This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Re: replace http://www.perl.com/CPAN/ with http://www.cpan.org/
[perl5.git] / pod / perlhack.pod
CommitLineData
e8cd7eae
GS
1=head1 NAME
2
3perlhack - How to hack at the Perl internals
4
5=head1 DESCRIPTION
6
7This document attempts to explain how Perl development takes place,
8and ends with some suggestions for people wanting to become bona fide
9porters.
10
11The perl5-porters mailing list is where the Perl standard distribution
12is maintained and developed. The list can get anywhere from 10 to 150
13messages a day, depending on the heatedness of the debate. Most days
14there are two or three patches, extensions, features, or bugs being
15discussed at a time.
16
f8e3975a 17A searchable archive of the list is at either:
e8cd7eae
GS
18
19 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
20
f8e3975a
IP
21or
22
23 http://archive.develooper.com/perl5-porters@perl.org/
24
e8cd7eae
GS
25List subscribers (the porters themselves) come in several flavours.
26Some are quiet curious lurkers, who rarely pitch in and instead watch
27the ongoing development to ensure they're forewarned of new changes or
28features in Perl. Some are representatives of vendors, who are there
29to make sure that Perl continues to compile and work on their
30platforms. Some patch any reported bug that they know how to fix,
31some are actively patching their pet area (threads, Win32, the regexp
32engine), while others seem to do nothing but complain. In other
33words, it's your usual mix of technical people.
34
35Over this group of porters presides Larry Wall. He has the final word
f6c51b38
GS
36in what does and does not change in the Perl language. Various
37releases of Perl are shepherded by a ``pumpking'', a porter
38responsible for gathering patches, deciding on a patch-by-patch
39feature-by-feature basis what will and will not go into the release.
40For instance, Gurusamy Sarathy is the pumpking for the 5.6 release of
41Perl.
e8cd7eae
GS
42
43In addition, various people are pumpkings for different things. For
44instance, Andy Dougherty and Jarkko Hietaniemi share the I<Configure>
45pumpkin, and Tom Christiansen is the documentation pumpking.
46
47Larry sees Perl development along the lines of the US government:
48there's the Legislature (the porters), the Executive branch (the
49pumpkings), and the Supreme Court (Larry). The legislature can
50discuss and submit patches to the executive branch all they like, but
51the executive branch is free to veto them. Rarely, the Supreme Court
52will side with the executive branch over the legislature, or the
53legislature over the executive branch. Mostly, however, the
54legislature and the executive branch are supposed to get along and
55work out their differences without impeachment or court cases.
56
57You might sometimes see reference to Rule 1 and Rule 2. Larry's power
58as Supreme Court is expressed in The Rules:
59
60=over 4
61
62=item 1
63
64Larry is always by definition right about how Perl should behave.
65This means he has final veto power on the core functionality.
66
67=item 2
68
69Larry is allowed to change his mind about any matter at a later date,
70regardless of whether he previously invoked Rule 1.
71
72=back
73
74Got that? Larry is always right, even when he was wrong. It's rare
75to see either Rule exercised, but they are often alluded to.
76
77New features and extensions to the language are contentious, because
78the criteria used by the pumpkings, Larry, and other porters to decide
79which features should be implemented and incorporated are not codified
80in a few small design goals as with some other languages. Instead,
81the heuristics are flexible and often difficult to fathom. Here is
82one person's list, roughly in decreasing order of importance, of
83heuristics that new features have to be weighed against:
84
85=over 4
86
87=item Does concept match the general goals of Perl?
88
89These haven't been written anywhere in stone, but one approximation
90is:
91
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.
97
98=item Where is the implementation?
99
100All the talk in the world is useless without an implementation. In
101almost every case, the person or people who argue for a new feature
102will be expected to be the ones who implement it. Porters capable
103of coding new features have their own agendas, and are not available
104to implement your (possibly good) idea.
105
106=item Backwards compatibility
107
108It's a cardinal sin to break existing Perl programs. New warnings are
109contentious--some say that a program that emits warnings is not
110broken, while others say it is. Adding keywords has the potential to
111break programs, changing the meaning of existing token sequences or
112functions might break programs.
113
114=item Could it be a module instead?
115
116Perl 5 has extension mechanisms, modules and XS, specifically to avoid
117the need to keep changing the Perl interpreter. You can write modules
118that export functions, you can give those functions prototypes so they
119can be called like built-in functions, you can even write XS code to
120mess with the runtime data structures of the Perl interpreter if you
121want to implement really complicated things. If it can be done in a
122module instead of in the core, it's highly unlikely to be added.
123
124=item Is the feature generic enough?
125
126Is this something that only the submitter wants added to the language,
127or would it be broadly useful? Sometimes, instead of adding a feature
128with a tight focus, the porters might decide to wait until someone
129implements the more generalized feature. For instance, instead of
130implementing a ``delayed evaluation'' feature, the porters are waiting
131for a macro system that would permit delayed evaluation and much more.
132
133=item Does it potentially introduce new bugs?
134
135Radical rewrites of large chunks of the Perl interpreter have the
136potential to introduce new bugs. The smaller and more localized the
137change, the better.
138
139=item Does it preclude other desirable features?
140
141A patch is likely to be rejected if it closes off future avenues of
142development. For instance, a patch that placed a true and final
143interpretation on prototypes is likely to be rejected because there
144are still options for the future of prototypes that haven't been
145addressed.
146
147=item Is the implementation robust?
148
149Good patches (tight code, complete, correct) stand more chance of
150going in. Sloppy or incorrect patches might be placed on the back
151burner until the pumpking has time to fix, or might be discarded
152altogether without further notice.
153
154=item Is the implementation generic enough to be portable?
155
156The worst patches make use of a system-specific features. It's highly
157unlikely that nonportable additions to the Perl language will be
158accepted.
159
160=item Is there enough documentation?
161
162Patches without documentation are probably ill-thought out or
163incomplete. Nothing can be added without documentation, so submitting
164a patch for the appropriate manpages as well as the source code is
165always a good idea. If appropriate, patches should add to the test
166suite as well.
167
168=item Is there another way to do it?
169
170Larry said ``Although the Perl Slogan is I<There's More Than One Way
171to Do It>, I hesitate to make 10 ways to do something''. This is a
172tricky heuristic to navigate, though--one man's essential addition is
173another man's pointless cruft.
174
175=item Does it create too much work?
176
177Work for the pumpking, work for Perl programmers, work for module
178authors, ... Perl is supposed to be easy.
179
f6c51b38
GS
180=item Patches speak louder than words
181
182Working code is always preferred to pie-in-the-sky ideas. A patch to
183add a feature stands a much higher chance of making it to the language
184than does a random feature request, no matter how fervently argued the
185request might be. This ties into ``Will it be useful?'', as the fact
186that someone took the time to make the patch demonstrates a strong
187desire for the feature.
188
e8cd7eae
GS
189=back
190
191If you're on the list, you might hear the word ``core'' bandied
192around. It refers to the standard distribution. ``Hacking on the
193core'' means you're changing the C source code to the Perl
194interpreter. ``A core module'' is one that ships with Perl.
195
a1f349fd
MB
196=head2 Keeping in sync
197
e8cd7eae
GS
198The source code to the Perl interpreter, in its different versions, is
199kept in a repository managed by a revision control system (which is
200currently the Perforce program, see http://perforce.com/). The
201pumpkings and a few others have access to the repository to check in
202changes. Periodically the pumpking for the development version of Perl
203will release a new version, so the rest of the porters can see what's
2be4c08b
GS
204changed. The current state of the main trunk of repository, and patches
205that describe the individual changes that have happened since the last
206public release are available at this location:
207
208 ftp://ftp.linux.activestate.com/pub/staff/gsar/APC/
209
a1f349fd
MB
210If you are a member of the perl5-porters mailing list, it is a good
211thing to keep in touch with the most recent changes. If not only to
212verify if what you would have posted as a bug report isn't already
213solved in the most recent available perl development branch, also
214known as perl-current, bleading edge perl, bleedperl or bleadperl.
2be4c08b
GS
215
216Needless to say, the source code in perl-current is usually in a perpetual
217state of evolution. You should expect it to be very buggy. Do B<not> use
218it for any purpose other than testing and development.
e8cd7eae 219
3e148164
JH
220Keeping in sync with the most recent branch can be done in several ways,
221but the most convenient and reliable way is using B<rsync>, available at
222ftp://rsync.samba.org/pub/rsync/ . (You can also get the most recent
223branch by FTP.)
a1f349fd
MB
224
225If you choose to keep in sync using rsync, there are two approaches
3e148164 226to doing so:
a1f349fd
MB
227
228=over 4
229
230=item rsync'ing the source tree
231
3e148164 232Presuming you are in the directory where your perl source resides
a1f349fd
MB
233and you have rsync installed and available, you can `upgrade' to
234the bleadperl using:
235
236 # rsync -avz rsync://ftp.linux.activestate.com/perl-current/ .
237
238This takes care of updating every single item in the source tree to
239the latest applied patch level, creating files that are new (to your
240distribution) and setting date/time stamps of existing files to
241reflect the bleadperl status.
242
c6d0653e
LC
243Note that this will not delete any files that were in '.' before
244the rsync. Once you are sure that the rsync is running correctly,
245run it with the --delete and the --dry-run options like this:
246
247 # rsync -avz --delete --dry-run rsync://ftp.linux.activestate.com/perl-current/ .
248
249This will I<simulate> an rsync run that also deletes files not
250present in the bleadperl master copy. Observe the results from
251this run closely. If you are sure that the actual run would delete
252no files precious to you, you could remove the '--dry-run' option.
253
a1f349fd
MB
254You can than check what patch was the latest that was applied by
255looking in the file B<.patch>, which will show the number of the
256latest patch.
257
258If you have more than one machine to keep in sync, and not all of
259them have access to the WAN (so you are not able to rsync all the
260source trees to the real source), there are some ways to get around
261this problem.
262
263=over 4
264
265=item Using rsync over the LAN
266
267Set up a local rsync server which makes the rsynced source tree
3e148164 268available to the LAN and sync the other machines against this
a1f349fd
MB
269directory.
