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 either:
19 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
23 http://archive.develooper.com/perl5-porters@perl.org/
25 List subscribers (the porters themselves) come in several flavours.
26 Some are quiet curious lurkers, who rarely pitch in and instead watch
27 the ongoing development to ensure they're forewarned of new changes or
28 features in Perl. Some are representatives of vendors, who are there
29 to make sure that Perl continues to compile and work on their
30 platforms. Some patch any reported bug that they know how to fix,
31 some are actively patching their pet area (threads, Win32, the regexp
32 engine), while others seem to do nothing but complain. In other
33 words, it's your usual mix of technical people.
35 Over this group of porters presides Larry Wall. He has the final word
36 in what does and does not change in the Perl language. Various
37 releases of Perl are shepherded by a ``pumpking'', a porter
38 responsible for gathering patches, deciding on a patch-by-patch
39 feature-by-feature basis what will and will not go into the release.
40 For instance, Gurusamy Sarathy is the pumpking for the 5.6 release of
43 In addition, various people are pumpkings for different things. For
44 instance, Andy Dougherty and Jarkko Hietaniemi share the I<Configure>
45 pumpkin, and Tom Christiansen is the documentation pumpking.
47 Larry sees Perl development along the lines of the US government:
48 there's the Legislature (the porters), the Executive branch (the
49 pumpkings), and the Supreme Court (Larry). The legislature can
50 discuss and submit patches to the executive branch all they like, but
51 the executive branch is free to veto them. Rarely, the Supreme Court
52 will side with the executive branch over the legislature, or the
53 legislature over the executive branch. Mostly, however, the
54 legislature and the executive branch are supposed to get along and
55 work out their differences without impeachment or court cases.
57 You might sometimes see reference to Rule 1 and Rule 2. Larry's power
58 as Supreme Court is expressed in The Rules:
64 Larry is always by definition right about how Perl should behave.
65 This means he has final veto power on the core functionality.
69 Larry is allowed to change his mind about any matter at a later date,
70 regardless of whether he previously invoked Rule 1.
74 Got that? Larry is always right, even when he was wrong. It's rare
75 to see either Rule exercised, but they are often alluded to.
77 New features and extensions to the language are contentious, because
78 the criteria used by the pumpkings, Larry, and other porters to decide
79 which features should be implemented and incorporated are not codified
80 in a few small design goals as with some other languages. Instead,
81 the heuristics are flexible and often difficult to fathom. Here is
82 one person's list, roughly in decreasing order of importance, of
83 heuristics that new features have to be weighed against:
87 =item Does concept match the general goals of Perl?
89 These haven't been written anywhere in stone, but one approximation
92 1. Keep it fast, simple, and useful.
93 2. Keep features/concepts as orthogonal as possible.
94 3. No arbitrary limits (platforms, data sizes, cultures).
95 4. Keep it open and exciting to use/patch/advocate Perl everywhere.
96 5. Either assimilate new technologies, or build bridges to them.
98 =item Where is the implementation?
100 All the talk in the world is useless without an implementation. In
101 almost every case, the person or people who argue for a new feature
102 will be expected to be the ones who implement it. Porters capable
103 of coding new features have their own agendas, and are not available
104 to implement your (possibly good) idea.
106 =item Backwards compatibility
108 It's a cardinal sin to break existing Perl programs. New warnings are
109 contentious--some say that a program that emits warnings is not
110 broken, while others say it is. Adding keywords has the potential to
111 break programs, changing the meaning of existing token sequences or
112 functions might break programs.
114 =item Could it be a module instead?
116 Perl 5 has extension mechanisms, modules and XS, specifically to avoid
117 the need to keep changing the Perl interpreter. You can write modules
118 that export functions, you can give those functions prototypes so they
119 can be called like built-in functions, you can even write XS code to
120 mess with the runtime data structures of the Perl interpreter if you
121 want to implement really complicated things. If it can be done in a
122 module instead of in the core, it's highly unlikely to be added.
124 =item Is the feature generic enough?
126 Is this something that only the submitter wants added to the language,
127 or would it be broadly useful? Sometimes, instead of adding a feature
128 with a tight focus, the porters might decide to wait until someone
129 implements the more generalized feature. For instance, instead of
130 implementing a ``delayed evaluation'' feature, the porters are waiting
131 for a macro system that would permit delayed evaluation and much more.
133 =item Does it potentially introduce new bugs?
135 Radical rewrites of large chunks of the Perl interpreter have the
136 potential to introduce new bugs. The smaller and more localized the
139 =item Does it preclude other desirable features?
141 A patch is likely to be rejected if it closes off future avenues of
142 development. For instance, a patch that placed a true and final
143 interpretation on prototypes is likely to be rejected because there
144 are still options for the future of prototypes that haven't been
147 =item Is the implementation robust?
149 Good patches (tight code, complete, correct) stand more chance of
150 going in. Sloppy or incorrect patches might be placed on the back
151 burner until the pumpking has time to fix, or might be discarded
152 altogether without further notice.
154 =item Is the implementation generic enough to be portable?
156 The worst patches make use of a system-specific features. It's highly
157 unlikely that nonportable additions to the Perl language will be
160 =item Is there enough documentation?
162 Patches without documentation are probably ill-thought out or
163 incomplete. Nothing can be added without documentation, so submitting
164 a patch for the appropriate manpages as well as the source code is
165 always a good idea. If appropriate, patches should add to the test
168 =item Is there another way to do it?
170 Larry said ``Although the Perl Slogan is I<There's More Than One Way
171 to Do It>, I hesitate to make 10 ways to do something''. This is a
172 tricky heuristic to navigate, though--one man's essential addition is
173 another man's pointless cruft.
175 =item Does it create too much work?
177 Work for the pumpking, work for Perl programmers, work for module
178 authors, ... Perl is supposed to be easy.
180 =item Patches speak louder than words
182 Working code is always preferred to pie-in-the-sky ideas. A patch to
183 add a feature stands a much higher chance of making it to the language
184 than does a random feature request, no matter how fervently argued the
185 request might be. This ties into ``Will it be useful?'', as the fact
186 that someone took the time to make the patch demonstrates a strong
187 desire for the feature.
191 If you're on the list, you might hear the word ``core'' bandied
192 around. It refers to the standard distribution. ``Hacking on the
193 core'' means you're changing the C source code to the Perl
194 interpreter. ``A core module'' is one that ships with Perl.
196 =head2 Keeping in sync
198 The source code to the Perl interpreter, in its different versions, is
199 kept in a repository managed by a revision control system (which is
200 currently the Perforce program, see http://perforce.com/). The
201 pumpkings and a few others have access to the repository to check in
202 changes. Periodically the pumpking for the development version of Perl
203 will release a new version, so the rest of the porters can see what's
204 changed. The current state of the main trunk of repository, and patches
205 that describe the individual changes that have happened since the last
206 public release are available at this location:
208 ftp://ftp.linux.activestate.com/pub/staff/gsar/APC/
210 If you are a member of the perl5-porters mailing list, it is a good
211 thing to keep in touch with the most recent changes. If not only to
212 verify if what you would have posted as a bug report isn't already
213 solved in the most recent available perl development branch, also
214 known as perl-current, bleading edge perl, bleedperl or bleadperl.
216 Needless to say, the source code in perl-current is usually in a perpetual
217 state of evolution. You should expect it to be very buggy. Do B<not> use
218 it for any purpose other than testing and development.
220 Keeping in sync with the most recent branch can be done in several ways,
221 but the most convenient and reliable way is using B<rsync>, available at
222 ftp://rsync.samba.org/pub/rsync/ . (You can also get the most recent
225 If you choose to keep in sync using rsync, there are two approaches
230 =item rsync'ing the source tree
232 Presuming you are in the directory where your perl source resides
233 and you have rsync installed and available, you can `upgrade' to
236 # rsync -avz rsync://ftp.linux.activestate.com/perl-current/ .
238 This takes care of updating every single item in the source tree to
239 the latest applied patch level, creating files that are new (to your
240 distribution) and setting date/time stamps of existing files to
241 reflect the bleadperl status.
243 Note that this will not delete any files that were in '.' before
244 the rsync. Once you are sure that the rsync is running correctly,
245 run it with the --delete and the --dry-run options like this:
247 # rsync -avz --delete --dry-run rsync://ftp.linux.activestate.com/perl-current/ .
249 This will I<simulate> an rsync run that also deletes files not
250 present in the bleadperl master copy. Observe the results from
251 this run closely. If you are sure that the actual run would delete
252 no files precious to you, you could remove the '--dry-run' option.
