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/"Options"> for description of options parsed by
195 C<DB::parse_options(string)> parses debugger options; see
196 L<pperldebug/Options> for a description of options recognized.
200 C<DB::dump_trace(skip[,count])> skips the specified number of frames
201 and returns a list containing information about the calling frames (all
202 of them, if C<count> is missing). Each entry is reference to a hash
203 with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine
204 name, or info about C<eval>), C<args> (C<undef> or a reference to
205 an array), C<file>, and C<line>.
209 C<DB::print_trace(FH, skip[, count[, short]])> prints
210 formatted info about caller frames. The last two functions may be
211 convenient as arguments to C<< < >>, C<< << >> commands.
215 Note that any variables and functions that are not documented in
216 this manpages (or in L<perldebug>) are considered for internal
217 use only, and as such are subject to change without notice.
219 =head1 Frame Listing Output Examples
221 The C<frame> option can be used to control the output of frame
222 information. For example, contrast this expression trace:
225 Stack dump during die enabled outside of evals.
227 Loading DB routines from perl5db.pl patch level 0.94
228 Emacs support available.
230 Enter h or `h h' for help.
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):
243 with this one, once the C<o>ption C<frame=2> has been set:
247 DB<5> t print foo() * bar()
257 By way of demonstration, we present below a laborious listing
258 resulting from setting your C<PERLDB_OPTS> environment variable to
259 the value C<f=n N>, and running I<perl -d -V> from the command line.
260 Examples use various values of C<n> are shown to give you a feel
261 for the difference between settings. Long those it may be, this
262 is not a complete listing, but only excerpts.
269 entering Config::BEGIN
270 Package lib/Exporter.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
285 entering Config::BEGIN
286 Package lib/Exporter.pm.
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
297 entering Config::myconfig
298 entering Config::FETCH
300 entering Config::FETCH
302 entering Config::FETCH
306 in $=main::BEGIN() from /dev/null:0
307 in $=Config::BEGIN() from lib/Config.pm:2
308 Package lib/Exporter.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
324 in $=main::BEGIN() from /dev/null:0
325 in $=Config::BEGIN() from lib/Config.pm:2
326 Package lib/Exporter.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
348 in $=main::BEGIN() from /dev/null:0
349 in $=Config::BEGIN() from lib/Config.pm:2
350 Package lib/Exporter.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
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 $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
383 scalar context return from Exporter::import: ''
387 In all cases shown above, the line indentation shows the call tree.
388 If bit 2 of C<frame> is set, a line is printed on exit from a
389 subroutine as well. If bit 4 is set, the arguments are printed
390 along with the caller info. If bit 8 is set, the arguments are
391 printed even if they are tied or references. If bit 16 is set, the
392 return value is printed, too.
394 When a package is compiled, a line like this
398 is printed with proper indentation.
400 =head1 Debugging regular expressions
402 There are two ways to enable debugging output for regular expressions.
404 If your perl is compiled with C<-DDEBUGGING>, you may use the
405 B<-Dr> flag on the command line.
407 Otherwise, one can C<use re 'debug'>, which has effects at
408 compile time and run time. It is not lexically scoped.
410 =head2 Compile-time output
412 The debugging output at compile time looks like this:
414 Compiling REx `[bc]d(ef*g)+h[ij]k$'
415 size 45 Got 364 bytes for offset annotations.
421 14: CURLYX[0] {1,32767}(28)
435 anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
436 stclass `ANYOF[bc]' minlen 7
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.
444 The first line shows the pre-compiled form of the regex. The second
445 shows the size of the compiled form (in arbitrary units, usually
446 4-byte words) and the total number of bytes allocated for the
447 offset/length table, usually 4+C<size>*8. The next line shows the
448 label I<id> of the first node that does a match.
452 anchored `de' at 1 floating `gh' at 3..2147483647 (checking floating)
453 stclass `ANYOF[bc]' minlen 7
455 line (split into two lines above) contains optimizer
456 information. In the example shown, the optimizer found that the match
457 should contain a substring C<de> at offset 1, plus substring C<gh>
458 at some offset between 3 and infinity. Moreover, when checking for
459 these substrings (to abandon impossible matches quickly), Perl will check
460 for the substring C<gh> before checking for the substring C<de>. The
461 optimizer may also use the knowledge that the match starts (at the
462 C<first> I<id>) with a character class, and no string
463 shorter than 7 characters can possibly match.
