1PERLDEBGUTS(1)         Perl Programmers Reference Guide         PERLDEBGUTS(1)
2
3
4

NAME

6       perldebguts - Guts of Perl debugging
7

DESCRIPTION

9       This is not perldebug, which tells you how to use the debugger.  This
10       manpage describes low-level details concerning the debugger's
11       internals, which range from difficult to impossible to understand for
12       anyone who isn't incredibly intimate with Perl's guts.  Caveat lector.
13

Debugger Internals

15       Perl has special debugging hooks at compile-time and run-time used to
16       create debugging environments.  These hooks are not to be confused with
17       the perl -Dxxx command described in perlrun, which is usable only if a
18       special Perl is built per the instructions in the INSTALL podpage in
19       the Perl source tree.
20
21       For example, whenever you call Perl's built-in "caller" function from
22       the package "DB", the arguments that the corresponding stack frame was
23       called with are copied to the @DB::args array.  These mechanisms are
24       enabled by calling Perl with the -d switch.  Specifically, the
25       following additional features are enabled (cf. "$^P" in perlvar):
26
27       ·   Perl inserts the contents of $ENV{PERL5DB} (or "BEGIN {require
28           'perl5db.pl'}" if not present) before the first line of your
29           program.
30
31       ·   Each array "@{"_<$filename"}" holds the lines of $filename for a
32           file compiled by Perl.  The same is also true for "eval"ed strings
33           that contain subroutines, or which are currently being executed.
34           The $filename for "eval"ed strings looks like "(eval 34)".
35
36           Values in this array are magical in numeric context: they compare
37           equal to zero only if the line is not breakable.
38
39       ·   Each hash "%{"_<$filename"}" contains breakpoints and actions keyed
40           by line number.  Individual entries (as opposed to the whole hash)
41           are settable.  Perl only cares about Boolean true here, although
42           the values used by perl5db.pl have the form
43           "$break_condition\0$action".
44
45           The same holds for evaluated strings that contain subroutines, or
46           which are currently being executed.  The $filename for "eval"ed
47           strings looks like "(eval 34)".
48
49       ·   Each scalar "${"_<$filename"}" contains "_<$filename".  This is
50           also the case for evaluated strings that contain subroutines, or
51           which are currently being executed.  The $filename for "eval"ed
52           strings looks like "(eval 34)".
53
54       ·   After each "require"d file is compiled, but before it is executed,
55           "DB::postponed(*{"_<$filename"})" is called if the subroutine
56           "DB::postponed" exists.  Here, the $filename is the expanded name
57           of the "require"d file, as found in the values of %INC.
58
59       ·   After each subroutine "subname" is compiled, the existence of
60           $DB::postponed{subname} is checked.  If this key exists,
61           "DB::postponed(subname)" is called if the "DB::postponed"
62           subroutine also exists.
63
64       ·   A hash %DB::sub is maintained, whose keys are subroutine names and
65           whose values have the form "filename:startline-endline".
66           "filename" has the form "(eval 34)" for subroutines defined inside
67           "eval"s.
68
69       ·   When the execution of your program reaches a point that can hold a
70           breakpoint, the "DB::DB()" subroutine is called if any of the
71           variables $DB::trace, $DB::single, or $DB::signal is true.  These
72           variables are not "local"izable.  This feature is disabled when
73           executing inside "DB::DB()", including functions called from it
74           unless "$^D & (1<<30)" is true.
75
76       ·   When execution of the program reaches a subroutine call, a call to
77           &DB::sub(args) is made instead, with $DB::sub set to identify the
78           called subroutine.  (This doesn't happen if the calling subroutine
79           was compiled in the "DB" package.)  $DB::sub normally holds the
80           name of the called subroutine, if it has a name by which it can be
81           looked up.  Failing that, $DB::sub will hold a reference to the
82           called subroutine.  Either way, the &DB::sub subroutine can use
83           $DB::sub as a reference by which to call the called subroutine,
84           which it will normally want to do.
85
86           If the call is to an lvalue subroutine, and &DB::lsub is defined
87           &DB::lsub(args) is called instead, otherwise falling back to
88           &DB::sub(args).
89
90       ·   When execution of the program uses "goto" to enter a non-XS
91           subroutine and the 0x80 bit is set in $^P, a call to &DB::goto is
92           made, with $DB::sub set to identify the subroutine being entered.
93           The call to &DB::goto does not replace the "goto"; the requested
94           subroutine will still be entered once &DB::goto has returned.
95           $DB::sub normally holds the name of the subroutine being entered,
96           if it has one.  Failing that, $DB::sub will hold a reference to the
97           subroutine being entered.  Unlike when &DB::sub is called, it is
98           not guaranteed that $DB::sub can be used as a reference to operate
99           on the subroutine being entered.