254 You can than check what patch was the latest that was applied by
255 looking in the file B<.patch>, which will show the number of the
258 If you have more than one machine to keep in sync, and not all of
259 them have access to the WAN (so you are not able to rsync all the
260 source trees to the real source), there are some ways to get around
265 =item Using rsync over the LAN
267 Set up a local rsync server which makes the rsynced source tree
268 available to the LAN and sync the other machines against this
271 From http://rsync.samba.org/README.html:
273 "Rsync uses rsh or ssh for communication. It does not need to be
274 setuid and requires no special privileges for installation. It
275 does not require an inetd entry or a daemon. You must, however,
276 have a working rsh or ssh system. Using ssh is recommended for
277 its security features."
279 =item Using pushing over the NFS
281 Having the other systems mounted over the NFS, you can take an
282 active pushing approach by checking the just updated tree against
283 the other not-yet synced trees. An example would be
292 $1 => [ (stat $1)[2, 7, 9] ]; # mode, size, mtime
295 my %remote = map { $_ => "/$_/pro/3gl/CPAN/perl-5.7.1" } qw(host1 host2);
297 foreach my $host (keys %remote) {
298 unless (-d $remote{$host}) {
299 print STDERR "Cannot Xsync for host $host\n";
302 foreach my $file (keys %MF) {
303 my $rfile = "$remote{$host}/$file";
304 my ($mode, $size, $mtime) = (stat $rfile)[2, 7, 9];
305 defined $size or ($mode, $size, $mtime) = (0, 0, 0);
306 $size == $MF{$file}[1] && $mtime == $MF{$file}[2] and next;
307 printf "%4s %-34s %8d %9d %8d %9d\n",
308 $host, $file, $MF{$file}[1], $MF{$file}[2], $size, $mtime;
310 copy ($file, $rfile);
311 utime time, $MF{$file}[2], $rfile;
312 chmod $MF{$file}[0], $rfile;
316 though this is not perfect. It could be improved with checking
317 file checksums before updating. Not all NFS systems support
318 reliable utime support (when used over the NFS).
322 =item rsync'ing the patches
324 The source tree is maintained by the pumpking who applies patches to
325 the files in the tree. These patches are either created by the
326 pumpking himself using C<diff -c> after updating the file manually or
327 by applying patches sent in by posters on the perl5-porters list.
328 These patches are also saved and rsync'able, so you can apply them
329 yourself to the source files.
331 Presuming you are in a directory where your patches reside, you can
332 get them in sync with
334 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
336 This makes sure the latest available patch is downloaded to your
339 It's then up to you to apply these patches, using something like
341 # last=`ls -rt1 *.gz | tail -1`
342 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
343 # find . -name '*.gz' -newer $last -exec gzcat {} \; >blead.patch
345 # patch -p1 -N <../perl-current-diffs/blead.patch
347 or, since this is only a hint towards how it works, use CPAN-patchaperl
348 from Andreas König to have better control over the patching process.
352 =head2 Why rsync the source tree
356 =item It's easier to rsync the source tree
358 Since you don't have to apply the patches yourself, you are sure all
359 files in the source tree are in the right state.
361 =item It's more recent
363 According to Gurusamy Sarathy:
365 "... The rsync mirror is automatic and syncs with the repository
368 "Updating the patch area still requires manual intervention
369 (with all the goofiness that implies, which you've noted) and
370 is typically on a daily cycle. Making this process automatic
371 is on my tuit list, but don't ask me when."
373 =item It's more reliable
375 Well, since the patches are updated by hand, I don't have to say any
376 more ... (see Sarathy's remark).
380 =head2 Why rsync the patches
384 =item It's easier to rsync the patches
386 If you have more than one machine that you want to keep in track with
387 bleadperl, it's easier to rsync the patches only once and then apply
388 them to all the source trees on the different machines.
390 In case you try to keep in pace on 5 different machines, for which
391 only one of them has access to the WAN, rsync'ing all the source
392 trees should than be done 5 times over the NFS. Having
393 rsync'ed the patches only once, I can apply them to all the source
394 trees automatically. Need you say more ;-)
396 =item It's a good reference
398 If you do not only like to have the most recent development branch,
399 but also like to B<fix> bugs, or extend features, you want to dive
400 into the sources. If you are a seasoned perl core diver, you don't
401 need no manuals, tips, roadmaps, perlguts.pod or other aids to find
402 your way around. But if you are a starter, the patches may help you
403 in finding where you should start and how to change the bits that
406 The file B<Changes> is updated on occasions the pumpking sees as his
407 own little sync points. On those occasions, he releases a tar-ball of
408 the current source tree (i.e. perl@7582.tar.gz), which will be an
409 excellent point to start with when choosing to use the 'rsync the
410 patches' scheme. Starting with perl@7582, which means a set of source
411 files on which the latest applied patch is number 7582, you apply all
412 succeeding patches available from then on (7583, 7584, ...).
414 You can use the patches later as a kind of search archive.
418 =item Finding a start point
420 If you want to fix/change the behaviour of function/feature Foo, just
421 scan the patches for patches that mention Foo either in the subject,
422 the comments, or the body of the fix. A good chance the patch shows
423 you the files that are affected by that patch which are very likely
424 to be the starting point of your journey into the guts of perl.
426 =item Finding how to fix a bug
428 If you've found I<where> the function/feature Foo misbehaves, but you
429 don't know how to fix it (but you do know the change you want to
430 make), you can, again, peruse the patches for similar changes and
431 look how others apply the fix.
433 =item Finding the source of misbehaviour
435 When you keep in sync with bleadperl, the pumpking would love to
436 I<see> that the community efforts really work. So after each of his
437 sync points, you are to 'make test' to check if everything is still
438 in working order. If it is, you do 'make ok', which will send an OK
439 report to perlbug@perl.org. (If you do not have access to a mailer
440 from the system you just finished successfully 'make test', you can
441 do 'make okfile', which creates the file C<perl.ok>, which you can
442 than take to your favourite mailer and mail yourself).
444 But of course, as always, things will not always lead to a success
445 path, and one or more test do not pass the 'make test'. Before
446 sending in a bug report (using 'make nok' or 'make nokfile'), check
447 the mailing list if someone else has reported the bug already and if
448 so, confirm it by replying to that message. If not, you might want to
449 trace the source of that misbehaviour B<before> sending in the bug,
450 which will help all the other porters in finding the solution.
452 Here the saved patches come in very handy. You can check the list of
453 patches to see which patch changed what file and what change caused
454 the misbehaviour. If you note that in the bug report, it saves the
455 one trying to solve it, looking for that point.
459 If searching the patches is too bothersome, you might consider using
460 perl's bugtron to find more information about discussions and
461 ramblings on posted bugs.
465 If you want to get the best of both worlds, rsync both the source
466 tree for convenience, reliability and ease and rsync the patches
469 =head2 Submitting patches
471 Always submit patches to I<perl5-porters@perl.org>. If you're
472 patching a core module and there's an author listed, send the author a
473 copy (see L<Patching a core module>). This lets other porters review
474 your patch, which catches a surprising number of errors in patches.
475 Either use the diff program (available in source code form from
476 I<ftp://ftp.gnu.org/pub/gnu/>), or use Johan Vromans' I<makepatch>
477 (available from I<CPAN/authors/id/JV/>). Unified diffs are preferred,
478 but context diffs are accepted. Do not send RCS-style diffs or diffs
479 without context lines. More information is given in the
480 I<Porting/patching.pod> file in the Perl source distribution. Please
481 patch against the latest B<development> version (e.g., if you're
482 fixing a bug in the 5.005 track, patch against the latest 5.005_5x
483 version). Only patches that survive the heat of the development
484 branch get applied to maintenance versions.
486 Your patch should update the documentation and test suite. See
489 To report a bug in Perl, use the program I<perlbug> which comes with
490 Perl (if you can't get Perl to work, send mail to the address
491 I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through
492 I<perlbug> feeds into the automated bug-tracking system, access to
493 which is provided through the web at I<http://bugs.perl.org/>. It
494 often pays to check the archives of the perl5-porters mailing list to
495 see whether the bug you're reporting has been reported before, and if
496 so whether it was considered a bug. See above for the location of
497 the searchable archives.
