Commit | Line | Data |
---|---|---|
055fd3a9 GS |
1 | =head1 NAME |
2 | ||
3 | perldebguts - Guts of Perl debugging | |
4 | ||
5 | =head1 DESCRIPTION | |
6 | ||
ba555bf5 | 7 | This is not L<perldebug>, which tells you how to use |
74410c12 JM |
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. | |
11 | Caveat lector. | |
055fd3a9 GS |
12 | |
13 | =head1 Debugger Internals | |
14 | ||
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 | |
4375e838 GS |
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 | |
055fd3a9 GS |
19 | F<INSTALL> podpage in the Perl source tree. |
20 | ||
21 | For example, whenever you call Perl's built-in C<caller> function | |
74410c12 JM |
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 | |
26 | (cf. L<perlvar/$^P>): | |
055fd3a9 | 27 | |
13a2d996 | 28 | =over 4 |
055fd3a9 GS |
29 | |
30 | =item * | |
31 | ||
32 | Perl inserts the contents of C<$ENV{PERL5DB}> (or C<BEGIN {require | |
33 | 'perl5db.pl'}> if not present) before the first line of your program. | |
34 | ||
35 | =item * | |
36 | ||
aa0b556f | 37 | Each array C<@{"_<$filename"}> holds the lines of $filename for a |
74410c12 JM |
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 $filename for C<eval>ed strings looks like C<(eval 34)>. | |
8894c26d MJD |
41 | |
42 | Values in this array are magical in numeric context: they compare | |
43 | equal to zero only if the line is not breakable. | |
055fd3a9 GS |
44 | |
45 | =item * | |
46 | ||
aa0b556f | 47 | Each hash C<%{"_<$filename"}> contains breakpoints and actions keyed |
055fd3a9 GS |
48 | by line number. Individual entries (as opposed to the whole hash) |
49 | are settable. Perl only cares about Boolean true here, although | |
50 | the values used by F<perl5db.pl> have the form | |
8894c26d | 51 | C<"$break_condition\0$action">. |
055fd3a9 GS |
52 | |
53 | The same holds for evaluated strings that contain subroutines, or | |
54 | which are currently being executed. The $filename for C<eval>ed strings | |
d24ca0c5 | 55 | looks like C<(eval 34)>. |
055fd3a9 GS |
56 | |
57 | =item * | |
58 | ||
aa0b556f | 59 | Each scalar C<${"_<$filename"}> contains C<"_<$filename">. This is |
055fd3a9 GS |
60 | also the case for evaluated strings that contain subroutines, or |
61 | which are currently being executed. The $filename for C<eval>ed | |
d24ca0c5 | 62 | strings looks like C<(eval 34)>. |
055fd3a9 GS |
63 | |
64 | =item * | |
65 | ||
66 | After each C<require>d file is compiled, but before it is executed, | |
67 | C<DB::postponed(*{"_<$filename"})> is called if the subroutine | |
68 | C<DB::postponed> exists. Here, the $filename is the expanded name of | |
69 | the C<require>d file, as found in the values of %INC. | |
70 | ||
71 | =item * | |
72 | ||
73 | After each subroutine C<subname> is compiled, the existence of | |
74 | C<$DB::postponed{subname}> is checked. If this key exists, | |
75 | C<DB::postponed(subname)> is called if the C<DB::postponed> subroutine | |
76 | also exists. | |
77 | ||
78 | =item * | |
79 | ||
80 | A hash C<%DB::sub> is maintained, whose keys are subroutine names | |
81 | and whose values have the form C<filename:startline-endline>. | |
82 | C<filename> has the form C<(eval 34)> for subroutines defined inside | |
d24ca0c5 | 83 | C<eval>s. |
055fd3a9 GS |
84 | |
85 | =item * | |
86 | ||
87 | When the execution of your program reaches a point that can hold a | |
74410c12 JM |
88 | breakpoint, the C<DB::DB()> subroutine is called if any of the variables |
89 | C<$DB::trace>, C<$DB::single>, or C<$DB::signal> is true. These variables | |
055fd3a9 GS |
90 | are not C<local>izable. This feature is disabled when executing |
91 | inside C<DB::DB()>, including functions called from it | |
92 | unless C<< $^D & (1<<30) >> is true. | |
93 | ||
94 | =item * | |
95 | ||
96 | When execution of the program reaches a subroutine call, a call to | |
97 | C<&DB::sub>(I<args>) is made instead, with C<$DB::sub> holding the | |
74410c12 | 98 | name of the called subroutine. (This doesn't happen if the subroutine |
055fd3a9 GS |
99 | was compiled in the C<DB> package.) |
100 | ||
77e42cd2 TC |
101 | X<&DB::lsub>If the call is to an lvalue subroutine, and C<&DB::lsub> |
102 | is defined C<&DB::lsub>(I<args>) is called instead, otherwise falling | |
103 | back to C<&DB::sub>(I<args>). | |
104 | ||
261cbad1 TC |
105 | =item * |
106 | ||
107 | When execution of the program uses C<goto> to enter a non-XS | |
108 | subroutine and the 0x80 bit is set in C<$^P>, a call to C<&DB::goto> | |
109 | is made, with C<$DB::sub> holding the name of the subroutine being | |
110 | entered. | |
111 | ||
055fd3a9 GS |
112 | =back |
113 | ||
114 | Note that if C<&DB::sub> needs external data for it to work, no | |
74410c12 JM |
115 | subroutine call is possible without it. As an example, the standard |
116 | debugger's C<&DB::sub> depends on the C<$DB::deep> variable | |
117 | (it defines how many levels of recursion deep into the debugger you can go | |
118 | before a mandatory break). If C<$DB::deep> is not defined, subroutine | |
119 | calls are not possible, even though C<&DB::sub> exists. | |
055fd3a9 GS |
120 | |
121 | =head2 Writing Your Own Debugger | |
122 | ||
74410c12 | 123 | =head3 Environment Variables |
666f95b9 | 124 | |
74410c12 JM |
125 | The C<PERL5DB> environment variable can be used to define a debugger. |
126 | For example, the minimal "working" debugger (it actually doesn't do anything) | |
127 | consists of one line: | |
666f95b9 | 128 | |
055fd3a9 GS |
129 | sub DB::DB {} |
130 | ||
74410c12 | 131 | It can easily be defined like this: |
666f95b9 | 132 | |
055fd3a9 GS |
133 | $ PERL5DB="sub DB::DB {}" perl -d your-script |
134 | ||
