3 perlhack - How to hack at the Perl internals
7 This document attempts to explain how Perl development takes place,
8 and ends with some suggestions for people wanting to become bona fide
11 The perl5-porters mailing list is where the Perl standard distribution
12 is maintained and developed. The list can get anywhere from 10 to 150
13 messages a day, depending on the heatedness of the debate. Most days
14 there are two or three patches, extensions, features, or bugs being
17 A searchable archive of the list is at:
19 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
21 The list is also archived under the usenet group name
22 C<perl.porters-gw> at:
26 List subscribers (the porters themselves) come in several flavours.
27 Some are quiet curious lurkers, who rarely pitch in and instead watch
28 the ongoing development to ensure they're forewarned of new changes or
29 features in Perl. Some are representatives of vendors, who are there
30 to make sure that Perl continues to compile and work on their
31 platforms. Some patch any reported bug that they know how to fix,
32 some are actively patching their pet area (threads, Win32, the regexp
33 engine), while others seem to do nothing but complain. In other
34 words, it's your usual mix of technical people.
36 Over this group of porters presides Larry Wall. He has the final word
37 in what does and does not change in the Perl language. Various
38 releases of Perl are shepherded by a ``pumpking'', a porter
39 responsible for gathering patches, deciding on a patch-by-patch
40 feature-by-feature basis what will and will not go into the release.
41 For instance, Gurusamy Sarathy is the pumpking for the 5.6 release of
44 In addition, various people are pumpkings for different things. For
45 instance, Andy Dougherty and Jarkko Hietaniemi share the I<Configure>
46 pumpkin, and Tom Christiansen is the documentation pumpking.
48 Larry sees Perl development along the lines of the US government:
49 there's the Legislature (the porters), the Executive branch (the
50 pumpkings), and the Supreme Court (Larry). The legislature can
51 discuss and submit patches to the executive branch all they like, but
52 the executive branch is free to veto them. Rarely, the Supreme Court
53 will side with the executive branch over the legislature, or the
54 legislature over the executive branch. Mostly, however, the
55 legislature and the executive branch are supposed to get along and
56 work out their differences without impeachment or court cases.
58 You might sometimes see reference to Rule 1 and Rule 2. Larry's power
59 as Supreme Court is expressed in The Rules:
65 Larry is always by definition right about how Perl should behave.
66 This means he has final veto power on the core functionality.
70 Larry is allowed to change his mind about any matter at a later date,
71 regardless of whether he previously invoked Rule 1.
75 Got that? Larry is always right, even when he was wrong. It's rare
76 to see either Rule exercised, but they are often alluded to.
78 New features and extensions to the language are contentious, because
79 the criteria used by the pumpkings, Larry, and other porters to decide
80 which features should be implemented and incorporated are not codified
81 in a few small design goals as with some other languages. Instead,
82 the heuristics are flexible and often difficult to fathom. Here is
83 one person's list, roughly in decreasing order of importance, of
84 heuristics that new features have to be weighed against:
88 =item Does concept match the general goals of Perl?
90 These haven't been written anywhere in stone, but one approximation
93 1. Keep it fast, simple, and useful.
94 2. Keep features/concepts as orthogonal as possible.
95 3. No arbitrary limits (platforms, data sizes, cultures).
96 4. Keep it open and exciting to use/patch/advocate Perl everywhere.
97 5. Either assimilate new technologies, or build bridges to them.
99 =item Where is the implementation?
101 All the talk in the world is useless without an implementation. In
102 almost every case, the person or people who argue for a new feature
103 will be expected to be the ones who implement it. Porters capable
104 of coding new features have their own agendas, and are not available
105 to implement your (possibly good) idea.
107 =item Backwards compatibility
109 It's a cardinal sin to break existing Perl programs. New warnings are
110 contentious--some say that a program that emits warnings is not
111 broken, while others say it is. Adding keywords has the potential to
112 break programs, changing the meaning of existing token sequences or
113 functions might break programs.
115 =item Could it be a module instead?
117 Perl 5 has extension mechanisms, modules and XS, specifically to avoid
118 the need to keep changing the Perl interpreter. You can write modules
119 that export functions, you can give those functions prototypes so they
120 can be called like built-in functions, you can even write XS code to
121 mess with the runtime data structures of the Perl interpreter if you
122 want to implement really complicated things. If it can be done in a
123 module instead of in the core, it's highly unlikely to be added.
125 =item Is the feature generic enough?
127 Is this something that only the submitter wants added to the language,
128 or would it be broadly useful? Sometimes, instead of adding a feature
129 with a tight focus, the porters might decide to wait until someone
130 implements the more generalized feature. For instance, instead of
131 implementing a ``delayed evaluation'' feature, the porters are waiting
132 for a macro system that would permit delayed evaluation and much more.
134 =item Does it potentially introduce new bugs?
136 Radical rewrites of large chunks of the Perl interpreter have the
137 potential to introduce new bugs. The smaller and more localized the
140 =item Does it preclude other desirable features?
142 A patch is likely to be rejected if it closes off future avenues of
143 development. For instance, a patch that placed a true and final
144 interpretation on prototypes is likely to be rejected because there
145 are still options for the future of prototypes that haven't been
148 =item Is the implementation robust?
150 Good patches (tight code, complete, correct) stand more chance of
151 going in. Sloppy or incorrect patches might be placed on the back
152 burner until the pumpking has time to fix, or might be discarded
153 altogether without further notice.
155 =item Is the implementation generic enough to be portable?
157 The worst patches make use of a system-specific features. It's highly
158 unlikely that nonportable additions to the Perl language will be
161 =item Is there enough documentation?
163 Patches without documentation are probably ill-thought out or
164 incomplete. Nothing can be added without documentation, so submitting
165 a patch for the appropriate manpages as well as the source code is
166 always a good idea. If appropriate, patches should add to the test
169 =item Is there another way to do it?
171 Larry said ``Although the Perl Slogan is I<There's More Than One Way
172 to Do It>, I hesitate to make 10 ways to do something''. This is a
173 tricky heuristic to navigate, though--one man's essential addition is
174 another man's pointless cruft.
176 =item Does it create too much work?
178 Work for the pumpking, work for Perl programmers, work for module
179 authors, ... Perl is supposed to be easy.
181 =item Patches speak louder than words
183 Working code is always preferred to pie-in-the-sky ideas. A patch to
184 add a feature stands a much higher chance of making it to the language
185 than does a random feature request, no matter how fervently argued the
186 request might be. This ties into ``Will it be useful?'', as the fact
187 that someone took the time to make the patch demonstrates a strong
188 desire for the feature.
192 If you're on the list, you might hear the word ``core'' bandied
193 around. It refers to the standard distribution. ``Hacking on the
194 core'' means you're changing the C source code to the Perl
195 interpreter. ``A core module'' is one that ships with Perl.
197 =head2 Keeping in sync
199 The source code to the Perl interpreter, in its different versions, is
200 kept in a repository managed by a revision control system (which is
201 currently the Perforce program, see http://perforce.com/). The
202 pumpkings and a few others have access to the repository to check in
203 changes. Periodically the pumpking for the development version of Perl
204 will release a new version, so the rest of the porters can see what's
205 changed. The current state of the main trunk of repository, and patches
206 that describe the individual changes that have happened since the last
207 public release are available at this location:
209 ftp://ftp.linux.activestate.com/pub/staff/gsar/APC/
211 If you are a member of the perl5-porters mailing list, it is a good
212 thing to keep in touch with the most recent changes. If not only to
213 verify if what you would have posted as a bug report isn't already
214 solved in the most recent available perl development branch, also
215 known as perl-current, bleading edge perl, bleedperl or bleadperl.
217 Needless to say, the source code in perl-current is usually in a perpetual
218 state of evolution. You should expect it to be very buggy. Do B<not> use
219 it for any purpose other than testing and development.
221 Keeping in sync with the most recent branch can be done in several ways,
222 but the most convenient and reliable way is using B<rsync>, available at
223 ftp://rsync.samba.org/pub/rsync/ . (You can also get the most recent
226 If you choose to keep in sync using rsync, there are two approaches
231 =item rsync'ing the source tree
233 Presuming you are in the directory where your perl source resides
234 and you have rsync installed and available, you can `upgrade' to
237 # rsync -avz rsync://ftp.linux.activestate.com/perl-current/ .
239 This takes care of updating every single item in the source tree to
240 the latest applied patch level, creating files that are new (to your
241 distribution) and setting date/time stamps of existing files to
242 reflect the bleadperl status.