100
101       Note that if &DB::sub needs external data for it to work, no subroutine
102       call is possible without it. As an example, the standard debugger's
103       &DB::sub depends on the $DB::deep variable (it defines how many levels
104       of recursion deep into the debugger you can go before a mandatory
105       break).  If $DB::deep is not defined, subroutine calls are not
106       possible, even though &DB::sub exists.
107
108   Writing Your Own Debugger
109       Environment Variables
110
111       The "PERL5DB" environment variable can be used to define a debugger.
112       For example, the minimal "working" debugger (it actually doesn't do
113       anything) consists of one line:
114
115         sub DB::DB {}
116
117       It can easily be defined like this:
118
119         $ PERL5DB="sub DB::DB {}" perl -d your-script
120
121       Another brief debugger, slightly more useful, can be created with only
122       the line:
123
124         sub DB::DB {print ++$i; scalar <STDIN>}
125
126       This debugger prints a number which increments for each statement
127       encountered and waits for you to hit a newline before continuing to the
128       next statement.
129
130       The following debugger is actually useful:
131
132         {
133           package DB;
134           sub DB  {}
135           sub sub {print ++$i, " $sub\n"; &$sub}
136         }
137
138       It prints the sequence number of each subroutine call and the name of
139       the called subroutine.  Note that &DB::sub is being compiled into the
140       package "DB" through the use of the "package" directive.
141
142       When it starts, the debugger reads your rc file (./.perldb or ~/.perldb
143       under Unix), which can set important options.  (A subroutine
144       (&afterinit) can be defined here as well; it is executed after the
145       debugger completes its own initialization.)
146
147       After the rc file is read, the debugger reads the PERLDB_OPTS
148       environment variable and uses it to set debugger options. The contents
149       of this variable are treated as if they were the argument of an "o ..."
150       debugger command (q.v. in "Configurable Options" in perldebug).
151
152       Debugger Internal Variables
153
154       In addition to the file and subroutine-related variables mentioned
155       above, the debugger also maintains various magical internal variables.
156
157       ·   @DB::dbline is an alias for "@{"::_<current_file"}", which holds
158           the lines of the currently-selected file (compiled by Perl), either
159           explicitly chosen with the debugger's "f" command, or implicitly by
160           flow of execution.
161
162           Values in this array are magical in numeric context: they compare
163           equal to zero only if the line is not breakable.
164
165       ·   %DB::dbline is an alias for "%{"::_<current_file"}", which contains
166           breakpoints and actions keyed by line number in the currently-
167           selected file, either explicitly chosen with the debugger's "f"
168           command, or implicitly by flow of execution.
169
170           As previously noted, individual entries (as opposed to the whole
171           hash) are settable.  Perl only cares about Boolean true here,
172           although the values used by perl5db.pl have the form
173           "$break_condition\0$action".
174
175       Debugger Customization Functions
176
177       Some functions are provided to simplify customization.
178
179       ·   See "Configurable Options" in perldebug for a description of
180           options parsed by "DB::parse_options(string)".
181
182       ·   "DB::dump_trace(skip[,count])" skips the specified number of frames
183           and returns a list containing information about the calling frames
184           (all of them, if "count" is missing).  Each entry is reference to a
185           hash with keys "context" (either ".", "$", or "@"), "sub"
186           (subroutine name, or info about "eval"), "args" ("undef" or a
187           reference to an array), "file", and "line".
188
189       ·   "DB::print_trace(FH, skip[, count[, short]])" prints formatted info
190           about caller frames.  The last two functions may be convenient as
191           arguments to "<", "<<" commands.
192
193       Note that any variables and functions that are not documented in this
194       manpages (or in perldebug) are considered for internal use only, and as
195       such are subject to change without notice.
196

Frame Listing Output Examples

198       The "frame" option can be used to control the output of frame
199       information.  For example, contrast this expression trace:
200
201        $ perl -de 42
202        Stack dump during die enabled outside of evals.
203
204        Loading DB routines from perl5db.pl patch level 0.94
205        Emacs support available.
206
207        Enter h or 'h h' for help.
208
209        main::(-e:1):   0
210          DB<1> sub foo { 14 }
211
212          DB<2> sub bar { 3 }
213
214          DB<3> t print foo() * bar()
215        main::((eval 172):3):   print foo() + bar();
216        main::foo((eval 168):2):
217        main::bar((eval 170):2):
218        42
219
220       with this one, once the "o"ption "frame=2" has been set:
221
222          DB<4> o f=2
223                       frame = '2'
224          DB<5> t print foo() * bar()
225        3:      foo() * bar()
226        entering main::foo
227         2:     sub foo { 14 };
228        exited main::foo
229        entering main::bar
230         2:     sub bar { 3 };
231        exited main::bar
232        42
233
234       By way of demonstration, we present below a laborious listing resulting
235       from setting your "PERLDB_OPTS" environment variable to the value "f=n
236       N", and running perl -d -V from the command line.  Examples using
237       various values of "n" are shown to give you a feel for the difference
238       between settings.  Long though it may be, this is not a complete
239       listing, but only excerpts.
