3 perldebguts - Guts of Perl debugging
7 This is not the perldebug(1) manpage, 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 $filename for C<eval>ed strings looks like C<(eval 34)>.
41 Code assertions in regexes look like C<(re_eval 19)>.
43 Values in this array are magical in numeric context: they compare
44 equal to zero only if the line is not breakable.
48 Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed
49 by line number. Individual entries (as opposed to the whole hash)
50 are settable. Perl only cares about Boolean true here, although
51 the values used by F<perl5db.pl> have the form
52 C<"$break_condition\0$action">.
54 The same holds for evaluated strings that contain subroutines, or
55 which are currently being executed. The $filename for C<eval>ed strings
56 looks like C<(eval 34)> or C<(re_eval 19)>.
60 Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is
61 also the case for evaluated strings that contain subroutines, or
62 which are currently being executed. The $filename for C<eval>ed
63 strings looks like C<(eval 34)> or C<(re_eval 19)>.
67 After each C<require>d file is compiled, but before it is executed,
68 C<DB::postponed(*{"_<$filename"})> is called if the subroutine
69 C<DB::postponed> exists. Here, the $filename is the expanded name of
70 the C<require>d file, as found in the values of %INC.
74 After each subroutine C<subname> is compiled, the existence of
75 C<$DB::postponed{subname}> is checked. If this key exists,
76 C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine
81 A hash C<%DB::sub> is maintained, whose keys are subroutine names
82 and whose values have the form C<filename:startline-endline>.
83 C<filename> has the form C<(eval 34)> for subroutines defined inside
84 C<eval>s, or C<(re_eval 19)> for those within regex code assertions.
88 When the execution of your program reaches a point that can hold a
89 breakpoint, the C<DB::DB()> subroutine is called if any of the variables
90 C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables
91 are not C<local>izable. This feature is disabled when executing
92 inside C<DB::DB()>, including functions called from it
93 unless C<< $^D & (1<<30) >> is true.
97 When execution of the program reaches a subroutine call, a call to
98 C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> holding the
99 name of the called subroutine. (This doesn't happen if the subroutine
100 was compiled in the C<DB> package.)
104 Note that if C<&DB::sub> needs external data for it to work, no
105 subroutine call is possible without it. As an example, the standard
106 debugger's C<&DB::sub> depends on the C<$DB::deep> variable
107 (it defines how many levels of recursion deep into the debugger you can go
108 before a mandatory break). If C<$DB::deep> is not defined, subroutine
109 calls are not possible, even though C<&DB::sub> exists.
111 =head2 Writing Your Own Debugger
113 =head3 Environment Variables
115 The C<PERL5DB> environment variable can be used to define a debugger.
116 For example, the minimal "working" debugger (it actually doesn't do anything)
117 consists of one line:
121 It can easily be defined like this:
123 $ PERL5DB="sub DB::DB {}" perl -d your-script
125 Another brief debugger, slightly more useful, can be created
128 sub DB::DB {print ++$i; scalar <STDIN>}
130 This debugger prints a number which increments for each statement
131 encountered and waits for you to hit a newline before continuing
132 to the next statement.
134 The following debugger is actually useful:
139 sub sub {print ++$i, " $sub\n"; &$sub}
142 It prints the sequence number of each subroutine call and the name of the
143 called subroutine. Note that C<&DB::sub> is being compiled into the
144 package C<DB> through the use of the C<package> directive.
146 When it starts, the debugger reads your rc file (F<./.perldb> or
147 F<~/.perldb> under Unix), which can set important options.
148 (A subroutine (C<&afterinit>) can be defined here as well; it is executed
149 after the debugger completes its own initialization.)
151 After the rc file is read, the debugger reads the PERLDB_OPTS
152 environment variable and uses it to set debugger options. The
153 contents of this variable are treated as if they were the argument
154 of an C<o ...> debugger command (q.v. in L<perldebug/Options>).
156 =head3 Debugger internal variables
157 In addition to the file and subroutine-related variables mentioned above,
158 the debugger also maintains various magical internal variables.