244 Note that this will not delete any files that were in '.' before
245 the rsync. Once you are sure that the rsync is running correctly,
246 run it with the --delete and the --dry-run options like this:
248 # rsync -avz --delete --dry-run rsync://ftp.linux.activestate.com/perl-current/ .
250 This will I<simulate> an rsync run that also deletes files not
251 present in the bleadperl master copy. Observe the results from
252 this run closely. If you are sure that the actual run would delete
253 no files precious to you, you could remove the '--dry-run' option.
255 You can than check what patch was the latest that was applied by
256 looking in the file B<.patch>, which will show the number of the
259 If you have more than one machine to keep in sync, and not all of
260 them have access to the WAN (so you are not able to rsync all the
261 source trees to the real source), there are some ways to get around
266 =item Using rsync over the LAN
268 Set up a local rsync server which makes the rsynced source tree
269 available to the LAN and sync the other machines against this
272 From http://rsync.samba.org/README.html:
274 "Rsync uses rsh or ssh for communication. It does not need to be
275 setuid and requires no special privileges for installation. It
276 does not require an inetd entry or a daemon. You must, however,
277 have a working rsh or ssh system. Using ssh is recommended for
278 its security features."
280 =item Using pushing over the NFS
282 Having the other systems mounted over the NFS, you can take an
283 active pushing approach by checking the just updated tree against
284 the other not-yet synced trees. An example would be
293 $1 => [ (stat $1)[2, 7, 9] ]; # mode, size, mtime
296 my %remote = map { $_ => "/$_/pro/3gl/CPAN/perl-5.7.1" } qw(host1 host2);
298 foreach my $host (keys %remote) {
299 unless (-d $remote{$host}) {
300 print STDERR "Cannot Xsync for host $host\n";
303 foreach my $file (keys %MF) {
304 my $rfile = "$remote{$host}/$file";
305 my ($mode, $size, $mtime) = (stat $rfile)[2, 7, 9];
306 defined $size or ($mode, $size, $mtime) = (0, 0, 0);
307 $size == $MF{$file}[1] && $mtime == $MF{$file}[2] and next;
308 printf "%4s %-34s %8d %9d %8d %9d\n",
309 $host, $file, $MF{$file}[1], $MF{$file}[2], $size, $mtime;
311 copy ($file, $rfile);
312 utime time, $MF{$file}[2], $rfile;
313 chmod $MF{$file}[0], $rfile;
317 though this is not perfect. It could be improved with checking
318 file checksums before updating. Not all NFS systems support
319 reliable utime support (when used over the NFS).
323 =item rsync'ing the patches
325 The source tree is maintained by the pumpking who applies patches to
326 the files in the tree. These patches are either created by the
327 pumpking himself using C<diff -c> after updating the file manually or
328 by applying patches sent in by posters on the perl5-porters list.
329 These patches are also saved and rsync'able, so you can apply them
330 yourself to the source files.
332 Presuming you are in a directory where your patches reside, you can
333 get them in sync with
335 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
337 This makes sure the latest available patch is downloaded to your
340 It's then up to you to apply these patches, using something like
342 # last=`ls -rt1 *.gz | tail -1`
343 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
344 # find . -name '*.gz' -newer $last -exec gzcat {} \; >blead.patch
346 # patch -p1 -N <../perl-current-diffs/blead.patch
348 or, since this is only a hint towards how it works, use CPAN-patchaperl
349 from Andreas König to have better control over the patching process.
353 =head2 Why rsync the source tree
357 =item It's easier to rsync the source tree
359 Since you don't have to apply the patches yourself, you are sure all
360 files in the source tree are in the right state.
362 =item It's more recent
364 According to Gurusamy Sarathy:
366 "... The rsync mirror is automatic and syncs with the repository
369 "Updating the patch area still requires manual intervention
370 (with all the goofiness that implies, which you've noted) and
371 is typically on a daily cycle. Making this process automatic
372 is on my tuit list, but don't ask me when."
374 =item It's more reliable
376 Well, since the patches are updated by hand, I don't have to say any
377 more ... (see Sarathy's remark).
381 =head2 Why rsync the patches
385 =item It's easier to rsync the patches
387 If you have more than one machine that you want to keep in track with
388 bleadperl, it's easier to rsync the patches only once and then apply
389 them to all the source trees on the different machines.
391 In case you try to keep in pace on 5 different machines, for which
392 only one of them has access to the WAN, rsync'ing all the source
393 trees should than be done 5 times over the NFS. Having
394 rsync'ed the patches only once, I can apply them to all the source
395 trees automatically. Need you say more ;-)
397 =item It's a good reference
399 If you do not only like to have the most recent development branch,
400 but also like to B<fix> bugs, or extend features, you want to dive
401 into the sources. If you are a seasoned perl core diver, you don't
402 need no manuals, tips, roadmaps, perlguts.pod or other aids to find
403 your way around. But if you are a starter, the patches may help you
404 in finding where you should start and how to change the bits that
407 The file B<Changes> is updated on occasions the pumpking sees as his
408 own little sync points. On those occasions, he releases a tar-ball of
409 the current source tree (i.e. perl@7582.tar.gz), which will be an
410 excellent point to start with when choosing to use the 'rsync the
411 patches' scheme. Starting with perl@7582, which means a set of source
412 files on which the latest applied patch is number 7582, you apply all
413 succeeding patches available from then on (7583, 7584, ...).
415 You can use the patches later as a kind of search archive.
419 =item Finding a start point
421 If you want to fix/change the behaviour of function/feature Foo, just
422 scan the patches for patches that mention Foo either in the subject,
423 the comments, or the body of the fix. A good chance the patch shows
424 you the files that are affected by that patch which are very likely
425 to be the starting point of your journey into the guts of perl.
427 =item Finding how to fix a bug
429 If you've found I<where> the function/feature Foo misbehaves, but you
430 don't know how to fix it (but you do know the change you want to
431 make), you can, again, peruse the patches for similar changes and
432 look how others apply the fix.
434 =item Finding the source of misbehaviour
436 When you keep in sync with bleadperl, the pumpking would love to
437 I<see> that the community efforts really work. So after each of his
438 sync points, you are to 'make test' to check if everything is still
439 in working order. If it is, you do 'make ok', which will send an OK
440 report to perlbug@perl.org. (If you do not have access to a mailer
441 from the system you just finished successfully 'make test', you can
442 do 'make okfile', which creates the file C<perl.ok>, which you can
443 than take to your favourite mailer and mail yourself).
445 But of course, as always, things will not always lead to a success
446 path, and one or more test do not pass the 'make test'. Before
447 sending in a bug report (using 'make nok' or 'make nokfile'), check
448 the mailing list if someone else has reported the bug already and if
449 so, confirm it by replying to that message. If not, you might want to
450 trace the source of that misbehaviour B<before> sending in the bug,
451 which will help all the other porters in finding the solution.
453 Here the saved patches come in very handy. You can check the list of
454 patches to see which patch changed what file and what change caused
455 the misbehaviour. If you note that in the bug report, it saves the
456 one trying to solve it, looking for that point.
460 If searching the patches is too bothersome, you might consider using
461 perl's bugtron to find more information about discussions and
462 ramblings on posted bugs.
466 If you want to get the best of both worlds, rsync both the source
467 tree for convenience, reliability and ease and rsync the patches
470 =head2 Submitting patches
472 Always submit patches to I<perl5-porters@perl.org>. If you're
473 patching a core module and there's an author listed, send the author a
474 copy (see L<Patching a core module>). This lets other porters review
475 your patch, which catches a surprising number of errors in patches.
476 Either use the diff program (available in source code form from
477 I<ftp://ftp.gnu.org/pub/gnu/>), or use Johan Vromans' I<makepatch>
478 (available from I<CPAN/authors/id/JV/>). Unified diffs are preferred,
479 but context diffs are accepted. Do not send RCS-style diffs or diffs
480 without context lines. More information is given in the
481 I<Porting/patching.pod> file in the Perl source distribution. Please
482 patch against the latest B<development> version (e.g., if you're
483 fixing a bug in the 5.005 track, patch against the latest 5.005_5x
484 version). Only patches that survive the heat of the development
485 branch get applied to maintenance versions.
487 Your patch should update the documentation and test suite. See
490 To report a bug in Perl, use the program I<perlbug> which comes with
491 Perl (if you can't get Perl to work, send mail to the address
492 I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through
493 I<perlbug> feeds into the automated bug-tracking system, access to
494 which is provided through the web at I<http://bugs.perl.org/>. It
495 often pays to check the archives of the perl5-porters mailing list to
496 see whether the bug you're reporting has been reported before, and if
497 so whether it was considered a bug. See above for the location of
498 the searchable archives.