499 The CPAN testers (I<http://testers.cpan.org/>) are a group of
500 volunteers who test CPAN modules on a variety of platforms. Perl Labs
501 (I<http://labs.perl.org/>) automatically tests Perl source releases on
502 platforms and gives feedback to the CPAN testers mailing list. Both
503 efforts welcome volunteers.
505 It's a good idea to read and lurk for a while before chipping in.
506 That way you'll get to see the dynamic of the conversations, learn the
507 personalities of the players, and hopefully be better prepared to make
508 a useful contribution when do you speak up.
510 If after all this you still think you want to join the perl5-porters
511 mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To
512 unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>.
514 To hack on the Perl guts, you'll need to read the following things:
520 This is of paramount importance, since it's the documentation of what
521 goes where in the Perl source. Read it over a couple of times and it
522 might start to make sense - don't worry if it doesn't yet, because the
523 best way to study it is to read it in conjunction with poking at Perl
524 source, and we'll do that later on.
526 You might also want to look at Gisle Aas's illustrated perlguts -
527 there's no guarantee that this will be absolutely up-to-date with the
528 latest documentation in the Perl core, but the fundamentals will be
529 right. (http://gisle.aas.no/perl/illguts/)
531 =item L<perlxstut> and L<perlxs>
533 A working knowledge of XSUB programming is incredibly useful for core
534 hacking; XSUBs use techniques drawn from the PP code, the portion of the
535 guts that actually executes a Perl program. It's a lot gentler to learn
536 those techniques from simple examples and explanation than from the core
541 The documentation for the Perl API explains what some of the internal
542 functions do, as well as the many macros used in the source.
544 =item F<Porting/pumpkin.pod>
546 This is a collection of words of wisdom for a Perl porter; some of it is
547 only useful to the pumpkin holder, but most of it applies to anyone
548 wanting to go about Perl development.
550 =item The perl5-porters FAQ
552 This is posted to perl5-porters at the beginning on every month, and
553 should be available from http://perlhacker.org/p5p-faq; alternatively,
554 you can get the FAQ emailed to you by sending mail to
555 C<perl5-porters-faq@perl.org>. It contains hints on reading
556 perl5-porters, information on how perl5-porters works and how Perl
557 development in general works.
561 =head2 Finding Your Way Around
563 Perl maintenance can be split into a number of areas, and certain people
564 (pumpkins) will have responsibility for each area. These areas sometimes
565 correspond to files or directories in the source kit. Among the areas are:
571 Modules shipped as part of the Perl core live in the F<lib/> and F<ext/>
572 subdirectories: F<lib/> is for the pure-Perl modules, and F<ext/>
573 contains the core XS modules.
577 There are tests for nearly all the modules, built-ins and major bits
578 of functionality. Test files all have a .t suffix. Module tests live
579 in the F<lib/> and F<ext/> directories next to the module being
580 tested. Others live in F<t/>. See L<Writing a test>
584 Documentation maintenance includes looking after everything in the
585 F<pod/> directory, (as well as contributing new documentation) and
586 the documentation to the modules in core.
590 The configure process is the way we make Perl portable across the
591 myriad of operating systems it supports. Responsibility for the
592 configure, build and installation process, as well as the overall
593 portability of the core code rests with the configure pumpkin - others
594 help out with individual operating systems.
596 The files involved are the operating system directories, (F<win32/>,
597 F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
598 and F<Makefile>, as well as the metaconfig files which generate
599 F<Configure>. (metaconfig isn't included in the core distribution.)
603 And of course, there's the core of the Perl interpreter itself. Let's
604 have a look at that in a little more detail.
608 Before we leave looking at the layout, though, don't forget that
609 F<MANIFEST> contains not only the file names in the Perl distribution,
610 but short descriptions of what's in them, too. For an overview of the
611 important files, try this:
613 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
615 =head2 Elements of the interpreter
617 The work of the interpreter has two main stages: compiling the code
618 into the internal representation, or bytecode, and then executing it.
619 L<perlguts/Compiled code> explains exactly how the compilation stage
622 Here is a short breakdown of perl's operation:
628 The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
629 This is very high-level code, enough to fit on a single screen, and it
630 resembles the code found in L<perlembed>; most of the real action takes
633 First, F<perlmain.c> allocates some memory and constructs a Perl
636 1 PERL_SYS_INIT3(&argc,&argv,&env);
638 3 if (!PL_do_undump) {
639 4 my_perl = perl_alloc();
642 7 perl_construct(my_perl);
643 8 PL_perl_destruct_level = 0;
646 Line 1 is a macro, and its definition is dependent on your operating
647 system. Line 3 references C<PL_do_undump>, a global variable - all
648 global variables in Perl start with C<PL_>. This tells you whether the
649 current running program was created with the C<-u> flag to perl and then
650 F<undump>, which means it's going to be false in any sane context.
652 Line 4 calls a function in F<perl.c> to allocate memory for a Perl
653 interpreter. It's quite a simple function, and the guts of it looks like
656 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
658 Here you see an example of Perl's system abstraction, which we'll see
659 later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
660 own C<malloc> as defined in F<malloc.c> if you selected that option at
663 Next, in line 7, we construct the interpreter; this sets up all the
664 special variables that Perl needs, the stacks, and so on.
666 Now we pass Perl the command line options, and tell it to go:
668 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
670 exitstatus = perl_run(my_perl);
674 C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
675 in F<perl.c>, which processes the command line options, sets up any
676 statically linked XS modules, opens the program and calls C<yyparse> to
681 The aim of this stage is to take the Perl source, and turn it into an op
682 tree. We'll see what one of those looks like later. Strictly speaking,
683 there's three things going on here.
685 C<yyparse>, the parser, lives in F<perly.c>, although you're better off
686 reading the original YACC input in F<perly.y>. (Yes, Virginia, there
687 B<is> a YACC grammar for Perl!) The job of the parser is to take your
688 code and `understand' it, splitting it into sentences, deciding which
689 operands go with which operators and so on.
691 The parser is nobly assisted by the lexer, which chunks up your input
692 into tokens, and decides what type of thing each token is: a variable
693 name, an operator, a bareword, a subroutine, a core function, and so on.
694 The main point of entry to the lexer is C<yylex>, and that and its
695 associated routines can be found in F<toke.c>. Perl isn't much like
696 other computer languages; it's highly context sensitive at times, it can
697 be tricky to work out what sort of token something is, or where a token
698 ends. As such, there's a lot of interplay between the tokeniser and the
699 parser, which can get pretty frightening if you're not used to it.
701 As the parser understands a Perl program, it builds up a tree of
702 operations for the interpreter to perform during execution. The routines
703 which construct and link together the various operations are to be found
704 in F<op.c>, and will be examined later.
708 Now the parsing stage is complete, and the finished tree represents
709 the operations that the Perl interpreter needs to perform to execute our
710 program. Next, Perl does a dry run over the tree looking for
711 optimisations: constant expressions such as C<3 + 4> will be computed
712 now, and the optimizer will also see if any multiple operations can be
713 replaced with a single one. For instance, to fetch the variable C<$foo>,
714 instead of grabbing the glob C<*foo> and looking at the scalar
715 component, the optimizer fiddles the op tree to use a function which
716 directly looks up the scalar in question. The main optimizer is C<peep>
717 in F<op.c>, and many ops have their own optimizing functions.
721 Now we're finally ready to go: we have compiled Perl byte code, and all
722 that's left to do is run it. The actual execution is done by the
723 C<runops_standard> function in F<run.c>; more specifically, it's done by
724 these three innocent looking lines:
726 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
730 You may be more comfortable with the Perl version of that:
732 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
734 Well, maybe not. Anyway, each op contains a function pointer, which
735 stipulates the function which will actually carry out the operation.
736 This function will return the next op in the sequence - this allows for
737 things like C<if> which choose the next op dynamically at run time.
738 The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
739 execution if required.
741 The actual functions called are known as PP code, and they're spread
742 between four files: F<pp_hot.c> contains the `hot' code, which is most
743 often used and highly optimized, F<pp_sys.c> contains all the
744 system-specific functions, F<pp_ctl.c> contains the functions which
745 implement control structures (C<if>, C<while> and the like) and F<pp.c>
746 contains everything else. These are, if you like, the C code for Perl's
747 built-in functions and operators.