164 C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which
165 holds the lines of the currently-selected file (compiled by Perl), either
166 explicitly chosen with the debugger's C<f> command, or implicitly by flow
169 Values in this array are magical in numeric context: they compare
170 equal to zero only if the line is not breakable.
174 C<%DB::dbline>, is an alias for C<%{"::_<current_file"}>, which
175 contains breakpoints and actions keyed by line number in
176 the currently-selected file, either explicitly chosen with the
177 debugger's C<f> command, or implicitly by flow of execution.
179 As previously noted, individual entries (as opposed to the whole hash)
180 are settable. Perl only cares about Boolean true here, although
181 the values used by F<perl5db.pl> have the form
182 C<"$break_condition\0$action">.
186 =head3 Debugger customization functions
188 Some functions are provided to simplify customization.
194 See L<perldebug/"Configurable Options"> for a description of options parsed by
195 C<DB::parse_options(string)>.
199 C<DB::dump_trace(skip[,count])> skips the specified number of frames
200 and returns a list containing information about the calling frames (all
201 of them, if C<count> is missing). Each entry is reference to a hash
202 with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
203 name, or info about C<eval>), C<args> (C<undef> or a reference to
204 an array), C<file>, and C<line>.
208 C<DB::print_trace(FH, skip[, count[, short]])> prints
209 formatted info about caller frames. The last two functions may be
210 convenient as arguments to C<< < >>, C<< << >> commands.
214 Note that any variables and functions that are not documented in
215 this manpages (or in L<perldebug>) are considered for internal
216 use only, and as such are subject to change without notice.
218 =head1 Frame Listing Output Examples
220 The C<frame> option can be used to control the output of frame
221 information. For example, contrast this expression trace:
224 Stack dump during die enabled outside of evals.
226 Loading DB routines from perl5db.pl patch level 0.94
227 Emacs support available.
229 Enter h or `h h' for help.
236 DB<3> t print foo() * bar()
237 main::((eval 172):3): print foo() + bar();
238 main::foo((eval 168):2):
239 main::bar((eval 170):2):
242 with this one, once the C<o>ption C<frame=2> has been set:
246 DB<5> t print foo() * bar()
256 By way of demonstration, we present below a laborious listing
257 resulting from setting your C<PERLDB_OPTS> environment variable to
258 the value C<f=n N>, and running I<perl -d -V> from the command line.
259 Examples use various values of C<n> are shown to give you a feel
260 for the difference between settings. Long those it may be, this
261 is not a complete listing, but only excerpts.
