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 | ||
101 | =back | |
102 | ||
103 | Note that if C<&DB::sub> needs external data for it to work, no | |
74410c12 JM |
104 | subroutine call is possible without it. As an example, the standard |
105 | debugger's C<&DB::sub> depends on the C<$DB::deep> variable | |
106 | (it defines how many levels of recursion deep into the debugger you can go | |
107 | before a mandatory break). If C<$DB::deep> is not defined, subroutine | |
108 | calls are not possible, even though C<&DB::sub> exists. | |
055fd3a9 GS |
109 | |
110 | =head2 Writing Your Own Debugger | |
111 | ||
74410c12 | 112 | =head3 Environment Variables |
666f95b9 | 113 | |
74410c12 JM |
114 | The C<PERL5DB> environment variable can be used to define a debugger. |
115 | For example, the minimal "working" debugger (it actually doesn't do anything) | |
116 | consists of one line: | |
666f95b9 | 117 | |
055fd3a9 GS |
118 | sub DB::DB {} |
119 | ||
74410c12 | 120 | It can easily be defined like this: |
666f95b9 | 121 | |
055fd3a9 GS |
122 | $ PERL5DB="sub DB::DB {}" perl -d your-script |
123 | ||
74410c12 | 124 | Another brief debugger, slightly more useful, can be created |
055fd3a9 GS |
125 | with only the line: |
126 | ||
127 | sub DB::DB {print ++$i; scalar <STDIN>} | |
128 | ||
74410c12 JM |
129 | This debugger prints a number which increments for each statement |
130 | encountered and waits for you to hit a newline before continuing | |
131 | to the next statement. | |
666f95b9 | 132 | |
74410c12 | 133 | The following debugger is actually useful: |
666f95b9 | 134 | |
055fd3a9 GS |
135 | { |
136 | package DB; | |
137 | sub DB {} | |
138 | sub sub {print ++$i, " $sub\n"; &$sub} | |
139 | } | |
140 | ||
74410c12 JM |
141 | It prints the sequence number of each subroutine call and the name of the |
142 | called subroutine. Note that C<&DB::sub> is being compiled into the | |
143 | package C<DB> through the use of the C<package> directive. | |
055fd3a9 | 144 | |
74410c12 JM |
145 | When it starts, the debugger reads your rc file (F<./.perldb> or |
146 | F<~/.perldb> under Unix), which can set important options. | |
147 | (A subroutine (C<&afterinit>) can be defined here as well; it is executed | |
148 | after the debugger completes its own initialization.) | |
055fd3a9 GS |
149 | |
150 | After the rc file is read, the debugger reads the PERLDB_OPTS | |
74410c12 JM |
151 | environment variable and uses it to set debugger options. The |
152 | contents of this variable are treated as if they were the argument | |
96090e4f | 153 | of an C<o ...> debugger command (q.v. in L<perldebug/"Configurable Options">). |
74410c12 | 154 | |
7b406369 | 155 | =head3 Debugger Internal Variables |
25cf7dea | 156 | |
74410c12 JM |
157 | In addition to the file and subroutine-related variables mentioned above, |
158 | the debugger also maintains various magical internal variables. | |
159 | ||
160 | =over 4 | |
161 | ||
162 | =item * | |
055fd3a9 | 163 | |
74410c12 JM |
164 | C<@DB::dbline> is an alias for C<@{"::_<current_file"}>, which |
165 | holds the lines of the currently-selected file (compiled by Perl), either | |
166 | explicitly chosen with the debugger's C<f> command, or implicitly by flow | |
167 | of execution. | |
168 | ||
169 | Values in this array are magical in numeric context: they compare | |
170 | equal to zero only if the line is not breakable. | |
171 | ||
172 | =item * | |
173 | ||
7b406369 | 174 | C<%DB::dbline> is an alias for C<%{"::_<current_file"}>, which |
74410c12 JM |
175 | contains breakpoints and actions keyed by line number in |
176 | the currently-selected file, either explicitly chosen with the | |
055fd3a9 GS |
177 | debugger's C<f> command, or implicitly by flow of execution. |
178 | ||
74410c12 JM |
179 | As previously noted, individual entries (as opposed to the whole hash) |
180 | are settable. Perl only cares about Boolean true here, although | |
181 | the values used by F<perl5db.pl> have the form | |
182 | C<"$break_condition\0$action">. | |
183 | ||
184 | =back | |
185 | ||
7b406369 | 186 | =head3 Debugger Customization Functions |
74410c12 JM |
187 | |
188 | Some functions are provided to simplify customization. | |
189 | ||
190 | =over 4 | |
191 | ||
192 | =item * | |
193 | ||
71110851 RGS |
194 | See L<perldebug/"Configurable Options"> for a description of options parsed by |
195 | C<DB::parse_options(string)>. | |
74410c12 JM |
196 | |
197 | =item * | |
198 | ||
199 | C<DB::dump_trace(skip[,count])> skips the specified number of frames | |
200 | and returns a list containing information about the calling frames (all | |
201 | of them, if C<count> is missing). Each entry is reference to a hash | |
202 | with keys C<context> (either C<.>, C<$>, or C<@>), C<sub> (subroutine | |
055fd3a9 GS |
203 | name, or info about C<eval>), C<args> (C<undef> or a reference to |
204 | an array), C<file>, and C<line>. | |
205 | ||
74410c12 JM |
206 | =item * |
207 | ||
208 | C<DB::print_trace(FH, skip[, count[, short]])> prints | |
055fd3a9 GS |
209 | formatted info about caller frames. The last two functions may be |