751 =head2 Internal Variable Types
753 You should by now have had a look at L<perlguts>, which tells you about
754 Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
757 These variables are used not only to represent Perl-space variables, but
758 also any constants in the code, as well as some structures completely
759 internal to Perl. The symbol table, for instance, is an ordinary Perl
760 hash. Your code is represented by an SV as it's read into the parser;
761 any program files you call are opened via ordinary Perl filehandles, and
764 The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
765 Perl program. Let's see, for instance, how Perl treats the constant
768 % perl -MDevel::Peek -e 'Dump("hello")'
769 1 SV = PV(0xa041450) at 0xa04ecbc
771 3 FLAGS = (POK,READONLY,pPOK)
772 4 PV = 0xa0484e0 "hello"\0
776 Reading C<Devel::Peek> output takes a bit of practise, so let's go
777 through it line by line.
779 Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
780 memory. SVs themselves are very simple structures, but they contain a
781 pointer to a more complex structure. In this case, it's a PV, a
782 structure which holds a string value, at location C<0xa041450>. Line 2
783 is the reference count; there are no other references to this data, so
786 Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
787 read-only SV (because it's a constant) and the data is a PV internally.
788 Next we've got the contents of the string, starting at location
791 Line 5 gives us the current length of the string - note that this does
792 B<not> include the null terminator. Line 6 is not the length of the
793 string, but the length of the currently allocated buffer; as the string
794 grows, Perl automatically extends the available storage via a routine
797 You can get at any of these quantities from C very easily; just add
798 C<Sv> to the name of the field shown in the snippet, and you've got a
799 macro which will return the value: C<SvCUR(sv)> returns the current
800 length of the string, C<SvREFCOUNT(sv)> returns the reference count,
801 C<SvPV(sv, len)> returns the string itself with its length, and so on.
802 More macros to manipulate these properties can be found in L<perlguts>.
804 Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
807 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
812 6 junk = SvPV_force(sv, tlen);
813 7 SvGROW(sv, tlen + len + 1);
816 10 Move(ptr,SvPVX(sv)+tlen,len,char);
818 12 *SvEND(sv) = '\0';
819 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
823 This is a function which adds a string, C<ptr>, of length C<len> onto
824 the end of the PV stored in C<sv>. The first thing we do in line 6 is
825 make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
826 macro to force a PV. As a side effect, C<tlen> gets set to the current
827 value of the PV, and the PV itself is returned to C<junk>.
829 In line 7, we make sure that the SV will have enough room to accommodate
830 the old string, the new string and the null terminator. If C<LEN> isn't
831 big enough, C<SvGROW> will reallocate space for us.
833 Now, if C<junk> is the same as the string we're trying to add, we can
834 grab the string directly from the SV; C<SvPVX> is the address of the PV
837 Line 10 does the actual catenation: the C<Move> macro moves a chunk of
838 memory around: we move the string C<ptr> to the end of the PV - that's
839 the start of the PV plus its current length. We're moving C<len> bytes
840 of type C<char>. After doing so, we need to tell Perl we've extended the
841 string, by altering C<CUR> to reflect the new length. C<SvEND> is a
842 macro which gives us the end of the string, so that needs to be a
845 Line 13 manipulates the flags; since we've changed the PV, any IV or NV
846 values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
847 want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF8-aware
848 version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
849 and turns on POK. The final C<SvTAINT> is a macro which launders tainted
850 data if taint mode is turned on.
852 AVs and HVs are more complicated, but SVs are by far the most common
853 variable type being thrown around. Having seen something of how we
854 manipulate these, let's go on and look at how the op tree is
859 First, what is the op tree, anyway? The op tree is the parsed
860 representation of your program, as we saw in our section on parsing, and
861 it's the sequence of operations that Perl goes through to execute your
862 program, as we saw in L</Running>.
864 An op is a fundamental operation that Perl can perform: all the built-in
865 functions and operators are ops, and there are a series of ops which
866 deal with concepts the interpreter needs internally - entering and
867 leaving a block, ending a statement, fetching a variable, and so on.
869 The op tree is connected in two ways: you can imagine that there are two
870 "routes" through it, two orders in which you can traverse the tree.
871 First, parse order reflects how the parser understood the code, and
872 secondly, execution order tells perl what order to perform the
875 The easiest way to examine the op tree is to stop Perl after it has
876 finished parsing, and get it to dump out the tree. This is exactly what
877 the compiler backends L<B::Terse|B::Terse> and L<B::Debug|B::Debug> do.
879 Let's have a look at how Perl sees C<$a = $b + $c>:
881 % perl -MO=Terse -e '$a=$b+$c'
882 1 LISTOP (0x8179888) leave
883 2 OP (0x81798b0) enter
884 3 COP (0x8179850) nextstate
885 4 BINOP (0x8179828) sassign
886 5 BINOP (0x8179800) add [1]
887 6 UNOP (0x81796e0) null [15]
888 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
889 8 UNOP (0x81797e0) null [15]
890 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
891 10 UNOP (0x816b4f0) null [15]
892 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
894 Let's start in the middle, at line 4. This is a BINOP, a binary
895 operator, which is at location C<0x8179828>. The specific operator in
896 question is C<sassign> - scalar assignment - and you can find the code
897 which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
898 binary operator, it has two children: the add operator, providing the
899 result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
902 Line 10 is the null op: this does exactly nothing. What is that doing
903 there? If you see the null op, it's a sign that something has been
904 optimized away after parsing. As we mentioned in L</Optimization>,
905 the optimization stage sometimes converts two operations into one, for
906 example when fetching a scalar variable. When this happens, instead of
907 rewriting the op tree and cleaning up the dangling pointers, it's easier
908 just to replace the redundant operation with the null op. Originally,
909 the tree would have looked like this:
911 10 SVOP (0x816b4f0) rv2sv [15]
912 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
914 That is, fetch the C<a> entry from the main symbol table, and then look
915 at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
916 happens to do both these things.
918 The right hand side, starting at line 5 is similar to what we've just
919 seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
922 Now, what's this about?
924 1 LISTOP (0x8179888) leave
925 2 OP (0x81798b0) enter
926 3 COP (0x8179850) nextstate
928 C<enter> and C<leave> are scoping ops, and their job is to perform any
929 housekeeping every time you enter and leave a block: lexical variables
930 are tidied up, unreferenced variables are destroyed, and so on. Every
931 program will have those first three lines: C<leave> is a list, and its
932 children are all the statements in the block. Statements are delimited
933 by C<nextstate>, so a block is a collection of C<nextstate> ops, with
934 the ops to be performed for each statement being the children of
935 C<nextstate>. C<enter> is a single op which functions as a marker.
937 That's how Perl parsed the program, from top to bottom:
950 However, it's impossible to B<perform> the operations in this order:
951 you have to find the values of C<$b> and C<$c> before you add them
952 together, for instance. So, the other thread that runs through the op
953 tree is the execution order: each op has a field C<op_next> which points
954 to the next op to be run, so following these pointers tells us how perl
955 executes the code. We can traverse the tree in this order using
956 the C<exec> option to C<B::Terse>:
958 % perl -MO=Terse,exec -e '$a=$b+$c'
959 1 OP (0x8179928) enter
960 2 COP (0x81798c8) nextstate
961 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
962 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
963 5 BINOP (0x8179878) add [1]
964 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
965 7 BINOP (0x81798a0) sassign
966 8 LISTOP (0x8179900) leave
968 This probably makes more sense for a human: enter a block, start a
969 statement. Get the values of C<$b> and C<$c>, and add them together.
970 Find C<$a>, and assign one to the other. Then leave.
972 The way Perl builds up these op trees in the parsing process can be
973 unravelled by examining F<perly.y>, the YACC grammar. Let's take the
974 piece we need to construct the tree for C<$a = $b + $c>
976 1 term : term ASSIGNOP term
977 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
979 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
981 If you're not used to reading BNF grammars, this is how it works: You're
982 fed certain things by the tokeniser, which generally end up in upper
983 case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
984 code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
985 `terminal symbols', because you can't get any simpler than them.
987 The grammar, lines one and three of the snippet above, tells you how to
988 build up more complex forms. These complex forms, `non-terminal symbols'
989 are generally placed in lower case. C<term> here is a non-terminal
990 symbol, representing a single expression.
992 The grammar gives you the following rule: you can make the thing on the
993 left of the colon if you see all the things on the right in sequence.
994 This is called a "reduction", and the aim of parsing is to completely
995 reduce the input. There are several different ways you can perform a
996 reduction, separated by vertical bars: so, C<term> followed by C<=>
997 followed by C<term> makes a C<term>, and C<term> followed by C<+>
998 followed by C<term> can also make a C<term>.