465 The fields of interest which may appear in this line are
469 =item C<anchored> I<STRING> C<at> I<POS>
471 =item C<floating> I<STRING> C<at> I<POS1..POS2>
475 =item C<matching floating/anchored>
477 Which substring to check first.
481 The minimal length of the match.
483 =item C<stclass> I<TYPE>
485 Type of first matching node.
489 Don't scan for the found substrings.
493 Means that the optimizer information is all that the regular
494 expression contains, and thus one does not need to enter the regex engine at
499 Set if the pattern contains C<\G>.
503 Set if the pattern starts with a repeated char (as in C<x+y>).
507 Set if the pattern starts with C<.*>.
511 Set if the pattern contain eval-groups, such as C<(?{ code })> and
514 =item C<anchored(TYPE)>
516 If the pattern may match only at a handful of places, (with C<TYPE>
517 being C<BOL>, C<MBOL>, or C<GPOS>. See the table below.
521 If a substring is known to match at end-of-line only, it may be
522 followed by C<$>, as in C<floating `k'$>.
524 The optimizer-specific information is used to avoid entering (a slow) regex
525 engine on strings that will not definitely match. If the C<isall> flag
526 is set, a call to the regex engine may be avoided even when the optimizer
527 found an appropriate place for the match.
529 Above the optimizer section is the list of I<nodes> of the compiled
530 form of the regex. Each line has format
532 C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>)
534 =head2 Types of nodes
536 Here are the possible types, with short descriptions:
538 # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
541 END no End of program.
542 SUCCEED no Return from a subroutine, basically.
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.
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
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.
581 BRANCH node Match this alternative, or the next...
583 # BACK Normal "next" pointers all implicitly point forward; BACK
584 # exists to make loop structures possible.
586 BACK no Match "", "next" ptr points backward.
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).
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.
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.
603 STAR node Match this (simple) thing 0 or more times.
604 PLUS node Match this (simple) thing 1 or more times.
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.
612 # This terminator creates a loop structure for CURLYX
613 WHILEM no Do curly processing and see if rest matches.
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.
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.
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.
630 # Support for long regex
631 LONGJMP off 1 1 Jump far away.
632 BRANCHJ off 1 1 BRANCH with long offset.
635 EVAL evl 1 Execute some Perl code.
638 MINMOD no Next operator is not greedy.
639 LOGICAL no Next opcode should set the flag only.
641 # This is not used yet
642 RENUM off 1 1 Group with independently numbered parens.
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.
648 =for unprinted-credits
649 Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421
651 Following the optimizer information is a dump of the offset/length
652 table, here split across several lines:
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]
660 The first line here indicates that the offset/length table contains 45
661 entries. Each entry is a pair of integers, denoted by C<offset[length]>.
662 Entries are numbered starting with 1, so entry #1 here is C<1[4]> and
663 entry #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
665 pre-compiled form of the regex, and has a length of 4 characters.
666 C<5[1]> in position 12
667 indicates that the node labeled C<12:>
668 (the C<< 12: EXACT <d> >>) begins at character position 5 in the
669 pre-compiled form of the regex, and has a length of 1 character.
670 C<12[1]> in position 14
671 indicates that the node labeled C<14:>
672 (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the
673 pre-compiled form of the regex, and has a length of 1 character---that
674 is, it corresponds to the C<+> symbol in the precompiled regex.
676 C<0[0]> items indicate that there is no corresponding node.
678 =head2 Run-time output
680 First of all, when doing a match, one may get no run-time output even
681 if debugging is enabled. This means that the regex engine was never
682 entered and that all of the job was therefore done by the optimizer.
684 If the regex engine was entered, the output may look like this:
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>
712 The most significant information in the output is about the particular I<node>
713 of the compiled regex that is currently being tested against the target string.
714 The format of these lines is
716 C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE>
718 The I<TYPE> info is indented with respect to the backtracking level.
719 Other incidental information appears interspersed within.
721 =head1 Debugging Perl memory usage
723 Perl is a profligate wastrel when it comes to memory use. There
724 is a saying that to estimate memory usage of Perl, assume a reasonable
725 algorithm for memory allocation, multiply that estimate by 10, and
726 while you still may miss the mark, at least you won't be quite so
727 astonished. This is not absolutely true, but may provide a good
728 grasp of what happens.