74410c12 | 135 | Another brief debugger, slightly more useful, can be created |
055fd3a9 GS |
136 | with only the line: |
137 | ||
138 | sub DB::DB {print ++$i; scalar <STDIN>} | |
139 | ||
74410c12 JM |
140 | This debugger prints a number which increments for each statement |
141 | encountered and waits for you to hit a newline before continuing | |
142 | to the next statement. | |
666f95b9 | 143 | |
74410c12 | 144 | The following debugger is actually useful: |
666f95b9 | 145 | |
055fd3a9 GS |
146 | { |
147 | package DB; | |
148 | sub DB {} | |
149 | sub sub {print ++$i, " $sub\n"; &$sub} | |
150 | } | |
151 | ||
74410c12 JM |
152 | It prints the sequence number of each subroutine call and the name of the |
153 | called subroutine. Note that C<&DB::sub> is being compiled into the | |
154 | package C<DB> through the use of the C<package> directive. | |
055fd3a9 | 155 | |
74410c12 JM |
156 | When it starts, the debugger reads your rc file (F<./.perldb> or |
157 | F<~/.perldb> under Unix), which can set important options. | |
158 | (A subroutine (C<&afterinit>) can be defined here as well; it is executed | |
159 | after the debugger completes its own initialization.) | |
055fd3a9 GS |
160 | |
161 | After the rc file is read, the debugger reads the PERLDB_OPTS | |
74410c12 JM |
162 | environment variable and uses it to set debugger options. The |
163 | contents of this variable are treated as if they were the argument | |
96090e4f | 164 | of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">). |
74410c12 | 165 | |
7b406369 | 166 | =head3 Debugger Internal Variables |
25cf7dea | 167 | |
74410c12 JM |
168 | In addition to the file and subroutine-related variables mentioned above, |
169 | the debugger also maintains various magical internal variables. | |
170 | ||
171 | =over 4 | |
172 | ||
173 | =item * | |
055fd3a9 | 174 | |
74410c12 JM |
175 | C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which |
176 | holds the lines of the currently-selected file (compiled by Perl), either | |
177 | explicitly chosen with the debugger's C<f> command, or implicitly by flow | |
178 | of execution. | |
179 | ||
180 | Values in this array are magical in numeric context: they compare | |
181 | equal to zero only if the line is not breakable. | |
182 | ||
183 | =item * | |
184 | ||
7b406369 | 185 | C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which |
74410c12 JM |
186 | contains breakpoints and actions keyed by line number in |
187 | the currently-selected file, either explicitly chosen with the | |
055fd3a9 GS |
188 | debugger's C<f> command, or implicitly by flow of execution. |
189 | ||
74410c12 JM |
190 | As previously noted, individual entries (as opposed to the whole hash) |
191 | are settable. Perl only cares about Boolean true here, although | |
192 | the values used by F<perl5db.pl> have the form | |
193 | C<"$break_condition\0$action">. | |
194 | ||
195 | =back | |
196 | ||
7b406369 | 197 | =head3 Debugger Customization Functions |
74410c12 JM |
198 | |
199 | Some functions are provided to simplify customization. | |
200 | ||
201 | =over 4 | |
202 | ||
203 | =item * | |
204 | ||
71110851 RGS |
205 | See L<perldebug/"Configurable Options"> for a description of options parsed by |
206 | C<DB::parse_options(string)>. | |
74410c12 JM |
207 | |
208 | =item * | |
209 | ||
210 | C<DB::dump_trace(skip[,count])> skips the specified number of frames | |
211 | and returns a list containing information about the calling frames (all | |
212 | of them, if C<count> is missing). Each entry is reference to a hash | |
213 | with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine | |
055fd3a9 GS |
214 | name, or info about C<eval>), C<args> (C<undef> or a reference to |
215 | an array), C<file>, and C<line>. | |
216 | ||
74410c12 JM |
217 | =item * |
218 | ||
219 | C<DB::print_trace(FH, skip[, count[, short]])> prints | |
055fd3a9 GS |
220 | formatted info about caller frames. The last two functions may be |
221 | convenient as arguments to C<< < >>, C<< << >> commands. | |
222 | ||
74410c12 JM |
223 | =back |
224 | ||
055fd3a9 GS |
225 | Note that any variables and functions that are not documented in |
226 | this manpages (or in L<perldebug>) are considered for internal | |
227 | use only, and as such are subject to change without notice. | |
228 | ||
229 | =head1 Frame Listing Output Examples | |
230 | ||
231 | The C<frame> option can be used to control the output of frame | |
232 | information. For example, contrast this expression trace: | |
233 | ||
234 | $ perl -de 42 | |
235 | Stack dump during die enabled outside of evals. | |
236 | ||
237 | Loading DB routines from perl5db.pl patch level 0.94 | |
238 | Emacs support available. | |
239 | ||
ccf3535a | 240 | Enter h or 'h h' for help. |
055fd3a9 GS |
241 | |
242 | main::(-e:1): 0 | |
243 | DB<1> sub foo { 14 } | |
244 | ||
245 | DB<2> sub bar { 3 } | |
246 | ||
247 | DB<3> t print foo() * bar() | |
248 | main::((eval 172):3): print foo() + bar(); | |
249 | main::foo((eval 168):2): | |
250 | main::bar((eval 170):2): | |
251 | 42 | |
252 | ||
492652be | 253 | with this one, once the C<o>ption C<frame=2> has been set: |
055fd3a9 | 254 | |
492652be | 255 | DB<4> o f=2 |
055fd3a9 GS |
256 | frame = '2' |
257 | DB<5> t print foo() * bar() | |
258 | 3: foo() * bar() | |
259 | entering main::foo | |
260 | 2: sub foo { 14 }; | |
261 | exited main::foo | |
262 | entering main::bar | |
263 | 2: sub bar { 3 }; | |
264 | exited main::bar | |
265 | 42 | |
266 | ||
267 | By way of demonstration, we present below a laborious listing | |
268 | resulting from setting your C<PERLDB_OPTS> environment variable to | |
269 | the value C<f=n N>, and running I<perl -d -V> from the command line. | |