268 entering Config::BEGIN
269 Package lib/Exporter.pm.
271 Package lib/Config.pm.
272 entering Config::TIEHASH
273 entering Exporter::import
274 entering Exporter::export
275 entering Config::myconfig
276 entering Config::FETCH
277 entering Config::FETCH
278 entering Config::FETCH
279 entering Config::FETCH
284 entering Config::BEGIN
285 Package lib/Exporter.pm.
288 Package lib/Config.pm.
289 entering Config::TIEHASH
290 exited Config::TIEHASH
291 entering Exporter::import
292 entering Exporter::export
293 exited Exporter::export
294 exited Exporter::import
296 entering Config::myconfig
297 entering Config::FETCH
299 entering Config::FETCH
301 entering Config::FETCH
305 in $=main::BEGIN() from /dev/null:0
306 in $=Config::BEGIN() from lib/Config.pm:2
307 Package lib/Exporter.pm.
309 Package lib/Config.pm.
310 in $=Config::TIEHASH('Config') from lib/Config.pm:644
311 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
312 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
313 in @=Config::myconfig() from /dev/null:0
314 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
315 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
316 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
317 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
318 in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
319 in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
323 in $=main::BEGIN() from /dev/null:0
324 in $=Config::BEGIN() from lib/Config.pm:2
325 Package lib/Exporter.pm.
327 out $=Config::BEGIN() from lib/Config.pm:0
328 Package lib/Config.pm.
329 in $=Config::TIEHASH('Config') from lib/Config.pm:644
330 out $=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 lib/
333 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
334 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
335 out $=main::BEGIN() from /dev/null:0
336 in @=Config::myconfig() from /dev/null:0
337 in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
338 out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
339 in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
340 out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
341 in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
342 out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
343 in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
347 in $=main::BEGIN() from /dev/null:0
348 in $=Config::BEGIN() from lib/Config.pm:2
349 Package lib/Exporter.pm.
351 out $=Config::BEGIN() from lib/Config.pm:0
352 Package lib/Config.pm.
353 in $=Config::TIEHASH('Config') from lib/Config.pm:644
354 out $=Config::TIEHASH('Config') from lib/Config.pm:644
355 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
356 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
357 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
358 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
359 out $=main::BEGIN() from /dev/null:0
360 in @=Config::myconfig() from /dev/null:0
361 in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
362 out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
363 in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
364 out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
368 in $=CODE(0x15eca4)() from /dev/null:0
369 in $=CODE(0x182528)() from lib/Config.pm:2
370 Package lib/Exporter.pm.
371 out $=CODE(0x182528)() from lib/Config.pm:0
372 scalar context return from CODE(0x182528): undef
373 Package lib/Config.pm.
374 in $=Config::TIEHASH('Config') from lib/Config.pm:628
375 out $=Config::TIEHASH('Config') from lib/Config.pm:628
376 scalar context return from Config::TIEHASH: empty hash
377 in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
378 in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
379 out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
380 scalar context return from Exporter::export: ''
381 out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
382 scalar context return from Exporter::import: ''
386 In all cases shown above, the line indentation shows the call tree.
387 If bit 2 of C<frame> is set, a line is printed on exit from a
388 subroutine as well. If bit 4 is set, the arguments are printed
389 along with the caller info. If bit 8 is set, the arguments are
390 printed even if they are tied or references. If bit 16 is set, the
391 return value is printed, too.
393 When a package is compiled, a line like this
397 is printed with proper indentation.
399 =head1 Debugging regular expressions
401 There are two ways to enable debugging output for regular expressions.
403 If your perl is compiled with C<-DDEBUGGING>, you may use the
404 B<-Dr> flag on the command line.
406 Otherwise, one can C<use re 'debug'>, which has effects at
407 compile time and run time. It is not lexically scoped.
409 =head2 Compile-time output
411 The debugging output at compile time looks like this:
413 Compiling REx `[bc]d(ef*g)+h[ij]k$'
414 size 45 Got 364 bytes for offset annotations.
420 14: CURLYX[0] {1,32767}(28)
434 anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
435 stclass `ANYOF[bc]' minlen 7
437 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
438 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
439 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
440 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
441 Omitting $` $& $' support.
443 The first line shows the pre-compiled form of the regex. The second
444 shows the size of the compiled form (in arbitrary units, usually
445 4-byte words) and the total number of bytes allocated for the
446 offset/length table, usually 4+C<size>*8. The next line shows the
447 label I<id> of the first node that does a match.
451 anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
452 stclass `ANYOF[bc]' minlen 7
454 line (split into two lines above) contains optimizer
455 information. In the example shown, the optimizer found that the match
456 should contain a substring C<de> at offset 1, plus substring C<gh>
457 at some offset between 3 and infinity. Moreover, when checking for
458 these substrings (to abandon impossible matches quickly), Perl will check
459 for the substring C<gh> before checking for the substring C<de>. The
460 optimizer may also use the knowledge that the match starts (at the
461 C<first> I<id>) with a character class, and no string
462 shorter than 7 characters can possibly match.
464 The fields of interest which may appear in this line are
468 =item C<anchored> I<STRING> C<at> I<POS>
470 =item C<floating> I<STRING> C<at> I<POS1..POS2>
474 =item C<matching floating/anchored>
476 Which substring to check first.
480 The minimal length of the match.
482 =item C<stclass> I<TYPE>
484 Type of first matching node.
488 Don't scan for the found substrings.
492 Means that the optimizer information is all that the regular
493 expression contains, and thus one does not need to enter the regex engine at
498 Set if the pattern contains C<\G>.
502 Set if the pattern starts with a repeated char (as in C<x+y>).
506 Set if the pattern starts with C<.*>.
510 Set if the pattern contain eval-groups, such as C<(?{ code })> and
513 =item C<anchored(TYPE)>
515 If the pattern may match only at a handful of places, (with C<TYPE>
516 being C<BOL>, C<MBOL>, or C<GPOS>. See the table below.