500 The CPAN testers (I<http://testers.cpan.org/>) are a group of
501 volunteers who test CPAN modules on a variety of platforms. Perl Labs
502 (I<http://labs.perl.org/>) automatically tests Perl source releases on
503 platforms and gives feedback to the CPAN testers mailing list. Both
504 efforts welcome volunteers.
506 It's a good idea to read and lurk for a while before chipping in.
507 That way you'll get to see the dynamic of the conversations, learn the
508 personalities of the players, and hopefully be better prepared to make
509 a useful contribution when do you speak up.
511 If after all this you still think you want to join the perl5-porters
512 mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To
513 unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>.
515 To hack on the Perl guts, you'll need to read the following things:
521 This is of paramount importance, since it's the documentation of what
522 goes where in the Perl source. Read it over a couple of times and it
523 might start to make sense - don't worry if it doesn't yet, because the
524 best way to study it is to read it in conjunction with poking at Perl
525 source, and we'll do that later on.
527 You might also want to look at Gisle Aas's illustrated perlguts -
528 there's no guarantee that this will be absolutely up-to-date with the
529 latest documentation in the Perl core, but the fundamentals will be
530 right. (http://gisle.aas.no/perl/illguts/)
532 =item L<perlxstut> and L<perlxs>
534 A working knowledge of XSUB programming is incredibly useful for core
535 hacking; XSUBs use techniques drawn from the PP code, the portion of the
536 guts that actually executes a Perl program. It's a lot gentler to learn
537 those techniques from simple examples and explanation than from the core
542 The documentation for the Perl API explains what some of the internal
543 functions do, as well as the many macros used in the source.
545 =item F<Porting/pumpkin.pod>
547 This is a collection of words of wisdom for a Perl porter; some of it is
548 only useful to the pumpkin holder, but most of it applies to anyone
549 wanting to go about Perl development.
551 =item The perl5-porters FAQ
553 This is posted to perl5-porters at the beginning on every month, and
554 should be available from http://perlhacker.org/p5p-faq; alternatively,
555 you can get the FAQ emailed to you by sending mail to
556 C<perl5-porters-faq@perl.org>. It contains hints on reading
557 perl5-porters, information on how perl5-porters works and how Perl
558 development in general works.
562 =head2 Finding Your Way Around
564 Perl maintenance can be split into a number of areas, and certain people
565 (pumpkins) will have responsibility for each area. These areas sometimes
566 correspond to files or directories in the source kit. Among the areas are:
572 Modules shipped as part of the Perl core live in the F<lib/> and F<ext/>
573 subdirectories: F<lib/> is for the pure-Perl modules, and F<ext/>
574 contains the core XS modules.
578 There are tests for nearly all the modules, built-ins and major bits
579 of functionality. Test files all have a .t suffix. Module tests live
580 in the F<lib/> and F<ext/> directories next to the module being
581 tested. Others live in F<t/>. See L<Writing a test>
585 Documentation maintenance includes looking after everything in the
586 F<pod/> directory, (as well as contributing new documentation) and
587 the documentation to the modules in core.
591 The configure process is the way we make Perl portable across the
592 myriad of operating systems it supports. Responsibility for the
593 configure, build and installation process, as well as the overall
594 portability of the core code rests with the configure pumpkin - others
595 help out with individual operating systems.
597 The files involved are the operating system directories, (F<win32/>,
598 F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
599 and F<Makefile>, as well as the metaconfig files which generate
600 F<Configure>. (metaconfig isn't included in the core distribution.)
604 And of course, there's the core of the Perl interpreter itself. Let's
605 have a look at that in a little more detail.
609 Before we leave looking at the layout, though, don't forget that
610 F<MANIFEST> contains not only the file names in the Perl distribution,
611 but short descriptions of what's in them, too. For an overview of the
612 important files, try this:
614 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
616 =head2 Elements of the interpreter
618 The work of the interpreter has two main stages: compiling the code
619 into the internal representation, or bytecode, and then executing it.
620 L<perlguts/Compiled code> explains exactly how the compilation stage
623 Here is a short breakdown of perl's operation:
629 The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
630 This is very high-level code, enough to fit on a single screen, and it
631 resembles the code found in L<perlembed>; most of the real action takes
634 First, F<perlmain.c> allocates some memory and constructs a Perl
637 1 PERL_SYS_INIT3(&argc,&argv,&env);
639 3 if (!PL_do_undump) {
640 4 my_perl = perl_alloc();
643 7 perl_construct(my_perl);
644 8 PL_perl_destruct_level = 0;
647 Line 1 is a macro, and its definition is dependent on your operating
648 system. Line 3 references C<PL_do_undump>, a global variable - all
649 global variables in Perl start with C<PL_>. This tells you whether the
650 current running program was created with the C<-u> flag to perl and then
651 F<undump>, which means it's going to be false in any sane context.
653 Line 4 calls a function in F<perl.c> to allocate memory for a Perl
654 interpreter. It's quite a simple function, and the guts of it looks like
657 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
659 Here you see an example of Perl's system abstraction, which we'll see
660 later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
661 own C<malloc> as defined in F<malloc.c> if you selected that option at
664 Next, in line 7, we construct the interpreter; this sets up all the
665 special variables that Perl needs, the stacks, and so on.
667 Now we pass Perl the command line options, and tell it to go:
669 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
671 exitstatus = perl_run(my_perl);
675 C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
676 in F<perl.c>, which processes the command line options, sets up any
677 statically linked XS modules, opens the program and calls C<yyparse> to
682 The aim of this stage is to take the Perl source, and turn it into an op
683 tree. We'll see what one of those looks like later. Strictly speaking,
684 there's three things going on here.
686 C<yyparse>, the parser, lives in F<perly.c>, although you're better off
687 reading the original YACC input in F<perly.y>. (Yes, Virginia, there
688 B<is> a YACC grammar for Perl!) The job of the parser is to take your
689 code and `understand' it, splitting it into sentences, deciding which
690 operands go with which operators and so on.
692 The parser is nobly assisted by the lexer, which chunks up your input
693 into tokens, and decides what type of thing each token is: a variable
694 name, an operator, a bareword, a subroutine, a core function, and so on.
695 The main point of entry to the lexer is C<yylex>, and that and its
696 associated routines can be found in F<toke.c>. Perl isn't much like
697 other computer languages; it's highly context sensitive at times, it can
698 be tricky to work out what sort of token something is, or where a token
699 ends. As such, there's a lot of interplay between the tokeniser and the
700 parser, which can get pretty frightening if you're not used to it.
702 As the parser understands a Perl program, it builds up a tree of
703 operations for the interpreter to perform during execution. The routines
704 which construct and link together the various operations are to be found
705 in F<op.c>, and will be examined later.
709 Now the parsing stage is complete, and the finished tree represents
710 the operations that the Perl interpreter needs to perform to execute our
711 program. Next, Perl does a dry run over the tree looking for
712 optimisations: constant expressions such as C<3 + 4> will be computed
713 now, and the optimizer will also see if any multiple operations can be
714 replaced with a single one. For instance, to fetch the variable C<$foo>,
715 instead of grabbing the glob C<*foo> and looking at the scalar
716 component, the optimizer fiddles the op tree to use a function which
717 directly looks up the scalar in question. The main optimizer is C<peep>
718 in F<op.c>, and many ops have their own optimizing functions.
722 Now we're finally ready to go: we have compiled Perl byte code, and all
723 that's left to do is run it. The actual execution is done by the
724 C<runops_standard> function in F<run.c>; more specifically, it's done by
725 these three innocent looking lines:
727 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
731 You may be more comfortable with the Perl version of that:
733 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
735 Well, maybe not. Anyway, each op contains a function pointer, which
736 stipulates the function which will actually carry out the operation.
737 This function will return the next op in the sequence - this allows for
738 things like C<if> which choose the next op dynamically at run time.
739 The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
740 execution if required.
742 The actual functions called are known as PP code, and they're spread
743 between four files: F<pp_hot.c> contains the `hot' code, which is most
744 often used and highly optimized, F<pp_sys.c> contains all the
745 system-specific functions, F<pp_ctl.c> contains the functions which
746 implement control structures (C<if>, C<while> and the like) and F<pp.c>
747 contains everything else. These are, if you like, the C code for Perl's
748 built-in functions and operators.