7b406369 FC |
270 | Examples using various values of C<n> are shown to give you a feel |
271 | for the difference between settings. Long though it may be, this | |
055fd3a9 GS |
272 | is not a complete listing, but only excerpts. |
273 | ||
274 | =over 4 | |
275 | ||
276 | =item 1 | |
277 | ||
278 | entering main::BEGIN | |
279 | entering Config::BEGIN | |
280 | Package lib/Exporter.pm. | |
281 | Package lib/Carp.pm. | |
282 | Package lib/Config.pm. | |
283 | entering Config::TIEHASH | |
284 | entering Exporter::import | |
285 | entering Exporter::export | |
286 | entering Config::myconfig | |
287 | entering Config::FETCH | |
288 | entering Config::FETCH | |
289 | entering Config::FETCH | |
290 | entering Config::FETCH | |
291 | ||
292 | =item 2 | |
293 | ||
294 | entering main::BEGIN | |
295 | entering Config::BEGIN | |
296 | Package lib/Exporter.pm. | |
297 | Package lib/Carp.pm. | |
298 | exited Config::BEGIN | |
299 | Package lib/Config.pm. | |
300 | entering Config::TIEHASH | |
301 | exited Config::TIEHASH | |
302 | entering Exporter::import | |
303 | entering Exporter::export | |
304 | exited Exporter::export | |
305 | exited Exporter::import | |
306 | exited main::BEGIN | |
307 | entering Config::myconfig | |
308 | entering Config::FETCH | |
309 | exited Config::FETCH | |
310 | entering Config::FETCH | |
311 | exited Config::FETCH | |
312 | entering Config::FETCH | |
313 | ||
d5e42f17 | 314 | =item 3 |
055fd3a9 GS |
315 | |
316 | in $=main::BEGIN() from /dev/null:0 | |
317 | in $=Config::BEGIN() from lib/Config.pm:2 | |
318 | Package lib/Exporter.pm. | |
319 | Package lib/Carp.pm. | |
320 | Package lib/Config.pm. | |
321 | in $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
322 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
323 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li | |
324 | in @=Config::myconfig() from /dev/null:0 | |
325 | in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 | |
326 | in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 | |
327 | in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 | |
328 | in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 | |
329 | in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574 | |
330 | in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574 | |
331 | ||
d5e42f17 | 332 | =item 4 |
055fd3a9 GS |
333 | |
334 | in $=main::BEGIN() from /dev/null:0 | |
335 | in $=Config::BEGIN() from lib/Config.pm:2 | |
336 | Package lib/Exporter.pm. | |
337 | Package lib/Carp.pm. | |
338 | out $=Config::BEGIN() from lib/Config.pm:0 | |
339 | Package lib/Config.pm. | |
340 | in $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
341 | out $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
342 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
343 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ | |
344 | out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ | |
345 | out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
346 | out $=main::BEGIN() from /dev/null:0 | |
347 | in @=Config::myconfig() from /dev/null:0 | |
348 | in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 | |
349 | out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 | |
350 | in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 | |
351 | out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 | |
352 | in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 | |
353 | out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 | |
354 | in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 | |
355 | ||
d5e42f17 | 356 | =item 5 |
055fd3a9 GS |
357 | |
358 | in $=main::BEGIN() from /dev/null:0 | |
359 | in $=Config::BEGIN() from lib/Config.pm:2 | |
360 | Package lib/Exporter.pm. | |
361 | Package lib/Carp.pm. | |
362 | out $=Config::BEGIN() from lib/Config.pm:0 | |
363 | Package lib/Config.pm. | |
364 | in $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
365 | out $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
366 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
367 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E | |
368 | out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E | |
369 | out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
370 | out $=main::BEGIN() from /dev/null:0 | |
371 | in @=Config::myconfig() from /dev/null:0 | |
372 | in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 | |
373 | out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 | |
374 | in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 | |
375 | out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 | |
376 | ||
d5e42f17 | 377 | =item 6 |
055fd3a9 GS |
378 | |
379 | in $=CODE(0x15eca4)() from /dev/null:0 | |
380 | in $=CODE(0x182528)() from lib/Config.pm:2 | |
381 | Package lib/Exporter.pm. | |
382 | out $=CODE(0x182528)() from lib/Config.pm:0 | |
383 | scalar context return from CODE(0x182528): undef | |
384 | Package lib/Config.pm. | |
385 | in $=Config::TIEHASH('Config') from lib/Config.pm:628 | |
386 | out $=Config::TIEHASH('Config') from lib/Config.pm:628 | |
387 | scalar context return from Config::TIEHASH: empty hash | |
388 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
389 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 | |
390 | out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 | |
391 | scalar context return from Exporter::export: '' | |
392 | out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
393 | scalar context return from Exporter::import: '' | |
394 | ||
395 | =back | |
396 | ||
397 | In all cases shown above, the line indentation shows the call tree. | |
398 | If bit 2 of C<frame> is set, a line is printed on exit from a | |