520 If a substring is known to match at end-of-line only, it may be
521 followed by C<$>, as in C<floating `k'$>.
523 The optimizer-specific information is used to avoid entering (a slow) regex
524 engine on strings that will not definitely match. If the C<isall> flag
525 is set, a call to the regex engine may be avoided even when the optimizer
526 found an appropriate place for the match.
528 Above the optimizer section is the list of I<nodes> of the compiled
529 form of the regex. Each line has format
531 C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
533 =head2 Types of nodes
535 Here are the possible types, with short descriptions:
537 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
540 END no End of program.
541 SUCCEED no Return from a subroutine, basically.
544 BOL no Match "" at beginning of line.
545 MBOL no Same, assuming multiline.
546 SBOL no Same, assuming singleline.
547 EOS no Match "" at end of string.
548 EOL no Match "" at end of line.
549 MEOL no Same, assuming multiline.
550 SEOL no Same, assuming singleline.
551 BOUND no Match "" at any word boundary
552 BOUNDL no Match "" at any word boundary
553 NBOUND no Match "" at any word non-boundary
554 NBOUNDL no Match "" at any word non-boundary
555 GPOS no Matches where last m//g left off.
557 # [Special] alternatives
558 ANY no Match any one character (except newline).
559 SANY no Match any one character.
560 ANYOF sv Match character in (or not in) this class.
561 ALNUM no Match any alphanumeric character
562 ALNUML no Match any alphanumeric char in locale
563 NALNUM no Match any non-alphanumeric character
564 NALNUML no Match any non-alphanumeric char in locale
565 SPACE no Match any whitespace character
566 SPACEL no Match any whitespace char in locale
567 NSPACE no Match any non-whitespace character
568 NSPACEL no Match any non-whitespace char in locale
569 DIGIT no Match any numeric character
570 NDIGIT no Match any non-numeric character
572 # BRANCH The set of branches constituting a single choice are hooked
573 # together with their "next" pointers, since precedence prevents
574 # anything being concatenated to any individual branch. The
575 # "next" pointer of the last BRANCH in a choice points to the
576 # thing following the whole choice. This is also where the
577 # final "next" pointer of each individual branch points; each
578 # branch starts with the operand node of a BRANCH node.
580 BRANCH node Match this alternative, or the next...
582 # BACK Normal "next" pointers all implicitly point forward; BACK
583 # exists to make loop structures possible.
585 BACK no Match "", "next" ptr points backward.
588 EXACT sv Match this string (preceded by length).
589 EXACTF sv Match this string, folded (prec. by length).
590 EXACTFL sv Match this string, folded in locale (w/len).
593 NOTHING no Match empty string.
594 # A variant of above which delimits a group, thus stops optimizations
595 TAIL no Match empty string. Can jump here from outside.
597 # STAR,PLUS '?', and complex '*' and '+', are implemented as circular
598 # BRANCH structures using BACK. Simple cases (one character
599 # per match) are implemented with STAR and PLUS for speed
600 # and to minimize recursive plunges.
602 STAR node Match this (simple) thing 0 or more times.
603 PLUS node Match this (simple) thing 1 or more times.
605 CURLY sv 2 Match this simple thing {n,m} times.
606 CURLYN no 2 Match next-after-this simple thing
607 # {n,m} times, set parens.
608 CURLYM no 2 Match this medium-complex thing {n,m} times.
609 CURLYX sv 2 Match this complex thing {n,m} times.
611 # This terminator creates a loop structure for CURLYX
612 WHILEM no Do curly processing and see if rest matches.
614 # OPEN,CLOSE,GROUPP ...are numbered at compile time.