270
271From http://rsync.samba.org/README.html:
272
273 "Rsync uses rsh or ssh for communication. It does not need to be
274 setuid and requires no special privileges for installation. It
3958b146 275 does not require an inetd entry or a daemon. You must, however,
a1f349fd
MB
276 have a working rsh or ssh system. Using ssh is recommended for
277 its security features."
278
279=item Using pushing over the NFS
280
281Having the other systems mounted over the NFS, you can take an
3e148164
JH
282active pushing approach by checking the just updated tree against
283the other not-yet synced trees. An example would be
284
285 #!/usr/bin/perl -w
286
287 use strict;
288 use File::Copy;
289
290 my %MF = map {
291 m/(\S+)/;
292 $1 => [ (stat $1)[2, 7, 9] ]; # mode, size, mtime
293 } `cat MANIFEST`;
294
295 my %remote = map { $_ => "/$_/pro/3gl/CPAN/perl-5.7.1" } qw(host1 host2);
296
297 foreach my $host (keys %remote) {
298 unless (-d $remote{$host}) {
299 print STDERR "Cannot Xsync for host $host\n";
300 next;
301 }
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;
309 unlink $rfile;
310 copy ($file, $rfile);
311 utime time, $MF{$file}[2], $rfile;
312 chmod $MF{$file}[0], $rfile;
313 }
314 }
315
316though this is not perfect. It could be improved with checking
a1f349fd
MB
317file checksums before updating. Not all NFS systems support
318reliable utime support (when used over the NFS).
319
320=back
321
322=item rsync'ing the patches
323
324The source tree is maintained by the pumpking who applies patches to
325the files in the tree. These patches are either created by the
326pumpking himself using C<diff -c> after updating the file manually or
327by applying patches sent in by posters on the perl5-porters list.
328These patches are also saved and rsync'able, so you can apply them
329yourself to the source files.
330
331Presuming you are in a directory where your patches reside, you can
3e148164 332get them in sync with
a1f349fd
MB
333
334 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
335
336This makes sure the latest available patch is downloaded to your
337patch directory.
338
3e148164 339It's then up to you to apply these patches, using something like
a1f349fd
MB
340
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
344 # cd ../perl-current
345 # patch -p1 -N <../perl-current-diffs/blead.patch
346
347or, since this is only a hint towards how it works, use CPAN-patchaperl
348from Andreas K├Ânig to have better control over the patching process.
349
350=back
351
f7e1e956 352=head2 Why rsync the source tree
a1f349fd
MB
353
354=over 4
355
10f58044 356=item It's easier to rsync the source tree
a1f349fd
MB
357
358Since you don't have to apply the patches yourself, you are sure all
359files in the source tree are in the right state.
360
361=item It's more recent
362
363According to Gurusamy Sarathy:
364
365 "... The rsync mirror is automatic and syncs with the repository
366 every five minutes.
367
3e148164 368 "Updating the patch area still requires manual intervention
a1f349fd
MB
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."
372
373=item It's more reliable
374
3e148164 375Well, since the patches are updated by hand, I don't have to say any
a1f349fd
MB
376more ... (see Sarathy's remark).
377
378=back
379
f7e1e956 380=head2 Why rsync the patches
a1f349fd
MB
381
382=over 4
383
10f58044 384=item It's easier to rsync the patches
a1f349fd
MB
385
386If you have more than one machine that you want to keep in track with
3e148164 387bleadperl, it's easier to rsync the patches only once and then apply
a1f349fd
MB
388them to all the source trees on the different machines.
389
390In case you try to keep in pace on 5 different machines, for which
391only one of them has access to the WAN, rsync'ing all the source
3e148164 392trees should than be done 5 times over the NFS. Having
a1f349fd 393rsync'ed the patches only once, I can apply them to all the source
3e148164 394trees automatically. Need you say more ;-)
a1f349fd
MB
395
396=item It's a good reference
397
398If you do not only like to have the most recent development branch,
399but also like to B<fix> bugs, or extend features, you want to dive
400into the sources. If you are a seasoned perl core diver, you don't
401need no manuals, tips, roadmaps, perlguts.pod or other aids to find
402your way around. But if you are a starter, the patches may help you
403in finding where you should start and how to change the bits that
404bug you.
405
406The file B<Changes> is updated on occasions the pumpking sees as his
407own little sync points. On those occasions, he releases a tar-ball of
408the current source tree (i.e. perl@7582.tar.gz), which will be an
409excellent point to start with when choosing to use the 'rsync the
410patches' scheme. Starting with perl@7582, which means a set of source
411files on which the latest applied patch is number 7582, you apply all
f18956b7 412succeeding patches available from then on (7583, 7584, ...).
a1f349fd
MB
413
414You can use the patches later as a kind of search archive.
415
416=over 4
417
418=item Finding a start point
419
420If you want to fix/change the behaviour of function/feature Foo, just
421scan the patches for patches that mention Foo either in the subject,
3e148164 422the comments, or the body of the fix. A good chance the patch shows
a1f349fd
MB
423you the files that are affected by that patch which are very likely
424to be the starting point of your journey into the guts of perl.
425
426=item Finding how to fix a bug
427
428If you've found I<where> the function/feature Foo misbehaves, but you
429don't know how to fix it (but you do know the change you want to
430make), you can, again, peruse the patches for similar changes and
431look how others apply the fix.
432
433=item Finding the source of misbehaviour
434
435When you keep in sync with bleadperl, the pumpking would love to
3958b146 436I<see> that the community efforts really work. So after each of his
a1f349fd
MB
437sync points, you are to 'make test' to check if everything is still
438in working order. If it is, you do 'make ok', which will send an OK
439report to perlbug@perl.org. (If you do not have access to a mailer
3e148164 440from the system you just finished successfully 'make test', you can
a1f349fd
MB
441do 'make okfile', which creates the file C<perl.ok>, which you can
442than take to your favourite mailer and mail yourself).
443
3958b146 444But of course, as always, things will not always lead to a success
a1f349fd
MB
445path, and one or more test do not pass the 'make test'. Before
446sending in a bug report (using 'make nok' or 'make nokfile'), check
447the mailing list if someone else has reported the bug already and if
448so, confirm it by replying to that message. If not, you might want to
449trace the source of that misbehaviour B<before> sending in the bug,
450which will help all the other porters in finding the solution.
451
3e148164
JH
452Here the saved patches come in very handy. You can check the list of
453patches to see which patch changed what file and what change caused
454the misbehaviour. If you note that in the bug report, it saves the
455one trying to solve it, looking for that point.
a1f349fd
MB
456
457=back
458
459If searching the patches is too bothersome, you might consider using
460perl's bugtron to find more information about discussions and
461ramblings on posted bugs.
462
3e148164
JH
463If you want to get the best of both worlds, rsync both the source
464tree for convenience, reliability and ease and rsync the patches
465for reference.
466
52315700
RF
467=back
468
469
470=head2 Perlbug remote interface
471
472=over 4
473
474There are three (3) remote administrative interfaces for modifying bug status, category, etc. In all cases an admin must be first registered with the Perlbug database by sending an email request to richard@perl.org or bugmongers@perl.org.
475
476The main requirement is the willingness to classify, (with the emphasis on closing where possible :), outstanding bugs. Further explanation can be garnered from the web at http://bugs.perl.org/, or by asking on the admin mailing list at: bugmongers@perl.org
477
478For more info on the web see
479
480 http://bugs.perl.org/perlbug.cgi?req=spec
481
482
483B<The interfaces:>
484
485
486=item 1 http://bugs.perl.org
487
488Login via the web, (remove B<admin/> if only browsing), where interested Cc's, tests, patches and change-ids, etc. may be assigned.
489
490 http://bugs.perl.org/admin/index.html
491
492
493=item 2 bugdb@perl.org
494
495Where the subject line is used for commands:
496
497 To: bugdb@perl.org
498 Subject: -a close bugid1 bugid2 aix install
499
500 To: bugdb@perl.org
501 Subject: -h
502
503
504=item 3 commands_and_bugdids@bugs.perl.org
505
506Where the address itself is the source for the commands:
507
508 To: close_bugid1_bugid2_aix@bugs.perl.org
509
510 To: help@bugs.perl.org
511
512
513=item notes, patches, tests
514
515For patches and tests, the message body is assigned to the appropriate bug/s and forwarded to p5p for their attention.
516
517 To: test_<bugid1>_aix_close@bugs.perl.org
518 Subject: this is a test for the (now closed) aix bug
519
520 Test is the body of the mail
521
522=back
523
a1f349fd
MB
524=head2 Submitting patches
525
f7e1e956
MS
526Always submit patches to I<perl5-porters@perl.org>. If you're
527patching a core module and there's an author listed, send the author a
528copy (see L<Patching a core module>). This lets other porters review
529your patch, which catches a surprising number of errors in patches.
530Either use the diff program (available in source code form from
531I<ftp://ftp.gnu.org/pub/gnu/>), or use Johan Vromans' I<makepatch>
532(available from I<CPAN/authors/id/JV/>). Unified diffs are preferred,
533but context diffs are accepted. Do not send RCS-style diffs or diffs
534without context lines. More information is given in the
535I<Porting/patching.pod> file in the Perl source distribution. Please
536patch against the latest B<development> version (e.g., if you're
537fixing a bug in the 5.005 track, patch against the latest 5.005_5x
538version). Only patches that survive the heat of the development
539branch get applied to maintenance versions.
540
541Your patch should update the documentation and test suite. See
542L<Writing a test>.
e8cd7eae
GS
543
544To report a bug in Perl, use the program I<perlbug> which comes with
545Perl (if you can't get Perl to work, send mail to the address
f18956b7 546I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through
e8cd7eae
GS
547I<perlbug> feeds into the automated bug-tracking system, access to
548which is provided through the web at I<http://bugs.perl.org/>. It
549often pays to check the archives of the perl5-porters mailing list to
550see whether the bug you're reporting has been reported before, and if
551so whether it was considered a bug. See above for the location of
552the searchable archives.