399 | subroutine as well. If bit 4 is set, the arguments are printed | |
400 | along with the caller info. If bit 8 is set, the arguments are | |
401 | printed even if they are tied or references. If bit 16 is set, the | |
402 | return value is printed, too. | |
403 | ||
404 | When a package is compiled, a line like this | |
405 | ||
406 | Package lib/Carp.pm. | |
407 | ||
408 | is printed with proper indentation. | |
409 | ||
7b406369 | 410 | =head1 Debugging Regular Expressions |
055fd3a9 GS |
411 | |
412 | There are two ways to enable debugging output for regular expressions. | |
413 | ||
414 | If your perl is compiled with C<-DDEBUGGING>, you may use the | |
415 | B<-Dr> flag on the command line. | |
416 | ||
417 | Otherwise, one can C<use re 'debug'>, which has effects at | |
3d71525d NJ |
418 | compile time and run time. Since Perl 5.9.5, this pragma is lexically |
419 | scoped. | |
055fd3a9 | 420 | |
7b406369 | 421 | =head2 Compile-time Output |
055fd3a9 GS |
422 | |
423 | The debugging output at compile time looks like this: | |
424 | ||
ccf3535a | 425 | Compiling REx '[bc]d(ef*g)+h[ij]k$' |
1c102323 MJD |
426 | size 45 Got 364 bytes for offset annotations. |
427 | first at 1 | |
428 | rarest char g at 0 | |
429 | rarest char d at 0 | |
430 | 1: ANYOF[bc](12) | |
431 | 12: EXACT <d>(14) | |
432 | 14: CURLYX[0] {1,32767}(28) | |
433 | 16: OPEN1(18) | |
434 | 18: EXACT <e>(20) | |
435 | 20: STAR(23) | |
436 | 21: EXACT <f>(0) | |
437 | 23: EXACT <g>(25) | |
438 | 25: CLOSE1(27) | |
439 | 27: WHILEM[1/1](0) | |
440 | 28: NOTHING(29) | |
441 | 29: EXACT <h>(31) | |
442 | 31: ANYOF[ij](42) | |
443 | 42: EXACT <k>(44) | |
444 | 44: EOL(45) | |
445 | 45: END(0) | |
ccf3535a JK |
446 | anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) |
447 | stclass 'ANYOF[bc]' minlen 7 | |
1c102323 MJD |
448 | Offsets: [45] |
449 | 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] | |
450 | 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] | |
451 | 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] | |
452 | 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] | |
453 | Omitting $` $& $' support. | |
055fd3a9 GS |
454 | |
455 | The first line shows the pre-compiled form of the regex. The second | |
456 | shows the size of the compiled form (in arbitrary units, usually | |
1c102323 MJD |
457 | 4-byte words) and the total number of bytes allocated for the |
458 | offset/length table, usually 4+C<size>*8. The next line shows the | |
459 | label I<id> of the first node that does a match. | |
055fd3a9 | 460 | |
1c102323 MJD |
461 | The |
462 | ||
ccf3535a JK |
463 | anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) |
464 | stclass 'ANYOF[bc]' minlen 7 | |
1c102323 MJD |
465 | |
466 | line (split into two lines above) contains optimizer | |
055fd3a9 GS |
467 | information. In the example shown, the optimizer found that the match |
468 | should contain a substring C<de> at offset 1, plus substring C<gh> | |
469 | at some offset between 3 and infinity. Moreover, when checking for | |
470 | these substrings (to abandon impossible matches quickly), Perl will check | |
471 | for the substring C<gh> before checking for the substring C<de>. The | |
472 | optimizer may also use the knowledge that the match starts (at the | |
1c102323 MJD |
473 | C<first> I<id>) with a character class, and no string |
474 | shorter than 7 characters can possibly match. | |
055fd3a9 | 475 | |
1c102323 | 476 | The fields of interest which may appear in this line are |
055fd3a9 | 477 | |
13a2d996 | 478 | =over 4 |
055fd3a9 GS |
479 | |
480 | =item C<anchored> I<STRING> C<at> I<POS> | |
481 | ||
482 | =item C<floating> I<STRING> C<at> I<POS1..POS2> | |
483 | ||
484 | See above. | |
485 | ||
486 | =item C<matching floating/anchored> | |
487 | ||
488 | Which substring to check first. | |
489 | ||
490 | =item C<minlen> | |
491 | ||
492 | The minimal length of the match. | |
493 | ||
494 | =item C<stclass> I<TYPE> | |
495 | ||
496 | Type of first matching node. | |
497 | ||
498 | =item C<noscan> | |
499 | ||
500 | Don't scan for the found substrings. | |
501 | ||
502 | =item C<isall> | |
503 | ||
1c102323 | 504 | Means that the optimizer information is all that the regular |
055fd3a9 GS |
505 | expression contains, and thus one does not need to enter the regex engine at |
506 | all. | |
507 | ||
508 | =item C<GPOS> | |
509 | ||
510 | Set if the pattern contains C<\G>. | |
511 | ||
512 | =item C<plus> | |
513 | ||
514 | Set if the pattern starts with a repeated char (as in C<x+y>). | |
515 | ||
516 | =item C<implicit> | |
517 | ||
518 | Set if the pattern starts with C<.*>. | |
519 | ||
520 | =item C<with eval> | |
521 | ||
522 | Set if the pattern contain eval-groups, such as C<(?{ code })> and | |
523 | C<(??{ code })>. | |
524 | ||
525 | =item C<anchored(TYPE)> | |
526 | ||
7b406369 | 527 | If the pattern may match only at a handful of places, with C<TYPE> |
d3d47aac | 528 | being C<SBOL>, C<MBOL>, or C<GPOS>. See the table below. |
055fd3a9 GS |
529 | |
530 | =back | |
531 | ||
532 | If a substring is known to match at end-of-line only, it may be | |
ccf3535a | 533 | followed by C<$>, as in C<floating 'k'$>. |
055fd3a9 | 534 | |
1c102323 MJD |
535 | The optimizer-specific information is used to avoid entering (a slow) regex |
536 | engine on strings that will not definitely match. If the C<isall> flag | |
055fd3a9 GS |
537 | is set, a call to the regex engine may be avoided even when the optimizer |
538 | found an appropriate place for the match. | |
539 | ||
1c102323 | 540 | Above the optimizer section is the list of I<nodes> of the compiled |
055fd3a9 GS |
541 | form of the regex. Each line has format |