730 Assume that an integer cannot take less than 20 bytes of memory, a
731 float cannot take less than 24 bytes, a string cannot take less
732 than 32 bytes (all these examples assume 32-bit architectures, the
733 result are quite a bit worse on 64-bit architectures). If a variable
734 is accessed in two of three different ways (which require an integer,
735 a float, or a string), the memory footprint may increase yet another
736 20 bytes. A sloppy malloc(3) implementation can inflate these
737 numbers dramatically.
739 On the opposite end of the scale, a declaration like
743 may take up to 500 bytes of memory, depending on which release of Perl
746 Anecdotal estimates of source-to-compiled code bloat suggest an
747 eightfold increase. This means that the compiled form of reasonable
748 (normally commented, properly indented etc.) code will take
749 about eight times more space in memory than the code took
752 The 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>).
754 The switch was used to track Perl's memory allocations and possible
755 memory leaks. These days the use of malloc debugging tools like
756 F<Purify> or F<valgrind> is suggested instead. See also
757 L<perlhack/PERL_MEM_LOG>.
759 One way to find out how much memory is being used by Perl data
760 structures is to install the Devel::Size module from CPAN: it gives
761 you the minimum number of bytes required to store a particular data
762 structure. Please be mindful of the difference between the size()
765 If Perl has been compiled using Perl's malloc you can analyze Perl
766 memory usage by setting the $ENV{PERL_DEBUG_MSTATS}.
768 =head2 Using C<$ENV{PERL_DEBUG_MSTATS}>
770 If your perl is using Perl's malloc() and was compiled with the
771 necessary switches (this is the default), then it will print memory
772 usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS}
773 > 1 >>, and before termination of the program when C<<
774 $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to
775 the following example:
777 $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
778 Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
779 14216 free: 130 117 28 7 9 0 2 2 1 0 0
781 60924 used: 125 137 161 55 7 8 6 16 2 0 1
783 Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
784 Memory allocation statistics after execution: (buckets 4(4)..8188(8192)
785 30888 free: 245 78 85 13 6 2 1 3 2 0 1
787 175816 used: 265 176 1112 111 26 22 11 27 2 1 1
789 Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
791 It is possible to ask for such a statistic at arbitrary points in
792 your execution using the mstat() function out of the standard
795 Here is some explanation of that format:
799 =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)>
801 Perl's malloc() uses bucketed allocations. Every request is rounded
802 up to the closest bucket size available, and a bucket is taken from
803 the pool of buckets of that size.
805 The line above describes the limits of buckets currently in use.
806 Each bucket has two sizes: memory footprint and the maximal size
807 of user data that can fit into this bucket. Suppose in the above
808 example that the smallest bucket were size 4. The biggest bucket
809 would have usable size 8188, and the memory footprint would be 8192.
811 In a Perl built for debugging, some buckets may have negative usable
812 size. This means that these buckets cannot (and will not) be used.
813 For larger buckets, the memory footprint may be one page greater
814 than a power of 2. If so, case the corresponding power of two is
815 printed in the C<APPROX> field above.
819 The 1 or 2 rows of numbers following that correspond to the number
820 of buckets of each size between C<SMALLEST> and C<GREATEST>. In
821 the first row, the sizes (memory footprints) of buckets are powers
822 of two--or possibly one page greater. In the second row, if present,
823 the memory footprints of the buckets are between the memory footprints
824 of two buckets "above".
826 For example, suppose under the previous example, the memory footprints
829 free: 8 16 32 64 128 256 512 1024 2048 4096 8192
832 With non-C<DEBUGGING> perl, the buckets starting from C<128> have
833 a 4-byte overhead, and thus an 8192-long bucket may take up to
834 8188-byte allocations.
836 =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS>
838 The first two fields give the total amount of memory perl sbrk(2)ed
839 (ess-broken? :-) and number of sbrk(2)s used. The third number is
840 what perl thinks about continuity of returned chunks. So long as
841 this number is positive, malloc() will assume that it is probable
842 that sbrk(2) will provide continuous memory.
844 Memory allocated by external libraries is not counted.
848 The amount of sbrk(2)ed memory needed to keep buckets aligned.
852 Although memory overhead of bigger buckets is kept inside the bucket, for
853 smaller buckets, it is kept in separate areas. This field gives the
854 total size of these areas.
858 malloc() may want to subdivide a bigger bucket into smaller buckets.
859 If only a part of the deceased bucket is left unsubdivided, the rest
860 is kept as an element of a linked list. This field gives the total
861 size of these chunks.
865 To minimize the number of sbrk(2)s, malloc() asks for more memory. This
866 field gives the size of the yet unused part, which is sbrk(2)ed, but