240
241       1.
242            entering main::BEGIN
243             entering Config::BEGIN
244              Package lib/Exporter.pm.
245              Package lib/Carp.pm.
246             Package lib/Config.pm.
247             entering Config::TIEHASH
248             entering Exporter::import
249              entering Exporter::export
250            entering Config::myconfig
251             entering Config::FETCH
252             entering Config::FETCH
253             entering Config::FETCH
254             entering Config::FETCH
255
256       2.
257            entering main::BEGIN
258             entering Config::BEGIN
259              Package lib/Exporter.pm.
260              Package lib/Carp.pm.
261             exited Config::BEGIN
262             Package lib/Config.pm.
263             entering Config::TIEHASH
264             exited Config::TIEHASH
265             entering Exporter::import
266              entering Exporter::export
267              exited Exporter::export
268             exited Exporter::import
269            exited main::BEGIN
270            entering Config::myconfig
271             entering Config::FETCH
272             exited Config::FETCH
273             entering Config::FETCH
274             exited Config::FETCH
275             entering Config::FETCH
276
277       3.
278            in  $=main::BEGIN() from /dev/null:0
279             in  $=Config::BEGIN() from lib/Config.pm:2
280              Package lib/Exporter.pm.
281              Package lib/Carp.pm.
282             Package lib/Config.pm.
283             in  $=Config::TIEHASH('Config') from lib/Config.pm:644
284             in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
285              in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li
286            in  @=Config::myconfig() from /dev/null:0
287             in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
288             in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
289             in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
290             in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
291             in  $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574
292             in  $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574
293
294       4.
295            in  $=main::BEGIN() from /dev/null:0
296             in  $=Config::BEGIN() from lib/Config.pm:2
297              Package lib/Exporter.pm.
298              Package lib/Carp.pm.
299             out $=Config::BEGIN() from lib/Config.pm:0
300             Package lib/Config.pm.
301             in  $=Config::TIEHASH('Config') from lib/Config.pm:644
302             out $=Config::TIEHASH('Config') from lib/Config.pm:644
303             in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
304              in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
305              out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/
306             out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
307            out $=main::BEGIN() from /dev/null:0
308            in  @=Config::myconfig() from /dev/null:0
309             in  $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
310             out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574
311             in  $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
312             out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574
313             in  $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
314             out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574
315             in  $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574
316
317       5.
318            in  $=main::BEGIN() from /dev/null:0
319             in  $=Config::BEGIN() from lib/Config.pm:2
320              Package lib/Exporter.pm.
321              Package lib/Carp.pm.
322             out $=Config::BEGIN() from lib/Config.pm:0
323             Package lib/Config.pm.
324             in  $=Config::TIEHASH('Config') from lib/Config.pm:644
325             out $=Config::TIEHASH('Config') from lib/Config.pm:644
326             in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
327              in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
328              out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E
329             out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
330            out $=main::BEGIN() from /dev/null:0
331            in  @=Config::myconfig() from /dev/null:0
332             in  $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
333             out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574
334             in  $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
335             out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574
336
337       6.
338            in  $=CODE(0x15eca4)() from /dev/null:0
339             in  $=CODE(0x182528)() from lib/Config.pm:2
340              Package lib/Exporter.pm.
341             out $=CODE(0x182528)() from lib/Config.pm:0
342             scalar context return from CODE(0x182528): undef
343             Package lib/Config.pm.
344             in  $=Config::TIEHASH('Config') from lib/Config.pm:628
345             out $=Config::TIEHASH('Config') from lib/Config.pm:628
346             scalar context return from Config::TIEHASH:   empty hash
347             in  $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
348              in  $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
349              out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171
350              scalar context return from Exporter::export: ''
351             out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0
352             scalar context return from Exporter::import: ''
353
354       In all cases shown above, the line indentation shows the call tree.  If
355       bit 2 of "frame" is set, a line is printed on exit from a subroutine as
356       well.  If bit 4 is set, the arguments are printed along with the caller
357       info.  If bit 8 is set, the arguments are printed even if they are tied
358       or references.  If bit 16 is set, the return value is printed, too.
359
360       When a package is compiled, a line like this
361
362           Package lib/Carp.pm.
363
364       is printed with proper indentation.
365

Debugging Regular Expressions

367       There are two ways to enable debugging output for regular expressions.
368
369       If your perl is compiled with "-DDEBUGGING", you may use the -Dr flag
370       on the command line, and "-Drv" for more verbose information.
371
372       Otherwise, one can "use re 'debug'", which has effects at both compile
373       time and run time.  Since Perl 5.9.5, this pragma is lexically scoped.
374
375   Compile-time Output
376       The debugging output at compile time looks like this:
377
378         Compiling REx '[bc]d(ef*g)+h[ij]k$'
379         size 45 Got 364 bytes for offset annotations.