210 | convenient as arguments to C<< < >>, C<< << >> commands. | |
211 | ||
74410c12 JM |
212 | =back |
213 | ||
055fd3a9 GS |
214 | Note that any variables and functions that are not documented in |
215 | this manpages (or in L<perldebug>) are considered for internal | |
216 | use only, and as such are subject to change without notice. | |
217 | ||
218 | =head1 Frame Listing Output Examples | |
219 | ||
220 | The C<frame> option can be used to control the output of frame | |
221 | information. For example, contrast this expression trace: | |
222 | ||
223 | $ perl -de 42 | |
224 | Stack dump during die enabled outside of evals. | |
225 | ||
226 | Loading DB routines from perl5db.pl patch level 0.94 | |
227 | Emacs support available. | |
228 | ||
ccf3535a | 229 | Enter h or 'h h' for help. |
055fd3a9 GS |
230 | |
231 | main::(-e:1): 0 | |
232 | DB<1> sub foo { 14 } | |
233 | ||
234 | DB<2> sub bar { 3 } | |
235 | ||
236 | DB<3> t print foo() * bar() | |
237 | main::((eval 172):3): print foo() + bar(); | |
238 | main::foo((eval 168):2): | |
239 | main::bar((eval 170):2): | |
240 | 42 | |
241 | ||
492652be | 242 | with this one, once the C<o>ption C<frame=2> has been set: |
055fd3a9 | 243 | |
492652be | 244 | DB<4> o f=2 |
055fd3a9 GS |
245 | frame = '2' |
246 | DB<5> t print foo() * bar() | |
247 | 3: foo() * bar() | |
248 | entering main::foo | |
249 | 2: sub foo { 14 }; | |
250 | exited main::foo | |
251 | entering main::bar | |
252 | 2: sub bar { 3 }; | |
253 | exited main::bar | |
254 | 42 | |
255 | ||
256 | By way of demonstration, we present below a laborious listing | |
257 | resulting from setting your C<PERLDB_OPTS> environment variable to | |
258 | the value C<f=n N>, and running I<perl -d -V> from the command line. | |
7b406369 FC |
259 | Examples using various values of C<n> are shown to give you a feel |
260 | for the difference between settings. Long though it may be, this | |
055fd3a9 GS |
261 | is not a complete listing, but only excerpts. |
262 | ||
263 | =over 4 | |
264 | ||
265 | =item 1 | |
266 | ||
267 | entering main::BEGIN | |
268 | entering Config::BEGIN | |
269 | Package lib/Exporter.pm. | |
270 | Package lib/Carp.pm. | |
271 | Package lib/Config.pm. | |
272 | entering Config::TIEHASH | |
273 | entering Exporter::import | |
274 | entering Exporter::export | |
275 | entering Config::myconfig | |
276 | entering Config::FETCH | |
277 | entering Config::FETCH | |
278 | entering Config::FETCH | |
279 | entering Config::FETCH | |
280 | ||
281 | =item 2 | |
282 | ||
283 | entering main::BEGIN | |
284 | entering Config::BEGIN | |
285 | Package lib/Exporter.pm. | |
286 | Package lib/Carp.pm. | |
287 | exited Config::BEGIN | |
288 | Package lib/Config.pm. | |
289 | entering Config::TIEHASH | |
290 | exited Config::TIEHASH | |
291 | entering Exporter::import | |
292 | entering Exporter::export | |
293 | exited Exporter::export | |
294 | exited Exporter::import | |
295 | exited main::BEGIN | |
296 | entering Config::myconfig | |
297 | entering Config::FETCH | |
298 | exited Config::FETCH | |
299 | entering Config::FETCH | |
300 | exited Config::FETCH | |
301 | entering Config::FETCH | |
302 | ||
d5e42f17 | 303 | =item 3 |
055fd3a9 GS |
304 | |
305 | in $=main::BEGIN() from /dev/null:0 | |
306 | in $=Config::BEGIN() from lib/Config.pm:2 | |
307 | Package lib/Exporter.pm. | |
308 | Package lib/Carp.pm. | |
309 | Package lib/Config.pm. | |
310 | in $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
311 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
312 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from li | |
313 | in @=Config::myconfig() from /dev/null:0 | |
314 | in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 | |
315 | in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 | |
316 | in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 | |
317 | in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 | |
318 | in $=Config::FETCH(ref(Config), 'osname') from lib/Config.pm:574 | |
319 | in $=Config::FETCH(ref(Config), 'osvers') from lib/Config.pm:574 | |
320 | ||
d5e42f17 | 321 | =item 4 |
055fd3a9 GS |
322 | |
323 | in $=main::BEGIN() from /dev/null:0 | |
324 | in $=Config::BEGIN() from lib/Config.pm:2 | |
325 | Package lib/Exporter.pm. | |
326 | Package lib/Carp.pm. | |
327 | out $=Config::BEGIN() from lib/Config.pm:0 | |
328 | Package lib/Config.pm. | |
329 | in $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
330 | out $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
331 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
332 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ | |
333 | out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/ | |
334 | out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
335 | out $=main::BEGIN() from /dev/null:0 | |
336 | in @=Config::myconfig() from /dev/null:0 | |
337 | in $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 | |
338 | out $=Config::FETCH(ref(Config), 'package') from lib/Config.pm:574 | |
339 | in $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 | |