752 =head2 Internal Variable Types
754 You should by now have had a look at L<perlguts>, which tells you about
755 Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
758 These variables are used not only to represent Perl-space variables, but
759 also any constants in the code, as well as some structures completely
760 internal to Perl. The symbol table, for instance, is an ordinary Perl
761 hash. Your code is represented by an SV as it's read into the parser;
762 any program files you call are opened via ordinary Perl filehandles, and
765 The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
766 Perl program. Let's see, for instance, how Perl treats the constant
769 % perl -MDevel::Peek -e 'Dump("hello")'
770 1 SV = PV(0xa041450) at 0xa04ecbc
772 3 FLAGS = (POK,READONLY,pPOK)
773 4 PV = 0xa0484e0 "hello"\0
777 Reading C<Devel::Peek> output takes a bit of practise, so let's go
778 through it line by line.
780 Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
781 memory. SVs themselves are very simple structures, but they contain a
782 pointer to a more complex structure. In this case, it's a PV, a
783 structure which holds a string value, at location C<0xa041450>. Line 2
784 is the reference count; there are no other references to this data, so
787 Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
788 read-only SV (because it's a constant) and the data is a PV internally.
789 Next we've got the contents of the string, starting at location
792 Line 5 gives us the current length of the string - note that this does
793 B<not> include the null terminator. Line 6 is not the length of the
794 string, but the length of the currently allocated buffer; as the string
795 grows, Perl automatically extends the available storage via a routine
798 You can get at any of these quantities from C very easily; just add
799 C<Sv> to the name of the field shown in the snippet, and you've got a
800 macro which will return the value: C<SvCUR(sv)> returns the current
801 length of the string, C<SvREFCOUNT(sv)> returns the reference count,
802 C<SvPV(sv, len)> returns the string itself with its length, and so on.
803 More macros to manipulate these properties can be found in L<perlguts>.
805 Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
808 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
813 6 junk = SvPV_force(sv, tlen);
814 7 SvGROW(sv, tlen + len + 1);
817 10 Move(ptr,SvPVX(sv)+tlen,len,char);
819 12 *SvEND(sv) = '\0';
820 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
824 This is a function which adds a string, C<ptr>, of length C<len> onto
825 the end of the PV stored in C<sv>. The first thing we do in line 6 is
826 make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
827 macro to force a PV. As a side effect, C<tlen> gets set to the current
828 value of the PV, and the PV itself is returned to C<junk>.
830 In line 7, we make sure that the SV will have enough room to accommodate
831 the old string, the new string and the null terminator. If C<LEN> isn't
832 big enough, C<SvGROW> will reallocate space for us.
834 Now, if C<junk> is the same as the string we're trying to add, we can
835 grab the string directly from the SV; C<SvPVX> is the address of the PV
838 Line 10 does the actual catenation: the C<Move> macro moves a chunk of
839 memory around: we move the string C<ptr> to the end of the PV - that's
840 the start of the PV plus its current length. We're moving C<len> bytes
841 of type C<char>. After doing so, we need to tell Perl we've extended the
842 string, by altering C<CUR> to reflect the new length. C<SvEND> is a
843 macro which gives us the end of the string, so that needs to be a
846 Line 13 manipulates the flags; since we've changed the PV, any IV or NV
847 values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
848 want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF8-aware
849 version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
850 and turns on POK. The final C<SvTAINT> is a macro which launders tainted
851 data if taint mode is turned on.
853 AVs and HVs are more complicated, but SVs are by far the most common
854 variable type being thrown around. Having seen something of how we
855 manipulate these, let's go on and look at how the op tree is
860 First, what is the op tree, anyway? The op tree is the parsed
861 representation of your program, as we saw in our section on parsing, and
862 it's the sequence of operations that Perl goes through to execute your
863 program, as we saw in L</Running>.
865 An op is a fundamental operation that Perl can perform: all the built-in
866 functions and operators are ops, and there are a series of ops which
867 deal with concepts the interpreter needs internally - entering and
868 leaving a block, ending a statement, fetching a variable, and so on.
870 The op tree is connected in two ways: you can imagine that there are two
871 "routes" through it, two orders in which you can traverse the tree.
872 First, parse order reflects how the parser understood the code, and
873 secondly, execution order tells perl what order to perform the
876 The easiest way to examine the op tree is to stop Perl after it has
877 finished parsing, and get it to dump out the tree. This is exactly what
878 the compiler backends L<B::Terse|B::Terse> and L<B::Debug|B::Debug> do.
880 Let's have a look at how Perl sees C<$a = $b + $c>:
882 % perl -MO=Terse -e '$a=$b+$c'
883 1 LISTOP (0x8179888) leave
884 2 OP (0x81798b0) enter
885 3 COP (0x8179850) nextstate
886 4 BINOP (0x8179828) sassign
887 5 BINOP (0x8179800) add [1]
888 6 UNOP (0x81796e0) null [15]
889 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
890 8 UNOP (0x81797e0) null [15]
891 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
892 10 UNOP (0x816b4f0) null [15]
893 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
895 Let's start in the middle, at line 4. This is a BINOP, a binary
896 operator, which is at location C<0x8179828>. The specific operator in
897 question is C<sassign> - scalar assignment - and you can find the code
898 which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
899 binary operator, it has two children: the add operator, providing the
900 result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
903 Line 10 is the null op: this does exactly nothing. What is that doing
904 there? If you see the null op, it's a sign that something has been
905 optimized away after parsing. As we mentioned in L</Optimization>,
906 the optimization stage sometimes converts two operations into one, for
907 example when fetching a scalar variable. When this happens, instead of
908 rewriting the op tree and cleaning up the dangling pointers, it's easier
909 just to replace the redundant operation with the null op. Originally,
910 the tree would have looked like this:
912 10 SVOP (0x816b4f0) rv2sv [15]
913 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
915 That is, fetch the C<a> entry from the main symbol table, and then look
916 at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
917 happens to do both these things.
919 The right hand side, starting at line 5 is similar to what we've just
920 seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
923 Now, what's this about?
925 1 LISTOP (0x8179888) leave
926 2 OP (0x81798b0) enter
927 3 COP (0x8179850) nextstate
929 C<enter> and C<leave> are scoping ops, and their job is to perform any
930 housekeeping every time you enter and leave a block: lexical variables
931 are tidied up, unreferenced variables are destroyed, and so on. Every
932 program will have those first three lines: C<leave> is a list, and its
933 children are all the statements in the block. Statements are delimited
934 by C<nextstate>, so a block is a collection of C<nextstate> ops, with
935 the ops to be performed for each statement being the children of
936 C<nextstate>. C<enter> is a single op which functions as a marker.
938 That's how Perl parsed the program, from top to bottom:
951 However, it's impossible to B<perform> the operations in this order:
952 you have to find the values of C<$b> and C<$c> before you add them
953 together, for instance. So, the other thread that runs through the op
954 tree is the execution order: each op has a field C<op_next> which points
955 to the next op to be run, so following these pointers tells us how perl
956 executes the code. We can traverse the tree in this order using
957 the C<exec> option to C<B::Terse>:
959 % perl -MO=Terse,exec -e '$a=$b+$c'
960 1 OP (0x8179928) enter
961 2 COP (0x81798c8) nextstate
962 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
963 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
964 5 BINOP (0x8179878) add [1]
965 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
966 7 BINOP (0x81798a0) sassign
967 8 LISTOP (0x8179900) leave
969 This probably makes more sense for a human: enter a block, start a
970 statement. Get the values of C<$b> and C<$c>, and add them together.
971 Find C<$a>, and assign one to the other. Then leave.
973 The way Perl builds up these op trees in the parsing process can be
974 unravelled by examining F<perly.y>, the YACC grammar. Let's take the
975 piece we need to construct the tree for C<$a = $b + $c>
977 1 term : term ASSIGNOP term
978 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
980 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
982 If you're not used to reading BNF grammars, this is how it works: You're
983 fed certain things by the tokeniser, which generally end up in upper
984 case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
985 code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
986 `terminal symbols', because you can't get any simpler than them.
988 The grammar, lines one and three of the snippet above, tells you how to
989 build up more complex forms. These complex forms, `non-terminal symbols'
990 are generally placed in lower case. C<term> here is a non-terminal
991 symbol, representing a single expression.
993 The grammar gives you the following rule: you can make the thing on the
994 left of the colon if you see all the things on the right in sequence.
