This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Indent fix
[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.
caf100c0 40For instance, Gurusamy Sarathy was the pumpking for the 5.6 release of
961f29c6
MB
41Perl, and Jarkko Hietaniemi was the pumpking for the 5.8 release, and
42Hugo van der Sanden and Rafael Garcia-Suarez share the pumpking for
43the 5.10 release.
e8cd7eae
GS
44
45In addition, various people are pumpkings for different things. For
961f29c6
MB
46instance, Andy Dougherty and Jarkko Hietaniemi did a grand job as the
47I<Configure> pumpkin up till the 5.8 release. For the 5.10 release
48H.Merijn Brand took over.
e8cd7eae
GS
49
50Larry sees Perl development along the lines of the US government:
51there's the Legislature (the porters), the Executive branch (the
52pumpkings), and the Supreme Court (Larry). The legislature can
53discuss and submit patches to the executive branch all they like, but
54the executive branch is free to veto them. Rarely, the Supreme Court
55will side with the executive branch over the legislature, or the
56legislature over the executive branch. Mostly, however, the
57legislature and the executive branch are supposed to get along and
58work out their differences without impeachment or court cases.
59
60You might sometimes see reference to Rule 1 and Rule 2. Larry's power
61as Supreme Court is expressed in The Rules:
62
63=over 4
64
65=item 1
66
67Larry is always by definition right about how Perl should behave.
68This means he has final veto power on the core functionality.
69
70=item 2
71
72Larry is allowed to change his mind about any matter at a later date,
73regardless of whether he previously invoked Rule 1.
74
75=back
76
77Got that? Larry is always right, even when he was wrong. It's rare
78to see either Rule exercised, but they are often alluded to.
79
80New features and extensions to the language are contentious, because
81the criteria used by the pumpkings, Larry, and other porters to decide
82which features should be implemented and incorporated are not codified
83in a few small design goals as with some other languages. Instead,
84the heuristics are flexible and often difficult to fathom. Here is
85one person's list, roughly in decreasing order of importance, of
86heuristics that new features have to be weighed against:
87
88=over 4
89
90=item Does concept match the general goals of Perl?
91
92These haven't been written anywhere in stone, but one approximation
93is:
94
95 1. Keep it fast, simple, and useful.
96 2. Keep features/concepts as orthogonal as possible.
97 3. No arbitrary limits (platforms, data sizes, cultures).
98 4. Keep it open and exciting to use/patch/advocate Perl everywhere.
99 5. Either assimilate new technologies, or build bridges to them.
100
101=item Where is the implementation?
102
103All the talk in the world is useless without an implementation. In
104almost every case, the person or people who argue for a new feature
105will be expected to be the ones who implement it. Porters capable
106of coding new features have their own agendas, and are not available
107to implement your (possibly good) idea.
108
109=item Backwards compatibility
110
111It's a cardinal sin to break existing Perl programs. New warnings are
112contentious--some say that a program that emits warnings is not
113broken, while others say it is. Adding keywords has the potential to
114break programs, changing the meaning of existing token sequences or
115functions might break programs.
116
117=item Could it be a module instead?
118
119Perl 5 has extension mechanisms, modules and XS, specifically to avoid
120the need to keep changing the Perl interpreter. You can write modules
121that export functions, you can give those functions prototypes so they
122can be called like built-in functions, you can even write XS code to
123mess with the runtime data structures of the Perl interpreter if you
124want to implement really complicated things. If it can be done in a
125module instead of in the core, it's highly unlikely to be added.
126
127=item Is the feature generic enough?
128
129Is this something that only the submitter wants added to the language,
130or would it be broadly useful? Sometimes, instead of adding a feature
131with a tight focus, the porters might decide to wait until someone
132implements the more generalized feature. For instance, instead of
133implementing a ``delayed evaluation'' feature, the porters are waiting
134for a macro system that would permit delayed evaluation and much more.
135
136=item Does it potentially introduce new bugs?
137
138Radical rewrites of large chunks of the Perl interpreter have the
139potential to introduce new bugs. The smaller and more localized the
140change, the better.
141
142=item Does it preclude other desirable features?
143
144A patch is likely to be rejected if it closes off future avenues of
145development. For instance, a patch that placed a true and final
146interpretation on prototypes is likely to be rejected because there
147are still options for the future of prototypes that haven't been
148addressed.
149
150=item Is the implementation robust?
151
152Good patches (tight code, complete, correct) stand more chance of
153going in. Sloppy or incorrect patches might be placed on the back
154burner until the pumpking has time to fix, or might be discarded
155altogether without further notice.
156
157=item Is the implementation generic enough to be portable?
158
159The worst patches make use of a system-specific features. It's highly
160unlikely that nonportable additions to the Perl language will be
161accepted.
162
a936dd3c
NC
163=item Is the implementation tested?
164
165Patches which change behaviour (fixing bugs or introducing new features)
166must include regression tests to verify that everything works as expected.
167Without tests provided by the original author, how can anyone else changing
168perl in the future be sure that they haven't unwittingly broken the behaviour
169the patch implements? And without tests, how can the patch's author be
9d077eaa 170confident that his/her hard work put into the patch won't be accidentally
a936dd3c
NC
171thrown away by someone in the future?