553
554The CPAN testers (I<http://testers.cpan.org/>) are a group of
555volunteers who test CPAN modules on a variety of platforms. Perl Labs
f6c51b38
GS
556(I<http://labs.perl.org/>) automatically tests Perl source releases on
557platforms and gives feedback to the CPAN testers mailing list. Both
558efforts welcome volunteers.
e8cd7eae 559
e8cd7eae
GS
560It's a good idea to read and lurk for a while before chipping in.
561That way you'll get to see the dynamic of the conversations, learn the
562personalities of the players, and hopefully be better prepared to make
563a useful contribution when do you speak up.
564
565If after all this you still think you want to join the perl5-porters
f6c51b38
GS
566mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To
567unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>.
e8cd7eae 568
a422fd2d
SC
569To hack on the Perl guts, you'll need to read the following things:
570
571=over 3
572
573=item L<perlguts>
574
575This is of paramount importance, since it's the documentation of what
576goes where in the Perl source. Read it over a couple of times and it
577might start to make sense - don't worry if it doesn't yet, because the
578best way to study it is to read it in conjunction with poking at Perl
579source, and we'll do that later on.
580
581You might also want to look at Gisle Aas's illustrated perlguts -
582there's no guarantee that this will be absolutely up-to-date with the
583latest documentation in the Perl core, but the fundamentals will be
584right. (http://gisle.aas.no/perl/illguts/)
585
586=item L<perlxstut> and L<perlxs>
587
588A working knowledge of XSUB programming is incredibly useful for core
589hacking; XSUBs use techniques drawn from the PP code, the portion of the
590guts that actually executes a Perl program. It's a lot gentler to learn
591those techniques from simple examples and explanation than from the core
592itself.
593
594=item L<perlapi>
595
596The documentation for the Perl API explains what some of the internal
597functions do, as well as the many macros used in the source.
598
599=item F<Porting/pumpkin.pod>
600
601This is a collection of words of wisdom for a Perl porter; some of it is
602only useful to the pumpkin holder, but most of it applies to anyone
603wanting to go about Perl development.
604
605=item The perl5-porters FAQ
606
607This is posted to perl5-porters at the beginning on every month, and
608should be available from http://perlhacker.org/p5p-faq; alternatively,
609you can get the FAQ emailed to you by sending mail to
610C<perl5-porters-faq@perl.org>. It contains hints on reading
611perl5-porters, information on how perl5-porters works and how Perl
612development in general works.
613
614=back
615
616=head2 Finding Your Way Around
617
618Perl maintenance can be split into a number of areas, and certain people
619(pumpkins) will have responsibility for each area. These areas sometimes
620correspond to files or directories in the source kit. Among the areas are:
621
622=over 3
623
624=item Core modules
625
626Modules shipped as part of the Perl core live in the F<lib/> and F<ext/>
627subdirectories: F<lib/> is for the pure-Perl modules, and F<ext/>
628contains the core XS modules.
629
f7e1e956
MS
630=item Tests
631
632There are tests for nearly all the modules, built-ins and major bits
633of functionality. Test files all have a .t suffix. Module tests live
634in the F<lib/> and F<ext/> directories next to the module being
635tested. Others live in F<t/>. See L<Writing a test>
636
a422fd2d
SC
637=item Documentation
638
639Documentation maintenance includes looking after everything in the
640F<pod/> directory, (as well as contributing new documentation) and
641the documentation to the modules in core.
642
643=item Configure
644
645The configure process is the way we make Perl portable across the
646myriad of operating systems it supports. Responsibility for the
647configure, build and installation process, as well as the overall
648portability of the core code rests with the configure pumpkin - others
649help out with individual operating systems.
650
651The files involved are the operating system directories, (F<win32/>,
652F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
653and F<Makefile>, as well as the metaconfig files which generate
654F<Configure>. (metaconfig isn't included in the core distribution.)
655
656=item Interpreter
657
658And of course, there's the core of the Perl interpreter itself. Let's
659have a look at that in a little more detail.
660
661=back
662
663Before we leave looking at the layout, though, don't forget that
664F<MANIFEST> contains not only the file names in the Perl distribution,
665but short descriptions of what's in them, too. For an overview of the
666important files, try this:
667
668 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
669
670=head2 Elements of the interpreter
671
672The work of the interpreter has two main stages: compiling the code
673into the internal representation, or bytecode, and then executing it.
674L<perlguts/Compiled code> explains exactly how the compilation stage
675happens.
676
677Here is a short breakdown of perl's operation:
678
679=over 3
680
681=item Startup
682
683The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
684This is very high-level code, enough to fit on a single screen, and it
685resembles the code found in L<perlembed>; most of the real action takes
686place in F<perl.c>
687
688First, F<perlmain.c> allocates some memory and constructs a Perl
689interpreter:
690
691 1 PERL_SYS_INIT3(&argc,&argv,&env);
692 2
693 3 if (!PL_do_undump) {
694 4 my_perl = perl_alloc();
695 5 if (!my_perl)
696 6 exit(1);
697 7 perl_construct(my_perl);
698 8 PL_perl_destruct_level = 0;
699 9 }
700
701Line 1 is a macro, and its definition is dependent on your operating
702system. Line 3 references C<PL_do_undump>, a global variable - all
703global variables in Perl start with C<PL_>. This tells you whether the
704current running program was created with the C<-u> flag to perl and then
705F<undump>, which means it's going to be false in any sane context.
706
707Line 4 calls a function in F<perl.c> to allocate memory for a Perl
708interpreter. It's quite a simple function, and the guts of it looks like
709this:
710
711 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
712
713Here you see an example of Perl's system abstraction, which we'll see
714later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
715own C<malloc> as defined in F<malloc.c> if you selected that option at
716configure time.
717
718Next, in line 7, we construct the interpreter; this sets up all the
719special variables that Perl needs, the stacks, and so on.
720
721Now we pass Perl the command line options, and tell it to go:
722
723 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
724 if (!exitstatus) {
725 exitstatus = perl_run(my_perl);
726 }
727
728
729C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
730in F<perl.c>, which processes the command line options, sets up any
731statically linked XS modules, opens the program and calls C<yyparse> to
732parse it.
733
734=item Parsing
735
736The aim of this stage is to take the Perl source, and turn it into an op
737tree. We'll see what one of those looks like later. Strictly speaking,
738there's three things going on here.
739
740C<yyparse>, the parser, lives in F<perly.c>, although you're better off
741reading the original YACC input in F<perly.y>. (Yes, Virginia, there
742B<is> a YACC grammar for Perl!) The job of the parser is to take your
743code and `understand' it, splitting it into sentences, deciding which
744operands go with which operators and so on.
745
746The parser is nobly assisted by the lexer, which chunks up your input
747into tokens, and decides what type of thing each token is: a variable
748name, an operator, a bareword, a subroutine, a core function, and so on.
749The main point of entry to the lexer is C<yylex>, and that and its
750associated routines can be found in F<toke.c>. Perl isn't much like
751other computer languages; it's highly context sensitive at times, it can
752be tricky to work out what sort of token something is, or where a token
753ends. As such, there's a lot of interplay between the tokeniser and the
754parser, which can get pretty frightening if you're not used to it.
755
756As the parser understands a Perl program, it builds up a tree of
757operations for the interpreter to perform during execution. The routines
758which construct and link together the various operations are to be found
759in F<op.c>, and will be examined later.
760
761=item Optimization
762
763Now the parsing stage is complete, and the finished tree represents
764the operations that the Perl interpreter needs to perform to execute our
765program. Next, Perl does a dry run over the tree looking for
766optimisations: constant expressions such as C<3 + 4> will be computed
767now, and the optimizer will also see if any multiple operations can be
768replaced with a single one. For instance, to fetch the variable C<$foo>,
769instead of grabbing the glob C<*foo> and looking at the scalar
770component, the optimizer fiddles the op tree to use a function which
771directly looks up the scalar in question. The main optimizer is C<peep>
772in F<op.c>, and many ops have their own optimizing functions.
773
774=item Running
775
776Now we're finally ready to go: we have compiled Perl byte code, and all
777that's left to do is run it. The actual execution is done by the
778C<runops_standard> function in F<run.c>; more specifically, it's done by
779these three innocent looking lines:
780
781 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
782 PERL_ASYNC_CHECK();
783 }
784
785You may be more comfortable with the Perl version of that:
786
787 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
788
789Well, maybe not. Anyway, each op contains a function pointer, which
790stipulates the function which will actually carry out the operation.
791This function will return the next op in the sequence - this allows for
792things like C<if> which choose the next op dynamically at run time.
793The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
794execution if required.
795
796The actual functions called are known as PP code, and they're spread
797between four files: F<pp_hot.c> contains the `hot' code, which is most
798often used and highly optimized, F<pp_sys.c> contains all the
799system-specific functions, F<pp_ctl.c> contains the functions which
800implement control structures (C<if>, C<while> and the like) and F<pp.c>
801contains everything else. These are, if you like, the C code for Perl's
802built-in functions and operators.
803
804=back
805
806=head2 Internal Variable Types
807
808You should by now have had a look at L<perlguts>, which tells you about
809Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
810that now.
811
812These variables are used not only to represent Perl-space variables, but
813also any constants in the code, as well as some structures completely
814internal to Perl. The symbol table, for instance, is an ordinary Perl
815hash. Your code is represented by an SV as it's read into the parser;
816any program files you call are opened via ordinary Perl filehandles, and
817so on.
818
819The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
820Perl program. Let's see, for instance, how Perl treats the constant
821C<"hello">.
822
823 % perl -MDevel::Peek -e 'Dump("hello")'
824 1 SV = PV(0xa041450) at 0xa04ecbc
825 2 REFCNT = 1
826 3 FLAGS = (POK,READONLY,pPOK)
827 4 PV = 0xa0484e0 "hello"\0
828 5 CUR = 5
829 6 LEN = 6
830
831Reading C<Devel::Peek> output takes a bit of practise, so let's go
832through it line by line.