995 This is called a "reduction", and the aim of parsing is to completely
996 reduce the input. There are several different ways you can perform a
997 reduction, separated by vertical bars: so, C<term> followed by C<=>
998 followed by C<term> makes a C<term>, and C<term> followed by C<+>
999 followed by C<term> can also make a C<term>.
1001 So, if you see two terms with an C<=> or C<+>, between them, you can
1002 turn them into a single expression. When you do this, you execute the
1003 code in the block on the next line: if you see C<=>, you'll do the code
1004 in line 2. If you see C<+>, you'll do the code in line 4. It's this code
1005 which contributes to the op tree.
1008 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1010 What this does is creates a new binary op, and feeds it a number of
1011 variables. The variables refer to the tokens: C<$1> is the first token in
1012 the input, C<$2> the second, and so on - think regular expression
1013 backreferences. C<$$> is the op returned from this reduction. So, we
1014 call C<newBINOP> to create a new binary operator. The first parameter to
1015 C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
1016 operator, so we want the type to be C<ADDOP>. We could specify this
1017 directly, but it's right there as the second token in the input, so we
1018 use C<$2>. The second parameter is the op's flags: 0 means `nothing
1019 special'. Then the things to add: the left and right hand side of our
1020 expression, in scalar context.
1024 When perl executes something like C<addop>, how does it pass on its
1025 results to the next op? The answer is, through the use of stacks. Perl
1026 has a number of stacks to store things it's currently working on, and
1027 we'll look at the three most important ones here.
1031 =item Argument stack
1033 Arguments are passed to PP code and returned from PP code using the
1034 argument stack, C<ST>. The typical way to handle arguments is to pop
1035 them off the stack, deal with them how you wish, and then push the result
1036 back onto the stack. This is how, for instance, the cosine operator
1041 value = Perl_cos(value);
1044 We'll see a more tricky example of this when we consider Perl's macros
1045 below. C<POPn> gives you the NV (floating point value) of the top SV on
1046 the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
1047 the result back as an NV. The C<X> in C<XPUSHn> means that the stack
1048 should be extended if necessary - it can't be necessary here, because we
1049 know there's room for one more item on the stack, since we've just
1050 removed one! The C<XPUSH*> macros at least guarantee safety.
1052 Alternatively, you can fiddle with the stack directly: C<SP> gives you
1053 the first element in your portion of the stack, and C<TOP*> gives you
1054 the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
1055 negation of an integer:
1059 Just set the integer value of the top stack entry to its negation.
1061 Argument stack manipulation in the core is exactly the same as it is in
1062 XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
1063 description of the macros used in stack manipulation.
1067 I say `your portion of the stack' above because PP code doesn't
1068 necessarily get the whole stack to itself: if your function calls
1069 another function, you'll only want to expose the arguments aimed for the
1070 called function, and not (necessarily) let it get at your own data. The
1071 way we do this is to have a `virtual' bottom-of-stack, exposed to each
1072 function. The mark stack keeps bookmarks to locations in the argument
1073 stack usable by each function. For instance, when dealing with a tied
1074 variable, (internally, something with `P' magic) Perl has to call
1075 methods for accesses to the tied variables. However, we need to separate
1076 the arguments exposed to the method to the argument exposed to the
1077 original function - the store or fetch or whatever it may be. Here's how
1078 the tied C<push> is implemented; see C<av_push> in F<av.c>:
1082 3 PUSHs(SvTIED_obj((SV*)av, mg));
1086 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1090 The lines which concern the mark stack are the first, fifth and last
1091 lines: they save away, restore and remove the current position of the
1094 Let's examine the whole implementation, for practice:
1098 Push the current state of the stack pointer onto the mark stack. This is
1099 so that when we've finished adding items to the argument stack, Perl
1100 knows how many things we've added recently.
1103 3 PUSHs(SvTIED_obj((SV*)av, mg));
1106 We're going to add two more items onto the argument stack: when you have
1107 a tied array, the C<PUSH> subroutine receives the object and the value
1108 to be pushed, and that's exactly what we have here - the tied object,
1109 retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
1113 Next we tell Perl to make the change to the global stack pointer: C<dSP>
1114 only gave us a local copy, not a reference to the global.
1117 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1120 C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
1121 variables are tidied up, everything that has been localised gets
1122 its previous value returned, and so on. Think of them as the C<{> and
1123 C<}> of a Perl block.
1125 To actually do the magic method call, we have to call a subroutine in
1126 Perl space: C<call_method> takes care of that, and it's described in
1127 L<perlcall>. We call the C<PUSH> method in scalar context, and we're
1128 going to discard its return value.
1132 Finally, we remove the value we placed on the mark stack, since we
1133 don't need it any more.
1137 C doesn't have a concept of local scope, so perl provides one. We've
1138 seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
1139 stack implements the C equivalent of, for example:
1146 See L<perlguts/Localising Changes> for how to use the save stack.
1150 =head2 Millions of Macros
1152 One thing you'll notice about the Perl source is that it's full of
1153 macros. Some have called the pervasive use of macros the hardest thing
1154 to understand, others find it adds to clarity. Let's take an example,
1155 the code which implements the addition operator:
1159 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
1162 6 SETn( left + right );
1167 Every line here (apart from the braces, of course) contains a macro. The
1168 first line sets up the function declaration as Perl expects for PP code;
1169 line 3 sets up variable declarations for the argument stack and the
1170 target, the return value of the operation. Finally, it tries to see if
1171 the addition operation is overloaded; if so, the appropriate subroutine
1174 Line 5 is another variable declaration - all variable declarations start
1175 with C<d> - which pops from the top of the argument stack two NVs (hence
1176 C<nn>) and puts them into the variables C<right> and C<left>, hence the
1177 C<rl>. These are the two operands to the addition operator. Next, we
1178 call C<SETn> to set the NV of the return value to the result of adding
1179 the two values. This done, we return - the C<RETURN> macro makes sure
1180 that our return value is properly handled, and we pass the next operator
1181 to run back to the main run loop.
1183 Most of these macros are explained in L<perlapi>, and some of the more
1184 important ones are explained in L<perlxs> as well. Pay special attention
1185 to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
1186 the C<[pad]THX_?> macros.
1189 =head2 Poking at Perl
1191 To really poke around with Perl, you'll probably want to build Perl for
1192 debugging, like this:
1194 ./Configure -d -D optimize=-g
1197 C<-g> is a flag to the C compiler to have it produce debugging
1198 information which will allow us to step through a running program.
1199 F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
1200 enables all the internal debugging code in Perl. There are a whole bunch
1201 of things you can debug with this: L<perlrun> lists them all, and the
1202 best way to find out about them is to play about with them. The most
1203 useful options are probably
1205 l Context (loop) stack processing
1207 o Method and overloading resolution
1208 c String/numeric conversions
1210 Some of the functionality of the debugging code can be achieved using XS
1213 -Dr => use re 'debug'
1214 -Dx => use O 'Debug'
1216 =head2 Using a source-level debugger
1218 If the debugging output of C<-D> doesn't help you, it's time to step
1219 through perl's execution with a source-level debugger.
1225 We'll use C<gdb> for our examples here; the principles will apply to any
1226 debugger, but check the manual of the one you're using.
1230 To fire up the debugger, type
1234 You'll want to do that in your Perl source tree so the debugger can read
1235 the source code. You should see the copyright message, followed by the
1240 C<help> will get you into the documentation, but here are the most
1247 Run the program with the given arguments.
1249 =item break function_name
1251 =item break source.c:xxx
1253 Tells the debugger that we'll want to pause execution when we reach
1254 either the named function (but see L<perlguts/Internal Functions>!) or the given
1255 line in the named source file.
1259 Steps through the program a line at a time.
1263 Steps through the program a line at a time, without descending into
1268 Run until the next breakpoint.
1272 Run until the end of the current function, then stop again.
1276 Just pressing Enter will do the most recent operation again - it's a
1277 blessing when stepping through miles of source code.
1281 Execute the given C code and print its results. B<WARNING>: Perl makes
1282 heavy use of macros, and F<gdb> is not aware of macros. You'll have to
1283 substitute them yourself. So, for instance, you can't say
1285 print SvPV_nolen(sv)
1289 print Perl_sv_2pv_nolen(sv)
1291 You may find it helpful to have a "macro dictionary", which you can
1292 produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
1293 recursively apply the macros for you.