172
e8cd7eae
GS
173=item Is there enough documentation?
174
175Patches without documentation are probably ill-thought out or
176incomplete. Nothing can be added without documentation, so submitting
177a patch for the appropriate manpages as well as the source code is
a936dd3c 178always a good idea.
e8cd7eae
GS
179
180=item Is there another way to do it?
181
182Larry said ``Although the Perl Slogan is I<There's More Than One Way
183to Do It>, I hesitate to make 10 ways to do something''. This is a
184tricky heuristic to navigate, though--one man's essential addition is
185another man's pointless cruft.
186
187=item Does it create too much work?
188
189Work for the pumpking, work for Perl programmers, work for module
190authors, ... Perl is supposed to be easy.
191
f6c51b38
GS
192=item Patches speak louder than words
193
194Working code is always preferred to pie-in-the-sky ideas. A patch to
195add a feature stands a much higher chance of making it to the language
196than does a random feature request, no matter how fervently argued the
197request might be. This ties into ``Will it be useful?'', as the fact
198that someone took the time to make the patch demonstrates a strong
199desire for the feature.
200
e8cd7eae
GS
201=back
202
203If you're on the list, you might hear the word ``core'' bandied
204around. It refers to the standard distribution. ``Hacking on the
205core'' means you're changing the C source code to the Perl
206interpreter. ``A core module'' is one that ships with Perl.
207
a1f349fd
MB
208=head2 Keeping in sync
209
e8cd7eae 210The source code to the Perl interpreter, in its different versions, is
f224927c
JH
211kept in a repository managed by a revision control system ( which is
212currently the Perforce program, see http://perforce.com/ ). The
e8cd7eae
GS
213pumpkings and a few others have access to the repository to check in
214changes. Periodically the pumpking for the development version of Perl
215will release a new version, so the rest of the porters can see what's
2be4c08b
GS
216changed. The current state of the main trunk of repository, and patches
217that describe the individual changes that have happened since the last
218public release are available at this location:
219
0cfb3454 220 http://public.activestate.com/gsar/APC/
2be4c08b
GS
221 ftp://ftp.linux.activestate.com/pub/staff/gsar/APC/
222
0cfb3454
GS
223If you're looking for a particular change, or a change that affected
224a particular set of files, you may find the B<Perl Repository Browser>
225useful:
226
227 http://public.activestate.com/cgi-bin/perlbrowse
228
229You may also want to subscribe to the perl5-changes mailing list to
230receive a copy of each patch that gets submitted to the maintenance
231and development "branches" of the perl repository. See
232http://lists.perl.org/ for subscription information.
233
a1f349fd
MB
234If you are a member of the perl5-porters mailing list, it is a good
235thing to keep in touch with the most recent changes. If not only to
236verify if what you would have posted as a bug report isn't already
237solved in the most recent available perl development branch, also
238known as perl-current, bleading edge perl, bleedperl or bleadperl.
2be4c08b
GS
239
240Needless to say, the source code in perl-current is usually in a perpetual
241state of evolution. You should expect it to be very buggy. Do B<not> use
242it for any purpose other than testing and development.
e8cd7eae 243
3e148164
JH
244Keeping in sync with the most recent branch can be done in several ways,
245but the most convenient and reliable way is using B<rsync>, available at
246ftp://rsync.samba.org/pub/rsync/ . (You can also get the most recent
247branch by FTP.)
a1f349fd
MB
248
249If you choose to keep in sync using rsync, there are two approaches
3e148164 250to doing so:
a1f349fd
MB
251
252=over 4
253
254=item rsync'ing the source tree
255
3e148164 256Presuming you are in the directory where your perl source resides
a1f349fd
MB
257and you have rsync installed and available, you can `upgrade' to
258the bleadperl using:
259
260 # rsync -avz rsync://ftp.linux.activestate.com/perl-current/ .
261
262This takes care of updating every single item in the source tree to
263the latest applied patch level, creating files that are new (to your
264distribution) and setting date/time stamps of existing files to
265reflect the bleadperl status.
266
c6d0653e
LC
267Note that this will not delete any files that were in '.' before
268the rsync. Once you are sure that the rsync is running correctly,
269run it with the --delete and the --dry-run options like this:
270
271 # rsync -avz --delete --dry-run rsync://ftp.linux.activestate.com/perl-current/ .
272
273This will I<simulate> an rsync run that also deletes files not
274present in the bleadperl master copy. Observe the results from
275this run closely. If you are sure that the actual run would delete
276no files precious to you, you could remove the '--dry-run' option.
277
a1f349fd
MB
278You can than check what patch was the latest that was applied by
279looking in the file B<.patch>, which will show the number of the
280latest patch.
281
282If you have more than one machine to keep in sync, and not all of
283them have access to the WAN (so you are not able to rsync all the
284source trees to the real source), there are some ways to get around
285this problem.
286
287=over 4
288
289=item Using rsync over the LAN
290
291Set up a local rsync server which makes the rsynced source tree
3e148164 292available to the LAN and sync the other machines against this
a1f349fd
MB
293directory.
