchpv()--which
928was itself called from newSUB().  Please stop to catch your breath now.
929
930B<NOTE>: To reach this point in the debugger and skip the calls to
931savepvn() during the compilation of the main program, you should
932set a C breakpoint
933in Perl_warn(), continue until this point is reached, and I<then> set
934a C breakpoint in Perl_savepvn().  Note that you may need to skip a
935handful of Perl_savepvn() calls that do not correspond to mass production
936of CVs (there are more C<903> allocations than 146 similar lines of
937F<lib/auto/POSIX/autosplit.ix>).  Note also that C<Perl_> prefixes are
938added by macroization code in perl header files to avoid conflicts
939with external libraries.
940
941Anyway, we see that C<903> ids correspond to creation of globs, twice
942per glob - for glob name, and glob stringification magic.
943
944Here are explanations for other I<Id>s above:
945
946=over 4
947
948=item C<717>
949
950Creates bigger C<XPV*> structures.  In the case above, it
951creates 3 C<AV>s per subroutine, one for a list of lexical variable
952names, one for a scratchpad (which contains lexical variables and
953C<targets>), and one for the array of scratchpads needed for
954recursion.
955
956It also creates a C<GV> and a C<CV> per subroutine, all called from
957start_subparse().
958
959=item C<002>
960
961Creates a C array corresponding to the C<AV> of scratchpads and the
962scratchpad itself.  The first fake entry of this scratchpad is
963created though the subroutine itself is not defined yet.
964
965It also creates C arrays to keep data for the stash.  This is one HV,
966but it grows; thus, there are 4 big allocations: the big chunks are not
967freed, but are kept as additional arenas for C<SV> allocations.
968
969=item C<054>
970
971Creates a C<HEK> for the name of the glob for the subroutine.  This
972name is a key in a I<stash>.
973
974Big allocations with this I<Id> correspond to allocations of new
975arenas to keep C<HE>.
976
977=item C<602>
978
979Creates a C<GP> for the glob for the subroutine.
980
981=item C<702>
982
983Creates the C<MAGIC> for the glob for the subroutine.
984
985=item C<704>
986
987Creates I<arenas> which keep SVs.
988
989=back
990
991=head2 B<-DL> details
992
993If Perl is run with B<-DL> option, then warn()s that start with `!'
994behave specially.  They print a list of I<categories> of memory
995allocations, and statistics of allocations of different sizes for
996these categories.
997
998If warn() string starts with
999
1000=over 4
1001
1002=item C<!!!>
1003
1004print changed categories only, print the differences in counts of allocations.
1005
1006=item C<!!>
1007
1008print grown categories only; print the absolute values of counts, and totals.
1009
1010=item C<!>
1011
1012print nonempty categories, print the absolute values of counts and totals.
1013
1014=back
1015
1016=head2 Limitations of B<-DL> statistics
1017
1018If an extension or external library does not use the Perl API to
1019allocate memory, such allocations are not counted.
1020
1021=head1 SEE ALSO
1022
1023L<perldebug>,
1024L<perlguts>,
1025L<perlrun>
1026L<re>,
1027and
1028L<Devel::DProf>.
1029