833
834Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
835memory. SVs themselves are very simple structures, but they contain a
836pointer to a more complex structure. In this case, it's a PV, a
837structure which holds a string value, at location C<0xa041450>. Line 2
838is the reference count; there are no other references to this data, so
839it's 1.
840
841Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
842read-only SV (because it's a constant) and the data is a PV internally.
843Next we've got the contents of the string, starting at location
844C<0xa0484e0>.
845
846Line 5 gives us the current length of the string - note that this does
847B<not> include the null terminator. Line 6 is not the length of the
848string, but the length of the currently allocated buffer; as the string
849grows, Perl automatically extends the available storage via a routine
850called C<SvGROW>.
851
852You can get at any of these quantities from C very easily; just add
853C<Sv> to the name of the field shown in the snippet, and you've got a
854macro which will return the value: C<SvCUR(sv)> returns the current
855length of the string, C<SvREFCOUNT(sv)> returns the reference count,
856C<SvPV(sv, len)> returns the string itself with its length, and so on.
857More macros to manipulate these properties can be found in L<perlguts>.
858
859Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
860
861 1 void
862 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
863 3 {
864 4 STRLEN tlen;
865 5 char *junk;
866
867 6 junk = SvPV_force(sv, tlen);
868 7 SvGROW(sv, tlen + len + 1);
869 8 if (ptr == junk)
870 9 ptr = SvPVX(sv);
871 10 Move(ptr,SvPVX(sv)+tlen,len,char);
872 11 SvCUR(sv) += len;
873 12 *SvEND(sv) = '\0';
874 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
875 14 SvTAINT(sv);
876 15 }
877
878This is a function which adds a string, C<ptr>, of length C<len> onto
879the end of the PV stored in C<sv>. The first thing we do in line 6 is
880make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
881macro to force a PV. As a side effect, C<tlen> gets set to the current
882value of the PV, and the PV itself is returned to C<junk>.
883
b1866b2d 884In line 7, we make sure that the SV will have enough room to accommodate
a422fd2d
SC
885the old string, the new string and the null terminator. If C<LEN> isn't
886big enough, C<SvGROW> will reallocate space for us.
887
888Now, if C<junk> is the same as the string we're trying to add, we can
889grab the string directly from the SV; C<SvPVX> is the address of the PV
890in the SV.
891
892Line 10 does the actual catenation: the C<Move> macro moves a chunk of
893memory around: we move the string C<ptr> to the end of the PV - that's
894the start of the PV plus its current length. We're moving C<len> bytes
895of type C<char>. After doing so, we need to tell Perl we've extended the
896string, by altering C<CUR> to reflect the new length. C<SvEND> is a
897macro which gives us the end of the string, so that needs to be a
898C<"\0">.
899
900Line 13 manipulates the flags; since we've changed the PV, any IV or NV
901values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
902want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF8-aware
903version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
904and turns on POK. The final C<SvTAINT> is a macro which launders tainted
905data if taint mode is turned on.
906
907AVs and HVs are more complicated, but SVs are by far the most common
908variable type being thrown around. Having seen something of how we
909manipulate these, let's go on and look at how the op tree is
910constructed.
911
912=head2 Op Trees
913
914First, what is the op tree, anyway? The op tree is the parsed
915representation of your program, as we saw in our section on parsing, and
916it's the sequence of operations that Perl goes through to execute your
917program, as we saw in L</Running>.
918
919An op is a fundamental operation that Perl can perform: all the built-in
920functions and operators are ops, and there are a series of ops which
921deal with concepts the interpreter needs internally - entering and
922leaving a block, ending a statement, fetching a variable, and so on.
923
924The op tree is connected in two ways: you can imagine that there are two
925"routes" through it, two orders in which you can traverse the tree.
926First, parse order reflects how the parser understood the code, and
927secondly, execution order tells perl what order to perform the
928operations in.
929
930The easiest way to examine the op tree is to stop Perl after it has
931finished parsing, and get it to dump out the tree. This is exactly what
932the compiler backends L<B::Terse|B::Terse> and L<B::Debug|B::Debug> do.
933
934Let's have a look at how Perl sees C<$a = $b + $c>:
935
936 % perl -MO=Terse -e '$a=$b+$c'
937 1 LISTOP (0x8179888) leave
938 2 OP (0x81798b0) enter
939 3 COP (0x8179850) nextstate
940 4 BINOP (0x8179828) sassign
941 5 BINOP (0x8179800) add [1]
942 6 UNOP (0x81796e0) null [15]
943 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
944 8 UNOP (0x81797e0) null [15]
945 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
946 10 UNOP (0x816b4f0) null [15]
947 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
948
949Let's start in the middle, at line 4. This is a BINOP, a binary
950operator, which is at location C<0x8179828>. The specific operator in
951question is C<sassign> - scalar assignment - and you can find the code
952which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
953binary operator, it has two children: the add operator, providing the
954result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
955line 10.
956
957Line 10 is the null op: this does exactly nothing. What is that doing
958there? If you see the null op, it's a sign that something has been
959optimized away after parsing. As we mentioned in L</Optimization>,
960the optimization stage sometimes converts two operations into one, for
961example when fetching a scalar variable. When this happens, instead of
962rewriting the op tree and cleaning up the dangling pointers, it's easier
963just to replace the redundant operation with the null op. Originally,
964the tree would have looked like this:
965
966 10 SVOP (0x816b4f0) rv2sv [15]
967 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
968
969That is, fetch the C<a> entry from the main symbol table, and then look
970at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
971happens to do both these things.
972
973The right hand side, starting at line 5 is similar to what we've just
974seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
975two C<gvsv>s.
976
977Now, what's this about?
978
979 1 LISTOP (0x8179888) leave
980 2 OP (0x81798b0) enter
981 3 COP (0x8179850) nextstate
982
983C<enter> and C<leave> are scoping ops, and their job is to perform any
984housekeeping every time you enter and leave a block: lexical variables
985are tidied up, unreferenced variables are destroyed, and so on. Every
986program will have those first three lines: C<leave> is a list, and its
987children are all the statements in the block. Statements are delimited
988by C<nextstate>, so a block is a collection of C<nextstate> ops, with
989the ops to be performed for each statement being the children of
990C<nextstate>. C<enter> is a single op which functions as a marker.
991
992That's how Perl parsed the program, from top to bottom:
993
994 Program
995 |
996 Statement
997 |
998 =
999 / \
1000 / \
1001 $a +
1002 / \
1003 $b $c
1004
1005However, it's impossible to B<perform> the operations in this order:
1006you have to find the values of C<$b> and C<$c> before you add them
1007together, for instance. So, the other thread that runs through the op
1008tree is the execution order: each op has a field C<op_next> which points
1009to the next op to be run, so following these pointers tells us how perl
1010executes the code. We can traverse the tree in this order using
1011the C<exec> option to C<B::Terse>:
1012
1013 % perl -MO=Terse,exec -e '$a=$b+$c'
1014 1 OP (0x8179928) enter
1015 2 COP (0x81798c8) nextstate
1016 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
1017 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
1018 5 BINOP (0x8179878) add [1]
1019 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
1020 7 BINOP (0x81798a0) sassign
1021 8 LISTOP (0x8179900) leave
1022
1023This probably makes more sense for a human: enter a block, start a
1024statement. Get the values of C<$b> and C<$c>, and add them together.
1025Find C<$a>, and assign one to the other. Then leave.
1026
1027The way Perl builds up these op trees in the parsing process can be
1028unravelled by examining F<perly.y>, the YACC grammar. Let's take the
1029piece we need to construct the tree for C<$a = $b + $c>
1030
1031 1 term : term ASSIGNOP term
1032 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
1033 3 | term ADDOP term
1034 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1035
1036If you're not used to reading BNF grammars, this is how it works: You're
1037fed certain things by the tokeniser, which generally end up in upper
1038case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
1039code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
1040`terminal symbols', because you can't get any simpler than them.
1041
1042The grammar, lines one and three of the snippet above, tells you how to
1043build up more complex forms. These complex forms, `non-terminal symbols'
1044are generally placed in lower case. C<term> here is a non-terminal
1045symbol, representing a single expression.
1046
1047The grammar gives you the following rule: you can make the thing on the
1048left of the colon if you see all the things on the right in sequence.
1049This is called a "reduction", and the aim of parsing is to completely
1050reduce the input. There are several different ways you can perform a
1051reduction, separated by vertical bars: so, C<term> followed by C<=>
1052followed by C<term> makes a C<term>, and C<term> followed by C<+>
1053followed by C<term> can also make a C<term>.
1054
1055So, if you see two terms with an C<=> or C<+>, between them, you can
1056turn them into a single expression. When you do this, you execute the
1057code in the block on the next line: if you see C<=>, you'll do the code
1058in line 2. If you see C<+>, you'll do the code in line 4. It's this code
1059which contributes to the op tree.
1060
1061 | term ADDOP term
1062 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1063
1064What this does is creates a new binary op, and feeds it a number of
1065variables. The variables refer to the tokens: C<$1> is the first token in
1066the input, C<$2> the second, and so on - think regular expression
1067backreferences. C<$$> is the op returned from this reduction. So, we
1068call C<newBINOP> to create a new binary operator. The first parameter to
1069C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
1070operator, so we want the type to be C<ADDOP>. We could specify this
1071directly, but it's right there as the second token in the input, so we
1072use C<$2>. The second parameter is the op's flags: 0 means `nothing
1073special'. Then the things to add: the left and right hand side of our
1074expression, in scalar context.
1075
1076=head2 Stacks
1077
1078When perl executes something like C<addop>, how does it pass on its
1079results to the next op? The answer is, through the use of stacks. Perl
1080has a number of stacks to store things it's currently working on, and
1081we'll look at the three most important ones here.
1082
1083=over 3
1084
1085=item Argument stack
1086
1087Arguments are passed to PP code and returned from PP code using the
1088argument stack, C<ST>. The typical way to handle arguments is to pop
1089them off the stack, deal with them how you wish, and then push the result
1090back onto the stack. This is how, for instance, the cosine operator
1091works:
1092
1093 NV value;
1094 value = POPn;
1095 value = Perl_cos(value);
1096 XPUSHn(value);
1097
1098We'll see a more tricky example of this when we consider Perl's macros
1099below. C<POPn> gives you the NV (floating point value) of the top SV on
1100the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
1101the result back as an NV. The C<X> in C<XPUSHn> means that the stack
1102should be extended if necessary - it can't be necessary here, because we
1103know there's room for one more item on the stack, since we've just
1104removed one! The C<XPUSH*> macros at least guarantee safety.