1000 So, if you see two terms with an C<=> or C<+>, between them, you can
1001 turn them into a single expression. When you do this, you execute the
1002 code in the block on the next line: if you see C<=>, you'll do the code
1003 in line 2. If you see C<+>, you'll do the code in line 4. It's this code
1004 which contributes to the op tree.
1007 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1009 What this does is creates a new binary op, and feeds it a number of
1010 variables. The variables refer to the tokens: C<$1> is the first token in
1011 the input, C<$2> the second, and so on - think regular expression
1012 backreferences. C<$$> is the op returned from this reduction. So, we
1013 call C<newBINOP> to create a new binary operator. The first parameter to
1014 C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
1015 operator, so we want the type to be C<ADDOP>. We could specify this
1016 directly, but it's right there as the second token in the input, so we
1017 use C<$2>. The second parameter is the op's flags: 0 means `nothing
1018 special'. Then the things to add: the left and right hand side of our
1019 expression, in scalar context.
1023 When perl executes something like C<addop>, how does it pass on its
1024 results to the next op? The answer is, through the use of stacks. Perl
1025 has a number of stacks to store things it's currently working on, and
1026 we'll look at the three most important ones here.
1030 =item Argument stack
1032 Arguments are passed to PP code and returned from PP code using the
1033 argument stack, C<ST>. The typical way to handle arguments is to pop
1034 them off the stack, deal with them how you wish, and then push the result
1035 back onto the stack. This is how, for instance, the cosine operator
1040 value = Perl_cos(value);
1043 We'll see a more tricky example of this when we consider Perl's macros
1044 below. C<POPn> gives you the NV (floating point value) of the top SV on
1045 the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
1046 the result back as an NV. The C<X> in C<XPUSHn> means that the stack
1047 should be extended if necessary - it can't be necessary here, because we
1048 know there's room for one more item on the stack, since we've just
1049 removed one! The C<XPUSH*> macros at least guarantee safety.
1051 Alternatively, you can fiddle with the stack directly: C<SP> gives you
1052 the first element in your portion of the stack, and C<TOP*> gives you
1053 the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
1054 negation of an integer:
1058 Just set the integer value of the top stack entry to its negation.
1060 Argument stack manipulation in the core is exactly the same as it is in
1061 XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
1062 description of the macros used in stack manipulation.
1066 I say `your portion of the stack' above because PP code doesn't
1067 necessarily get the whole stack to itself: if your function calls
1068 another function, you'll only want to expose the arguments aimed for the
1069 called function, and not (necessarily) let it get at your own data. The
1070 way we do this is to have a `virtual' bottom-of-stack, exposed to each
1071 function. The mark stack keeps bookmarks to locations in the argument
1072 stack usable by each function. For instance, when dealing with a tied
1073 variable, (internally, something with `P' magic) Perl has to call
1074 methods for accesses to the tied variables. However, we need to separate
1075 the arguments exposed to the method to the argument exposed to the
1076 original function - the store or fetch or whatever it may be. Here's how
1077 the tied C<push> is implemented; see C<av_push> in F<av.c>:
1081 3 PUSHs(SvTIED_obj((SV*)av, mg));
1085 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1089 The lines which concern the mark stack are the first, fifth and last
1090 lines: they save away, restore and remove the current position of the
1093 Let's examine the whole implementation, for practice:
1097 Push the current state of the stack pointer onto the mark stack. This is
1098 so that when we've finished adding items to the argument stack, Perl
1099 knows how many things we've added recently.
1102 3 PUSHs(SvTIED_obj((SV*)av, mg));
1105 We're going to add two more items onto the argument stack: when you have
1106 a tied array, the C<PUSH> subroutine receives the object and the value
1107 to be pushed, and that's exactly what we have here - the tied object,
1108 retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
1112 Next we tell Perl to make the change to the global stack pointer: C<dSP>
1113 only gave us a local copy, not a reference to the global.
1116 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1119 C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
1120 variables are tidied up, everything that has been localised gets
1121 its previous value returned, and so on. Think of them as the C<{> and
1122 C<}> of a Perl block.
1124 To actually do the magic method call, we have to call a subroutine in
1125 Perl space: C<call_method> takes care of that, and it's described in
1126 L<perlcall>. We call the C<PUSH> method in scalar context, and we're
1127 going to discard its return value.
1131 Finally, we remove the value we placed on the mark stack, since we
1132 don't need it any more.
1136 C doesn't have a concept of local scope, so perl provides one. We've
1137 seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
1138 stack implements the C equivalent of, for example:
1145 See L<perlguts/Localising Changes> for how to use the save stack.
1149 =head2 Millions of Macros
1151 One thing you'll notice about the Perl source is that it's full of
1152 macros. Some have called the pervasive use of macros the hardest thing
1153 to understand, others find it adds to clarity. Let's take an example,
1154 the code which implements the addition operator:
1158 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
1161 6 SETn( left + right );
1166 Every line here (apart from the braces, of course) contains a macro. The
1167 first line sets up the function declaration as Perl expects for PP code;
1168 line 3 sets up variable declarations for the argument stack and the
1169 target, the return value of the operation. Finally, it tries to see if
1170 the addition operation is overloaded; if so, the appropriate subroutine
1173 Line 5 is another variable declaration - all variable declarations start
1174 with C<d> - which pops from the top of the argument stack two NVs (hence
1175 C<nn>) and puts them into the variables C<right> and C<left>, hence the
1176 C<rl>. These are the two operands to the addition operator. Next, we
1177 call C<SETn> to set the NV of the return value to the result of adding
1178 the two values. This done, we return - the C<RETURN> macro makes sure
1179 that our return value is properly handled, and we pass the next operator
1180 to run back to the main run loop.
1182 Most of these macros are explained in L<perlapi>, and some of the more
1183 important ones are explained in L<perlxs> as well. Pay special attention
1184 to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
1185 the C<[pad]THX_?> macros.
1188 =head2 Poking at Perl
1190 To really poke around with Perl, you'll probably want to build Perl for
1191 debugging, like this:
1193 ./Configure -d -D optimize=-g
1196 C<-g> is a flag to the C compiler to have it produce debugging
1197 information which will allow us to step through a running program.
1198 F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
1199 enables all the internal debugging code in Perl. There are a whole bunch
1200 of things you can debug with this: L<perlrun> lists them all, and the
1201 best way to find out about them is to play about with them. The most
1202 useful options are probably
1204 l Context (loop) stack processing
1206 o Method and overloading resolution
1207 c String/numeric conversions
1209 Some of the functionality of the debugging code can be achieved using XS
1212 -Dr => use re 'debug'
1213 -Dx => use O 'Debug'
1215 =head2 Using a source-level debugger
1217 If the debugging output of C<-D> doesn't help you, it's time to step
1218 through perl's execution with a source-level debugger.
1224 We'll use C<gdb> for our examples here; the principles will apply to any
1225 debugger, but check the manual of the one you're using.
1229 To fire up the debugger, type
1233 You'll want to do that in your Perl source tree so the debugger can read
1234 the source code. You should see the copyright message, followed by the
1239 C<help> will get you into the documentation, but here are the most
1246 Run the program with the given arguments.
1248 =item break function_name
1250 =item break source.c:xxx
1252 Tells the debugger that we'll want to pause execution when we reach
1253 either the named function (but see L<perlguts/Internal Functions>!) or the given
1254 line in the named source file.
1258 Steps through the program a line at a time.
1262 Steps through the program a line at a time, without descending into
1267 Run until the next breakpoint.
1271 Run until the end of the current function, then stop again.
1275 Just pressing Enter will do the most recent operation again - it's a
1276 blessing when stepping through miles of source code.
1280 Execute the given C code and print its results. B<WARNING>: Perl makes
1281 heavy use of macros, and F<gdb> is not aware of macros. You'll have to
1282 substitute them yourself. So, for instance, you can't say
1284 print SvPV_nolen(sv)
1288 print Perl_sv_2pv_nolen(sv)
1290 You may find it helpful to have a "macro dictionary", which you can
1291 produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
1292 recursively apply the macros for you.
1296 =head2 Dumping Perl Data Structures
1298 One way to get around this macro hell is to use the dumping functions in
1299 F<dump.c>; these work a little like an internal
1300 L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
1301 that you can't get at from Perl. Let's take an example. We'll use the
1302 C<$a = $b + $c> we used before, but give it a bit of context:
1303 C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
1305 What about C<pp_add>, the function we examined earlier to implement the
1308 (gdb) break Perl_pp_add
1309 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
1311 Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>.