380         first at 1
381         rarest char g at 0
382         rarest char d at 0
383            1: ANYOF[bc](12)
384           12: EXACT <d>(14)
385           14: CURLYX[0] {1,32767}(28)
386           16:   OPEN1(18)
387           18:     EXACT <e>(20)
388           20:     STAR(23)
389           21:       EXACT <f>(0)
390           23:     EXACT <g>(25)
391           25:   CLOSE1(27)
392           27:   WHILEM[1/1](0)
393           28: NOTHING(29)
394           29: EXACT <h>(31)
395           31: ANYOF[ij](42)
396           42: EXACT <k>(44)
397           44: EOL(45)
398           45: END(0)
399         anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
400               stclass 'ANYOF[bc]' minlen 7
401         Offsets: [45]
402               1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
403               0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
404               11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
405               0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
406         Omitting $` $& $' support.
407
408       The first line shows the pre-compiled form of the regex.  The second
409       shows the size of the compiled form (in arbitrary units, usually 4-byte
410       words) and the total number of bytes allocated for the offset/length
411       table, usually 4+"size"*8.  The next line shows the label id of the
412       first node that does a match.
413
414       The
415
416         anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating)
417               stclass 'ANYOF[bc]' minlen 7
418
419       line (split into two lines above) contains optimizer information.  In
420       the example shown, the optimizer found that the match should contain a
421       substring "de" at offset 1, plus substring "gh" at some offset between
422       3 and infinity.  Moreover, when checking for these substrings (to
423       abandon impossible matches quickly), Perl will check for the substring
424       "gh" before checking for the substring "de".  The optimizer may also
425       use the knowledge that the match starts (at the "first" id) with a
426       character class, and no string shorter than 7 characters can possibly
427       match.
428
429       The fields of interest which may appear in this line are
430
431       "anchored" STRING "at" POS
432       "floating" STRING "at" POS1..POS2
433           See above.
434
435       "matching floating/anchored"
436           Which substring to check first.
437
438       "minlen"
439           The minimal length of the match.
440
441       "stclass" TYPE
442           Type of first matching node.
443
444       "noscan"
445           Don't scan for the found substrings.
446
447       "isall"
448           Means that the optimizer information is all that the regular
449           expression contains, and thus one does not need to enter the regex
450           engine at all.
451
452       "GPOS"
453           Set if the pattern contains "\G".
454
455       "plus"
456           Set if the pattern starts with a repeated char (as in "x+y").
457
458       "implicit"
459           Set if the pattern starts with ".*".
460
461       "with eval"
462           Set if the pattern contain eval-groups, such as "(?{ code })" and
463           "(??{ code })".
464
465       "anchored(TYPE)"
466           If the pattern may match only at a handful of places, with "TYPE"
467           being "SBOL", "MBOL", or "GPOS".  See the table below.
468
469       If a substring is known to match at end-of-line only, it may be
470       followed by "$", as in "floating 'k'$".
471
472       The optimizer-specific information is used to avoid entering (a slow)
473       regex engine on strings that will not definitely match.  If the "isall"
474       flag is set, a call to the regex engine may be avoided even when the
475       optimizer found an appropriate place for the match.
476
477       Above the optimizer section is the list of nodes of the compiled form
478       of the regex.  Each line has format
479
480       "   "id: TYPE OPTIONAL-INFO (next-id)
481
482   Types of Nodes
483       Here are the current possible types, with short descriptions:
484
485        # TYPE arg-description [num-args] [longjump-len] DESCRIPTION
486
487        # Exit points
488
489        END              no         End of program.
490        SUCCEED          no         Return from a subroutine, basically.
491
492        # Line Start Anchors:
493        SBOL             no         Match "" at beginning of line: /^/, /\A/
494        MBOL             no         Same, assuming multiline: /^/m
495
496        # Line End Anchors:
497        SEOL             no         Match "" at end of line: /$/
498        MEOL             no         Same, assuming multiline: /$/m
499        EOS              no         Match "" at end of string: /\z/
500
501        # Match Start Anchors:
502        GPOS             no         Matches where last m//g left off.
503
504        # Word Boundary Opcodes:
505        BOUND            no         Like BOUNDA for non-utf8, otherwise match
506                                    "" between any Unicode \w\W or \W\w
507        BOUNDL           no         Like BOUND/BOUNDU, but \w and \W are
508                                    defined by current locale
509        BOUNDU           no         Match "" at any boundary of a given type
510                                    using /u rules.
511        BOUNDA           no         Match "" at any boundary between \w\W or
512                                    \W\w, where \w is [_a-zA-Z0-9]
513        NBOUND           no         Like NBOUNDA for non-utf8, otherwise match
514                                    "" between any Unicode \w\w or \W\W
515        NBOUNDL          no         Like NBOUND/NBOUNDU, but \w and \W are
516                                    defined by current locale
517        NBOUNDU          no         Match "" at any non-boundary of a given
518                                    type using using /u rules.