1105
1106Alternatively, you can fiddle with the stack directly: C<SP> gives you
1107the first element in your portion of the stack, and C<TOP*> gives you
1108the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
1109negation of an integer:
1110
1111 SETi(-TOPi);
1112
1113Just set the integer value of the top stack entry to its negation.
1114
1115Argument stack manipulation in the core is exactly the same as it is in
1116XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
1117description of the macros used in stack manipulation.
1118
1119=item Mark stack
1120
1121I say `your portion of the stack' above because PP code doesn't
1122necessarily get the whole stack to itself: if your function calls
1123another function, you'll only want to expose the arguments aimed for the
1124called function, and not (necessarily) let it get at your own data. The
1125way we do this is to have a `virtual' bottom-of-stack, exposed to each
1126function. The mark stack keeps bookmarks to locations in the argument
1127stack usable by each function. For instance, when dealing with a tied
1128variable, (internally, something with `P' magic) Perl has to call
1129methods for accesses to the tied variables. However, we need to separate
1130the arguments exposed to the method to the argument exposed to the
1131original function - the store or fetch or whatever it may be. Here's how
1132the tied C<push> is implemented; see C<av_push> in F<av.c>:
1133
1134 1 PUSHMARK(SP);
1135 2 EXTEND(SP,2);
1136 3 PUSHs(SvTIED_obj((SV*)av, mg));
1137 4 PUSHs(val);
1138 5 PUTBACK;
1139 6 ENTER;
1140 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1141 8 LEAVE;
1142 9 POPSTACK;
13a2d996 1143
a422fd2d
SC
1144The lines which concern the mark stack are the first, fifth and last
1145lines: they save away, restore and remove the current position of the
1146argument stack.
1147
1148Let's examine the whole implementation, for practice:
1149
1150 1 PUSHMARK(SP);
1151
1152Push the current state of the stack pointer onto the mark stack. This is
1153so that when we've finished adding items to the argument stack, Perl
1154knows how many things we've added recently.
1155
1156 2 EXTEND(SP,2);
1157 3 PUSHs(SvTIED_obj((SV*)av, mg));
1158 4 PUSHs(val);
1159
1160We're going to add two more items onto the argument stack: when you have
1161a tied array, the C<PUSH> subroutine receives the object and the value
1162to be pushed, and that's exactly what we have here - the tied object,
1163retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
1164
1165 5 PUTBACK;
1166
1167Next we tell Perl to make the change to the global stack pointer: C<dSP>
1168only gave us a local copy, not a reference to the global.
1169
1170 6 ENTER;
1171 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1172 8 LEAVE;
1173
1174C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
1175variables are tidied up, everything that has been localised gets
1176its previous value returned, and so on. Think of them as the C<{> and
1177C<}> of a Perl block.
1178
1179To actually do the magic method call, we have to call a subroutine in
1180Perl space: C<call_method> takes care of that, and it's described in
1181L<perlcall>. We call the C<PUSH> method in scalar context, and we're
1182going to discard its return value.
1183
1184 9 POPSTACK;
1185
1186Finally, we remove the value we placed on the mark stack, since we
1187don't need it any more.
1188
1189=item Save stack
1190
1191C doesn't have a concept of local scope, so perl provides one. We've
1192seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
1193stack implements the C equivalent of, for example:
1194
1195 {
1196 local $foo = 42;
1197 ...
1198 }
1199
1200See L<perlguts/Localising Changes> for how to use the save stack.
1201
1202=back
1203
1204=head2 Millions of Macros
1205
1206One thing you'll notice about the Perl source is that it's full of
1207macros. Some have called the pervasive use of macros the hardest thing
1208to understand, others find it adds to clarity. Let's take an example,
1209the code which implements the addition operator:
1210
1211 1 PP(pp_add)
1212 2 {
39644a26 1213 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
a422fd2d
SC
1214 4 {
1215 5 dPOPTOPnnrl_ul;
1216 6 SETn( left + right );
1217 7 RETURN;
1218 8 }
1219 9 }
1220
1221Every line here (apart from the braces, of course) contains a macro. The
1222first line sets up the function declaration as Perl expects for PP code;
1223line 3 sets up variable declarations for the argument stack and the
1224target, the return value of the operation. Finally, it tries to see if
1225the addition operation is overloaded; if so, the appropriate subroutine
1226is called.
1227
1228Line 5 is another variable declaration - all variable declarations start
1229with C<d> - which pops from the top of the argument stack two NVs (hence
1230C<nn>) and puts them into the variables C<right> and C<left>, hence the
1231C<rl>. These are the two operands to the addition operator. Next, we
1232call C<SETn> to set the NV of the return value to the result of adding
1233the two values. This done, we return - the C<RETURN> macro makes sure
1234that our return value is properly handled, and we pass the next operator
1235to run back to the main run loop.
1236
1237Most of these macros are explained in L<perlapi>, and some of the more
1238important ones are explained in L<perlxs> as well. Pay special attention
1239to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
1240the C<[pad]THX_?> macros.
1241
1242
1243=head2 Poking at Perl
1244
1245To really poke around with Perl, you'll probably want to build Perl for
1246debugging, like this:
1247
1248 ./Configure -d -D optimize=-g
1249 make
1250
1251C<-g> is a flag to the C compiler to have it produce debugging
1252information which will allow us to step through a running program.
1253F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
1254enables all the internal debugging code in Perl. There are a whole bunch
1255of things you can debug with this: L<perlrun> lists them all, and the
1256best way to find out about them is to play about with them. The most
1257useful options are probably
1258
1259 l Context (loop) stack processing
1260 t Trace execution
1261 o Method and overloading resolution
1262 c String/numeric conversions
1263
1264Some of the functionality of the debugging code can be achieved using XS
1265modules.
13a2d996 1266
a422fd2d
SC
1267 -Dr => use re 'debug'
1268 -Dx => use O 'Debug'
1269
1270=head2 Using a source-level debugger
1271
1272If the debugging output of C<-D> doesn't help you, it's time to step
1273through perl's execution with a source-level debugger.
1274
1275=over 3
1276
1277=item *
1278
1279We'll use C<gdb> for our examples here; the principles will apply to any
1280debugger, but check the manual of the one you're using.
1281
1282=back
1283
1284To fire up the debugger, type
1285
1286 gdb ./perl
1287
1288You'll want to do that in your Perl source tree so the debugger can read
1289the source code. You should see the copyright message, followed by the
1290prompt.
1291
1292 (gdb)
1293
1294C<help> will get you into the documentation, but here are the most
1295useful commands:
1296
1297=over 3
1298
1299=item run [args]
1300
1301Run the program with the given arguments.
1302
1303=item break function_name
1304
1305=item break source.c:xxx
1306
1307Tells the debugger that we'll want to pause execution when we reach
cea6626f 1308either the named function (but see L<perlguts/Internal Functions>!) or the given
a422fd2d
SC
1309line in the named source file.
1310
1311=item step
1312
1313Steps through the program a line at a time.
1314
1315=item next
1316
1317Steps through the program a line at a time, without descending into
1318functions.
1319
1320=item continue
1321
1322Run until the next breakpoint.
1323
1324=item finish
1325
1326Run until the end of the current function, then stop again.
1327
13a2d996 1328=item 'enter'
a422fd2d
SC
1329
1330Just pressing Enter will do the most recent operation again - it's a
1331blessing when stepping through miles of source code.
1332
1333=item print
1334
1335Execute the given C code and print its results. B<WARNING>: Perl makes
1336heavy use of macros, and F<gdb> is not aware of macros. You'll have to
1337substitute them yourself. So, for instance, you can't say
1338
1339 print SvPV_nolen(sv)
1340
1341but you have to say
1342
1343 print Perl_sv_2pv_nolen(sv)
1344
1345You may find it helpful to have a "macro dictionary", which you can
1346produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
1347recursively apply the macros for you.
1348
1349=back
1350
1351=head2 Dumping Perl Data Structures
1352
1353One way to get around this macro hell is to use the dumping functions in
1354F<dump.c>; these work a little like an internal
1355L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
1356that you can't get at from Perl. Let's take an example. We'll use the
1357C<$a = $b + $c> we used before, but give it a bit of context:
1358C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
1359
1360What about C<pp_add>, the function we examined earlier to implement the
1361C<+> operator:
1362
1363 (gdb) break Perl_pp_add
1364 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
1365
cea6626f 1366Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>.
a422fd2d
SC
1367With the breakpoint in place, we can run our program:
1368
1369 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
1370
1371Lots of junk will go past as gdb reads in the relevant source files and
1372libraries, and then:
1373
1374 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
39644a26 1375 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
a422fd2d
SC
1376 (gdb) step
1377 311 dPOPTOPnnrl_ul;
1378 (gdb)
1379
1380We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
1381arranges for two C<NV>s to be placed into C<left> and C<right> - let's
1382slightly expand it:
1383
1384 #define dPOPTOPnnrl_ul NV right = POPn; \
1385 SV *leftsv = TOPs; \
1386 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
1387
1388C<POPn> takes the SV from the top of the stack and obtains its NV either
1389directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
1390C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
1391C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
1392C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
1393
1394Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
1395convert it. If we step again, we'll find ourselves there:
1396
1397 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1398 1669 if (!sv)
1399 (gdb)
1400
1401We can now use C<Perl_sv_dump> to investigate the SV:
1402
1403 SV = PV(0xa057cc0) at 0xa0675d0
1404 REFCNT = 1
1405 FLAGS = (POK,pPOK)
1406 PV = 0xa06a510 "6XXXX"\0
1407 CUR = 5
1408 LEN = 6
1409 $1 = void
1410
1411We know we're going to get C<6> from this, so let's finish the
1412subroutine:
1413
1414 (gdb) finish
1415 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
1416 0x462669 in Perl_pp_add () at pp_hot.c:311
1417 311 dPOPTOPnnrl_ul;
1418
1419We can also dump out this op: the current op is always stored in
1420C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
1421similar output to L<B::Debug|B::Debug>.