340 | out $=Config::FETCH(ref(Config), 'baserev') from lib/Config.pm:574 | |
341 | in $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 | |
342 | out $=Config::FETCH(ref(Config), 'PERL_VERSION') from lib/Config.pm:574 | |
343 | in $=Config::FETCH(ref(Config), 'PERL_SUBVERSION') from lib/Config.pm:574 | |
344 | ||
d5e42f17 | 345 | =item 5 |
055fd3a9 GS |
346 | |
347 | in $=main::BEGIN() from /dev/null:0 | |
348 | in $=Config::BEGIN() from lib/Config.pm:2 | |
349 | Package lib/Exporter.pm. | |
350 | Package lib/Carp.pm. | |
351 | out $=Config::BEGIN() from lib/Config.pm:0 | |
352 | Package lib/Config.pm. | |
353 | in $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
354 | out $=Config::TIEHASH('Config') from lib/Config.pm:644 | |
355 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
356 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E | |
357 | out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/E | |
358 | out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
359 | out $=main::BEGIN() from /dev/null:0 | |
360 | in @=Config::myconfig() from /dev/null:0 | |
361 | in $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 | |
362 | out $=Config::FETCH('Config=HASH(0x1aa444)', 'package') from lib/Config.pm:574 | |
363 | in $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 | |
364 | out $=Config::FETCH('Config=HASH(0x1aa444)', 'baserev') from lib/Config.pm:574 | |
365 | ||
d5e42f17 | 366 | =item 6 |
055fd3a9 GS |
367 | |
368 | in $=CODE(0x15eca4)() from /dev/null:0 | |
369 | in $=CODE(0x182528)() from lib/Config.pm:2 | |
370 | Package lib/Exporter.pm. | |
371 | out $=CODE(0x182528)() from lib/Config.pm:0 | |
372 | scalar context return from CODE(0x182528): undef | |
373 | Package lib/Config.pm. | |
374 | in $=Config::TIEHASH('Config') from lib/Config.pm:628 | |
375 | out $=Config::TIEHASH('Config') from lib/Config.pm:628 | |
376 | scalar context return from Config::TIEHASH: empty hash | |
377 | in $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
378 | in $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 | |
379 | out $=Exporter::export('Config', 'main', 'myconfig', 'config_vars') from lib/Exporter.pm:171 | |
380 | scalar context return from Exporter::export: '' | |
381 | out $=Exporter::import('Config', 'myconfig', 'config_vars') from /dev/null:0 | |
382 | scalar context return from Exporter::import: '' | |
383 | ||
384 | =back | |
385 | ||
386 | In all cases shown above, the line indentation shows the call tree. | |
387 | If bit 2 of C<frame> is set, a line is printed on exit from a | |
388 | subroutine as well. If bit 4 is set, the arguments are printed | |
389 | along with the caller info. If bit 8 is set, the arguments are | |
390 | printed even if they are tied or references. If bit 16 is set, the | |
391 | return value is printed, too. | |
392 | ||
393 | When a package is compiled, a line like this | |
394 | ||
395 | Package lib/Carp.pm. | |
396 | ||
397 | is printed with proper indentation. | |
398 | ||
7b406369 | 399 | =head1 Debugging Regular Expressions |
055fd3a9 GS |
400 | |
401 | There are two ways to enable debugging output for regular expressions. | |
402 | ||
403 | If your perl is compiled with C<-DDEBUGGING>, you may use the | |
404 | B<-Dr> flag on the command line. | |
405 | ||
406 | Otherwise, one can C<use re 'debug'>, which has effects at | |
3d71525d NJ |
407 | compile time and run time. Since Perl 5.9.5, this pragma is lexically |
408 | scoped. | |
055fd3a9 | 409 | |
7b406369 | 410 | =head2 Compile-time Output |
055fd3a9 GS |
411 | |
412 | The debugging output at compile time looks like this: | |
413 | ||
ccf3535a | 414 | Compiling REx '[bc]d(ef*g)+h[ij]k$' |
1c102323 MJD |
415 | size 45 Got 364 bytes for offset annotations. |
416 | first at 1 | |
417 | rarest char g at 0 | |
418 | rarest char d at 0 | |
419 | 1: ANYOF[bc](12) | |
420 | 12: EXACT <d>(14) | |
421 | 14: CURLYX[0] {1,32767}(28) | |
422 | 16: OPEN1(18) | |
423 | 18: EXACT <e>(20) | |
424 | 20: STAR(23) | |
425 | 21: EXACT <f>(0) | |
426 | 23: EXACT <g>(25) | |
427 | 25: CLOSE1(27) | |
428 | 27: WHILEM[1/1](0) | |
429 | 28: NOTHING(29) | |
430 | 29: EXACT <h>(31) | |
431 | 31: ANYOF[ij](42) | |
432 | 42: EXACT <k>(44) | |
433 | 44: EOL(45) | |
434 | 45: END(0) | |
ccf3535a JK |
435 | anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) |
436 | stclass 'ANYOF[bc]' minlen 7 | |
1c102323 MJD |
437 | Offsets: [45] |
438 | 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] | |
439 | 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] | |
440 | 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] | |
441 | 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] | |
442 | Omitting $` $& $' support. | |
055fd3a9 GS |
443 | |
444 | The first line shows the pre-compiled form of the regex. The second | |
445 | shows the size of the compiled form (in arbitrary units, usually | |
1c102323 MJD |
446 | 4-byte words) and the total number of bytes allocated for the |
447 | offset/length table, usually 4+C<size>*8. The next line shows the | |
448 | label I<id> of the first node that does a match. | |