542 | ||
543 | C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>) | |
544 | ||
7b406369 | 545 | =head2 Types of Nodes |
055fd3a9 | 546 | |
78465a4b | 547 | Here are the current possible types, with short descriptions: |
055fd3a9 | 548 | |
65aa4ca7 FC |
549 | =for comment |
550 | This table is generated by regen/regcomp.pl. Any changes made here | |
551 | will be lost. | |
552 | ||
553 | =for regcomp.pl begin | |
554 | ||
5da6b59a KW |
555 | # TYPE arg-description [num-args] [longjump-len] DESCRIPTION |
556 | ||
557 | # Exit points | |
65aa4ca7 | 558 | |
7bc66b18 KW |
559 | END no End of program. |
560 | SUCCEED no Return from a subroutine, basically. | |
5da6b59a | 561 | |
d3d47aac YO |
562 | # Line Start Anchors: |
563 | SBOL no Match "" at beginning of line: /^/, /\A/ | |
564 | MBOL no Same, assuming multiline: /^/m | |
5da6b59a | 565 | |
d3d47aac YO |
566 | # Line End Anchors: |
567 | SEOL no Match "" at end of line: /$/ | |
568 | MEOL no Same, assuming multiline: /$/m | |
569 | EOS no Match "" at end of string: /\z/ | |
570 | ||
571 | # Match Start Anchors: | |
572 | GPOS no Matches where last m//g left off. | |
573 | ||
574 | # Word Boundary Opcodes: | |
7bc66b18 | 575 | BOUND no Match "" at any word boundary using native |
850b7ec9 | 576 | charset rules for non-utf8 |
7bc66b18 KW |
577 | BOUNDL no Match "" at any locale word boundary |
578 | BOUNDU no Match "" at any word boundary using Unicode | |
850b7ec9 | 579 | rules |
7bc66b18 | 580 | BOUNDA no Match "" at any word boundary using ASCII |
850b7ec9 | 581 | rules |
7bc66b18 | 582 | NBOUND no Match "" at any word non-boundary using |
850b7ec9 | 583 | native charset rules for non-utf8 |
7bc66b18 KW |
584 | NBOUNDL no Match "" at any locale word non-boundary |
585 | NBOUNDU no Match "" at any word non-boundary using | |
850b7ec9 | 586 | Unicode rules |
7bc66b18 | 587 | NBOUNDA no Match "" at any word non-boundary using |
850b7ec9 | 588 | ASCII rules |
5da6b59a KW |
589 | |
590 | # [Special] alternatives: | |
7bc66b18 KW |
591 | REG_ANY no Match any one character (except newline). |
592 | SANY no Match any one character. | |
593 | CANY no Match any one byte. | |
975a06f7 | 594 | ANYOF sv 1 Match character in (or not in) this class, |
7bc66b18 | 595 | single char match only |
a4525e78 | 596 | ANYOFL sv 1 Like ANYOF, but /l is in effect |
7bc66b18 | 597 | |
d3d47aac | 598 | # POSIX Character Classes: |
7bc66b18 KW |
599 | POSIXD none Some [[:class:]] under /d; the FLAGS field |
600 | gives which one | |
601 | POSIXL none Some [[:class:]] under /l; the FLAGS field | |
602 | gives which one | |
603 | POSIXU none Some [[:class:]] under /u; the FLAGS field | |
604 | gives which one | |
605 | POSIXA none Some [[:class:]] under /a; the FLAGS field | |
606 | gives which one | |
607 | NPOSIXD none complement of POSIXD, [[:^class:]] | |
608 | NPOSIXL none complement of POSIXL, [[:^class:]] | |
609 | NPOSIXU none complement of POSIXU, [[:^class:]] | |
610 | NPOSIXA none complement of POSIXA, [[:^class:]] | |
611 | ||
612 | CLUMP no Match any extended grapheme cluster sequence | |
5da6b59a KW |
613 | |
614 | # Alternation | |
615 | ||
65aa4ca7 FC |
616 | # BRANCH The set of branches constituting a single choice are |
617 | # hooked together with their "next" pointers, since | |
618 | # precedence prevents anything being concatenated to | |
619 | # any individual branch. The "next" pointer of the last | |
620 | # BRANCH in a choice points to the thing following the | |
621 | # whole choice. This is also where the final "next" | |
622 | # pointer of each individual branch points; each branch | |
623 | # starts with the operand node of a BRANCH node. | |
5da6b59a | 624 | # |
7bc66b18 | 625 | BRANCH node Match this alternative, or the next... |
5da6b59a | 626 | |
5da6b59a KW |
627 | # Literals |
628 | ||
7bc66b18 | 629 | EXACT str Match this string (preceded by length). |
a4525e78 | 630 | EXACTL str Like EXACT, but /l is in effect. |
7bc66b18 KW |
631 | EXACTF str Match this non-UTF-8 string (not guaranteed |
632 | to be folded) using /id rules (w/len). | |
633 | EXACTFL str Match this string (not guaranteed to be | |
634 | folded) using /il rules (w/len). | |
635 | EXACTFU str Match this string (folded iff in UTF-8, | |
636 | length in folding doesn't change if not in | |
637 | UTF-8) using /iu rules (w/len). | |
638 | EXACTFA str Match this string (not guaranteed to be | |
639 | folded) using /iaa rules (w/len). | |
a4525e78 | 640 | |
7bc66b18 KW |
641 | EXACTFU_SS str Match this string (folded iff in UTF-8, |
642 | length in folding may change even if not in | |
643 | UTF-8) using /iu rules (w/len). | |
a4525e78 KW |
644 | EXACTFLU8 str Rare cirucmstances: like EXACTFU, but is |
645 | under /l, UTF-8, folded, and everything in | |
646 | it is above 255. | |
7bc66b18 KW |
647 | EXACTFA_NO_TRIE str Match this string (which is not trie-able; |
648 | not guaranteed to be folded) using /iaa | |
649 | rules (w/len). | |
5da6b59a KW |
650 | |
651 | # Do nothing types | |
652 | ||
7bc66b18 | 653 | NOTHING no Match empty string. |
5da6b59a | 654 | # A variant of above which delimits a group, thus stops optimizations |
7bc66b18 KW |
655 | TAIL no Match empty string. Can jump here from |
656 | outside. | |
5da6b59a KW |
657 | |
658 | # Loops | |
659 | ||
65aa4ca7 | 660 | # STAR,PLUS '?', and complex '*' and '+', are implemented as |
62e6ef33 | 661 | # circular BRANCH structures. Simple cases |
65aa4ca7 FC |
662 | # (one character per match) are implemented with STAR |
663 | # and PLUS for speed and to minimize recursive plunges. | |
5da6b59a | 664 | # |
7bc66b18 KW |
665 | STAR node Match this (simple) thing 0 or more times. |