1422
1423 {
1424 13 TYPE = add ===> 14
1425 TARG = 1
1426 FLAGS = (SCALAR,KIDS)
1427 {
1428 TYPE = null ===> (12)
1429 (was rv2sv)
1430 FLAGS = (SCALAR,KIDS)
1431 {
1432 11 TYPE = gvsv ===> 12
1433 FLAGS = (SCALAR)
1434 GV = main::b
1435 }
1436 }
1437
10f58044 1438# finish this later #
a422fd2d
SC
1439
1440=head2 Patching
1441
1442All right, we've now had a look at how to navigate the Perl sources and
1443some things you'll need to know when fiddling with them. Let's now get
1444on and create a simple patch. Here's something Larry suggested: if a
1445C<U> is the first active format during a C<pack>, (for example,
1446C<pack "U3C8", @stuff>) then the resulting string should be treated as
1447UTF8 encoded.
1448
1449How do we prepare to fix this up? First we locate the code in question -
1450the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1451files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1452altering this file, let's copy it to F<pp.c~>.
1453
a6ec74c1
JH
1454[Well, it was in F<pp.c> when this tutorial was written. It has now been
1455split off with C<pp_unpack> to its own file, F<pp_pack.c>]
1456
a422fd2d
SC
1457Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1458loop over the pattern, taking each format character in turn into
1459C<datum_type>. Then for each possible format character, we swallow up
1460the other arguments in the pattern (a field width, an asterisk, and so
1461on) and convert the next chunk input into the specified format, adding
1462it onto the output SV C<cat>.
1463
1464How do we know if the C<U> is the first format in the C<pat>? Well, if
1465we have a pointer to the start of C<pat> then, if we see a C<U> we can
1466test whether we're still at the start of the string. So, here's where
1467C<pat> is set up:
1468
1469 STRLEN fromlen;
1470 register char *pat = SvPVx(*++MARK, fromlen);
1471 register char *patend = pat + fromlen;
1472 register I32 len;
1473 I32 datumtype;
1474 SV *fromstr;
1475
1476We'll have another string pointer in there:
1477
1478 STRLEN fromlen;
1479 register char *pat = SvPVx(*++MARK, fromlen);
1480 register char *patend = pat + fromlen;
1481 + char *patcopy;
1482 register I32 len;
1483 I32 datumtype;
1484 SV *fromstr;
1485
1486And just before we start the loop, we'll set C<patcopy> to be the start
1487of C<pat>:
1488
1489 items = SP - MARK;
1490 MARK++;
1491 sv_setpvn(cat, "", 0);
1492 + patcopy = pat;
1493 while (pat < patend) {
1494
1495Now if we see a C<U> which was at the start of the string, we turn on
1496the UTF8 flag for the output SV, C<cat>:
1497
1498 + if (datumtype == 'U' && pat==patcopy+1)
1499 + SvUTF8_on(cat);
1500 if (datumtype == '#') {
1501 while (pat < patend && *pat != '\n')
1502 pat++;
1503
1504Remember that it has to be C<patcopy+1> because the first character of
1505the string is the C<U> which has been swallowed into C<datumtype!>
1506
1507Oops, we forgot one thing: what if there are spaces at the start of the
1508pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1509character, even though it's not the first thing in the pattern. In this
1510case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1511
1512 if (isSPACE(datumtype))
1513 continue;
1514
1515needs to become
1516
1517 if (isSPACE(datumtype)) {
1518 patcopy++;
1519 continue;
1520 }
1521
1522OK. That's the C part done. Now we must do two additional things before
1523this patch is ready to go: we've changed the behaviour of Perl, and so
1524we must document that change. We must also provide some more regression
1525tests to make sure our patch works and doesn't create a bug somewhere
1526else along the line.
1527
b23b8711
MS
1528The regression tests for each operator live in F<t/op/>, and so we
1529make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our
1530tests to the end. First, we'll test that the C<U> does indeed create
1531Unicode strings.
1532
1533t/op/pack.t has a sensible ok() function, but if it didn't we could
1534write one easily.
1535
1536 my $test = 1;
1537 sub ok {
812f5127 1538 my($ok, $name) = @_;
c274e827
MS
1539
1540 # You have to do it this way or VMS will get confused.
812f5127 1541 print $ok ? "ok $test - $name\n" : "not ok $test - $name\n";
c274e827 1542
17639bde
MS
1543 printf "# Failed test at line %d\n", (caller)[2] unless $ok;
1544
b23b8711
MS
1545 $test++;
1546 return $ok;
1547 }
1548
1549so instead of this:
a422fd2d
SC
1550
1551 print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000);
1552 print "ok $test\n"; $test++;
1553
b23b8711
MS
1554we can write the (somewhat) more sensible:
1555
812f5127
MS
1556 ok( "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000),
1557 "U* produces unicode" );
b23b8711 1558
a422fd2d
SC
1559Now we'll test that we got that space-at-the-beginning business right:
1560
812f5127
MS
1561 ok( "1.20.300.4000" eq sprintf "%vd", pack(" U*",1,20,300,4000),
1562 " with spaces at the beginning" );
a422fd2d
SC
1563
1564And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1565the first active format:
1566
812f5127
MS
1567 ok( v1.20.300.4000 ne sprintf "%vd", pack("C0U*",1,20,300,4000),
1568 "U* not first isn't unicode" );
a422fd2d 1569
b1866b2d 1570Mustn't forget to change the number of tests which appears at the top, or
a422fd2d
SC
1571else the automated tester will get confused:
1572
1573 -print "1..156\n";
1574 +print "1..159\n";
1575
1576We now compile up Perl, and run it through the test suite. Our new
1577tests pass, hooray!
1578
1579Finally, the documentation. The job is never done until the paperwork is
1580over, so let's describe the change we've just made. The relevant place
1581is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1582this text in the description of C<pack>:
1583
1584 =item *
1585
1586 If the pattern begins with a C<U>, the resulting string will be treated
1587 as Unicode-encoded. You can force UTF8 encoding on in a string with an
1588 initial C<U0>, and the bytes that follow will be interpreted as Unicode
1589 characters. If you don't want this to happen, you can begin your pattern
1590 with C<C0> (or anything else) to force Perl not to UTF8 encode your
1591 string, and then follow this with a C<U*> somewhere in your pattern.
1592
1593All done. Now let's create the patch. F<Porting/patching.pod> tells us
1594that if we're making major changes, we should copy the entire directory
1595to somewhere safe before we begin fiddling, and then do
13a2d996 1596
a422fd2d
SC
1597 diff -ruN old new > patch
1598
1599However, we know which files we've changed, and we can simply do this:
1600
1601 diff -u pp.c~ pp.c > patch
1602 diff -u t/op/pack.t~ t/op/pack.t >> patch
1603 diff -u pod/perlfunc.pod~ pod/perlfunc.pod >> patch
1604
1605We end up with a patch looking a little like this:
1606
1607 --- pp.c~ Fri Jun 02 04:34:10 2000
1608 +++ pp.c Fri Jun 16 11:37:25 2000
1609 @@ -4375,6 +4375,7 @@
1610 register I32 items;
1611 STRLEN fromlen;
1612 register char *pat = SvPVx(*++MARK, fromlen);
1613 + char *patcopy;
1614 register char *patend = pat + fromlen;
1615 register I32 len;
1616 I32 datumtype;
1617 @@ -4405,6 +4406,7 @@
1618 ...
1619
1620And finally, we submit it, with our rationale, to perl5-porters. Job
1621done!
1622
f7e1e956
MS
1623=head2 Patching a core module
1624
1625This works just like patching anything else, with an extra
1626consideration. Many core modules also live on CPAN. If this is so,
1627patch the CPAN version instead of the core and send the patch off to
1628the module maintainer (with a copy to p5p). This will help the module
1629maintainer keep the CPAN version in sync with the core version without
1630constantly scanning p5p.
1631
acbe17fc
JP
1632=head2 Adding a new function to the core
1633
1634If, as part of a patch to fix a bug, or just because you have an
1635especially good idea, you decide to add a new function to the core,
1636discuss your ideas on p5p well before you start work. It may be that
1637someone else has already attempted to do what you are considering and
1638can give lots of good advice or even provide you with bits of code
1639that they already started (but never finished).
1640
1641You have to follow all of the advice given above for patching. It is
1642extremely important to test any addition thoroughly and add new tests
1643to explore all boundary conditions that your new function is expected
1644to handle. If your new function is used only by one module (e.g. toke),
1645then it should probably be named S_your_function (for static); on the
1646other hand, if you expect it to accessable from other functions in
1647Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions>
1648for more details.
1649
1650The location of any new code is also an important consideration. Don't
1651just create a new top level .c file and put your code there; you would
1652have to make changes to Configure (so the Makefile is created properly),
1653as well as possibly lots of include files. This is strictly pumpking
1654business.
1655
1656It is better to add your function to one of the existing top level
1657source code files, but your choice is complicated by the nature of
1658the Perl distribution. Only the files that are marked as compiled
1659static are located in the perl executable. Everything else is located
1660in the shared library (or DLL if you are running under WIN32). So,
1661for example, if a function was only used by functions located in
1662toke.c, then your code can go in toke.c. If, however, you want to call
1663the function from universal.c, then you should put your code in another
1664location, for example util.c.
1665
1666In addition to writing your c-code, you will need to create an
1667appropriate entry in embed.pl describing your function, then run
1668'make regen_headers' to create the entries in the numerous header
1669files that perl needs to compile correctly. See L<perlguts/Internal Functions>
1670for information on the various options that you can set in embed.pl.
1671You will forget to do this a few (or many) times and you will get
1672warnings during the compilation phase. Make sure that you mention
1673this when you post your patch to P5P; the pumpking needs to know this.