055fd3a9 | 449 | |
1c102323 MJD |
450 | The |
451 | ||
ccf3535a JK |
452 | anchored 'de' at 1 floating 'gh' at 3..2147483647 (checking floating) |
453 | stclass 'ANYOF[bc]' minlen 7 | |
1c102323 MJD |
454 | |
455 | line (split into two lines above) contains optimizer | |
055fd3a9 GS |
456 | information. In the example shown, the optimizer found that the match |
457 | should contain a substring C<de> at offset 1, plus substring C<gh> | |
458 | at some offset between 3 and infinity. Moreover, when checking for | |
459 | these substrings (to abandon impossible matches quickly), Perl will check | |
460 | for the substring C<gh> before checking for the substring C<de>. The | |
461 | optimizer may also use the knowledge that the match starts (at the | |
1c102323 MJD |
462 | C<first> I<id>) with a character class, and no string |
463 | shorter than 7 characters can possibly match. | |
055fd3a9 | 464 | |
1c102323 | 465 | The fields of interest which may appear in this line are |
055fd3a9 | 466 | |
13a2d996 | 467 | =over 4 |
055fd3a9 GS |
468 | |
469 | =item C<anchored> I<STRING> C<at> I<POS> | |
470 | ||
471 | =item C<floating> I<STRING> C<at> I<POS1..POS2> | |
472 | ||
473 | See above. | |
474 | ||
475 | =item C<matching floating/anchored> | |
476 | ||
477 | Which substring to check first. | |
478 | ||
479 | =item C<minlen> | |
480 | ||
481 | The minimal length of the match. | |
482 | ||
483 | =item C<stclass> I<TYPE> | |
484 | ||
485 | Type of first matching node. | |
486 | ||
487 | =item C<noscan> | |
488 | ||
489 | Don't scan for the found substrings. | |
490 | ||
491 | =item C<isall> | |
492 | ||
1c102323 | 493 | Means that the optimizer information is all that the regular |
055fd3a9 GS |
494 | expression contains, and thus one does not need to enter the regex engine at |
495 | all. | |
496 | ||
497 | =item C<GPOS> | |
498 | ||
499 | Set if the pattern contains C<\G>. | |
500 | ||
501 | =item C<plus> | |
502 | ||
503 | Set if the pattern starts with a repeated char (as in C<x+y>). | |
504 | ||
505 | =item C<implicit> | |
506 | ||
507 | Set if the pattern starts with C<.*>. | |
508 | ||
509 | =item C<with eval> | |
510 | ||
511 | Set if the pattern contain eval-groups, such as C<(?{ code })> and | |
512 | C<(??{ code })>. | |
513 | ||
514 | =item C<anchored(TYPE)> | |
515 | ||
7b406369 | 516 | If the pattern may match only at a handful of places, with C<TYPE> |
055fd3a9 GS |
517 | being C<BOL>, C<MBOL>, or C<GPOS>. See the table below. |
518 | ||
519 | =back | |
520 | ||
521 | If a substring is known to match at end-of-line only, it may be | |
ccf3535a | 522 | followed by C<$>, as in C<floating 'k'$>. |
055fd3a9 | 523 | |
1c102323 MJD |
524 | The optimizer-specific information is used to avoid entering (a slow) regex |
525 | engine on strings that will not definitely match. If the C<isall> flag | |
055fd3a9 GS |
526 | is set, a call to the regex engine may be avoided even when the optimizer |
527 | found an appropriate place for the match. | |
528 | ||
1c102323 | 529 | Above the optimizer section is the list of I<nodes> of the compiled |
055fd3a9 GS |
530 | form of the regex. Each line has format |
531 | ||
532 | C< >I<id>: I<TYPE> I<OPTIONAL-INFO> (I<next-id>) | |
533 | ||
7b406369 | 534 | =head2 Types of Nodes |
055fd3a9 GS |
535 | |
536 | Here are the possible types, with short descriptions: | |
537 | ||
5da6b59a KW |
538 | # TYPE arg-description [num-args] [longjump-len] DESCRIPTION |
539 | ||
540 | # Exit points | |
541 | END no End of program. | |
542 | SUCCEED no Return from a subroutine, basically. | |
543 | ||
544 | # Anchors: | |
545 | ||
546 | BOL no Match "" at beginning of line. | |
547 | MBOL no Same, assuming multiline. | |
548 | SBOL no Same, assuming singleline. | |
549 | EOS no Match "" at end of string. | |
550 | EOL no Match "" at end of line. | |
551 | MEOL no Same, assuming multiline. | |
552 | SEOL no Same, assuming singleline. | |
553 | BOUND no Match "" at any word boundary using native charset | |
554 | semantics for non-utf8 | |
555 | BOUNDL no Match "" at any locale word boundary | |
556 | BOUNDU no Match "" at any word boundary using Unicode semantics | |
557 | BOUNDA no Match "" at any word boundary using ASCII semantics | |
558 | NBOUND no Match "" at any word non-boundary using native charset | |
559 | semantics for non-utf8 | |
560 | NBOUNDL no Match "" at any locale word non-boundary | |
561 | NBOUNDU no Match "" at any word non-boundary using Unicode semantics | |
562 | NBOUNDA no Match "" at any word non-boundary using ASCII semantics | |
563 | GPOS no Matches where last m//g left off. | |
564 | ||
565 | # [Special] alternatives: | |
566 | ||
567 | REG_ANY no Match any one character (except newline). | |
568 | SANY no Match any one character. | |
569 | CANY no Match any one byte. | |
570 | ANYOF sv Match character in (or not in) this class, single char | |
571 | match only | |
572 | ANYOFV sv Match character in (or not in) this class, can | |
573 | match-multiple chars | |
574 | ALNUM no Match any alphanumeric character using native charset | |
575 | semantics for non-utf8 | |