294
1577cd80 295From http://rsync.samba.org/README.html :
a1f349fd
MB
296
297 "Rsync uses rsh or ssh for communication. It does not need to be
298 setuid and requires no special privileges for installation. It
3958b146 299 does not require an inetd entry or a daemon. You must, however,
a1f349fd
MB
300 have a working rsh or ssh system. Using ssh is recommended for
301 its security features."
302
303=item Using pushing over the NFS
304
305Having the other systems mounted over the NFS, you can take an
3e148164
JH
306active pushing approach by checking the just updated tree against
307the other not-yet synced trees. An example would be
308
309 #!/usr/bin/perl -w
310
311 use strict;
312 use File::Copy;
313
314 my %MF = map {
315 m/(\S+)/;
316 $1 => [ (stat $1)[2, 7, 9] ]; # mode, size, mtime
317 } `cat MANIFEST`;
318
319 my %remote = map { $_ => "/$_/pro/3gl/CPAN/perl-5.7.1" } qw(host1 host2);
320
321 foreach my $host (keys %remote) {
322 unless (-d $remote{$host}) {
323 print STDERR "Cannot Xsync for host $host\n";
324 next;
325 }
326 foreach my $file (keys %MF) {
327 my $rfile = "$remote{$host}/$file";
328 my ($mode, $size, $mtime) = (stat $rfile)[2, 7, 9];
329 defined $size or ($mode, $size, $mtime) = (0, 0, 0);
330 $size == $MF{$file}[1] && $mtime == $MF{$file}[2] and next;
331 printf "%4s %-34s %8d %9d %8d %9d\n",
332 $host, $file, $MF{$file}[1], $MF{$file}[2], $size, $mtime;
333 unlink $rfile;
334 copy ($file, $rfile);
335 utime time, $MF{$file}[2], $rfile;
336 chmod $MF{$file}[0], $rfile;
337 }
338 }
339
340though this is not perfect. It could be improved with checking
a1f349fd
MB
341file checksums before updating. Not all NFS systems support
342reliable utime support (when used over the NFS).
343
344=back
345
346=item rsync'ing the patches
347
348The source tree is maintained by the pumpking who applies patches to
349the files in the tree. These patches are either created by the
350pumpking himself using C<diff -c> after updating the file manually or
351by applying patches sent in by posters on the perl5-porters list.
352These patches are also saved and rsync'able, so you can apply them
353yourself to the source files.
354
355Presuming you are in a directory where your patches reside, you can
3e148164 356get them in sync with
a1f349fd
MB
357
358 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
359
360This makes sure the latest available patch is downloaded to your
361patch directory.
362
3e148164 363It's then up to you to apply these patches, using something like
a1f349fd 364
df3477ff 365 # last=`ls -t *.gz | sed q`
a1f349fd
MB
366 # rsync -avz rsync://ftp.linux.activestate.com/perl-current-diffs/ .
367 # find . -name '*.gz' -newer $last -exec gzcat {} \; >blead.patch
368 # cd ../perl-current
369 # patch -p1 -N <../perl-current-diffs/blead.patch
370
371or, since this is only a hint towards how it works, use CPAN-patchaperl
372from Andreas K├Ânig to have better control over the patching process.
373
374=back
375
f7e1e956 376=head2 Why rsync the source tree
a1f349fd
MB
377
378=over 4
379
10f58044 380=item It's easier to rsync the source tree
a1f349fd
MB
381
382Since you don't have to apply the patches yourself, you are sure all
383files in the source tree are in the right state.
384
a1f349fd
MB
385=item It's more reliable
386
0cfb3454
GS
387While both the rsync-able source and patch areas are automatically
388updated every few minutes, keep in mind that applying patches may
389sometimes mean careful hand-holding, especially if your version of
390the C<patch> program does not understand how to deal with new files,
391files with 8-bit characters, or files without trailing newlines.
a1f349fd
MB
392
393=back
394
f7e1e956 395=head2 Why rsync the patches
a1f349fd
MB
396
397=over 4
398
10f58044 399=item It's easier to rsync the patches
a1f349fd
MB
400
401If you have more than one machine that you want to keep in track with
3e148164 402bleadperl, it's easier to rsync the patches only once and then apply
a1f349fd
MB
403them to all the source trees on the different machines.
404
405In case you try to keep in pace on 5 different machines, for which
406only one of them has access to the WAN, rsync'ing all the source
3e148164 407trees should than be done 5 times over the NFS. Having
a1f349fd 408rsync'ed the patches only once, I can apply them to all the source
3e148164 409trees automatically. Need you say more ;-)
a1f349fd
MB
410
411=item It's a good reference
412
413If you do not only like to have the most recent development branch,
414but also like to B<fix> bugs, or extend features, you want to dive
415into the sources. If you are a seasoned perl core diver, you don't
416need no manuals, tips, roadmaps, perlguts.pod or other aids to find
417your way around. But if you are a starter, the patches may help you
418in finding where you should start and how to change the bits that
419bug you.