666 | PLUS node Match this (simple) thing 1 or more times. | |
667 | ||
668 | CURLY sv 2 Match this simple thing {n,m} times. | |
669 | CURLYN no 2 Capture next-after-this simple thing | |
670 | CURLYM no 2 Capture this medium-complex thing {n,m} | |
671 | times. | |
672 | CURLYX sv 2 Match this complex thing {n,m} times. | |
5da6b59a KW |
673 | |
674 | # This terminator creates a loop structure for CURLYX | |
7bc66b18 | 675 | WHILEM no Do curly processing and see if rest matches. |
5da6b59a KW |
676 | |
677 | # Buffer related | |
678 | ||
679 | # OPEN,CLOSE,GROUPP ...are numbered at compile time. | |
7bc66b18 KW |
680 | OPEN num 1 Mark this point in input as start of #n. |
681 | CLOSE num 1 Analogous to OPEN. | |
682 | ||
683 | REF num 1 Match some already matched string | |
684 | REFF num 1 Match already matched string, folded using | |
850b7ec9 | 685 | native charset rules for non-utf8 |
7bc66b18 KW |
686 | REFFL num 1 Match already matched string, folded in loc. |
687 | REFFU num 1 Match already matched string, folded using | |
850b7ec9 | 688 | unicode rules for non-utf8 |
7bc66b18 | 689 | REFFA num 1 Match already matched string, folded using |
850b7ec9 KW |
690 | unicode rules for non-utf8, no mixing ASCII, |
691 | non-ASCII | |
65aa4ca7 FC |
692 | |
693 | # Named references. Code in regcomp.c assumes that these all are after | |
694 | # the numbered references | |
7bc66b18 KW |
695 | NREF no-sv 1 Match some already matched string |
696 | NREFF no-sv 1 Match already matched string, folded using | |
850b7ec9 | 697 | native charset rules for non-utf8 |
7bc66b18 KW |
698 | NREFFL no-sv 1 Match already matched string, folded in loc. |
699 | NREFFU num 1 Match already matched string, folded using | |
850b7ec9 | 700 | unicode rules for non-utf8 |
7bc66b18 | 701 | NREFFA num 1 Match already matched string, folded using |
850b7ec9 KW |
702 | unicode rules for non-utf8, no mixing ASCII, |
703 | non-ASCII | |
7bc66b18 | 704 | |
d3d47aac YO |
705 | # Support for long RE |
706 | LONGJMP off 1 1 Jump far away. | |
707 | BRANCHJ off 1 1 BRANCH with long offset. | |
708 | ||
709 | # Special Case Regops | |
7bc66b18 KW |
710 | IFMATCH off 1 2 Succeeds if the following matches. |
711 | UNLESSM off 1 2 Fails if the following matches. | |
712 | SUSPEND off 1 1 "Independent" sub-RE. | |
713 | IFTHEN off 1 1 Switch, should be preceded by switcher. | |
714 | GROUPP num 1 Whether the group matched. | |
5da6b59a | 715 | |
5da6b59a KW |
716 | # The heavy worker |
717 | ||
13f27704 KW |
718 | EVAL evl/flags Execute some Perl code. |
719 | 2L | |
5da6b59a KW |
720 | |
721 | # Modifiers | |
722 | ||
7bc66b18 KW |
723 | MINMOD no Next operator is not greedy. |
724 | LOGICAL no Next opcode should set the flag only. | |
5da6b59a KW |
725 | |
726 | # This is not used yet | |
7bc66b18 | 727 | RENUM off 1 1 Group with independently numbered parens. |
5da6b59a KW |
728 | |
729 | # Trie Related | |
730 | ||
65aa4ca7 FC |
731 | # Behave the same as A|LIST|OF|WORDS would. The '..C' variants |
732 | # have inline charclass data (ascii only), the 'C' store it in the | |
733 | # structure. | |
5da6b59a | 734 | |
7bc66b18 KW |
735 | TRIE trie 1 Match many EXACT(F[ALU]?)? at once. |
736 | flags==type | |
737 | TRIEC trie Same as TRIE, but with embedded charclass | |
738 | charclass data | |
5da6b59a | 739 | |
7bc66b18 KW |
740 | AHOCORASICK trie 1 Aho Corasick stclass. flags==type |
741 | AHOCORASICKC trie Same as AHOCORASICK, but with embedded | |
742 | charclass charclass data | |
5da6b59a KW |
743 | |
744 | # Regex Subroutines | |
7bc66b18 KW |
745 | GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2 |
746 | GOSTART no recurse to start of pattern | |
5da6b59a KW |
747 | |
748 | # Special conditionals | |
7bc66b18 KW |
749 | NGROUPP no-sv 1 Whether the group matched. |
750 | INSUBP num 1 Whether we are in a specific recurse. | |
751 | DEFINEP none 1 Never execute directly. | |
5da6b59a KW |
752 | |
753 | # Backtracking Verbs | |
7bc66b18 KW |
754 | ENDLIKE none Used only for the type field of verbs |
755 | OPFAIL none Same as (?!) | |
756 | ACCEPT parno 1 Accepts the current matched string. | |
5da6b59a KW |
757 | |
758 | # Verbs With Arguments | |
7bc66b18 KW |
759 | VERB no-sv 1 Used only for the type field of verbs |
760 | PRUNE no-sv 1 Pattern fails at this startpoint if no- | |
761 | backtracking through this | |
762 | MARKPOINT no-sv 1 Push the current location for rollback by | |
763 | cut. | |
764 | SKIP no-sv 1 On failure skip forward (to the mark) before | |
765 | retrying | |
766 | COMMIT no-sv 1 Pattern fails outright if backtracking | |
767 | through this | |
768 | CUTGROUP no-sv 1 On failure go to the next alternation in the | |
769 | group | |
5da6b59a KW |
770 | |
771 | # Control what to keep in $&. | |
7bc66b18 | 772 | KEEPS no $& begins here. |
5da6b59a KW |
773 | |
774 | # New charclass like patterns | |
7bc66b18 | 775 | LNBREAK none generic newline pattern |
5da6b59a KW |
776 | |
777 | # SPECIAL REGOPS | |
778 | ||
65aa4ca7 FC |
779 | # This is not really a node, but an optimized away piece of a "long" |
780 | # node. To simplify debugging output, we mark it as if it were a node | |
7bc66b18 | 781 | OPTIMIZED off Placeholder for dump. |
5da6b59a KW |
782 | |
783 | # Special opcode with the property that no opcode in a compiled program | |
784 | # will ever be of this type. Thus it can be used as a flag value that | |
785 | # no other opcode has been seen. END is used similarly, in that an END | |
65aa4ca7 FC |
786 | # node cant be optimized. So END implies "unoptimizable" and PSEUDO |
787 | # mean "not seen anything to optimize yet". | |
7bc66b18 | 788 | PSEUDO off Pseudo opcode for internal use. |
65aa4ca7 FC |