576 | ALNUML no Match any alphanumeric char in locale | |
577 | ALNUMU no Match any alphanumeric char using Unicode semantics | |
578 | ALNUMA no Match [A-Za-z_0-9] | |
579 | NALNUM no Match any non-alphanumeric character using native charset | |
580 | semantics for non-utf8 | |
581 | NALNUML no Match any non-alphanumeric char in locale | |
582 | NALNUMU no Match any non-alphanumeric char using Unicode semantics | |
583 | NALNUMA no Match [^A-Za-z_0-9] | |
584 | SPACE no Match any whitespace character using native charset | |
585 | semantics for non-utf8 | |
586 | SPACEL no Match any whitespace char in locale | |
587 | SPACEU no Match any whitespace char using Unicode semantics | |
588 | SPACEA no Match [ \t\n\f\r] | |
589 | NSPACE no Match any non-whitespace character using native charset | |
590 | semantics for non-utf8 | |
591 | NSPACEL no Match any non-whitespace char in locale | |
592 | NSPACEU no Match any non-whitespace char using Unicode semantics | |
593 | NSPACEA no Match [^ \t\n\f\r] | |
594 | DIGIT no Match any numeric character using native charset semantics | |
595 | for non-utf8 | |
596 | DIGITL no Match any numeric character in locale | |
597 | DIGITA no Match [0-9] | |
598 | NDIGIT no Match any non-numeric character using native charset | |
599 | i semantics for non-utf8 | |
600 | NDIGITL no Match any non-numeric character in locale | |
601 | NDIGITA no Match [^0-9] | |
602 | CLUMP no Match any extended grapheme cluster sequence | |
603 | ||
604 | # Alternation | |
605 | ||
606 | # BRANCH The set of branches constituting a single choice are hooked | |
607 | # together with their "next" pointers, since precedence prevents | |
608 | # anything being concatenated to any individual branch. The | |
609 | # "next" pointer of the last BRANCH in a choice points to the | |
610 | # thing following the whole choice. This is also where the | |
611 | # final "next" pointer of each individual branch points; each | |
612 | # branch starts with the operand node of a BRANCH node. | |
613 | # | |
614 | BRANCH node Match this alternative, or the next... | |
615 | ||
616 | # Back pointer | |
617 | ||
618 | # BACK Normal "next" pointers all implicitly point forward; BACK | |
619 | # exists to make loop structures possible. | |
620 | # not used | |
621 | BACK no Match "", "next" ptr points backward. | |
622 | ||
623 | # Literals | |
624 | ||
625 | EXACT str Match this string (preceded by length). | |
626 | EXACTF str Match this string, folded, native charset semantics for | |
627 | non-utf8 (prec. by length). | |
628 | EXACTFL str Match this string, folded in locale (w/len). | |
629 | EXACTFU str Match this string, folded, Unicode semantics for non-utf8 | |
630 | (prec. by length). | |
631 | EXACTFA str Match this string, folded, Unicode semantics for non-utf8, | |
632 | but no ASCII-range character matches outside ASCII (prec. | |
633 | by length),. | |
634 | ||
635 | # Do nothing types | |
636 | ||
637 | NOTHING no Match empty string. | |
638 | # A variant of above which delimits a group, thus stops optimizations | |
639 | TAIL no Match empty string. Can jump here from outside. | |
640 | ||
641 | # Loops | |
642 | ||
643 | # STAR,PLUS '?', and complex '*' and '+', are implemented as circular | |
644 | # BRANCH structures using BACK. Simple cases (one character | |
645 | # per match) are implemented with STAR and PLUS for speed | |
646 | # and to minimize recursive plunges. | |
647 | # | |
648 | STAR node Match this (simple) thing 0 or more times. | |
649 | PLUS node Match this (simple) thing 1 or more times. | |
650 | ||
651 | CURLY sv 2 Match this simple thing {n,m} times. | |
652 | CURLYN no 2 Capture next-after-this simple thing | |
653 | CURLYM no 2 Capture this medium-complex thing {n,m} times. | |
654 | CURLYX sv 2 Match this complex thing {n,m} times. | |
655 | ||
656 | # This terminator creates a loop structure for CURLYX | |
657 | WHILEM no Do curly processing and see if rest matches. | |
658 | ||
659 | # Buffer related | |
660 | ||
661 | # OPEN,CLOSE,GROUPP ...are numbered at compile time. | |
662 | OPEN num 1 Mark this point in input as start of #n. | |
663 | CLOSE num 1 Analogous to OPEN. | |
664 | ||
665 | REF num 1 Match some already matched string | |
666 | REFF num 1 Match already matched string, folded using native charset | |
667 | semantics for non-utf8 | |
668 | REFFL num 1 Match already matched string, folded in loc. | |
669 | REFFU num 1 Match already matched string, folded using unicode | |
670 | semantics for non-utf8 | |
671 | REFFA num 1 Match already matched string, folded using unicode | |
672 | semantics for non-utf8, no mixing ASCII, non-ASCII | |
673 | ||
674 | # Named references. Code in regcomp.c assumes that these all are after the | |
675 | # numbered references | |
676 | NREF no-sv 1 Match some already matched string | |
677 | NREFF no-sv 1 Match already matched string, folded using native charset | |
678 | semantics for non-utf8 | |
679 | NREFFL no-sv 1 Match already matched string, folded in loc. | |
680 | NREFFU num 1 Match already matched string, folded using unicode | |