420
421The file B<Changes> is updated on occasions the pumpking sees as his
422own little sync points. On those occasions, he releases a tar-ball of
423the current source tree (i.e. perl@7582.tar.gz), which will be an
424excellent point to start with when choosing to use the 'rsync the
425patches' scheme. Starting with perl@7582, which means a set of source
426files on which the latest applied patch is number 7582, you apply all
f18956b7 427succeeding patches available from then on (7583, 7584, ...).
a1f349fd
MB
428
429You can use the patches later as a kind of search archive.
430
431=over 4
432
433=item Finding a start point
434
435If you want to fix/change the behaviour of function/feature Foo, just
436scan the patches for patches that mention Foo either in the subject,
3e148164 437the comments, or the body of the fix. A good chance the patch shows
a1f349fd
MB
438you the files that are affected by that patch which are very likely
439to be the starting point of your journey into the guts of perl.
440
441=item Finding how to fix a bug
442
443If you've found I<where> the function/feature Foo misbehaves, but you
444don't know how to fix it (but you do know the change you want to
445make), you can, again, peruse the patches for similar changes and
446look how others apply the fix.
447
448=item Finding the source of misbehaviour
449
450When you keep in sync with bleadperl, the pumpking would love to
3958b146 451I<see> that the community efforts really work. So after each of his
a1f349fd
MB
452sync points, you are to 'make test' to check if everything is still
453in working order. If it is, you do 'make ok', which will send an OK
454report to perlbug@perl.org. (If you do not have access to a mailer
3e148164 455from the system you just finished successfully 'make test', you can
a1f349fd
MB
456do 'make okfile', which creates the file C<perl.ok>, which you can
457than take to your favourite mailer and mail yourself).
458
3958b146 459But of course, as always, things will not always lead to a success
a1f349fd
MB
460path, and one or more test do not pass the 'make test'. Before
461sending in a bug report (using 'make nok' or 'make nokfile'), check
462the mailing list if someone else has reported the bug already and if
463so, confirm it by replying to that message. If not, you might want to
464trace the source of that misbehaviour B<before> sending in the bug,
465which will help all the other porters in finding the solution.
466
3e148164
JH
467Here the saved patches come in very handy. You can check the list of
468patches to see which patch changed what file and what change caused
469the misbehaviour. If you note that in the bug report, it saves the
470one trying to solve it, looking for that point.
a1f349fd
MB
471
472=back
473
474If searching the patches is too bothersome, you might consider using
475perl's bugtron to find more information about discussions and
476ramblings on posted bugs.
477
3e148164
JH
478If you want to get the best of both worlds, rsync both the source
479tree for convenience, reliability and ease and rsync the patches
480for reference.
481
52315700
RF
482=back
483
fcc89a64
DR
484=head2 Working with the source
485
486Because you cannot use the Perforce client, you cannot easily generate
487diffs against the repository, nor will merges occur when you update
488via rsync. If you edit a file locally and then rsync against the
489latest source, changes made in the remote copy will I<overwrite> your
490local versions!
491
492The best way to deal with this is to maintain a tree of symlinks to
493the rsync'd source. Then, when you want to edit a file, you remove
494the symlink, copy the real file into the other tree, and edit it. You
495can then diff your edited file against the original to generate a
496patch, and you can safely update the original tree.
497
498Perl's F<Configure> script can generate this tree of symlinks for you.
499The following example assumes that you have used rsync to pull a copy
500of the Perl source into the F<perl-rsync> directory. In the directory
501above that one, you can execute the following commands:
502
503 mkdir perl-dev
504 cd perl-dev
505 ../perl-rsync/Configure -Dmksymlinks -Dusedevel -D"optimize=-g"
506
507This will start the Perl configuration process. After a few prompts,
508you should see something like this:
509
510 Symbolic links are supported.
511
512 Checking how to test for symbolic links...
513 Your builtin 'test -h' may be broken.
514 Trying external '/usr/bin/test -h'.
515 You can test for symbolic links with '/usr/bin/test -h'.
516
517 Creating the symbolic links...
518 (First creating the subdirectories...)
519 (Then creating the symlinks...)
520
521The specifics may vary based on your operating system, of course.
522After you see this, you can abort the F<Configure> script, and you
523will see that the directory you are in has a tree of symlinks to the
524F<perl-rsync> directories and files.
525
526If you plan to do a lot of work with the Perl source, here are some
527Bourne shell script functions that can make your life easier:
528
529 function edit {
530 if [ -L $1 ]; then
531 mv $1 $1.orig
532 cp $1.orig $1
533 vi $1
534 else
535 /bin/vi $1
536 fi
537 }
538
539 function unedit {
540 if [ -L $1.orig ]; then
541 rm $1
542 mv $1.orig $1
543 fi
544 }
545
546Replace "vi" with your favorite flavor of editor.