519        NBOUNDA          no         Match "" betweeen any \w\w or \W\W, where
520                                    \w is [_a-zA-Z0-9]
521
522        # [Special] alternatives:
523        REG_ANY          no         Match any one character (except newline).
524        SANY             no         Match any one character.
525        ANYOF            sv         Match character in (or not in) this class,
526                         charclass  single char match only
527        ANYOFD           sv         Like ANYOF, but /d is in effect
528                         charclass
529        ANYOFL           sv         Like ANYOF, but /l is in effect
530                         charclass
531        ANYOFPOSIXL      sv         Like ANYOFL, but matches [[:posix:]]
532                         charclass_ classes
533                         posixl
534        ANYOFH           sv 1       Like ANYOF, but only has "High" matches,
535                                    none in the bitmap; non-zero flags "f"
536                                    means "f" is the first UTF-8 byte shared in
537                                    common by all code points matched
538        ANYOFM           byte 1     Like ANYOF, but matches an invariant byte
539                                    as determined by the mask and arg
540        NANYOFM          byte 1     complement of ANYOFM
541
542        # POSIX Character Classes:
543        POSIXD           none       Some [[:class:]] under /d; the FLAGS field
544                                    gives which one
545        POSIXL           none       Some [[:class:]] under /l; the FLAGS field
546                                    gives which one
547        POSIXU           none       Some [[:class:]] under /u; the FLAGS field
548                                    gives which one
549        POSIXA           none       Some [[:class:]] under /a; the FLAGS field
550                                    gives which one
551        NPOSIXD          none       complement of POSIXD, [[:^class:]]
552        NPOSIXL          none       complement of POSIXL, [[:^class:]]
553        NPOSIXU          none       complement of POSIXU, [[:^class:]]
554        NPOSIXA          none       complement of POSIXA, [[:^class:]]
555
556        CLUMP            no         Match any extended grapheme cluster
557                                    sequence
558
559        # Alternation
560
561        # BRANCH        The set of branches constituting a single choice are
562        #               hooked together with their "next" pointers, since
563        #               precedence prevents anything being concatenated to
564        #               any individual branch.  The "next" pointer of the last
565        #               BRANCH in a choice points to the thing following the
566        #               whole choice.  This is also where the final "next"
567        #               pointer of each individual branch points; each branch
568        #               starts with the operand node of a BRANCH node.
569        #
570        BRANCH           node       Match this alternative, or the next...
571
572        # Literals
573
574        EXACT            str        Match this string (preceded by length).
575        EXACTL           str        Like EXACT, but /l is in effect (used so
576                                    locale-related warnings can be checked
577                                    for).
578        EXACTF           str        Match this string using /id rules (w/len);
579                                    (string not UTF-8, not guaranteed to be
580                                    folded).
581        EXACTFL          str        Match this string using /il rules (w/len);
582                                    (string not guaranteed to be folded).
583        EXACTFU          str        Match this string using /iu rules (w/len);
584                                    (string folded iff in UTF-8; non-UTF8
585                                    folded length <= unfolded).
586        EXACTFAA         str        Match this string using /iaa rules (w/len)
587                                    (string folded iff in UTF-8; non-UTF8
588                                    folded length <= unfolded).
589
590        EXACTFUP         str        Match this string using /iu rules (w/len);
591                                    (string not UTF-8, not guaranteed to be
592                                    folded; and its Problematic).
593
594        EXACTFLU8        str        Like EXACTFU, but use /il, UTF-8, folded,
595                                    and everything in it is above 255.
596        EXACTFAA_NO_TRIE str        Match this string using /iaa rules (w/len)
597                                    (string not UTF-8, not guaranteed to be
598                                    folded, not currently trie-able).
599
600        EXACT_ONLY8      str        Like EXACT, but only UTF-8 encoded targets
601                                    can match
602        EXACTFU_ONLY8    str        Like EXACTFU, but only UTF-8 encoded
603                                    targets can match
604
605        EXACTFU_S_EDGE   str        /di rules, but nothing in it precludes /ui,
606                                    except begins and/or ends with [Ss];
607                                    (string not UTF-8; compile-time only).
608
609        # Do nothing types
610
611        NOTHING          no         Match empty string.
612        # A variant of above which delimits a group, thus stops optimizations
613        TAIL             no         Match empty string. Can jump here from
614                                    outside.
615
616        # Loops
617
618        # STAR,PLUS    '?', and complex '*' and '+', are implemented as
619        #               circular BRANCH structures.  Simple cases
620        #               (one character per match) are implemented with STAR
621        #               and PLUS for speed and to minimize recursive plunges.
622        #
623        STAR             node       Match this (simple) thing 0 or more times.
624        PLUS             node       Match this (simple) thing 1 or more times.
625
626        CURLY            sv 2       Match this simple thing {n,m} times.
627        CURLYN           no 2       Capture next-after-this simple thing
628        CURLYM           no 2       Capture this medium-complex thing {n,m}
629                                    times.