681 | semantics for non-utf8 | |
682 | NREFFA num 1 Match already matched string, folded using unicode | |
683 | semantics for non-utf8, no mixing ASCII, non-ASCII | |
684 | ||
685 | IFMATCH off 1 2 Succeeds if the following matches. | |
686 | UNLESSM off 1 2 Fails if the following matches. | |
687 | SUSPEND off 1 1 "Independent" sub-RE. | |
688 | IFTHEN off 1 1 Switch, should be preceded by switcher. | |
689 | GROUPP num 1 Whether the group matched. | |
690 | ||
691 | # Support for long RE | |
692 | ||
693 | LONGJMP off 1 1 Jump far away. | |
694 | BRANCHJ off 1 1 BRANCH with long offset. | |
695 | ||
696 | # The heavy worker | |
697 | ||
698 | EVAL evl 1 Execute some Perl code. | |
699 | ||
700 | # Modifiers | |
701 | ||
702 | MINMOD no Next operator is not greedy. | |
703 | LOGICAL no Next opcode should set the flag only. | |
704 | ||
705 | # This is not used yet | |
706 | RENUM off 1 1 Group with independently numbered parens. | |
707 | ||
708 | # Trie Related | |
709 | ||
710 | # Behave the same as A|LIST|OF|WORDS would. The '..C' variants have | |
711 | # inline charclass data (ascii only), the 'C' store it in the structure. | |
712 | # NOTE: the relative order of the TRIE-like regops is significant | |
713 | ||
714 | TRIE trie 1 Match many EXACT(F[ALU]?)? at once. flags==type | |
715 | TRIEC charclass Same as TRIE, but with embedded charclass data | |
716 | ||
717 | # For start classes, contains an added fail table. | |
718 | AHOCORASICK trie 1 Aho Corasick stclass. flags==type | |
719 | AHOCORASICKC charclass Same as AHOCORASICK, but with embedded charclass data | |
720 | ||
721 | # Regex Subroutines | |
722 | GOSUB num/ofs 2L recurse to paren arg1 at (signed) ofs arg2 | |
723 | GOSTART no recurse to start of pattern | |
724 | ||
725 | # Special conditionals | |
726 | NGROUPP no-sv 1 Whether the group matched. | |
727 | INSUBP num 1 Whether we are in a specific recurse. | |
728 | DEFINEP none 1 Never execute directly. | |
729 | ||
730 | # Backtracking Verbs | |
731 | ENDLIKE none Used only for the type field of verbs | |
732 | OPFAIL none Same as (?!) | |
733 | ACCEPT parno 1 Accepts the current matched string. | |
734 | ||
735 | ||
736 | # Verbs With Arguments | |
737 | VERB no-sv 1 Used only for the type field of verbs | |
738 | PRUNE no-sv 1 Pattern fails at this startpoint if no-backtracking through this | |
739 | MARKPOINT no-sv 1 Push the current location for rollback by cut. | |
740 | SKIP no-sv 1 On failure skip forward (to the mark) before retrying | |
741 | COMMIT no-sv 1 Pattern fails outright if backtracking through this | |
742 | CUTGROUP no-sv 1 On failure go to the next alternation in the group | |
743 | ||
744 | # Control what to keep in $&. | |
745 | KEEPS no $& begins here. | |
746 | ||
747 | # New charclass like patterns | |
748 | LNBREAK none generic newline pattern | |
749 | VERTWS none vertical whitespace (Perl 6) | |
750 | NVERTWS none not vertical whitespace (Perl 6) | |
751 | HORIZWS none horizontal whitespace (Perl 6) | |
752 | NHORIZWS none not horizontal whitespace (Perl 6) | |
753 | ||
754 | FOLDCHAR codepoint 1 codepoint with tricky case folding properties. | |
755 | ||
756 | # SPECIAL REGOPS | |
757 | ||
758 | # This is not really a node, but an optimized away piece of a "long" node. | |
759 | # To simplify debugging output, we mark it as if it were a node | |
760 | OPTIMIZED off Placeholder for dump. | |
761 | ||
762 | # Special opcode with the property that no opcode in a compiled program | |
763 | # will ever be of this type. Thus it can be used as a flag value that | |
764 | # no other opcode has been seen. END is used similarly, in that an END | |
765 | # node cant be optimized. So END implies "unoptimizable" and PSEUDO mean | |
766 | # "not seen anything to optimize yet". | |
767 | PSEUDO off Pseudo opcode for internal use. | |
055fd3a9 | 768 | |
1c102323 MJD |
769 | =for unprinted-credits |
770 | Next section M-J. Dominus (mjd-perl-patch+@plover.com) 20010421 | |
771 | ||
772 | Following the optimizer information is a dump of the offset/length | |
773 | table, here split across several lines: | |
774 | ||
775 | Offsets: [45] | |
776 | 1[4] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 5[1] | |
777 | 0[0] 12[1] 0[0] 6[1] 0[0] 7[1] 0[0] 9[1] 8[1] 0[0] 10[1] 0[0] | |
778 | 11[1] 0[0] 12[0] 12[0] 13[1] 0[0] 14[4] 0[0] 0[0] 0[0] 0[0] | |
779 | 0[0] 0[0] 0[0] 0[0] 0[0] 0[0] 18[1] 0[0] 19[1] 20[0] | |
780 | ||
781 | The first line here indicates that the offset/length table contains 45 | |
782 | entries. Each entry is a pair of integers, denoted by C<offset[length]>. | |
17c338f3 | 783 | Entries are numbered starting with 1, so entry #1 here is C<1[4]> and |
1c102323 MJD |
784 | entry #12 is C<5[1]>. C<1[4]> indicates that the node labeled C<1:> |
785 | (the C<1: ANYOF[bc]>) begins at character position 1 in the | |
786 | pre-compiled form of the regex, and has a length of 4 characters. | |
787 | C<5[1]> in position 12 | |
788 | indicates that the node labeled C<12:> | |
789 | (the C<< 12: EXACT <d> >>) begins at character position 5 in the | |