789 | |
790 | =for regcomp.pl end | |
055fd3a9 | 791 | |
1c102323 MJD |
792 | =for unprinted-credits |
793 | Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421 | |
794 | ||
795 | Following the optimizer information is a dump of the offset/length | |
796 | table, here split across several lines: | |
797 | ||
798 | Offsets: [45] | |
799 | 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] | |
800 | 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] | |
801 | 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] | |
802 | 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] | |
803 | ||
804 | The first line here indicates that the offset/length table contains 45 | |
805 | entries. Each entry is a pair of integers, denoted by C<offset[length]>. | |
17c338f3 | 806 | Entries are numbered starting with 1, so entry #1 here is C<1[4]> and |
1c102323 MJD |
807 | entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:> |
808 | (the C<1: ANYOF[bc]>) begins at character position 1 in the | |
809 | pre-compiled form of the regex, and has a length of 4 characters. | |
810 | C<5[1]> in position 12 | |
811 | indicates that the node labeled C<12:> | |
812 | (the C<< 12: EXACT <d> >>) begins at character position 5 in the | |
813 | pre-compiled form of the regex, and has a length of 1 character. | |
814 | C<12[1]> in position 14 | |
815 | indicates that the node labeled C<14:> | |
816 | (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the | |
817 | pre-compiled form of the regex, and has a length of 1 character---that | |
818 | is, it corresponds to the C<+> symbol in the precompiled regex. | |
819 | ||
820 | C<0[0]> items indicate that there is no corresponding node. | |
821 | ||
7b406369 | 822 | =head2 Run-time Output |
055fd3a9 GS |
823 | |
824 | First of all, when doing a match, one may get no run-time output even | |
825 | if debugging is enabled. This means that the regex engine was never | |
826 | entered and that all of the job was therefore done by the optimizer. | |
827 | ||
828 | If the regex engine was entered, the output may look like this: | |
829 | ||
ccf3535a | 830 | Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__' |
055fd3a9 GS |
831 | Setting an EVAL scope, savestack=3 |
832 | 2 <ab> <cdefg__gh_> | 1: ANYOF | |
833 | 3 <abc> <defg__gh_> | 11: EXACT <d> | |
834 | 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767} | |
835 | 4 <abcd> <efg__gh_> | 26: WHILEM | |
836 | 0 out of 1..32767 cc=effff31c | |
837 | 4 <abcd> <efg__gh_> | 15: OPEN1 | |
838 | 4 <abcd> <efg__gh_> | 17: EXACT <e> | |
839 | 5 <abcde> <fg__gh_> | 19: STAR | |
840 | EXACT <f> can match 1 times out of 32767... | |
841 | Setting an EVAL scope, savestack=3 | |
842 | 6 <bcdef> <g__gh__> | 22: EXACT <g> | |
843 | 7 <bcdefg> <__gh__> | 24: CLOSE1 | |
844 | 7 <bcdefg> <__gh__> | 26: WHILEM | |
845 | 1 out of 1..32767 cc=effff31c | |
846 | Setting an EVAL scope, savestack=12 | |
847 | 7 <bcdefg> <__gh__> | 15: OPEN1 | |
848 | 7 <bcdefg> <__gh__> | 17: EXACT <e> | |
849 | restoring \1 to 4(4)..7 | |
850 | failed, try continuation... | |
851 | 7 <bcdefg> <__gh__> | 27: NOTHING | |
852 | 7 <bcdefg> <__gh__> | 28: EXACT <h> | |
853 | failed... | |
854 | failed... | |
855 | ||
856 | The most significant information in the output is about the particular I<node> | |
857 | of the compiled regex that is currently being tested against the target string. | |
858 | The format of these lines is | |
859 | ||
860 | C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE> | |
861 | ||
862 | The I<TYPE> info is indented with respect to the backtracking level. | |
863 | Other incidental information appears interspersed within. | |
864 | ||
7b406369 | 865 | =head1 Debugging Perl Memory Usage |
055fd3a9 GS |
866 | |
867 | Perl is a profligate wastrel when it comes to memory use. There | |
868 | is a saying that to estimate memory usage of Perl, assume a reasonable | |
869 | algorithm for memory allocation, multiply that estimate by 10, and | |
870 | while you still may miss the mark, at least you won't be quite so | |
4375e838 | 871 | astonished. This is not absolutely true, but may provide a good |
055fd3a9 GS |
872 | grasp of what happens. |
873 | ||
874 | Assume that an integer cannot take less than 20 bytes of memory, a | |
875 | float cannot take less than 24 bytes, a string cannot take less | |
876 | than 32 bytes (all these examples assume 32-bit architectures, the | |
877 | result are quite a bit worse on 64-bit architectures). If a variable | |
878 | is accessed in two of three different ways (which require an integer, | |
879 | a float, or a string), the memory footprint may increase yet another | |
b9449ee0 | 880 | 20 bytes. A sloppy malloc(3) implementation can inflate these |
055fd3a9 GS |
881 | numbers dramatically. |
882 | ||
883 | On the opposite end of the scale, a declaration like | |
884 | ||
885 | sub foo; | |
886 | ||
887 | may take up to 500 bytes of memory, depending on which release of Perl | |
888 | you're running. | |
889 | ||
890 | Anecdotal estimates of source-to-compiled code bloat suggest an | |
891 | eightfold increase. This means that the compiled form of reasonable | |
892 | (normally commented, properly indented etc.) code will take | |
893 | about eight times more space in memory than the code took | |
894 | on disk. | |
895 | ||
b30f304a JH |
896 | The B<-DL> command-line switch is obsolete since circa Perl 5.6.0 |
897 | (it was available only if Perl was built with C<-DDEBUGGING>). | |
898 | The switch was used to track Perl's memory allocations and possible | |
899 | memory leaks. These days the use of malloc debugging tools like | |
5b6a3331 | 900 | F<Purify> or F<valgrind> is suggested instead. See also |