1297 =head2 Dumping Perl Data Structures
1299 One way to get around this macro hell is to use the dumping functions in
1300 F<dump.c>; these work a little like an internal
1301 L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
1302 that you can't get at from Perl. Let's take an example. We'll use the
1303 C<$a = $b + $c> we used before, but give it a bit of context:
1304 C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
1306 What about C<pp_add>, the function we examined earlier to implement the
1309 (gdb) break Perl_pp_add
1310 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
1312 Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>.
1313 With the breakpoint in place, we can run our program:
1315 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
1317 Lots of junk will go past as gdb reads in the relevant source files and
1318 libraries, and then:
1320 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
1321 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
1326 We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
1327 arranges for two C<NV>s to be placed into C<left> and C<right> - let's
1330 #define dPOPTOPnnrl_ul NV right = POPn; \
1331 SV *leftsv = TOPs; \
1332 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
1334 C<POPn> takes the SV from the top of the stack and obtains its NV either
1335 directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
1336 C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
1337 C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
1338 C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
1340 Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
1341 convert it. If we step again, we'll find ourselves there:
1343 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1347 We can now use C<Perl_sv_dump> to investigate the SV:
1349 SV = PV(0xa057cc0) at 0xa0675d0
1352 PV = 0xa06a510 "6XXXX"\0
1357 We know we're going to get C<6> from this, so let's finish the
1361 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
1362 0x462669 in Perl_pp_add () at pp_hot.c:311
1365 We can also dump out this op: the current op is always stored in
1366 C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
1367 similar output to L<B::Debug|B::Debug>.
1370 13 TYPE = add ===> 14
1372 FLAGS = (SCALAR,KIDS)
1374 TYPE = null ===> (12)
1376 FLAGS = (SCALAR,KIDS)
1378 11 TYPE = gvsv ===> 12
1384 # finish this later #
1388 All right, we've now had a look at how to navigate the Perl sources and
1389 some things you'll need to know when fiddling with them. Let's now get
1390 on and create a simple patch. Here's something Larry suggested: if a
1391 C<U> is the first active format during a C<pack>, (for example,
1392 C<pack "U3C8", @stuff>) then the resulting string should be treated as
1395 How do we prepare to fix this up? First we locate the code in question -
1396 the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1397 files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1398 altering this file, let's copy it to F<pp.c~>.
1400 [Well, it was in F<pp.c> when this tutorial was written. It has now been
1401 split off with C<pp_unpack> to its own file, F<pp_pack.c>]
1403 Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1404 loop over the pattern, taking each format character in turn into
1405 C<datum_type>. Then for each possible format character, we swallow up
1406 the other arguments in the pattern (a field width, an asterisk, and so
1407 on) and convert the next chunk input into the specified format, adding
1408 it onto the output SV C<cat>.
1410 How do we know if the C<U> is the first format in the C<pat>? Well, if
1411 we have a pointer to the start of C<pat> then, if we see a C<U> we can
1412 test whether we're still at the start of the string. So, here's where
1416 register char *pat = SvPVx(*++MARK, fromlen);
1417 register char *patend = pat + fromlen;
1422 We'll have another string pointer in there:
1425 register char *pat = SvPVx(*++MARK, fromlen);
1426 register char *patend = pat + fromlen;
1432 And just before we start the loop, we'll set C<patcopy> to be the start
1437 sv_setpvn(cat, "", 0);
1439 while (pat < patend) {
1441 Now if we see a C<U> which was at the start of the string, we turn on
1442 the UTF8 flag for the output SV, C<cat>:
1444 + if (datumtype == 'U' && pat==patcopy+1)
1446 if (datumtype == '#') {
1447 while (pat < patend && *pat != '\n')
1450 Remember that it has to be C<patcopy+1> because the first character of
1451 the string is the C<U> which has been swallowed into C<datumtype!>
1453 Oops, we forgot one thing: what if there are spaces at the start of the
1454 pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1455 character, even though it's not the first thing in the pattern. In this
1456 case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1458 if (isSPACE(datumtype))
1463 if (isSPACE(datumtype)) {
1468 OK. That's the C part done. Now we must do two additional things before
1469 this patch is ready to go: we've changed the behaviour of Perl, and so
1470 we must document that change. We must also provide some more regression
1471 tests to make sure our patch works and doesn't create a bug somewhere
1472 else along the line.
1474 The regression tests for each operator live in F<t/op/>, and so we
1475 make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our
1476 tests to the end. First, we'll test that the C<U> does indeed create
1479 t/op/pack.t has a sensible ok() function, but if it didn't we could
1485 print "not " unless $ok;
1493 print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000);
1494 print "ok $test\n"; $test++;
1496 we can write the (somewhat) more sensible:
1498 ok( "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000) );
1500 Now we'll test that we got that space-at-the-beginning business right:
1502 ok( "1.20.300.4000" eq sprintf "%vd", pack(" U*",1,20,300,4000) );
1504 And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1505 the first active format:
1507 ok( v1.20.300.4000 ne sprintf "%vd", pack("C0U*",1,20,300,4000) );
1509 Mustn't forget to change the number of tests which appears at the top, or
1510 else the automated tester will get confused:
1515 We now compile up Perl, and run it through the test suite. Our new
1518 Finally, the documentation. The job is never done until the paperwork is
1519 over, so let's describe the change we've just made. The relevant place
1520 is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1521 this text in the description of C<pack>:
1525 If the pattern begins with a C<U>, the resulting string will be treated
1526 as Unicode-encoded. You can force UTF8 encoding on in a string with an
1527 initial C<U0>, and the bytes that follow will be interpreted as Unicode
1528 characters. If you don't want this to happen, you can begin your pattern
1529 with C<C0> (or anything else) to force Perl not to UTF8 encode your
1530 string, and then follow this with a C<U*> somewhere in your pattern.
1532 All done. Now let's create the patch. F<Porting/patching.pod> tells us
1533 that if we're making major changes, we should copy the entire directory
1534 to somewhere safe before we begin fiddling, and then do
1536 diff -ruN old new > patch
1538 However, we know which files we've changed, and we can simply do this:
1540 diff -u pp.c~ pp.c > patch
1541 diff -u t/op/pack.t~ t/op/pack.t >> patch
1542 diff -u pod/perlfunc.pod~ pod/perlfunc.pod >> patch
1544 We end up with a patch looking a little like this:
1546 --- pp.c~ Fri Jun 02 04:34:10 2000
1547 +++ pp.c Fri Jun 16 11:37:25 2000
1548 @@ -4375,6 +4375,7 @@
1551 register char *pat = SvPVx(*++MARK, fromlen);
1553 register char *patend = pat + fromlen;
1556 @@ -4405,6 +4406,7 @@
1559 And finally, we submit it, with our rationale, to perl5-porters. Job
1562 =head2 Patching a core module
1564 This works just like patching anything else, with an extra
1565 consideration. Many core modules also live on CPAN. If this is so,
1566 patch the CPAN version instead of the core and send the patch off to
1567 the module maintainer (with a copy to p5p). This will help the module
1568 maintainer keep the CPAN version in sync with the core version without
1569 constantly scanning p5p.
1571 =head2 Adding a new function to the core
1573 If, as part of a patch to fix a bug, or just because you have an
1574 especially good idea, you decide to add a new function to the core,
1575 discuss your ideas on p5p well before you start work. It may be that
1576 someone else has already attempted to do what you are considering and
1577 can give lots of good advice or even provide you with bits of code
1578 that they already started (but never finished).
1580 You have to follow all of the advice given above for patching. It is
1581 extremely important to test any addition thoroughly and add new tests
1582 to explore all boundary conditions that your new function is expected
1583 to handle. If your new function is used only by one module (e.g. toke),
1584 then it should probably be named S_your_function (for static); on the
1585 other hand, if you expect it to accessable from other functions in
1586 Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions>
1589 The location of any new code is also an important consideration. Don't
1590 just create a new top level .c file and put your code there; you would
1591 have to make changes to Configure (so the Makefile is created properly),
1592 as well as possibly lots of include files. This is strictly pumpking
1595 It is better to add your function to one of the existing top level
1596 source code files, but your choice is complicated by the nature of
1597 the Perl distribution. Only the files that are marked as compiled
1598 static are located in the perl executable. Everything else is located
1599 in the shared library (or DLL if you are running under WIN32). So,
1600 for example, if a function was only used by functions located in
1601 toke.c, then your code can go in toke.c. If, however, you want to call
1602 the function from universal.c, then you should put your code in another
1603 location, for example util.c.