790 | pre-compiled form of the regex, and has a length of 1 character. | |
791 | C<12[1]> in position 14 | |
792 | indicates that the node labeled C<14:> | |
793 | (the C<< 14: CURLYX[0] {1,32767} >>) begins at character position 12 in the | |
794 | pre-compiled form of the regex, and has a length of 1 character---that | |
795 | is, it corresponds to the C<+> symbol in the precompiled regex. | |
796 | ||
797 | C<0[0]> items indicate that there is no corresponding node. | |
798 | ||
7b406369 | 799 | =head2 Run-time Output |
055fd3a9 GS |
800 | |
801 | First of all, when doing a match, one may get no run-time output even | |
802 | if debugging is enabled. This means that the regex engine was never | |
803 | entered and that all of the job was therefore done by the optimizer. | |
804 | ||
805 | If the regex engine was entered, the output may look like this: | |
806 | ||
ccf3535a | 807 | Matching '[bc]d(ef*g)+h[ij]k$' against 'abcdefg__gh__' |
055fd3a9 GS |
808 | Setting an EVAL scope, savestack=3 |
809 | 2 <ab> <cdefg__gh_> | 1: ANYOF | |
810 | 3 <abc> <defg__gh_> | 11: EXACT <d> | |
811 | 4 <abcd> <efg__gh_> | 13: CURLYX {1,32767} | |
812 | 4 <abcd> <efg__gh_> | 26: WHILEM | |
813 | 0 out of 1..32767 cc=effff31c | |
814 | 4 <abcd> <efg__gh_> | 15: OPEN1 | |
815 | 4 <abcd> <efg__gh_> | 17: EXACT <e> | |
816 | 5 <abcde> <fg__gh_> | 19: STAR | |
817 | EXACT <f> can match 1 times out of 32767... | |
818 | Setting an EVAL scope, savestack=3 | |
819 | 6 <bcdef> <g__gh__> | 22: EXACT <g> | |
820 | 7 <bcdefg> <__gh__> | 24: CLOSE1 | |
821 | 7 <bcdefg> <__gh__> | 26: WHILEM | |
822 | 1 out of 1..32767 cc=effff31c | |
823 | Setting an EVAL scope, savestack=12 | |
824 | 7 <bcdefg> <__gh__> | 15: OPEN1 | |
825 | 7 <bcdefg> <__gh__> | 17: EXACT <e> | |
826 | restoring \1 to 4(4)..7 | |
827 | failed, try continuation... | |
828 | 7 <bcdefg> <__gh__> | 27: NOTHING | |
829 | 7 <bcdefg> <__gh__> | 28: EXACT <h> | |
830 | failed... | |
831 | failed... | |
832 | ||
833 | The most significant information in the output is about the particular I<node> | |
834 | of the compiled regex that is currently being tested against the target string. | |
835 | The format of these lines is | |
836 | ||
837 | C< >I<STRING-OFFSET> <I<PRE-STRING>> <I<POST-STRING>> |I<ID>: I<TYPE> | |
838 | ||
839 | The I<TYPE> info is indented with respect to the backtracking level. | |
840 | Other incidental information appears interspersed within. | |
841 | ||
7b406369 | 842 | =head1 Debugging Perl Memory Usage |
055fd3a9 GS |
843 | |
844 | Perl is a profligate wastrel when it comes to memory use. There | |
845 | is a saying that to estimate memory usage of Perl, assume a reasonable | |
846 | algorithm for memory allocation, multiply that estimate by 10, and | |
847 | while you still may miss the mark, at least you won't be quite so | |
4375e838 | 848 | astonished. This is not absolutely true, but may provide a good |
055fd3a9 GS |
849 | grasp of what happens. |
850 | ||
851 | Assume that an integer cannot take less than 20 bytes of memory, a | |
852 | float cannot take less than 24 bytes, a string cannot take less | |
853 | than 32 bytes (all these examples assume 32-bit architectures, the | |
854 | result are quite a bit worse on 64-bit architectures). If a variable | |
855 | is accessed in two of three different ways (which require an integer, | |
856 | a float, or a string), the memory footprint may increase yet another | |
b9449ee0 | 857 | 20 bytes. A sloppy malloc(3) implementation can inflate these |
055fd3a9 GS |
858 | numbers dramatically. |
859 | ||
860 | On the opposite end of the scale, a declaration like | |
861 | ||
862 | sub foo; | |
863 | ||
864 | may take up to 500 bytes of memory, depending on which release of Perl | |
865 | you're running. | |
866 | ||
867 | Anecdotal estimates of source-to-compiled code bloat suggest an | |
868 | eightfold increase. This means that the compiled form of reasonable | |
869 | (normally commented, properly indented etc.) code will take | |
870 | about eight times more space in memory than the code took | |
871 | on disk. | |
872 | ||
b30f304a JH |
873 | The B<-DL> command-line switch is obsolete since circa Perl 5.6.0 |
874 | (it was available only if Perl was built with C<-DDEBUGGING>). | |
875 | The switch was used to track Perl's memory allocations and possible | |
876 | memory leaks. These days the use of malloc debugging tools like | |
5b6a3331 | 877 | F<Purify> or F<valgrind> is suggested instead. See also |
7b406369 | 878 | L<perlhacktips/PERL_MEM_LOG>. |
b30f304a JH |
879 | |
880 | One way to find out how much memory is being used by Perl data | |
881 | structures is to install the Devel::Size module from CPAN: it gives | |
882 | you the minimum number of bytes required to store a particular data | |
883 | structure. Please be mindful of the difference between the size() | |
884 | and total_size(). | |
885 | ||
886 | If Perl has been compiled using Perl's malloc you can analyze Perl | |
7b406369 | 887 | memory usage by setting $ENV{PERL_DEBUG_MSTATS}. |
055fd3a9 GS |
888 | |