615 OPEN num 1 Mark this point in input as start of #n.
616 CLOSE num 1 Analogous to OPEN.
618 REF num 1 Match some already matched string
619 REFF num 1 Match already matched string, folded
620 REFFL num 1 Match already matched string, folded in loc.
622 # grouping assertions
623 IFMATCH off 1 2 Succeeds if the following matches.
624 UNLESSM off 1 2 Fails if the following matches.
625 SUSPEND off 1 1 "Independent" sub-regex.
626 IFTHEN off 1 1 Switch, should be preceded by switcher .
627 GROUPP num 1 Whether the group matched.
629 # Support for long regex
630 LONGJMP off 1 1 Jump far away.
631 BRANCHJ off 1 1 BRANCH with long offset.
634 EVAL evl 1 Execute some Perl code.
637 MINMOD no Next operator is not greedy.
638 LOGICAL no Next opcode should set the flag only.
640 # This is not used yet
641 RENUM off 1 1 Group with independently numbered parens.
643 # This is not really a node, but an optimized away piece of a "long" node.
644 # To simplify debugging output, we mark it as if it were a node
645 OPTIMIZED off Placeholder for dump.
647 =for unprinted-credits
648 Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
650 Following the optimizer information is a dump of the offset/length
651 table, here split across several lines:
654 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
655 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
656 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
657 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
659 The first line here indicates that the offset/length table contains 45
660 entries. Each entry is a pair of integers, denoted by C<offset[length]>.
661 Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
662 entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:>
663 (the C<1: ANYOF[bc]>) begins at character position 1 in the
664 pre-compiled form of the regex, and has a length of 4 characters.
665 C<5[1]> in position 12
666 indicates that the node labeled C<12:>
667 (the C<< 12: EXACT <d> >>) begins at character position 5 in the
668 pre-compiled form of the regex, and has a length of 1 character.
669 C<12[1]> in position 14
670 indicates that the node labeled C<14:>
671 (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
672 pre-compiled form of the regex, and has a length of 1 character---that
673 is, it corresponds to the C<+> symbol in the precompiled regex.
675 C<0[0]> items indicate that there is no corresponding node.
677 =head2 Run-time output
679 First of all, when doing a match, one may get no run-time output even
680 if debugging is enabled. This means that the regex engine was never
681 entered and that all of the job was therefore done by the optimizer.
683 If the regex engine was entered, the output may look like this:
685 Matching `[bc]d(ef*g)+h[ij]k$' against `abcdefg__gh__'
686 Setting an EVAL scope, savestack=3
687 2 <ab> <cdefg__gh_> | 1: ANYOF
688 3 <abc> <defg__gh_> | 11: EXACT <d>
689 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767}
690 4 <abcd> <efg__gh_> | 26: WHILEM
691 0 out of 1..32767 cc=effff31c
692 4 <abcd> <efg__gh_> | 15: OPEN1
693 4 <abcd> <efg__gh_> | 17: EXACT <e>
694 5 <abcde> <fg__gh_> | 19: STAR
695 EXACT <f> can match 1 times out of 32767...
696 Setting an EVAL scope, savestack=3
697 6 <bcdef> <g__gh__> | 22: EXACT <g>
698 7 <bcdefg> <__gh__> | 24: CLOSE1
699 7 <bcdefg> <__gh__> | 26: WHILEM
700 1 out of 1..32767 cc=effff31c
701 Setting an EVAL scope, savestack=12
702 7 <bcdefg> <__gh__> | 15: OPEN1
703 7 <bcdefg> <__gh__> | 17: EXACT <e>
704 restoring \1 to 4(4)..7
705 failed, try continuation...
706 7 <bcdefg> <__gh__> | 27: NOTHING
707 7 <bcdefg> <__gh__> | 28: EXACT <h>
711 The most significant information in the output is about the particular I<node>
712 of the compiled regex that is currently being tested against the target string.
713 The format of these lines is
715 C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
717 The I<TYPE> info is indented with respect to the backtracking level.
718 Other incidental information appears interspersed within.
720 =head1 Debugging Perl memory usage
722 Perl is a profligate wastrel when it comes to memory use. There
723 is a saying that to estimate memory usage of Perl, assume a reasonable
724 algorithm for memory allocation, multiply that estimate by 10, and
725 while you still may miss the mark, at least you won't be quite so
726 astonished. This is not absolutely true, but may provide a good
727 grasp of what happens.