1312 With the breakpoint in place, we can run our program:
1314 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
1316 Lots of junk will go past as gdb reads in the relevant source files and
1317 libraries, and then:
1319 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
1320 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
1325 We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
1326 arranges for two C<NV>s to be placed into C<left> and C<right> - let's
1329 #define dPOPTOPnnrl_ul NV right = POPn; \
1330 SV *leftsv = TOPs; \
1331 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
1333 C<POPn> takes the SV from the top of the stack and obtains its NV either
1334 directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
1335 C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
1336 C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
1337 C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
1339 Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
1340 convert it. If we step again, we'll find ourselves there:
1342 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1346 We can now use C<Perl_sv_dump> to investigate the SV:
1348 SV = PV(0xa057cc0) at 0xa0675d0
1351 PV = 0xa06a510 "6XXXX"\0
1356 We know we're going to get C<6> from this, so let's finish the
1360 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
1361 0x462669 in Perl_pp_add () at pp_hot.c:311
1364 We can also dump out this op: the current op is always stored in
1365 C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
1366 similar output to L<B::Debug|B::Debug>.
1369 13 TYPE = add ===> 14
1371 FLAGS = (SCALAR,KIDS)
1373 TYPE = null ===> (12)
1375 FLAGS = (SCALAR,KIDS)
1377 11 TYPE = gvsv ===> 12
1383 # finish this later #
1387 All right, we've now had a look at how to navigate the Perl sources and
1388 some things you'll need to know when fiddling with them. Let's now get
1389 on and create a simple patch. Here's something Larry suggested: if a
1390 C<U> is the first active format during a C<pack>, (for example,
1391 C<pack "U3C8", @stuff>) then the resulting string should be treated as
1394 How do we prepare to fix this up? First we locate the code in question -
1395 the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1396 files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1397 altering this file, let's copy it to F<pp.c~>.
1399 [Well, it was in F<pp.c> when this tutorial was written. It has now been
1400 split off with C<pp_unpack> to its own file, F<pp_pack.c>]
1402 Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1403 loop over the pattern, taking each format character in turn into
1404 C<datum_type>. Then for each possible format character, we swallow up
1405 the other arguments in the pattern (a field width, an asterisk, and so
1406 on) and convert the next chunk input into the specified format, adding
1407 it onto the output SV C<cat>.
1409 How do we know if the C<U> is the first format in the C<pat>? Well, if
1410 we have a pointer to the start of C<pat> then, if we see a C<U> we can
1411 test whether we're still at the start of the string. So, here's where
1415 register char *pat = SvPVx(*++MARK, fromlen);
1416 register char *patend = pat + fromlen;
1421 We'll have another string pointer in there:
1424 register char *pat = SvPVx(*++MARK, fromlen);
1425 register char *patend = pat + fromlen;
1431 And just before we start the loop, we'll set C<patcopy> to be the start
1436 sv_setpvn(cat, "", 0);
1438 while (pat < patend) {
1440 Now if we see a C<U> which was at the start of the string, we turn on
1441 the UTF8 flag for the output SV, C<cat>:
1443 + if (datumtype == 'U' && pat==patcopy+1)
1445 if (datumtype == '#') {
1446 while (pat < patend && *pat != '\n')
1449 Remember that it has to be C<patcopy+1> because the first character of
1450 the string is the C<U> which has been swallowed into C<datumtype!>
1452 Oops, we forgot one thing: what if there are spaces at the start of the
1453 pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1454 character, even though it's not the first thing in the pattern. In this
1455 case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1457 if (isSPACE(datumtype))
1462 if (isSPACE(datumtype)) {
1467 OK. That's the C part done. Now we must do two additional things before
1468 this patch is ready to go: we've changed the behaviour of Perl, and so
1469 we must document that change. We must also provide some more regression
1470 tests to make sure our patch works and doesn't create a bug somewhere
1471 else along the line.
1473 The regression tests for each operator live in F<t/op/>, and so we
1474 make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our
1475 tests to the end. First, we'll test that the C<U> does indeed create
1478 t/op/pack.t has a sensible ok() function, but if it didn't we could
1483 my($ok, $name) = @_;
1485 # You have to do it this way or VMS will get confused.
1486 print $ok ? "ok $test - $name\n" : "not ok $test - $name\n";
1488 printf "# Failed test at line %d\n", (caller)[2] unless $ok;
1496 print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000);
1497 print "ok $test\n"; $test++;
1499 we can write the (somewhat) more sensible:
1501 ok( "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000),
1502 "U* produces unicode" );
1504 Now we'll test that we got that space-at-the-beginning business right:
1506 ok( "1.20.300.4000" eq sprintf "%vd", pack(" U*",1,20,300,4000),
1507 " with spaces at the beginning" );
1509 And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1510 the first active format:
1512 ok( v1.20.300.4000 ne sprintf "%vd", pack("C0U*",1,20,300,4000),
1513 "U* not first isn't unicode" );
1515 Mustn't forget to change the number of tests which appears at the top, or
1516 else the automated tester will get confused:
1521 We now compile up Perl, and run it through the test suite. Our new
1524 Finally, the documentation. The job is never done until the paperwork is
1525 over, so let's describe the change we've just made. The relevant place
1526 is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1527 this text in the description of C<pack>:
1531 If the pattern begins with a C<U>, the resulting string will be treated
1532 as Unicode-encoded. You can force UTF8 encoding on in a string with an
1533 initial C<U0>, and the bytes that follow will be interpreted as Unicode
1534 characters. If you don't want this to happen, you can begin your pattern
1535 with C<C0> (or anything else) to force Perl not to UTF8 encode your
1536 string, and then follow this with a C<U*> somewhere in your pattern.
1538 All done. Now let's create the patch. F<Porting/patching.pod> tells us
1539 that if we're making major changes, we should copy the entire directory
1540 to somewhere safe before we begin fiddling, and then do
1542 diff -ruN old new > patch
1544 However, we know which files we've changed, and we can simply do this:
1546 diff -u pp.c~ pp.c > patch
1547 diff -u t/op/pack.t~ t/op/pack.t >> patch
1548 diff -u pod/perlfunc.pod~ pod/perlfunc.pod >> patch
1550 We end up with a patch looking a little like this:
1552 --- pp.c~ Fri Jun 02 04:34:10 2000
1553 +++ pp.c Fri Jun 16 11:37:25 2000
1554 @@ -4375,6 +4375,7 @@
1557 register char *pat = SvPVx(*++MARK, fromlen);
1559 register char *patend = pat + fromlen;
1562 @@ -4405,6 +4406,7 @@
1565 And finally, we submit it, with our rationale, to perl5-porters. Job
1568 =head2 Patching a core module
1570 This works just like patching anything else, with an extra
1571 consideration. Many core modules also live on CPAN. If this is so,
1572 patch the CPAN version instead of the core and send the patch off to
1573 the module maintainer (with a copy to p5p). This will help the module
1574 maintainer keep the CPAN version in sync with the core version without
1575 constantly scanning p5p.
1577 =head2 Adding a new function to the core
1579 If, as part of a patch to fix a bug, or just because you have an
1580 especially good idea, you decide to add a new function to the core,
1581 discuss your ideas on p5p well before you start work. It may be that
1582 someone else has already attempted to do what you are considering and
1583 can give lots of good advice or even provide you with bits of code
1584 that they already started (but never finished).
1586 You have to follow all of the advice given above for patching. It is
1587 extremely important to test any addition thoroughly and add new tests
1588 to explore all boundary conditions that your new function is expected
1589 to handle. If your new function is used only by one module (e.g. toke),
1590 then it should probably be named S_your_function (for static); on the
1591 other hand, if you expect it to accessable from other functions in
1592 Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions>
1595 The location of any new code is also an important consideration. Don't
1596 just create a new top level .c file and put your code there; you would
1597 have to make changes to Configure (so the Makefile is created properly),
1598 as well as possibly lots of include files. This is strictly pumpking
1601 It is better to add your function to one of the existing top level
1602 source code files, but your choice is complicated by the nature of
1603 the Perl distribution. Only the files that are marked as compiled
1604 static are located in the perl executable. Everything else is located
1605 in the shared library (or DLL if you are running under WIN32). So,
1606 for example, if a function was only used by functions located in
1607 toke.c, then your code can go in toke.c. If, however, you want to call
1608 the function from universal.c, then you should put your code in another
1609 location, for example util.c.