889 | =head2 Using C<$ENV{PERL_DEBUG_MSTATS}> | |
890 | ||
891 | If your perl is using Perl's malloc() and was compiled with the | |
892 | necessary switches (this is the default), then it will print memory | |
4375e838 | 893 | usage statistics after compiling your code when C<< $ENV{PERL_DEBUG_MSTATS} |
055fd3a9 GS |
894 | > 1 >>, and before termination of the program when C<< |
895 | $ENV{PERL_DEBUG_MSTATS} >= 1 >>. The report format is similar to | |
896 | the following example: | |
897 | ||
898 | $ PERL_DEBUG_MSTATS=2 perl -e "require Carp" | |
899 | Memory allocation statistics after compilation: (buckets 4(4)..8188(8192) | |
900 | 14216 free: 130 117 28 7 9 0 2 2 1 0 0 | |
901 | 437 61 36 0 5 | |
902 | 60924 used: 125 137 161 55 7 8 6 16 2 0 1 | |
903 | 74 109 304 84 20 | |
904 | Total sbrk(): 77824/21:119. Odd ends: pad+heads+chain+tail: 0+636+0+2048. | |
905 | Memory allocation statistics after execution: (buckets 4(4)..8188(8192) | |
906 | 30888 free: 245 78 85 13 6 2 1 3 2 0 1 | |
907 | 315 162 39 42 11 | |
908 | 175816 used: 265 176 1112 111 26 22 11 27 2 1 1 | |
909 | 196 178 1066 798 39 | |
910 | Total sbrk(): 215040/47:145. Odd ends: pad+heads+chain+tail: 0+2192+0+6144. | |
911 | ||
912 | It is possible to ask for such a statistic at arbitrary points in | |
b9449ee0 | 913 | your execution using the mstat() function out of the standard |
055fd3a9 GS |
914 | Devel::Peek module. |
915 | ||
916 | Here is some explanation of that format: | |
917 | ||
13a2d996 | 918 | =over 4 |
055fd3a9 GS |
919 | |
920 | =item C<buckets SMALLEST(APPROX)..GREATEST(APPROX)> | |
921 | ||
922 | Perl's malloc() uses bucketed allocations. Every request is rounded | |
923 | up to the closest bucket size available, and a bucket is taken from | |
924 | the pool of buckets of that size. | |
925 | ||
926 | The line above describes the limits of buckets currently in use. | |
927 | Each bucket has two sizes: memory footprint and the maximal size | |
928 | of user data that can fit into this bucket. Suppose in the above | |
929 | example that the smallest bucket were size 4. The biggest bucket | |
930 | would have usable size 8188, and the memory footprint would be 8192. | |
931 | ||
932 | In a Perl built for debugging, some buckets may have negative usable | |
933 | size. This means that these buckets cannot (and will not) be used. | |
934 | For larger buckets, the memory footprint may be one page greater | |
7b406369 | 935 | than a power of 2. If so, the corresponding power of two is |
055fd3a9 GS |
936 | printed in the C<APPROX> field above. |
937 | ||
938 | =item Free/Used | |
939 | ||
940 | The 1 or 2 rows of numbers following that correspond to the number | |
941 | of buckets of each size between C<SMALLEST> and C<GREATEST>. In | |
942 | the first row, the sizes (memory footprints) of buckets are powers | |
943 | of two--or possibly one page greater. In the second row, if present, | |
944 | the memory footprints of the buckets are between the memory footprints | |
945 | of two buckets "above". | |
946 | ||
4375e838 | 947 | For example, suppose under the previous example, the memory footprints |
055fd3a9 GS |
948 | were |
949 | ||
950 | free: 8 16 32 64 128 256 512 1024 2048 4096 8192 | |
951 | 4 12 24 48 80 | |
952 | ||
7b406369 | 953 | With a non-C<DEBUGGING> perl, the buckets starting from C<128> have |
d1be9408 | 954 | a 4-byte overhead, and thus an 8192-long bucket may take up to |
055fd3a9 GS |
955 | 8188-byte allocations. |
956 | ||
957 | =item C<Total sbrk(): SBRKed/SBRKs:CONTINUOUS> | |
958 | ||
959 | The first two fields give the total amount of memory perl sbrk(2)ed | |
960 | (ess-broken? :-) and number of sbrk(2)s used. The third number is | |
961 | what perl thinks about continuity of returned chunks. So long as | |
962 | this number is positive, malloc() will assume that it is probable | |
963 | that sbrk(2) will provide continuous memory. | |
964 | ||
965 | Memory allocated by external libraries is not counted. | |
966 | ||
967 | =item C<pad: 0> | |
968 | ||
969 | The amount of sbrk(2)ed memory needed to keep buckets aligned. | |
970 | ||
971 | =item C<heads: 2192> | |
972 | ||
973 | Although memory overhead of bigger buckets is kept inside the bucket, for | |
974 | smaller buckets, it is kept in separate areas. This field gives the | |
975 | total size of these areas. | |
976 | ||
977 | =item C<chain: 0> | |
978 | ||
979 | malloc() may want to subdivide a bigger bucket into smaller buckets. | |
980 | If only a part of the deceased bucket is left unsubdivided, the rest | |
981 | is kept as an element of a linked list. This field gives the total | |
982 | size of these chunks. | |
983 | ||
984 | =item C<tail: 6144> | |
985 | ||
986 | To minimize the number of sbrk(2)s, malloc() asks for more memory. This | |
987 | field gives the size of the yet unused part, which is sbrk(2)ed, but | |
988 | never touched. | |
989 | ||
990 | =back | |
991 | ||
055fd3a9 GS |
992 | =head1 SEE ALSO |
993 | ||
994 | L<perldebug>, | |
995 | L<perlguts>, | |
996 | L<perlrun> | |
997 | L<re>, | |
998 | and | |
fe854a6f | 999 | L<Devel::DProf>. |