630        CURLYX           sv 2       Match this complex thing {n,m} times.
631
632        # This terminator creates a loop structure for CURLYX
633        WHILEM           no         Do curly processing and see if rest
634                                    matches.
635
636        # Buffer related
637
638        # OPEN,CLOSE,GROUPP     ...are numbered at compile time.
639        OPEN             num 1      Mark this point in input as start of #n.
640        CLOSE            num 1      Close corresponding OPEN of #n.
641        SROPEN           none       Same as OPEN, but for script run
642        SRCLOSE          none       Close preceding SROPEN
643
644        REF              num 1      Match some already matched string
645        REFF             num 1      Match already matched string, using /di
646                                    rules.
647        REFFL            num 1      Match already matched string, using /li
648                                    rules.
649        REFFU            num 1      Match already matched string, usng /ui.
650        REFFA            num 1      Match already matched string, using /aai
651                                    rules.
652
653        # Named references.  Code in regcomp.c assumes that these all are after
654        # the numbered references
655        NREF             no-sv 1    Match some already matched string
656        NREFF            no-sv 1    Match already matched string, using /di
657                                    rules.
658        NREFFL           no-sv 1    Match already matched string, using /li
659                                    rules.
660        NREFFU           num 1      Match already matched string, using /ui
661                                    rules.
662        NREFFA           num 1      Match already matched string, using /aai
663                                    rules.
664
665        # Support for long RE
666        LONGJMP          off 1 1    Jump far away.
667        BRANCHJ          off 1 1    BRANCH with long offset.
668
669        # Special Case Regops
670        IFMATCH          off 1 1    Succeeds if the following matches; non-zero
671                                    flags "f", next_off "o" means lookbehind
672                                    assertion starting "f..(f-o)" characters
673                                    before current
674        UNLESSM          off 1 1    Fails if the following matches; non-zero
675                                    flags "f", next_off "o" means lookbehind
676                                    assertion starting "f..(f-o)" characters
677                                    before current
678        SUSPEND          off 1 1    "Independent" sub-RE.
679        IFTHEN           off 1 1    Switch, should be preceded by switcher.
680        GROUPP           num 1      Whether the group matched.
681
682        # The heavy worker
683
684        EVAL             evl/flags  Execute some Perl code.
685                         2L
686
687        # Modifiers
688
689        MINMOD           no         Next operator is not greedy.
690        LOGICAL          no         Next opcode should set the flag only.
691
692        # This is not used yet
693        RENUM            off 1 1    Group with independently numbered parens.
694
695        # Trie Related
696
697        # Behave the same as A|LIST|OF|WORDS would. The '..C' variants
698        # have inline charclass data (ascii only), the 'C' store it in the
699        # structure.
700
701        TRIE             trie 1     Match many EXACT(F[ALU]?)? at once.
702                                    flags==type
703        TRIEC            trie       Same as TRIE, but with embedded charclass
704                         charclass  data
705
706        AHOCORASICK      trie 1     Aho Corasick stclass. flags==type
707        AHOCORASICKC     trie       Same as AHOCORASICK, but with embedded
708                         charclass  charclass data
709
710        # Regex Subroutines
711        GOSUB            num/ofs 2L recurse to paren arg1 at (signed) ofs arg2
712
713        # Special conditionals
714        NGROUPP          no-sv 1    Whether the group matched.
715        INSUBP           num 1      Whether we are in a specific recurse.
716        DEFINEP          none 1     Never execute directly.
717
718        # Backtracking Verbs
719        ENDLIKE          none       Used only for the type field of verbs
720        OPFAIL           no-sv 1    Same as (?!), but with verb arg
721        ACCEPT           no-sv/num  Accepts the current matched string, with
722                         2L         verbar
723
724        # Verbs With Arguments
725        VERB             no-sv 1    Used only for the type field of verbs
726        PRUNE            no-sv 1    Pattern fails at this startpoint if no-
727                                    backtracking through this
728        MARKPOINT        no-sv 1    Push the current location for rollback by
729                                    cut.
730        SKIP             no-sv 1    On failure skip forward (to the mark)
731                                    before retrying
732        COMMIT           no-sv 1    Pattern fails outright if backtracking
733                                    through this
734        CUTGROUP         no-sv 1    On failure go to the next alternation in
735                                    the group
736
737        # Control what to keep in $&.
738        KEEPS            no         $& begins here.
739
740        # New charclass like patterns
741        LNBREAK          none       generic newline pattern
742
743        # SPECIAL  REGOPS
744
745        # This is not really a node, but an optimized away piece of a "long"
746        # node.  To simplify debugging output, we mark it as if it were a node
747        OPTIMIZED        off        Placeholder for dump.
748
749        # Special opcode with the property that no opcode in a compiled program
750        # will ever be of this type. Thus it can be used as a flag value that
751        # no other opcode has been seen. END is used similarly, in that an END
752        # node cant be optimized. So END implies "unoptimizable" and PSEUDO
753        # mean "not seen anything to optimize yet".