547
548Here is another function which will quickly generate a patch for the
549files which have been edited in your symlink tree:
550
551 mkpatchorig() {
552 local diffopts
553 for f in `find . -name '*.orig' | sed s,^\./,,`
554 do
555 case `echo $f | sed 's,.orig$,,;s,.*\.,,'` in
556 c) diffopts=-p ;;
557 pod) diffopts='-F^=' ;;
558 *) diffopts= ;;
559 esac
560 diff -du $diffopts $f `echo $f | sed 's,.orig$,,'`
561 done
562 }
563
564This function produces patches which include enough context to make
565your changes obvious. This makes it easier for the Perl pumpking(s)
566to review them when you send them to the perl5-porters list, and that
567means they're more likely to get applied.
568
569This function assumed a GNU diff, and may require some tweaking for
570other diff variants.
52315700 571
3fd28c4e 572=head2 Perlbug administration
52315700 573
3fd28c4e
RF
574There is a single remote administrative interface for modifying bug status,
575category, open issues etc. using the B<RT> I<bugtracker> system, maintained
576by I<Robert Spier>. Become an administrator, and close any bugs you can get
577your sticky mitts on:
52315700 578
3fd28c4e 579 http://rt.perl.org
52315700 580
3fd28c4e 581The bugtracker mechanism for B<perl5> bugs in particular is at:
52315700 582
3fd28c4e 583 http://bugs6.perl.org/perlbug
52315700 584
3fd28c4e 585To email the bug system administrators:
52315700 586
3fd28c4e 587 "perlbug-admin" <perlbug-admin@perl.org>
52315700 588
52315700 589
a1f349fd
MB
590=head2 Submitting patches
591
f7e1e956
MS
592Always submit patches to I<perl5-porters@perl.org>. If you're
593patching a core module and there's an author listed, send the author a
594copy (see L<Patching a core module>). This lets other porters review
595your patch, which catches a surprising number of errors in patches.
596Either use the diff program (available in source code form from
f224927c 597ftp://ftp.gnu.org/pub/gnu/ , or use Johan Vromans' I<makepatch>
f7e1e956
MS
598(available from I<CPAN/authors/id/JV/>). Unified diffs are preferred,
599but context diffs are accepted. Do not send RCS-style diffs or diffs
600without context lines. More information is given in the
601I<Porting/patching.pod> file in the Perl source distribution. Please
602patch against the latest B<development> version (e.g., if you're
603fixing a bug in the 5.005 track, patch against the latest 5.005_5x
604version). Only patches that survive the heat of the development
605branch get applied to maintenance versions.
606
607Your patch should update the documentation and test suite. See
608L<Writing a test>.
e8cd7eae
GS
609
610To report a bug in Perl, use the program I<perlbug> which comes with
611Perl (if you can't get Perl to work, send mail to the address
f18956b7 612I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through
e8cd7eae 613I<perlbug> feeds into the automated bug-tracking system, access to
f224927c 614which is provided through the web at http://bugs.perl.org/ . It
e8cd7eae
GS
615often pays to check the archives of the perl5-porters mailing list to
616see whether the bug you're reporting has been reported before, and if
617so whether it was considered a bug. See above for the location of
618the searchable archives.
619
f224927c 620The CPAN testers ( http://testers.cpan.org/ ) are a group of
ba139f7d
JH
621volunteers who test CPAN modules on a variety of platforms. Perl
622Smokers ( http://archives.develooper.com/daily-build@perl.org/ )
623automatically tests Perl source releases on platforms with various
624configurations. Both efforts welcome volunteers.
e8cd7eae 625
e8cd7eae
GS
626It's a good idea to read and lurk for a while before chipping in.
627That way you'll get to see the dynamic of the conversations, learn the
628personalities of the players, and hopefully be better prepared to make
629a useful contribution when do you speak up.
630
631If after all this you still think you want to join the perl5-porters
f6c51b38
GS
632mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To
633unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>.
e8cd7eae 634
a422fd2d
SC
635To hack on the Perl guts, you'll need to read the following things:
636
637=over 3
638
639=item L<perlguts>
640
641This is of paramount importance, since it's the documentation of what
642goes where in the Perl source. Read it over a couple of times and it
643might start to make sense - don't worry if it doesn't yet, because the
644best way to study it is to read it in conjunction with poking at Perl
645source, and we'll do that later on.
646
647You might also want to look at Gisle Aas's illustrated perlguts -
648there's no guarantee that this will be absolutely up-to-date with the
649latest documentation in the Perl core, but the fundamentals will be
1577cd80 650right. ( http://gisle.aas.no/perl/illguts/ )
a422fd2d
SC
651
652=item L<perlxstut> and L<perlxs>
653
654A working knowledge of XSUB programming is incredibly useful for core
655hacking; XSUBs use techniques drawn from the PP code, the portion of the
656guts that actually executes a Perl program. It's a lot gentler to learn
657those techniques from simple examples and explanation than from the core
658itself.
659
660=item L<perlapi>
661
662The documentation for the Perl API explains what some of the internal
663functions do, as well as the many macros used in the source.
664
665=item F<Porting/pumpkin.pod>
666
667This is a collection of words of wisdom for a Perl porter; some of it is
668only useful to the pumpkin holder, but most of it applies to anyone
669wanting to go about Perl development.