1611 In addition to writing your c-code, you will need to create an
1612 appropriate entry in embed.pl describing your function, then run
1613 'make regen_headers' to create the entries in the numerous header
1614 files that perl needs to compile correctly. See L<perlguts/Internal Functions>
1615 for information on the various options that you can set in embed.pl.
1616 You will forget to do this a few (or many) times and you will get
1617 warnings during the compilation phase. Make sure that you mention
1618 this when you post your patch to P5P; the pumpking needs to know this.
1620 When you write your new code, please be conscious of existing code
1621 conventions used in the perl source files. See <perlstyle> for
1622 details. Although most of the guidelines discussed seem to focus on
1623 Perl code, rather than c, they all apply (except when they don't ;).
1624 See also I<Porting/patching.pod> file in the Perl source distribution
1625 for lots of details about both formatting and submitting patches of
1628 Lastly, TEST TEST TEST TEST TEST any code before posting to p5p.
1629 Test on as many platforms as you can find. Test as many perl
1630 Configure options as you can (e.g. MULTIPLICITY). If you have
1631 profiling or memory tools, see L<EXTERNAL TOOLS FOR DEBUGGING PERL>
1632 below for how to use them to futher test your code. Remember that
1633 most of the people on P5P are doing this on their own time and
1634 don't have the time to debug your code.
1636 =head2 Writing a test
1638 Every module and built-in function has an associated test file (or
1639 should...). If you add or change functionality, you have to write a
1640 test. If you fix a bug, you have to write a test so that bug never
1641 comes back. If you alter the docs, it would be nice to test what the
1642 new documentation says.
1644 In short, if you submit a patch you probably also have to patch the
1647 For modules, the test file is right next to the module itself.
1648 F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation,
1649 so there are some snags (and it would be wonderful for you to brush
1650 them out), but it basically works that way. Everything else lives in
1657 Testing of the absolute basic functionality of Perl. Things like
1658 C<if>, basic file reads and writes, simple regexes, etc. These are
1659 run first in the test suite and if any of them fail, something is
1664 These test the basic control structures, C<if/else>, C<while>,
1669 Tests basic issues of how Perl parses and compiles itself.
1673 Tests for built-in IO functions, including command line arguments.
1677 The old home for the module tests, you shouldn't put anything new in
1678 here. There are still some bits and pieces hanging around in here
1679 that need to be moved. Perhaps you could move them? Thanks!
1683 Tests for perl's built in functions that don't fit into any of the
1688 Tests for POD directives. There are still some tests for the Pod
1689 modules hanging around in here that need to be moved out into F<lib/>.
1693 Testing features of how perl actually runs, including exit codes and
1694 handling of PERL* environment variables.
1698 The core uses the same testing style as the rest of Perl, a simple
1699 "ok/not ok" run through Test::Harness, but there are a few special
1702 For most libraries and extensions, you'll want to use the Test::More
1703 library rather than rolling your own test functions. If a module test
1704 doesn't use Test::More, consider rewriting it so it does. For the
1705 rest it's best to use a simple C<print "ok $test_num\n"> style to avoid
1706 broken core functionality from causing the whole test to collapse.
1708 When you say "make test" Perl uses the F<t/TEST> program to run the
1709 test suite. All tests are run from the F<t/> directory, B<not> the
1710 directory which contains the test. This causes some problems with the
1711 tests in F<lib/>, so here's some opportunity for some patching.
1713 You must be triply conscious of cross-platform concerns. This usually
1714 boils down to using File::Spec and avoiding things like C<fork()> and
1715 C<system()> unless absolutely necessary.
1718 =head1 EXTERNAL TOOLS FOR DEBUGGING PERL
1720 Sometimes it helps to use external tools while debugging and
1721 testing Perl. This section tries to guide you through using
1722 some common testing and debugging tools with Perl. This is
1723 meant as a guide to interfacing these tools with Perl, not
1724 as any kind of guide to the use of the tools themselves.
1726 =head2 Rational Software's Purify
1728 Purify is a commercial tool that is helpful in identifying
1729 memory overruns, wild pointers, memory leaks and other such
1730 badness. Perl must be compiled in a specific way for
1731 optimal testing with Purify. Purify is available under
1732 Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.
1734 The only currently known leaks happen when there are
1735 compile-time errors within eval or require. (Fixing these
1736 is non-trivial, unfortunately, but they must be fixed
1739 =head2 Purify on Unix
1741 On Unix, Purify creates a new Perl binary. To get the most
1742 benefit out of Purify, you should create the perl to Purify
1745 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
1746 -Uusemymalloc -Dusemultiplicity
1748 where these arguments mean:
1752 =item -Accflags=-DPURIFY
1754 Disables Perl's arena memory allocation functions, as well as
1755 forcing use of memory allocation functions derived from the
1758 =item -Doptimize='-g'
1760 Adds debugging information so that you see the exact source
1761 statements where the problem occurs. Without this flag, all
1762 you will see is the source filename of where the error occurred.
1766 Disable Perl's malloc so that Purify can more closely monitor
1767 allocations and leaks. Using Perl's malloc will make Purify
1768 report most leaks in the "potential" leaks category.
1770 =item -Dusemultiplicity
1772 Enabling the multiplicity option allows perl to clean up
1773 thoroughly when the interpreter shuts down, which reduces the
1774 number of bogus leak reports from Purify.
1778 Once you've compiled a perl suitable for Purify'ing, then you
1783 which creates a binary named 'pureperl' that has been Purify'ed.
1784 This binary is used in place of the standard 'perl' binary
1785 when you want to debug Perl memory problems.
1787 As an example, to show any memory leaks produced during the
1788 standard Perl testset you would create and run the Purify'ed
1793 ../pureperl -I../lib harness
1795 which would run Perl on test.pl and report any memory problems.
1797 Purify outputs messages in "Viewer" windows by default. If
1798 you don't have a windowing environment or if you simply
1799 want the Purify output to unobtrusively go to a log file
1800 instead of to the interactive window, use these following
1801 options to output to the log file "perl.log":
1803 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
1804 -log-file=perl.log -append-logfile=yes"
1806 If you plan to use the "Viewer" windows, then you only need this option:
1808 setenv PURIFYOPTIONS "-chain-length=25"
1812 Purify on Windows NT instruments the Perl binary 'perl.exe'
1813 on the fly. There are several options in the makefile you
1814 should change to get the most use out of Purify:
1820 You should add -DPURIFY to the DEFINES line so the DEFINES
1821 line looks something like:
1823 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1
1825 to disable Perl's arena memory allocation functions, as
1826 well as to force use of memory allocation functions derived
1827 from the system malloc.
1829 =item USE_MULTI = define
1831 Enabling the multiplicity option allows perl to clean up
1832 thoroughly when the interpreter shuts down, which reduces the
1833 number of bogus leak reports from Purify.
1835 =item #PERL_MALLOC = define
1837 Disable Perl's malloc so that Purify can more closely monitor
1838 allocations and leaks. Using Perl's malloc will make Purify
1839 report most leaks in the "potential" leaks category.
1843 Adds debugging information so that you see the exact source
1844 statements where the problem occurs. Without this flag, all
1845 you will see is the source filename of where the error occurred.
1849 As an example, to show any memory leaks produced during the
1850 standard Perl testset you would create and run Purify as:
1855 purify ../perl -I../lib harness
1857 which would instrument Perl in memory, run Perl on test.pl,
1858 then finally report any memory problems.
1860 =head2 Compaq's/Digital's Third Degree
1862 Third Degree is a tool for memory leak detection and memory access checks.
1863 It is one of the many tools in the ATOM toolkit. The toolkit is only
1864 available on Tru64 (formerly known as Digital UNIX formerly known as
1867 When building Perl, you must first run Configure with -Doptimize=-g
1868 and -Uusemymalloc flags, after that you can use the make targets
1869 "perl.third" and "test.third". (What is required is that Perl must be
1870 compiled using the C<-g> flag, you may need to re-Configure.)
1872 The short story is that with "atom" you can instrument the Perl
1873 executable to create a new executable called F<perl.third>. When the
1874 instrumented executable is run, it creates a log of dubious memory
1875 traffic in file called F<perl.3log>. See the manual pages of atom and
1876 third for more information. The most extensive Third Degree
1877 documentation is available in the Compaq "Tru64 UNIX Programmer's
1878 Guide", chapter "Debugging Programs with Third Degree".
1880 The "test.third" leaves a lot of files named F<perl.3log.*> in the t/
1881 subdirectory. There is a problem with these files: Third Degree is so
1882 effective that it finds problems also in the system libraries.