1674
1675When you write your new code, please be conscious of existing code
1676conventions used in the perl source files. See <perlstyle> for
1677details. Although most of the guidelines discussed seem to focus on
1678Perl code, rather than c, they all apply (except when they don't ;).
1679See also I<Porting/patching.pod> file in the Perl source distribution
1680for lots of details about both formatting and submitting patches of
1681your changes.
1682
1683Lastly, TEST TEST TEST TEST TEST any code before posting to p5p.
1684Test on as many platforms as you can find. Test as many perl
1685Configure options as you can (e.g. MULTIPLICITY). If you have
1686profiling or memory tools, see L<EXTERNAL TOOLS FOR DEBUGGING PERL>
1687below for how to use them to futher test your code. Remember that
1688most of the people on P5P are doing this on their own time and
1689don't have the time to debug your code.
f7e1e956
MS
1690
1691=head2 Writing a test
1692
1693Every module and built-in function has an associated test file (or
1694should...). If you add or change functionality, you have to write a
1695test. If you fix a bug, you have to write a test so that bug never
1696comes back. If you alter the docs, it would be nice to test what the
1697new documentation says.
1698
1699In short, if you submit a patch you probably also have to patch the
1700tests.
1701
1702For modules, the test file is right next to the module itself.
1703F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation,
1704so there are some snags (and it would be wonderful for you to brush
1705them out), but it basically works that way. Everything else lives in
1706F<t/>.
1707
1708=over 3
1709
1710=item F<t/base/>
1711
1712Testing of the absolute basic functionality of Perl. Things like
1713C<if>, basic file reads and writes, simple regexes, etc. These are
1714run first in the test suite and if any of them fail, something is
1715I<really> broken.
1716
1717=item F<t/cmd/>
1718
1719These test the basic control structures, C<if/else>, C<while>,
1720subroutines, etc...
1721
1722=item F<t/comp/>
1723
1724Tests basic issues of how Perl parses and compiles itself.
1725
1726=item F<t/io/>
1727
1728Tests for built-in IO functions, including command line arguments.
1729
1730=item F<t/lib/>
1731
1732The old home for the module tests, you shouldn't put anything new in
1733here. There are still some bits and pieces hanging around in here
1734that need to be moved. Perhaps you could move them? Thanks!
1735
1736=item F<t/op/>
1737
1738Tests for perl's built in functions that don't fit into any of the
1739other directories.
1740
1741=item F<t/pod/>
1742
1743Tests for POD directives. There are still some tests for the Pod
1744modules hanging around in here that need to be moved out into F<lib/>.
1745
1746=item F<t/run/>
1747
1748Testing features of how perl actually runs, including exit codes and
1749handling of PERL* environment variables.
1750
1751=back
1752
1753The core uses the same testing style as the rest of Perl, a simple
1754"ok/not ok" run through Test::Harness, but there are a few special
1755considerations.
1756
1757For most libraries and extensions, you'll want to use the Test::More
1758library rather than rolling your own test functions. If a module test
1759doesn't use Test::More, consider rewriting it so it does. For the
23acf682 1760rest it's best to use a simple C<print "ok $test_num\n"> style to avoid
f7e1e956
MS
1761broken core functionality from causing the whole test to collapse.
1762
1763When you say "make test" Perl uses the F<t/TEST> program to run the
1764test suite. All tests are run from the F<t/> directory, B<not> the
1765directory which contains the test. This causes some problems with the
1766tests in F<lib/>, so here's some opportunity for some patching.
1767
1768You must be triply conscious of cross-platform concerns. This usually
1769boils down to using File::Spec and avoiding things like C<fork()> and
1770C<system()> unless absolutely necessary.
1771
1772
902b9dbf
MF
1773=head1 EXTERNAL TOOLS FOR DEBUGGING PERL
1774
1775Sometimes it helps to use external tools while debugging and
1776testing Perl. This section tries to guide you through using
1777some common testing and debugging tools with Perl. This is
1778meant as a guide to interfacing these tools with Perl, not
1779as any kind of guide to the use of the tools themselves.
1780
1781=head2 Rational Software's Purify
1782
1783Purify is a commercial tool that is helpful in identifying
1784memory overruns, wild pointers, memory leaks and other such
1785badness. Perl must be compiled in a specific way for
1786optimal testing with Purify. Purify is available under
1787Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.
1788
1789The only currently known leaks happen when there are
1790compile-time errors within eval or require. (Fixing these
1791is non-trivial, unfortunately, but they must be fixed
1792eventually.)
1793
1794=head2 Purify on Unix
1795
1796On Unix, Purify creates a new Perl binary. To get the most
1797benefit out of Purify, you should create the perl to Purify
1798using:
1799
1800 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
1801 -Uusemymalloc -Dusemultiplicity
1802
1803where these arguments mean:
1804
1805=over 4
1806
1807=item -Accflags=-DPURIFY
1808
1809Disables Perl's arena memory allocation functions, as well as
1810forcing use of memory allocation functions derived from the
1811system malloc.
1812
1813=item -Doptimize='-g'
1814
1815Adds debugging information so that you see the exact source
1816statements where the problem occurs. Without this flag, all
1817you will see is the source filename of where the error occurred.
1818
1819=item -Uusemymalloc
1820
1821Disable Perl's malloc so that Purify can more closely monitor
1822allocations and leaks. Using Perl's malloc will make Purify
1823report most leaks in the "potential" leaks category.
1824
1825=item -Dusemultiplicity
1826
1827Enabling the multiplicity option allows perl to clean up
1828thoroughly when the interpreter shuts down, which reduces the
1829number of bogus leak reports from Purify.
1830
1831=back
1832
1833Once you've compiled a perl suitable for Purify'ing, then you
1834can just:
1835
1836 make pureperl
1837
1838which creates a binary named 'pureperl' that has been Purify'ed.
1839This binary is used in place of the standard 'perl' binary
1840when you want to debug Perl memory problems.
1841
1842As an example, to show any memory leaks produced during the
1843standard Perl testset you would create and run the Purify'ed
1844perl as:
1845
1846 make pureperl
1847 cd t
1848 ../pureperl -I../lib harness
1849
1850which would run Perl on test.pl and report any memory problems.
1851
1852Purify outputs messages in "Viewer" windows by default. If
1853you don't have a windowing environment or if you simply
1854want the Purify output to unobtrusively go to a log file
1855instead of to the interactive window, use these following
1856options to output to the log file "perl.log":
1857
1858 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
1859 -log-file=perl.log -append-logfile=yes"
1860
1861If you plan to use the "Viewer" windows, then you only need this option:
1862
1863 setenv PURIFYOPTIONS "-chain-length=25"
1864
1865=head2 Purify on NT
1866
1867Purify on Windows NT instruments the Perl binary 'perl.exe'
1868on the fly. There are several options in the makefile you
1869should change to get the most use out of Purify:
1870
1871=over 4
1872
1873=item DEFINES
1874
1875You should add -DPURIFY to the DEFINES line so the DEFINES
1876line looks something like:
1877
1878 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1
1879
1880to disable Perl's arena memory allocation functions, as
1881well as to force use of memory allocation functions derived
1882from the system malloc.
1883
1884=item USE_MULTI = define
1885
1886Enabling the multiplicity option allows perl to clean up
1887thoroughly when the interpreter shuts down, which reduces the
1888number of bogus leak reports from Purify.
1889
1890=item #PERL_MALLOC = define
1891
1892Disable Perl's malloc so that Purify can more closely monitor
1893allocations and leaks. Using Perl's malloc will make Purify
1894report most leaks in the "potential" leaks category.
1895
1896=item CFG = Debug
1897
1898Adds debugging information so that you see the exact source
1899statements where the problem occurs. Without this flag, all
1900you will see is the source filename of where the error occurred.
1901
1902=back
1903
1904As an example, to show any memory leaks produced during the
1905standard Perl testset you would create and run Purify as:
1906
1907 cd win32
1908 make
1909 cd ../t
1910 purify ../perl -I../lib harness
1911
1912which would instrument Perl in memory, run Perl on test.pl,
1913then finally report any memory problems.
1914
09187cb1
JH
1915=head2 Compaq's/Digital's Third Degree
1916
1917Third Degree is a tool for memory leak detection and memory access checks.
1918It is one of the many tools in the ATOM toolkit. The toolkit is only
1919available on Tru64 (formerly known as Digital UNIX formerly known as
1920DEC OSF/1).
1921
1922When building Perl, you must first run Configure with -Doptimize=-g
1923and -Uusemymalloc flags, after that you can use the make targets
51a35ef1
JH
1924"perl.third" and "test.third". (What is required is that Perl must be
1925compiled using the C<-g> flag, you may need to re-Configure.)
09187cb1 1926
64cea5fd 1927The short story is that with "atom" you can instrument the Perl
83f0ef60 1928executable to create a new executable called F<perl.third>. When the
4ae3d70a 1929instrumented executable is run, it creates a log of dubious memory
83f0ef60 1930traffic in file called F<perl.3log>. See the manual pages of atom and
4ae3d70a
JH
1931third for more information. The most extensive Third Degree
1932documentation is available in the Compaq "Tru64 UNIX Programmer's
1933Guide", chapter "Debugging Programs with Third Degree".
64cea5fd 1934
83f0ef60 1935The "test.third" leaves a lot of files named F<perl.3log.*> in the t/
64cea5fd
JH
1936subdirectory. There is a problem with these files: Third Degree is so
1937effective that it finds problems also in the system libraries.
83f0ef60
JH
1938Therefore there are certain types of errors that you should ignore in
1939your debugging. Errors with stack traces matching
64cea5fd
JH
1940
1941 __actual_atof|__catgets|_doprnt|__exc_|__exec|_findio|__localtime|setlocale|__sia_|__strxfrm
1942
1943(all in libc.so) are known to be non-serious. You can also
1944ignore the combinations
1945
1946 Perl_gv_fetchfile() calling strcpy()
1947 S_doopen_pmc() calling strcmp()
1948
1949causing "rih" (reading invalid heap) errors.
1950
1951There are also leaks that for given certain definition of a leak,
1952aren't. See L</PERL_DESTRUCT_LEVEL> for more information.