670
671=item The perl5-porters FAQ
672
7d7d5695
RGS
673This should be available from http://simon-cozens.org/writings/p5p-faq ;
674alternatively, you can get the FAQ emailed to you by sending mail to
675C<perl5-porters-faq@perl.org>. It contains hints on reading perl5-porters,
676information on how perl5-porters works and how Perl development in general
677works.
a422fd2d
SC
678
679=back
680
681=head2 Finding Your Way Around
682
683Perl maintenance can be split into a number of areas, and certain people
684(pumpkins) will have responsibility for each area. These areas sometimes
685correspond to files or directories in the source kit. Among the areas are:
686
687=over 3
688
689=item Core modules
690
691Modules shipped as part of the Perl core live in the F<lib/> and F<ext/>
692subdirectories: F<lib/> is for the pure-Perl modules, and F<ext/>
693contains the core XS modules.
694
f7e1e956
MS
695=item Tests
696
697There are tests for nearly all the modules, built-ins and major bits
698of functionality. Test files all have a .t suffix. Module tests live
699in the F<lib/> and F<ext/> directories next to the module being
700tested. Others live in F<t/>. See L<Writing a test>
701
a422fd2d
SC
702=item Documentation
703
704Documentation maintenance includes looking after everything in the
705F<pod/> directory, (as well as contributing new documentation) and
706the documentation to the modules in core.
707
708=item Configure
709
710The configure process is the way we make Perl portable across the
711myriad of operating systems it supports. Responsibility for the
712configure, build and installation process, as well as the overall
713portability of the core code rests with the configure pumpkin - others
714help out with individual operating systems.
715
716The files involved are the operating system directories, (F<win32/>,
717F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
718and F<Makefile>, as well as the metaconfig files which generate
719F<Configure>. (metaconfig isn't included in the core distribution.)
720
721=item Interpreter
722
723And of course, there's the core of the Perl interpreter itself. Let's
724have a look at that in a little more detail.
725
726=back
727
728Before we leave looking at the layout, though, don't forget that
729F<MANIFEST> contains not only the file names in the Perl distribution,
730but short descriptions of what's in them, too. For an overview of the
731important files, try this:
732
733 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
734
735=head2 Elements of the interpreter
736
737The work of the interpreter has two main stages: compiling the code
738into the internal representation, or bytecode, and then executing it.
739L<perlguts/Compiled code> explains exactly how the compilation stage
740happens.
741
742Here is a short breakdown of perl's operation:
743
744=over 3
745
746=item Startup
747
748The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
749This is very high-level code, enough to fit on a single screen, and it
750resembles the code found in L<perlembed>; most of the real action takes
751place in F<perl.c>
752
753First, F<perlmain.c> allocates some memory and constructs a Perl
754interpreter:
755
756 1 PERL_SYS_INIT3(&argc,&argv,&env);
757 2
758 3 if (!PL_do_undump) {
759 4 my_perl = perl_alloc();
760 5 if (!my_perl)
761 6 exit(1);
762 7 perl_construct(my_perl);
763 8 PL_perl_destruct_level = 0;
764 9 }
765
766Line 1 is a macro, and its definition is dependent on your operating
767system. Line 3 references C<PL_do_undump>, a global variable - all
768global variables in Perl start with C<PL_>. This tells you whether the
769current running program was created with the C<-u> flag to perl and then
770F<undump>, which means it's going to be false in any sane context.
771
772Line 4 calls a function in F<perl.c> to allocate memory for a Perl
773interpreter. It's quite a simple function, and the guts of it looks like
774this:
775
776 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
777
778Here you see an example of Perl's system abstraction, which we'll see
779later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
780own C<malloc> as defined in F<malloc.c> if you selected that option at
781configure time.
782
783Next, in line 7, we construct the interpreter; this sets up all the
784special variables that Perl needs, the stacks, and so on.
785
786Now we pass Perl the command line options, and tell it to go:
787
788 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
789 if (!exitstatus) {
790 exitstatus = perl_run(my_perl);
791 }
792
793
794C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
795in F<perl.c>, which processes the command line options, sets up any
796statically linked XS modules, opens the program and calls C<yyparse> to
797parse it.
798
799=item Parsing
800
801The aim of this stage is to take the Perl source, and turn it into an op
802tree. We'll see what one of those looks like later. Strictly speaking,
803there's three things going on here.
804
805C<yyparse>, the parser, lives in F<perly.c>, although you're better off
806reading the original YACC input in F<perly.y>. (Yes, Virginia, there
807B<is> a YACC grammar for Perl!) The job of the parser is to take your
808code and `understand' it, splitting it into sentences, deciding which
809operands go with which operators and so on.
810
811The parser is nobly assisted by the lexer, which chunks up your input
812into tokens, and decides what type of thing each token is: a variable
813name, an operator, a bareword, a subroutine, a core function, and so on.
814The main point of entry to the lexer is C<yylex>, and that and its
815associated routines can be found in F<toke.c>. Perl isn't much like
816other computer languages; it's highly context sensitive at times, it can
817be tricky to work out what sort of token something is, or where a token
818ends. As such, there's a lot of interplay between the tokeniser and the
819parser, which can get pretty frightening if you're not used to it.