1883 Therefore there are certain types of errors that you should ignore in
1884 your debugging. Errors with stack traces matching
1886 __actual_atof|__catgets|_doprnt|__exc_|__exec|_findio|__localtime|setlocale|__sia_|__strxfrm
1888 (all in libc.so) are known to be non-serious. You can also
1889 ignore the combinations
1891 Perl_gv_fetchfile() calling strcpy()
1892 S_doopen_pmc() calling strcmp()
1894 causing "rih" (reading invalid heap) errors.
1896 There are also leaks that for given certain definition of a leak,
1897 aren't. See L</PERL_DESTRUCT_LEVEL> for more information.
1899 =head2 PERL_DESTRUCT_LEVEL
1901 If you want to run any of the tests yourself manually using the
1902 pureperl or perl.third executables, please note that by default
1903 perl B<does not> explicitly cleanup all the memory it has allocated
1904 (such as global memory arenas) but instead lets the exit() of
1905 the whole program "take care" of such allocations, also known
1906 as "global destruction of objects".
1908 There is a way to tell perl to do complete cleanup: set the
1909 environment variable PERL_DESTRUCT_LEVEL to a non-zero value.
1910 The t/TEST wrapper does set this to 2, and this is what you
1911 need to do too, if you don't want to see the "global leaks":
1913 PERL_DESTRUCT_LEVEL=2 ./perl.third t/foo/bar.t
1917 Depending on your platform there are various of profiling Perl.
1919 There are two commonly used techniques of profiling executables:
1920 I<statistical time-sampling> and I<basic-block counting>.
1922 The first method takes periodically samples of the CPU program
1923 counter, and since the program counter can be correlated with the code
1924 generated for functions, we get a statistical view of in which
1925 functions the program is spending its time. The caveats are that very
1926 small/fast functions have lower probability of showing up in the
1927 profile, and that periodically interrupting the program (this is
1928 usually done rather frequently, in the scale of milliseconds) imposes
1929 an additional overhead that may skew the results. The first problem
1930 can be alleviated by running the code for longer (in general this is a
1931 good idea for profiling), the second problem is usually kept in guard
1932 by the profiling tools themselves.
1934 The second method divides up the generated code into I<basic blocks>.
1935 Basic blocks are sections of code that are entered only in the
1936 beginning and exited only at the end. For example, a conditional jump
1937 starts a basic block. Basic block profiling usually works by
1938 I<instrumenting> the code by adding I<enter basic block #nnnn>
1939 book-keeping code to the generated code. During the execution of the
1940 code the basic block counters are then updated appropriately. The
1941 caveat is that the added extra code can skew the results: again, the
1942 profiling tools usually try to factor their own effects out of the
1945 =head2 Gprof Profiling
1947 gprof is a profiling tool available in many UNIX platforms,
1948 it uses F<statistical time-sampling>.
1950 You can build a profiled version of perl called "perl.gprof" by
1951 invoking the make target "perl.gprof" (What is required is that Perl
1952 must be compiled using the C<-pg> flag, you may need to re-Configure).
1953 Running the profiled version of Perl will create an output file called
1954 F<gmon.out> is created which contains the profiling data collected
1955 during the execution.
1957 The gprof tool can then display the collected data in various ways.
1958 Usually gprof understands the following options:
1964 Suppress statically defined functions from the profile.
1968 Suppress the verbose descriptions in the profile.
1972 Exclude the given routine and its descendants from the profile.
1976 Display only the given routine and its descendants in the profile.
1980 Generate a summary file called F<gmon.sum> which then may be given
1981 to subsequent gprof runs to accumulate data over several runs.
1985 Display routines that have zero usage.
1989 For more detailed explanation of the available commands and output
1990 formats, see your own local documentation of gprof.
1992 =head2 GCC gcov Profiling
1994 Starting from GCC 3.0 I<basic block profiling> is officially available
1997 You can build a profiled version of perl called F<perl.gcov> by
1998 invoking the make target "perl.gcov" (what is required that Perl must
1999 be compiled using gcc with the flags C<-fprofile-arcs
2000 -ftest-coverage>, you may need to re-Configure).
2002 Running the profiled version of Perl will cause profile output to be
2003 generated. For each source file an accompanying ".da" file will be
2006 To display the results you use the "gcov" utility (which should
2007 be installed if you have gcc 3.0 or newer installed). F<gcov> is
2008 run on source code files, like this
2012 which will cause F<sv.c.gcov> to be created. The F<.gcov> files
2013 contain the source code annotated with relative frequencies of
2014 execution indicated by "#" markers.
2016 Useful options of F<gcov> include C<-b> which will summarise the
2017 basic block, branch, and function call coverage, and C<-c> which
2018 instead of relative frequencies will use the actual counts. For
2019 more information on the use of F<gcov> and basic block profiling
2020 with gcc, see the latest GNU CC manual, as of GCC 3.0 see
2022 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html
2024 and its section titled "8. gcov: a Test Coverage Program"
2026 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132
2028 =head2 Pixie Profiling
2030 Pixie is a profiling tool available on IRIX and Tru64 (aka Digital
2031 UNIX aka DEC OSF/1) platforms. Pixie does its profiling using
2032 I<basic-block counting>.
2034 You can build a profiled version of perl called F<perl.pixie> by
2035 invoking the make target "perl.pixie" (what is required is that Perl
2036 must be compiled using the C<-g> flag, you may need to re-Configure).
2038 In Tru64 a file called F<perl.Addrs> will also be silently created,
2039 this file contains the addresses of the basic blocks. Running the
2040 profiled version of Perl will create a new file called "perl.Counts"
2041 which contains the counts for the basic block for that particular
2044 To display the results you use the F<prof> utility. The exact
2045 incantation depends on your operating system, "prof perl.Counts" in
2046 IRIX, and "prof -pixie -all -L. perl" in Tru64.
2048 In IRIX the following prof options are available:
2054 Reports the most heavily used lines in descending order of use.
2055 Useful for finding the hotspot lines.
2059 Groups lines by procedure, with procedures sorted in descending order of use.
2060 Within a procedure, lines are listed in source order.
2061 Useful for finding the hotspots of procedures.
2065 In Tru64 the following options are available:
2071 Procedures sorted in descending order by the number of cycles executed
2072 in each procedure. Useful for finding the hotspot procedures.
2073 (This is the default option.)
2077 Lines sorted in descending order by the number of cycles executed in
2078 each line. Useful for finding the hotspot lines.
2080 =item -i[nvocations]
2082 The called procedures are sorted in descending order by number of calls
2083 made to the procedures. Useful for finding the most used procedures.
2087 Grouped by procedure, sorted by cycles executed per procedure.
2088 Useful for finding the hotspots of procedures.
2092 The compiler emitted code for these lines, but the code was unexecuted.
2096 Unexecuted procedures.
2100 For further information, see your system's manual pages for pixie and prof.
2104 We've had a brief look around the Perl source, an overview of the stages
2105 F<perl> goes through when it's running your code, and how to use a
2106 debugger to poke at the Perl guts. We took a very simple problem and
2107 demonstrated how to solve it fully - with documentation, regression
2108 tests, and finally a patch for submission to p5p. Finally, we talked
2109 about how to use external tools to debug and test Perl.
2111 I'd now suggest you read over those references again, and then, as soon
2112 as possible, get your hands dirty. The best way to learn is by doing,
2119 Subscribe to perl5-porters, follow the patches and try and understand
2120 them; don't be afraid to ask if there's a portion you're not clear on -
2121 who knows, you may unearth a bug in the patch...
2125 Keep up to date with the bleeding edge Perl distributions and get
2126 familiar with the changes. Try and get an idea of what areas people are
2127 working on and the changes they're making.
2131 Do read the README associated with your operating system, e.g. README.aix
2132 on the IBM AIX OS. Don't hesitate to supply patches to that README if
2133 you find anything missing or changed over a new OS release.
2137 Find an area of Perl that seems interesting to you, and see if you can
2138 work out how it works. Scan through the source, and step over it in the
2139 debugger. Play, poke, investigate, fiddle! You'll probably get to
2140 understand not just your chosen area but a much wider range of F<perl>'s
2141 activity as well, and probably sooner than you'd think.
2147 =item I<The Road goes ever on and on, down from the door where it began.>
2151 If you can do these things, you've started on the long road to Perl porting.
2152 Thanks for wanting to help make Perl better - and happy hacking!
2156 This document was written by Nathan Torkington, and is maintained by
2157 the perl5-porters mailing list.