729 Assume that an integer cannot take less than 20 bytes of memory, a
730 float cannot take less than 24 bytes, a string cannot take less
731 than 32 bytes (all these examples assume 32-bit architectures, the
732 result are quite a bit worse on 64-bit architectures). If a variable
733 is accessed in two of three different ways (which require an integer,
734 a float, or a string), the memory footprint may increase yet another
735 20 bytes. A sloppy malloc(3) implementation can inflate these
736 numbers dramatically.
738 On the opposite end of the scale, a declaration like
742 may take up to 500 bytes of memory, depending on which release of Perl
745 Anecdotal estimates of source-to-compiled code bloat suggest an
746 eightfold increase. This means that the compiled form of reasonable
747 (normally commented, properly indented etc.) code will take
748 about eight times more space in memory than the code took
751 The B<-DL> command-line switch is obsolete since circa Perl 5.6.0
752 (it was available only if Perl was built with C<-DDEBUGGING>).
753 The switch was used to track Perl's memory allocations and possible
754 memory leaks. These days the use of malloc debugging tools like
755 F<Purify> or F<valgrind> is suggested instead. See also
756 L<perlhack/PERL_MEM_LOG>.
758 One way to find out how much memory is being used by Perl data
759 structures is to install the Devel::Size module from CPAN: it gives
760 you the minimum number of bytes required to store a particular data
761 structure. Please be mindful of the difference between the size()
764 If Perl has been compiled using Perl's malloc you can analyze Perl
765 memory usage by setting the $ENV{PERL_DEBUG_MSTATS}.
767 =head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
769 If your perl is using Perl's malloc() and was compiled with the
770 necessary switches (this is the default), then it will print memory
771 usage 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
774 the following example:
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
780 60924 used: 125 137 161 55 7 8 6 16 2 0 1
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
786 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
788 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
790 It is possible to ask for such a statistic at arbitrary points in
791 your execution using the mstat() function out of the standard
794 Here is some explanation of that format:
798 =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
800 Perl's malloc() uses bucketed allocations. Every request is rounded
801 up to the closest bucket size available, and a bucket is taken from
802 the pool of buckets of that size.
804 The line above describes the limits of buckets currently in use.
805 Each bucket has two sizes: memory footprint and the maximal size
806 of user data that can fit into this bucket. Suppose in the above
807 example that the smallest bucket were size 4. The biggest bucket
808 would have usable size 8188, and the memory footprint would be 8192.
810 In a Perl built for debugging, some buckets may have negative usable
811 size. This means that these buckets cannot (and will not) be used.
812 For larger buckets, the memory footprint may be one page greater
813 than a power of 2. If so, case the corresponding power of two is
814 printed in the C<APPROX> field above.
818 The 1 or 2 rows of numbers following that correspond to the number
819 of buckets of each size between C<SMALLEST> and C<GREATEST>. In
820 the first row, the sizes (memory footprints) of buckets are powers
821 of two--or possibly one page greater. In the second row, if present,
822 the memory footprints of the buckets are between the memory footprints
823 of two buckets "above".
825 For example, suppose under the previous example, the memory footprints
828 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
831 With non-C<DEBUGGING> perl, the buckets starting from C<128> have
832 a 4-byte overhead, and thus an 8192-long bucket may take up to
833 8188-byte allocations.
835 =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
837 The 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
839 what perl thinks about continuity of returned chunks. So long as
840 this number is positive, malloc() will assume that it is probable
841 that sbrk(2) will provide continuous memory.
843 Memory allocated by external libraries is not counted.
847 The amount of sbrk(2)ed memory needed to keep buckets aligned.
851 Although memory overhead of bigger buckets is kept inside the bucket, for
852 smaller buckets, it is kept in separate areas. This field gives the
853 total size of these areas.
857 malloc() may want to subdivide a bigger bucket into smaller buckets.
858 If only a part of the deceased bucket is left unsubdivided, the rest
859 is kept as an element of a linked list. This field gives the total
860 size of these chunks.
864 To minimize the number of sbrk(2)s, malloc() asks for more memory. This
865 field gives the size of the yet unused part, which is sbrk(2)ed, but