820
821As the parser understands a Perl program, it builds up a tree of
822operations for the interpreter to perform during execution. The routines
823which construct and link together the various operations are to be found
824in F<op.c>, and will be examined later.
825
826=item Optimization
827
828Now the parsing stage is complete, and the finished tree represents
829the operations that the Perl interpreter needs to perform to execute our
830program. Next, Perl does a dry run over the tree looking for
831optimisations: constant expressions such as C<3 + 4> will be computed
832now, and the optimizer will also see if any multiple operations can be
833replaced with a single one. For instance, to fetch the variable C<$foo>,
834instead of grabbing the glob C<*foo> and looking at the scalar
835component, the optimizer fiddles the op tree to use a function which
836directly looks up the scalar in question. The main optimizer is C<peep>
837in F<op.c>, and many ops have their own optimizing functions.
838
839=item Running
840
841Now we're finally ready to go: we have compiled Perl byte code, and all
842that's left to do is run it. The actual execution is done by the
843C<runops_standard> function in F<run.c>; more specifically, it's done by
844these three innocent looking lines:
845
846 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
847 PERL_ASYNC_CHECK();
848 }
849
850You may be more comfortable with the Perl version of that:
851
852 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
853
854Well, maybe not. Anyway, each op contains a function pointer, which
855stipulates the function which will actually carry out the operation.
856This function will return the next op in the sequence - this allows for
857things like C<if> which choose the next op dynamically at run time.
858The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
859execution if required.
860
861The actual functions called are known as PP code, and they're spread
862between four files: F<pp_hot.c> contains the `hot' code, which is most
863often used and highly optimized, F<pp_sys.c> contains all the
864system-specific functions, F<pp_ctl.c> contains the functions which
865implement control structures (C<if>, C<while> and the like) and F<pp.c>
866contains everything else. These are, if you like, the C code for Perl's
867built-in functions and operators.
868
dfc98234
DM
869Note that each C<pp_> function is expected to return a pointer to the next
870op. Calls to perl subs (and eval blocks) are handled within the same
871runops loop, and do not consume extra space on the C stack. For example,
872C<pp_entersub> and C<pp_entertry> just push a C<CxSUB> or C<CxEVAL> block
873struct onto the context stack which contain the address of the op
874following the sub call or eval. They then return the first op of that sub
875or eval block, and so execution continues of that sub or block. Later, a
876C<pp_leavesub> or C<pp_leavetry> op pops the C<CxSUB> or C<CxEVAL>,
877retrieves the return op from it, and returns it.
878
879=item Exception handing
880
881Perl's exception handing (ie C<die> etc) is built on top of the low-level
882C<setjmp()>/C<longjmp()> C-library functions. These basically provide a
883way to capture the current PC and SP registers and later restore them; ie
884a C<longjmp()> continues at the point in code where a previous C<setjmp()>
885was done, with anything further up on the C stack being lost. This is why
886code should always save values using C<SAVE_FOO> rather than in auto
887variables.
888
889The perl core wraps C<setjmp()> etc in the macros C<JMPENV_PUSH> and
890C<JMPENV_JUMP>. The basic rule of perl exceptions is that C<exit>, and
891C<die> (in the absence of C<eval>) perform a C<JMPENV_JUMP(2)>, while
892C<die> within C<eval> does a C<JMPENV_JUMP(3)>.
893
894At entry points to perl, such as C<perl_parse()>, C<perl_run()> and
895C<call_sv(cv, G_EVAL)> each does a C<JMPENV_PUSH>, then enter a runops
896loop or whatever, and handle possible exception returns. For a 2 return,
897final cleanup is performed, such as popping stacks and calling C<CHECK> or
898C<END> blocks. Amongst other things, this is how scope cleanup still
899occurs during an C<exit>.
900
901If a C<die> can find a C<CxEVAL> block on the context stack, then the
902stack is popped to that level and the return op in that block is assigned
903to C<PL_restartop>; then a C<JMPENV_JUMP(3)> is performed. This normally
904passes control back to the guard. In the case of C<perl_run> and
905C<call_sv>, a non-null C<PL_restartop> triggers re-entry to the runops
906loop. The is the normal way that C<die> or C<croak> is handled within an
907C<eval>.
908
909Sometimes ops are executed within an inner runops loop, such as tie, sort
910or overload code. In this case, something like
911
912 sub FETCH { eval { die } }
913
914would cause a longjmp right back to the guard in C<perl_run>, popping both
915runops loops, which is clearly incorrect. One way to avoid this is for the
916tie code to do a C<JMPENV_PUSH> before executing C<FETCH> in the inner
917runops loop, but for efficiency reasons, perl in fact just sets a flag,
918using C<CATCH_SET(TRUE)>. The C<pp_require>, C<pp_entereval> and
919C<pp_entertry> ops check this flag, and if true, they call C<docatch>,
920which does a C<JMPENV_PUSH> and starts a new runops level to execute the
921code, rather than doing it on the current loop.