754        PSEUDO           off        Pseudo opcode for internal use.
755
756       Following the optimizer information is a dump of the offset/length
757       table, here split across several lines:
758
759         Offsets: [45]
760               1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1]
761               0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0]
762               11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0]
763               0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0]
764
765       The first line here indicates that the offset/length table contains 45
766       entries.  Each entry is a pair of integers, denoted by
767       "offset[length]".  Entries are numbered starting with 1, so entry #1
768       here is "1[4]" and entry #12 is "5[1]".  "1[4]" indicates that the node
769       labeled "1:" (the "1: ANYOF[bc]") begins at character position 1 in the
770       pre-compiled form of the regex, and has a length of 4 characters.
771       "5[1]" in position 12 indicates that the node labeled "12:" (the "12:
772       EXACT <d>") begins at character position 5 in the pre-compiled form of
773       the regex, and has a length of 1 character.  "12[1]" in position 14
774       indicates that the node labeled "14:" (the "14: CURLYX[0] {1,32767}")
775       begins at character position 12 in the pre-compiled form of the regex,
776       and has a length of 1 character---that is, it corresponds to the "+"
777       symbol in the precompiled regex.
778
779       "0[0]" items indicate that there is no corresponding node.
780
781   Run-time Output
782       First of all, when doing a match, one may get no run-time output even
783       if debugging is enabled.  This means that the regex engine was never
784       entered and that all of the job was therefore done by the optimizer.
785
786       If the regex engine was entered, the output may look like this:
787
788         Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__'
789           Setting an EVAL scope, savestack=3
790            2 <ab> <cdefg__gh_>    |  1: ANYOF
791            3 <abc> <defg__gh_>    | 11: EXACT <d>
792            4 <abcd> <efg__gh_>    | 13: CURLYX {1,32767}
793            4 <abcd> <efg__gh_>    | 26:   WHILEM
794                                       0 out of 1..32767  cc=effff31c
795            4 <abcd> <efg__gh_>    | 15:     OPEN1
796            4 <abcd> <efg__gh_>    | 17:     EXACT <e>
797            5 <abcde> <fg__gh_>    | 19:     STAR
798                                    EXACT <f> can match 1 times out of 32767...
799           Setting an EVAL scope, savestack=3
800            6 <bcdef> <g__gh__>    | 22:       EXACT <g>
801            7 <bcdefg> <__gh__>    | 24:       CLOSE1
802            7 <bcdefg> <__gh__>    | 26:       WHILEM
803                                           1 out of 1..32767  cc=effff31c
804           Setting an EVAL scope, savestack=12
805            7 <bcdefg> <__gh__>    | 15:         OPEN1
806            7 <bcdefg> <__gh__>    | 17:         EXACT <e>
807              restoring \1 to 4(4)..7
808                                           failed, try continuation...
809            7 <bcdefg> <__gh__>    | 27:         NOTHING
810            7 <bcdefg> <__gh__>    | 28:         EXACT <h>
811                                           failed...
812                                       failed...
813
814       The most significant information in the output is about the particular
815       node of the compiled regex that is currently being tested against the
816       target string.  The format of these lines is
817
818       "    "STRING-OFFSET <PRE-STRING> <POST-STRING>   |ID:  TYPE
819
820       The TYPE info is indented with respect to the backtracking level.
821       Other incidental information appears interspersed within.
822

Debugging Perl Memory Usage

824       Perl is a profligate wastrel when it comes to memory use.  There is a
825       saying that to estimate memory usage of Perl, assume a reasonable
826       algorithm for memory allocation, multiply that estimate by 10, and
827       while you still may miss the mark, at least you won't be quite so
828       astonished.  This is not absolutely true, but may provide a good grasp
829       of what happens.
830
831       Assume that an integer cannot take less than 20 bytes of memory, a
832       float cannot take less than 24 bytes, a string cannot take less than 32
833       bytes (all these examples assume 32-bit architectures, the result are
834       quite a bit worse on 64-bit architectures).  If a variable is accessed
835       in two of three different ways (which require an integer, a float, or a
836       string), the memory footprint may increase yet another 20 bytes.  A
837       sloppy malloc(3) implementation can inflate these numbers dramatically.
838
839       On the opposite end of the scale, a declaration like
840
841         sub foo;
842
843       may take up to 500 bytes of memory, depending on which release of Perl
844       you're running.
845
846       Anecdotal estimates of source-to-compiled code bloat suggest an
847       eightfold increase.  This means that the compiled form of reasonable
848       (normally commented, properly indented etc.) code will take about eight
849       times more space in memory than the code took on disk.
850
851       The -DL command-line switch is obsolete since circa Perl 5.6.0 (it was
852       available only if Perl was built with "-DDEBUGGING").  The switch was
853       used to track Perl's memory allocations and possible memory leaks.