1953
1954=head2 PERL_DESTRUCT_LEVEL
1955
1956If you want to run any of the tests yourself manually using the
1957pureperl or perl.third executables, please note that by default
1958perl B<does not> explicitly cleanup all the memory it has allocated
1959(such as global memory arenas) but instead lets the exit() of
1960the whole program "take care" of such allocations, also known
1961as "global destruction of objects".
1962
1963There is a way to tell perl to do complete cleanup: set the
1964environment variable PERL_DESTRUCT_LEVEL to a non-zero value.
1965The t/TEST wrapper does set this to 2, and this is what you
1966need to do too, if you don't want to see the "global leaks":
1967
1968 PERL_DESTRUCT_LEVEL=2 ./perl.third t/foo/bar.t
09187cb1 1969
51a35ef1
JH
1970=head2 Profiling
1971
1972Depending on your platform there are various of profiling Perl.
1973
1974There are two commonly used techniques of profiling executables:
10f58044 1975I<statistical time-sampling> and I<basic-block counting>.
51a35ef1
JH
1976
1977The first method takes periodically samples of the CPU program
1978counter, and since the program counter can be correlated with the code
1979generated for functions, we get a statistical view of in which
1980functions the program is spending its time. The caveats are that very
1981small/fast functions have lower probability of showing up in the
1982profile, and that periodically interrupting the program (this is
1983usually done rather frequently, in the scale of milliseconds) imposes
1984an additional overhead that may skew the results. The first problem
1985can be alleviated by running the code for longer (in general this is a
1986good idea for profiling), the second problem is usually kept in guard
1987by the profiling tools themselves.
1988
10f58044 1989The second method divides up the generated code into I<basic blocks>.
51a35ef1
JH
1990Basic blocks are sections of code that are entered only in the
1991beginning and exited only at the end. For example, a conditional jump
1992starts a basic block. Basic block profiling usually works by
10f58044 1993I<instrumenting> the code by adding I<enter basic block #nnnn>
51a35ef1
JH
1994book-keeping code to the generated code. During the execution of the
1995code the basic block counters are then updated appropriately. The
1996caveat is that the added extra code can skew the results: again, the
1997profiling tools usually try to factor their own effects out of the
1998results.
1999
83f0ef60
JH
2000=head2 Gprof Profiling
2001
51a35ef1
JH
2002gprof is a profiling tool available in many UNIX platforms,
2003it uses F<statistical time-sampling>.
83f0ef60
JH
2004
2005You can build a profiled version of perl called "perl.gprof" by
51a35ef1
JH
2006invoking the make target "perl.gprof" (What is required is that Perl
2007must be compiled using the C<-pg> flag, you may need to re-Configure).
2008Running the profiled version of Perl will create an output file called
2009F<gmon.out> is created which contains the profiling data collected
2010during the execution.
83f0ef60
JH
2011
2012The gprof tool can then display the collected data in various ways.
2013Usually gprof understands the following options:
2014
2015=over 4
2016
2017=item -a
2018
2019Suppress statically defined functions from the profile.
2020
2021=item -b
2022
2023Suppress the verbose descriptions in the profile.
2024
2025=item -e routine
2026
2027Exclude the given routine and its descendants from the profile.
2028
2029=item -f routine
2030
2031Display only the given routine and its descendants in the profile.
2032
2033=item -s
2034
2035Generate a summary file called F<gmon.sum> which then may be given
2036to subsequent gprof runs to accumulate data over several runs.
2037
2038=item -z
2039
2040Display routines that have zero usage.
2041
2042=back
2043
2044For more detailed explanation of the available commands and output
2045formats, see your own local documentation of gprof.
2046
51a35ef1
JH
2047=head2 GCC gcov Profiling
2048
10f58044 2049Starting from GCC 3.0 I<basic block profiling> is officially available
51a35ef1
JH
2050for the GNU CC.
2051
2052You can build a profiled version of perl called F<perl.gcov> by
2053invoking the make target "perl.gcov" (what is required that Perl must
2054be compiled using gcc with the flags C<-fprofile-arcs
2055-ftest-coverage>, you may need to re-Configure).
2056
2057Running the profiled version of Perl will cause profile output to be
2058generated. For each source file an accompanying ".da" file will be
2059created.
2060
2061To display the results you use the "gcov" utility (which should
2062be installed if you have gcc 3.0 or newer installed). F<gcov> is
2063run on source code files, like this
2064
2065 gcov sv.c
2066
2067which will cause F<sv.c.gcov> to be created. The F<.gcov> files
2068contain the source code annotated with relative frequencies of
2069execution indicated by "#" markers.
2070
2071Useful options of F<gcov> include C<-b> which will summarise the
2072basic block, branch, and function call coverage, and C<-c> which
2073instead of relative frequencies will use the actual counts. For
2074more information on the use of F<gcov> and basic block profiling
2075with gcc, see the latest GNU CC manual, as of GCC 3.0 see
2076
2077 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html
2078
2079and its section titled "8. gcov: a Test Coverage Program"
2080
2081 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132
2082
4ae3d70a
JH
2083=head2 Pixie Profiling
2084
51a35ef1
JH
2085Pixie is a profiling tool available on IRIX and Tru64 (aka Digital
2086UNIX aka DEC OSF/1) platforms. Pixie does its profiling using
10f58044 2087I<basic-block counting>.
4ae3d70a 2088
83f0ef60 2089You can build a profiled version of perl called F<perl.pixie> by
51a35ef1
JH
2090invoking the make target "perl.pixie" (what is required is that Perl
2091must be compiled using the C<-g> flag, you may need to re-Configure).
2092
2093In Tru64 a file called F<perl.Addrs> will also be silently created,
2094this file contains the addresses of the basic blocks. Running the
2095profiled version of Perl will create a new file called "perl.Counts"
2096which contains the counts for the basic block for that particular
2097program execution.
4ae3d70a 2098
51a35ef1 2099To display the results you use the F<prof> utility. The exact
4ae3d70a
JH
2100incantation depends on your operating system, "prof perl.Counts" in
2101IRIX, and "prof -pixie -all -L. perl" in Tru64.
2102
6c41479b
JH
2103In IRIX the following prof options are available:
2104
2105=over 4
2106
2107=item -h
2108
2109Reports the most heavily used lines in descending order of use.
6e36760b 2110Useful for finding the hotspot lines.
6c41479b
JH
2111
2112=item -l
2113
2114Groups lines by procedure, with procedures sorted in descending order of use.
2115Within a procedure, lines are listed in source order.
6e36760b 2116Useful for finding the hotspots of procedures.
6c41479b
JH
2117
2118=back
2119
2120In Tru64 the following options are available:
2121
2122=over 4
2123
3958b146 2124=item -p[rocedures]
6c41479b 2125
3958b146 2126Procedures sorted in descending order by the number of cycles executed
6e36760b 2127in each procedure. Useful for finding the hotspot procedures.
6c41479b
JH
2128(This is the default option.)
2129
24000d2f 2130=item -h[eavy]
6c41479b 2131
6e36760b
JH
2132Lines sorted in descending order by the number of cycles executed in
2133each line. Useful for finding the hotspot lines.
6c41479b 2134
24000d2f 2135=item -i[nvocations]
6c41479b 2136
6e36760b
JH
2137The called procedures are sorted in descending order by number of calls
2138made to the procedures. Useful for finding the most used procedures.
6c41479b 2139
24000d2f 2140=item -l[ines]
6c41479b
JH
2141
2142Grouped by procedure, sorted by cycles executed per procedure.
6e36760b 2143Useful for finding the hotspots of procedures.
6c41479b
JH
2144
2145=item -testcoverage
2146
2147The compiler emitted code for these lines, but the code was unexecuted.
2148
24000d2f 2149=item -z[ero]
6c41479b
JH
2150
2151Unexecuted procedures.
2152
aa500c9e 2153=back
6c41479b
JH
2154
2155For further information, see your system's manual pages for pixie and prof.
4ae3d70a 2156
a422fd2d
SC
2157=head2 CONCLUSION
2158
2159We've had a brief look around the Perl source, an overview of the stages
2160F<perl> goes through when it's running your code, and how to use a
902b9dbf
MF
2161debugger to poke at the Perl guts. We took a very simple problem and
2162demonstrated how to solve it fully - with documentation, regression
2163tests, and finally a patch for submission to p5p. Finally, we talked
2164about how to use external tools to debug and test Perl.
a422fd2d
SC
2165
2166I'd now suggest you read over those references again, and then, as soon
2167as possible, get your hands dirty. The best way to learn is by doing,
2168so:
2169
2170=over 3
2171
2172=item *
2173
2174Subscribe to perl5-porters, follow the patches and try and understand
2175them; don't be afraid to ask if there's a portion you're not clear on -
2176who knows, you may unearth a bug in the patch...
2177
2178=item *
2179
2180Keep up to date with the bleeding edge Perl distributions and get
2181familiar with the changes. Try and get an idea of what areas people are
2182working on and the changes they're making.
2183
2184=item *
2185
3e148164 2186Do read the README associated with your operating system, e.g. README.aix
a1f349fd
MB
2187on the IBM AIX OS. Don't hesitate to supply patches to that README if
2188you find anything missing or changed over a new OS release.
2189
2190=item *
2191
a422fd2d
SC
2192Find an area of Perl that seems interesting to you, and see if you can
2193work out how it works. Scan through the source, and step over it in the
2194debugger. Play, poke, investigate, fiddle! You'll probably get to
2195understand not just your chosen area but a much wider range of F<perl>'s
2196activity as well, and probably sooner than you'd think.
2197
2198=back
2199
2200=over 3
2201
2202=item I<The Road goes ever on and on, down from the door where it began.>
2203
2204=back
2205
2206If you can do these things, you've started on the long road to Perl porting.
2207Thanks for wanting to help make Perl better - and happy hacking!
2208
e8cd7eae
GS
2209=head1 AUTHOR
2210
2211This document was written by Nathan Torkington, and is maintained by
2212the perl5-porters mailing list.
2213