1605 In addition to writing your c-code, you will need to create an
1606 appropriate entry in embed.pl describing your function, then run
1607 'make regen_headers' to create the entries in the numerous header
1608 files that perl needs to compile correctly. See L<perlguts/Internal Functions>
1609 for information on the various options that you can set in embed.pl.
1610 You will forget to do this a few (or many) times and you will get
1611 warnings during the compilation phase. Make sure that you mention
1612 this when you post your patch to P5P; the pumpking needs to know this.
1614 When you write your new code, please be conscious of existing code
1615 conventions used in the perl source files. See <perlstyle> for
1616 details. Although most of the guidelines discussed seem to focus on
1617 Perl code, rather than c, they all apply (except when they don't ;).
1618 See also I<Porting/patching.pod> file in the Perl source distribution
1619 for lots of details about both formatting and submitting patches of
1622 Lastly, TEST TEST TEST TEST TEST any code before posting to p5p.
1623 Test on as many platforms as you can find. Test as many perl
1624 Configure options as you can (e.g. MULTIPLICITY). If you have
1625 profiling or memory tools, see L<EXTERNAL TOOLS FOR DEBUGGING PERL>
1626 below for how to use them to futher test your code. Remember that
1627 most of the people on P5P are doing this on their own time and
1628 don't have the time to debug your code.
1630 =head2 Writing a test
1632 Every module and built-in function has an associated test file (or
1633 should...). If you add or change functionality, you have to write a
1634 test. If you fix a bug, you have to write a test so that bug never
1635 comes back. If you alter the docs, it would be nice to test what the
1636 new documentation says.
1638 In short, if you submit a patch you probably also have to patch the
1641 For modules, the test file is right next to the module itself.
1642 F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation,
1643 so there are some snags (and it would be wonderful for you to brush
1644 them out), but it basically works that way. Everything else lives in
1651 Testing of the absolute basic functionality of Perl. Things like
1652 C<if>, basic file reads and writes, simple regexes, etc. These are
1653 run first in the test suite and if any of them fail, something is
1658 These test the basic control structures, C<if/else>, C<while>,
1663 Tests basic issues of how Perl parses and compiles itself.
1667 Tests for built-in IO functions, including command line arguments.
1671 The old home for the module tests, you shouldn't put anything new in
1672 here. There are still some bits and pieces hanging around in here
1673 that need to be moved. Perhaps you could move them? Thanks!
1677 Tests for perl's built in functions that don't fit into any of the
1682 Tests for POD directives. There are still some tests for the Pod
1683 modules hanging around in here that need to be moved out into F<lib/>.
1687 Testing features of how perl actually runs, including exit codes and
1688 handling of PERL* environment variables.
1692 The core uses the same testing style as the rest of Perl, a simple
1693 "ok/not ok" run through Test::Harness, but there are a few special
1696 For most libraries and extensions, you'll want to use the Test::More
1697 library rather than rolling your own test functions. If a module test
1698 doesn't use Test::More, consider rewriting it so it does. For the
1699 rest it's best to use a simple C<print "ok $test_num\n"> style to avoid
1700 broken core functionality from causing the whole test to collapse.
1702 When you say "make test" Perl uses the F<t/TEST> program to run the
1703 test suite. All tests are run from the F<t/> directory, B<not> the
1704 directory which contains the test. This causes some problems with the
1705 tests in F<lib/>, so here's some opportunity for some patching.
1707 You must be triply conscious of cross-platform concerns. This usually
1708 boils down to using File::Spec and avoiding things like C<fork()> and
1709 C<system()> unless absolutely necessary.
1712 =head1 EXTERNAL TOOLS FOR DEBUGGING PERL
1714 Sometimes it helps to use external tools while debugging and
1715 testing Perl. This section tries to guide you through using
1716 some common testing and debugging tools with Perl. This is
1717 meant as a guide to interfacing these tools with Perl, not
1718 as any kind of guide to the use of the tools themselves.
1720 =head2 Rational Software's Purify
1722 Purify is a commercial tool that is helpful in identifying
1723 memory overruns, wild pointers, memory leaks and other such
1724 badness. Perl must be compiled in a specific way for
1725 optimal testing with Purify. Purify is available under
1726 Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.
1728 The only currently known leaks happen when there are
1729 compile-time errors within eval or require. (Fixing these
1730 is non-trivial, unfortunately, but they must be fixed
1733 =head2 Purify on Unix
1735 On Unix, Purify creates a new Perl binary. To get the most
1736 benefit out of Purify, you should create the perl to Purify
1739 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
1740 -Uusemymalloc -Dusemultiplicity
1742 where these arguments mean:
1746 =item -Accflags=-DPURIFY
1748 Disables Perl's arena memory allocation functions, as well as
1749 forcing use of memory allocation functions derived from the
1752 =item -Doptimize='-g'
1754 Adds debugging information so that you see the exact source
1755 statements where the problem occurs. Without this flag, all
1756 you will see is the source filename of where the error occurred.
1760 Disable Perl's malloc so that Purify can more closely monitor
1761 allocations and leaks. Using Perl's malloc will make Purify
1762 report most leaks in the "potential" leaks category.
1764 =item -Dusemultiplicity
1766 Enabling the multiplicity option allows perl to clean up
1767 thoroughly when the interpreter shuts down, which reduces the
1768 number of bogus leak reports from Purify.
1772 Once you've compiled a perl suitable for Purify'ing, then you
1777 which creates a binary named 'pureperl' that has been Purify'ed.
1778 This binary is used in place of the standard 'perl' binary
1779 when you want to debug Perl memory problems.
1781 As an example, to show any memory leaks produced during the
1782 standard Perl testset you would create and run the Purify'ed
1787 ../pureperl -I../lib harness
1789 which would run Perl on test.pl and report any memory problems.
1791 Purify outputs messages in "Viewer" windows by default. If
1792 you don't have a windowing environment or if you simply
1793 want the Purify output to unobtrusively go to a log file
1794 instead of to the interactive window, use these following
1795 options to output to the log file "perl.log":
1797 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
1798 -log-file=perl.log -append-logfile=yes"
1800 If you plan to use the "Viewer" windows, then you only need this option:
1802 setenv PURIFYOPTIONS "-chain-length=25"
1806 Purify on Windows NT instruments the Perl binary 'perl.exe'
1807 on the fly. There are several options in the makefile you
1808 should change to get the most use out of Purify:
1814 You should add -DPURIFY to the DEFINES line so the DEFINES
1815 line looks something like:
1817 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1
1819 to disable Perl's arena memory allocation functions, as
1820 well as to force use of memory allocation functions derived
1821 from the system malloc.
1823 =item USE_MULTI = define
1825 Enabling the multiplicity option allows perl to clean up
1826 thoroughly when the interpreter shuts down, which reduces the
1827 number of bogus leak reports from Purify.
1829 =item #PERL_MALLOC = define
1831 Disable Perl's malloc so that Purify can more closely monitor
1832 allocations and leaks. Using Perl's malloc will make Purify
1833 report most leaks in the "potential" leaks category.
1837 Adds debugging information so that you see the exact source
1838 statements where the problem occurs. Without this flag, all
1839 you will see is the source filename of where the error occurred.
1843 As an example, to show any memory leaks produced during the
1844 standard Perl testset you would create and run Purify as:
1849 purify ../perl -I../lib harness
1851 which would instrument Perl in memory, run Perl on test.pl,
1852 then finally report any memory problems.
1854 =head2 Compaq's/Digital's Third Degree
1856 Third Degree is a tool for memory leak detection and memory access checks.
1857 It is one of the many tools in the ATOM toolkit. The toolkit is only
1858 available on Tru64 (formerly known as Digital UNIX formerly known as
1861 When building Perl, you must first run Configure with -Doptimize=-g
1862 and -Uusemymalloc flags, after that you can use the make targets
1863 "perl.third" and "test.third". (What is required is that Perl must be
1864 compiled using the C<-g> flag, you may need to re-Configure.)
1866 The short story is that with "atom" you can instrument the Perl
1867 executable to create a new executable called F<perl.third>. When the
1868 instrumented executable is run, it creates a log of dubious memory
1869 traffic in file called F<perl.3log>. See the manual pages of atom and
1870 third for more information. The most extensive Third Degree
1871 documentation is available in the Compaq "Tru64 UNIX Programmer's
1872 Guide", chapter "Debugging Programs with Third Degree".
1874 The "test.third" leaves a lot of files named F<perl.3log.*> in the t/
1875 subdirectory. There is a problem with these files: Third Degree is so
1876 effective that it finds problems also in the system libraries.