7b406369 | 901 | L<perlhacktips/PERL_MEM_LOG>. |
b30f304a JH |
902 | |
903 | One way to find out how much memory is being used by Perl data | |
904 | structures is to install the Devel::Size module from CPAN: it gives | |
905 | you the minimum number of bytes required to store a particular data | |
906 | structure. Please be mindful of the difference between the size() | |
907 | and total_size(). | |
908 | ||
909 | If Perl has been compiled using Perl's malloc you can analyze Perl | |
7b406369 | 910 | memory usage by setting $ENV{PERL_DEBUG_MSTATS}. |
055fd3a9 GS |
911 | |
912 | =head2 Using C<$ENV{PERL_DEBUG_MSTATS}> | |
913 | ||
914 | If your perl is using Perl's malloc() and was compiled with the | |
915 | necessary switches (this is the default), then it will print memory | |
4375e838 | 916 | usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS} |
055fd3a9 GS |
917 | > 1 >>, and before termination of the program when C<< |
918 | $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to | |
919 | the following example: | |
920 | ||
921 | $ PERL_DEBUG_MSTATS=2 perl -e "require Carp" | |
922 | Memory allocation statistics after compilation: (buckets 4(4)..8188(8192) | |
923 | 14216 free: 130 117 28 7 9 0 2 2 1 0 0 | |
924 | 437 61 36 0 5 | |
925 | 60924 used: 125 137 161 55 7 8 6 16 2 0 1 | |
926 | 74 109 304 84 20 | |
927 | Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048. | |
928 | Memory allocation statistics after execution: (buckets 4(4)..8188(8192) | |
929 | 30888 free: 245 78 85 13 6 2 1 3 2 0 1 | |
930 | 315 162 39 42 11 | |
931 | 175816 used: 265 176 1112 111 26 22 11 27 2 1 1 | |
932 | 196 178 1066 798 39 | |
933 | Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144. | |
934 | ||
935 | It is possible to ask for such a statistic at arbitrary points in | |
b9449ee0 | 936 | your execution using the mstat() function out of the standard |
055fd3a9 GS |
937 | Devel::Peek module. |
938 | ||
939 | Here is some explanation of that format: | |
940 | ||
13a2d996 | 941 | =over 4 |
055fd3a9 GS |
942 | |
943 | =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)> | |
944 | ||
945 | Perl's malloc() uses bucketed allocations. Every request is rounded | |
946 | up to the closest bucket size available, and a bucket is taken from | |
947 | the pool of buckets of that size. | |
948 | ||
949 | The line above describes the limits of buckets currently in use. | |
950 | Each bucket has two sizes: memory footprint and the maximal size | |
951 | of user data that can fit into this bucket. Suppose in the above | |
952 | example that the smallest bucket were size 4. The biggest bucket | |
953 | would have usable size 8188, and the memory footprint would be 8192. | |
954 | ||
955 | In a Perl built for debugging, some buckets may have negative usable | |
956 | size. This means that these buckets cannot (and will not) be used. | |
957 | For larger buckets, the memory footprint may be one page greater | |
7b406369 | 958 | than a power of 2. If so, the corresponding power of two is |
055fd3a9 GS |
959 | printed in the C<APPROX> field above. |
960 | ||
961 | =item Free/Used | |
962 | ||
963 | The 1 or 2 rows of numbers following that correspond to the number | |
964 | of buckets of each size between C<SMALLEST> and C<GREATEST>. In | |
965 | the first row, the sizes (memory footprints) of buckets are powers | |
966 | of two--or possibly one page greater. In the second row, if present, | |
967 | the memory footprints of the buckets are between the memory footprints | |
968 | of two buckets "above". | |
969 | ||
4375e838 | 970 | For example, suppose under the previous example, the memory footprints |
055fd3a9 GS |
971 | were |
972 | ||
973 | free: 8 16 32 64 128 256 512 1024 2048 4096 8192 | |
974 | 4 12 24 48 80 | |
975 | ||
7b406369 | 976 | With a non-C<DEBUGGING> perl, the buckets starting from C<128> have |
d1be9408 | 977 | a 4-byte overhead, and thus an 8192-long bucket may take up to |
055fd3a9 GS |
978 | 8188-byte allocations. |
979 | ||
980 | =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS> | |
981 | ||
982 | The first two fields give the total amount of memory perl sbrk(2)ed | |
983 | (ess-broken? :-) and number of sbrk(2)s used. The third number is | |
984 | what perl thinks about continuity of returned chunks. So long as | |
985 | this number is positive, malloc() will assume that it is probable | |
986 | that sbrk(2) will provide continuous memory. | |
987 | ||
988 | Memory allocated by external libraries is not counted. | |
989 | ||
990 | =item C<pad: 0> | |
991 | ||
992 | The amount of sbrk(2)ed memory needed to keep buckets aligned. | |
993 | ||
994 | =item C<heads: 2192> | |
995 | ||
996 | Although memory overhead of bigger buckets is kept inside the bucket, for | |
997 | smaller buckets, it is kept in separate areas. This field gives the | |
998 | total size of these areas. | |
999 | ||
1000 | =item C<chain: 0> | |
1001 | ||
1002 | malloc() may want to subdivide a bigger bucket into smaller buckets. | |
1003 | If only a part of the deceased bucket is left unsubdivided, the rest | |
1004 | is kept as an element of a linked list. This field gives the total | |
1005 | size of these chunks. | |
1006 | ||
1007 | =item C<tail: 6144> | |
1008 | ||
1009 | To minimize the number of sbrk(2)s, malloc() asks for more memory. This | |
1010 | field gives the size of the yet unused part, which is sbrk(2)ed, but | |
1011 | never touched. | |
1012 | ||
1013 | =back | |
1014 | ||
055fd3a9 GS |
1015 | =head1 SEE ALSO |
1016 | ||
1017 | L<perldebug>, | |
1018 | L<perlguts>, | |
1019 | L<perlrun> | |
1020 | L<re>, | |
1021 | and | |
fe854a6f | 1022 | L<Devel::DProf>. |