854       These days the use of malloc debugging tools like Purify or valgrind is
855       suggested instead.  See also "PERL_MEM_LOG" in perlhacktips.
856
857       One way to find out how much memory is being used by Perl data
858       structures is to install the Devel::Size module from CPAN: it gives you
859       the minimum number of bytes required to store a particular data
860       structure.  Please be mindful of the difference between the size() and
861       total_size().
862
863       If Perl has been compiled using Perl's malloc you can analyze Perl
864       memory usage by setting $ENV{PERL_DEBUG_MSTATS}.
865
866   Using $ENV{PERL_DEBUG_MSTATS}
867       If your perl is using Perl's malloc() and was compiled with the
868       necessary switches (this is the default), then it will print memory
869       usage statistics after compiling your code when
870       "$ENV{PERL_DEBUG_MSTATS} > 1", and before termination of the program
871       when "$ENV{PERL_DEBUG_MSTATS} >= 1".  The report format is similar to
872       the following example:
873
874        $ PERL_DEBUG_MSTATS=2 perl -e "require Carp"
875        Memory allocation statistics after compilation: (buckets 4(4)..8188(8192)
876           14216 free:   130   117    28     7     9   0   2     2   1 0 0
877                       437    61    36     0     5
878           60924 used:   125   137   161    55     7   8   6    16   2 0 1
879                        74   109   304    84    20
880        Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048.
881        Memory allocation statistics after execution:   (buckets 4(4)..8188(8192)
882           30888 free:   245    78    85    13     6   2   1     3   2 0 1
883                       315   162    39    42    11
884          175816 used:   265   176  1112   111    26  22  11    27   2 1 1
885                       196   178  1066   798    39
886        Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144.
887
888       It is possible to ask for such a statistic at arbitrary points in your
889       execution using the mstat() function out of the standard Devel::Peek
890       module.
891
892       Here is some explanation of that format:
893
894       "buckets SMALLEST(APPROX)..GREATEST(APPROX)"
895           Perl's malloc() uses bucketed allocations.  Every request is
896           rounded up to the closest bucket size available, and a bucket is
897           taken from the pool of buckets of that size.
898
899           The line above describes the limits of buckets currently in use.
900           Each bucket has two sizes: memory footprint and the maximal size of
901           user data that can fit into this bucket.  Suppose in the above
902           example that the smallest bucket were size 4.  The biggest bucket
903           would have usable size 8188, and the memory footprint would be
904           8192.
905
906           In a Perl built for debugging, some buckets may have negative
907           usable size.  This means that these buckets cannot (and will not)
908           be used.  For larger buckets, the memory footprint may be one page
909           greater than a power of 2.  If so, the corresponding power of two
910           is printed in the "APPROX" field above.
911
912       Free/Used
913           The 1 or 2 rows of numbers following that correspond to the number
914           of buckets of each size between "SMALLEST" and "GREATEST".  In the
915           first row, the sizes (memory footprints) of buckets are powers of
916           two--or possibly one page greater.  In the second row, if present,
917           the memory footprints of the buckets are between the memory
918           footprints of two buckets "above".
919
920           For example, suppose under the previous example, the memory
921           footprints were
922
923              free:    8     16    32    64    128  256 512 1024 2048 4096 8192
924                      4     12    24    48    80
925
926           With a non-"DEBUGGING" perl, the buckets starting from 128 have a
927           4-byte overhead, and thus an 8192-long bucket may take up to
928           8188-byte allocations.
929
930       "Total sbrk(): SBRKed/SBRKs:CONTINUOUS"
931           The first two fields give the total amount of memory perl sbrk(2)ed
932           (ess-broken? :-) and number of sbrk(2)s used.  The third number is
933           what perl thinks about continuity of returned chunks.  So long as
934           this number is positive, malloc() will assume that it is probable
935           that sbrk(2) will provide continuous memory.
936
937           Memory allocated by external libraries is not counted.
938
939       "pad: 0"
940           The amount of sbrk(2)ed memory needed to keep buckets aligned.
941
942       "heads: 2192"
943           Although memory overhead of bigger buckets is kept inside the
944           bucket, for smaller buckets, it is kept in separate areas.  This
945           field gives the total size of these areas.
946
947       "chain: 0"
948           malloc() may want to subdivide a bigger bucket into smaller
949           buckets.  If only a part of the deceased bucket is left
950           unsubdivided, the rest is kept as an element of a linked list.
951           This field gives the total size of these chunks.
952
953       "tail: 6144"
954           To minimize the number of sbrk(2)s, malloc() asks for more memory.
955           This field gives the size of the yet unused part, which is
956           sbrk(2)ed, but never touched.
957

SEE ALSO

959       perldebug, perlguts, perlrun re, and Devel::DProf.
960
961
962
963perl v5.30.2                      2020-03-27                    PERLDEBGUTS(1)
Impressum