1877 Therefore there are certain types of errors that you should ignore in
1878 your debugging. Errors with stack traces matching
1880 __actual_atof|__catgets|_doprnt|__exc_|__exec|_findio|__localtime|setlocale|__sia_|__strxfrm
1882 (all in libc.so) are known to be non-serious. You can also
1883 ignore the combinations
1885 Perl_gv_fetchfile() calling strcpy()
1886 S_doopen_pmc() calling strcmp()
1888 causing "rih" (reading invalid heap) errors.
1890 There are also leaks that for given certain definition of a leak,
1891 aren't. See L</PERL_DESTRUCT_LEVEL> for more information.
1893 =head2 PERL_DESTRUCT_LEVEL
1895 If you want to run any of the tests yourself manually using the
1896 pureperl or perl.third executables, please note that by default
1897 perl B<does not> explicitly cleanup all the memory it has allocated
1898 (such as global memory arenas) but instead lets the exit() of
1899 the whole program "take care" of such allocations, also known
1900 as "global destruction of objects".
1902 There is a way to tell perl to do complete cleanup: set the
1903 environment variable PERL_DESTRUCT_LEVEL to a non-zero value.
1904 The t/TEST wrapper does set this to 2, and this is what you
1905 need to do too, if you don't want to see the "global leaks":
1907 PERL_DESTRUCT_LEVEL=2 ./perl.third t/foo/bar.t
1911 Depending on your platform there are various of profiling Perl.
1913 There are two commonly used techniques of profiling executables:
1914 I<statistical time-sampling> and I<basic-block counting>.
1916 The first method takes periodically samples of the CPU program
1917 counter, and since the program counter can be correlated with the code
1918 generated for functions, we get a statistical view of in which
1919 functions the program is spending its time. The caveats are that very
1920 small/fast functions have lower probability of showing up in the
1921 profile, and that periodically interrupting the program (this is
1922 usually done rather frequently, in the scale of milliseconds) imposes
1923 an additional overhead that may skew the results. The first problem
1924 can be alleviated by running the code for longer (in general this is a
1925 good idea for profiling), the second problem is usually kept in guard
1926 by the profiling tools themselves.
1928 The second method divides up the generated code into I<basic blocks>.
1929 Basic blocks are sections of code that are entered only in the
1930 beginning and exited only at the end. For example, a conditional jump
1931 starts a basic block. Basic block profiling usually works by
1932 I<instrumenting> the code by adding I<enter basic block #nnnn>
1933 book-keeping code to the generated code. During the execution of the
1934 code the basic block counters are then updated appropriately. The
1935 caveat is that the added extra code can skew the results: again, the
1936 profiling tools usually try to factor their own effects out of the
1939 =head2 Gprof Profiling
1941 gprof is a profiling tool available in many UNIX platforms,
1942 it uses F<statistical time-sampling>.
1944 You can build a profiled version of perl called "perl.gprof" by
1945 invoking the make target "perl.gprof" (What is required is that Perl
1946 must be compiled using the C<-pg> flag, you may need to re-Configure).
1947 Running the profiled version of Perl will create an output file called
1948 F<gmon.out> is created which contains the profiling data collected
1949 during the execution.
1951 The gprof tool can then display the collected data in various ways.
1952 Usually gprof understands the following options:
1958 Suppress statically defined functions from the profile.
1962 Suppress the verbose descriptions in the profile.
1966 Exclude the given routine and its descendants from the profile.
1970 Display only the given routine and its descendants in the profile.
1974 Generate a summary file called F<gmon.sum> which then may be given
1975 to subsequent gprof runs to accumulate data over several runs.
1979 Display routines that have zero usage.
1983 For more detailed explanation of the available commands and output
1984 formats, see your own local documentation of gprof.
1986 =head2 GCC gcov Profiling
1988 Starting from GCC 3.0 I<basic block profiling> is officially available
1991 You can build a profiled version of perl called F<perl.gcov> by
1992 invoking the make target "perl.gcov" (what is required that Perl must
1993 be compiled using gcc with the flags C<-fprofile-arcs
1994 -ftest-coverage>, you may need to re-Configure).
1996 Running the profiled version of Perl will cause profile output to be
1997 generated. For each source file an accompanying ".da" file will be
2000 To display the results you use the "gcov" utility (which should
2001 be installed if you have gcc 3.0 or newer installed). F<gcov> is
2002 run on source code files, like this
2006 which will cause F<sv.c.gcov> to be created. The F<.gcov> files
2007 contain the source code annotated with relative frequencies of
2008 execution indicated by "#" markers.
2010 Useful options of F<gcov> include C<-b> which will summarise the
2011 basic block, branch, and function call coverage, and C<-c> which
2012 instead of relative frequencies will use the actual counts. For
2013 more information on the use of F<gcov> and basic block profiling
2014 with gcc, see the latest GNU CC manual, as of GCC 3.0 see
2016 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html
2018 and its section titled "8. gcov: a Test Coverage Program"
2020 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132
2022 =head2 Pixie Profiling
2024 Pixie is a profiling tool available on IRIX and Tru64 (aka Digital
2025 UNIX aka DEC OSF/1) platforms. Pixie does its profiling using
2026 I<basic-block counting>.
2028 You can build a profiled version of perl called F<perl.pixie> by
2029 invoking the make target "perl.pixie" (what is required is that Perl
2030 must be compiled using the C<-g> flag, you may need to re-Configure).
2032 In Tru64 a file called F<perl.Addrs> will also be silently created,
2033 this file contains the addresses of the basic blocks. Running the
2034 profiled version of Perl will create a new file called "perl.Counts"
2035 which contains the counts for the basic block for that particular
2038 To display the results you use the F<prof> utility. The exact
2039 incantation depends on your operating system, "prof perl.Counts" in
2040 IRIX, and "prof -pixie -all -L. perl" in Tru64.
2042 In IRIX the following prof options are available:
2048 Reports the most heavily used lines in descending order of use.
2049 Useful for finding the hotspot lines.
2053 Groups lines by procedure, with procedures sorted in descending order of use.
2054 Within a procedure, lines are listed in source order.
2055 Useful for finding the hotspots of procedures.
2059 In Tru64 the following options are available:
2065 Procedures sorted in descending order by the number of cycles executed
2066 in each procedure. Useful for finding the hotspot procedures.
2067 (This is the default option.)
2071 Lines sorted in descending order by the number of cycles executed in
2072 each line. Useful for finding the hotspot lines.
2074 =item -i[nvocations]
2076 The called procedures are sorted in descending order by number of calls
2077 made to the procedures. Useful for finding the most used procedures.
2081 Grouped by procedure, sorted by cycles executed per procedure.
2082 Useful for finding the hotspots of procedures.
2086 The compiler emitted code for these lines, but the code was unexecuted.
2090 Unexecuted procedures.
2094 For further information, see your system's manual pages for pixie and prof.
2098 We've had a brief look around the Perl source, an overview of the stages
2099 F<perl> goes through when it's running your code, and how to use a
2100 debugger to poke at the Perl guts. We took a very simple problem and
2101 demonstrated how to solve it fully - with documentation, regression
2102 tests, and finally a patch for submission to p5p. Finally, we talked
2103 about how to use external tools to debug and test Perl.
2105 I'd now suggest you read over those references again, and then, as soon
2106 as possible, get your hands dirty. The best way to learn is by doing,
2113 Subscribe to perl5-porters, follow the patches and try and understand
2114 them; don't be afraid to ask if there's a portion you're not clear on -
2115 who knows, you may unearth a bug in the patch...
2119 Keep up to date with the bleeding edge Perl distributions and get
2120 familiar with the changes. Try and get an idea of what areas people are
2121 working on and the changes they're making.
2125 Do read the README associated with your operating system, e.g. README.aix
2126 on the IBM AIX OS. Don't hesitate to supply patches to that README if
2127 you find anything missing or changed over a new OS release.
2131 Find an area of Perl that seems interesting to you, and see if you can
2132 work out how it works. Scan through the source, and step over it in the
2133 debugger. Play, poke, investigate, fiddle! You'll probably get to
2134 understand not just your chosen area but a much wider range of F<perl>'s
2135 activity as well, and probably sooner than you'd think.
2141 =item I<The Road goes ever on and on, down from the door where it began.>
2145 If you can do these things, you've started on the long road to Perl porting.
2146 Thanks for wanting to help make Perl better - and happy hacking!
2150 This document was written by Nathan Torkington, and is maintained by
2151 the perl5-porters mailing list.