922
923As a further optimisation, on exit from the eval block in the C<FETCH>,
924execution of the code following the block is still carried on in the inner
925loop. When an exception is raised, C<docatch> compares the C<JMPENV>
926level of the C<CxEVAL> with C<PL_top_env> and if they differ, just
927re-throws the exception. In this way any inner loops get popped.
928
929Here's an example.
930
931 1: eval { tie @a, 'A' };
932 2: sub A::TIEARRAY {
933 3: eval { die };
934 4: die;
935 5: }
936
937To run this code, C<perl_run> is called, which does a C<JMPENV_PUSH> then
938enters a runops loop. This loop executes the eval and tie ops on line 1,
939with the eval pushing a C<CxEVAL> onto the context stack.
940
941The C<pp_tie> does a C<CATCH_SET(TRUE)>, then starts a second runops loop
942to execute the body of C<TIEARRAY>. When it executes the entertry op on
943line 3, C<CATCH_GET> is true, so C<pp_entertry> calls C<docatch> which
944does a C<JMPENV_PUSH> and starts a third runops loop, which then executes
945the die op. At this point the C call stack looks like this:
946
947 Perl_pp_die
948 Perl_runops # third loop
949 S_docatch_body
950 S_docatch
951 Perl_pp_entertry
952 Perl_runops # second loop
953 S_call_body
954 Perl_call_sv
955 Perl_pp_tie
956 Perl_runops # first loop
957 S_run_body
958 perl_run
959 main
960
961and the context and data stacks, as shown by C<-Dstv>, look like:
962
963 STACK 0: MAIN
964 CX 0: BLOCK =>
965 CX 1: EVAL => AV() PV("A"\0)
966 retop=leave
967 STACK 1: MAGIC
968 CX 0: SUB =>
969 retop=(null)
970 CX 1: EVAL => *
971 retop=nextstate
972
973The die pops the first C<CxEVAL> off the context stack, sets
974C<PL_restartop> from it, does a C<JMPENV_JUMP(3)>, and control returns to
975the top C<docatch>. This then starts another third-level runops level,
976which executes the nextstate, pushmark and die ops on line 4. At the point
977that the second C<pp_die> is called, the C call stack looks exactly like
978that above, even though we are no longer within an inner eval; this is
979because of the optimization mentioned earlier. However, the context stack
980now looks like this, ie with the top CxEVAL popped:
981
982 STACK 0: MAIN
983 CX 0: BLOCK =>
984 CX 1: EVAL => AV() PV("A"\0)
985 retop=leave
986 STACK 1: MAGIC
987 CX 0: SUB =>
988 retop=(null)
989
990The die on line 4 pops the context stack back down to the CxEVAL, leaving
991it as:
992
993 STACK 0: MAIN
994 CX 0: BLOCK =>
995
996As usual, C<PL_restartop> is extracted from the C<CxEVAL>, and a
997C<JMPENV_JUMP(3)> done, which pops the C stack back to the docatch:
998
999 S_docatch
1000 Perl_pp_entertry
1001 Perl_runops # second loop
1002 S_call_body
1003 Perl_call_sv
1004 Perl_pp_tie
1005 Perl_runops # first loop
1006 S_run_body
1007 perl_run
1008 main
1009
1010In this case, because the C<JMPENV> level recorded in the C<CxEVAL>
1011differs from the current one, C<docatch> just does a C<JMPENV_JUMP(3)>
1012and the C stack unwinds to:
1013
1014 perl_run
1015 main
1016
1017Because C<PL_restartop> is non-null, C<run_body> starts a new runops loop
1018and execution continues.
1019
a422fd2d
SC
1020=back
1021
1022=head2 Internal Variable Types
1023
1024You should by now have had a look at L<perlguts>, which tells you about
1025Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
1026that now.
1027
1028These variables are used not only to represent Perl-space variables, but
1029also any constants in the code, as well as some structures completely
1030internal to Perl. The symbol table, for instance, is an ordinary Perl
1031hash. Your code is represented by an SV as it's read into the parser;
1032any program files you call are opened via ordinary Perl filehandles, and
1033so on.
1034
1035The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
1036Perl program. Let's see, for instance, how Perl treats the constant
1037C<"hello">.
1038
1039 % perl -MDevel::Peek -e 'Dump("hello")'
1040 1 SV = PV(0xa041450) at 0xa04ecbc
1041 2 REFCNT = 1
1042 3 FLAGS = (POK,READONLY,pPOK)
1043 4 PV = 0xa0484e0 "hello"\0
1044 5 CUR = 5
1045 6 LEN = 6
1046
1047Reading C<Devel::Peek> output takes a bit of practise, so let's go
1048through it line by line.
1049
1050Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
1051memory. SVs themselves are very simple structures, but they contain a
1052pointer to a more complex structure. In this case, it's a PV, a
1053structure which holds a string value, at location C<0xa041450>. Line 2
1054is the reference count; there are no other references to this data, so
1055it's 1.
1056
1057Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
1058read-only SV (because it's a constant) and the data is a PV internally.
1059Next we've got the contents of the string, starting at location
1060C<0xa0484e0>.
1061
1062Line 5 gives us the current length of the string - note that this does
1063B<not> include the null terminator. Line 6 is not the length of the
1064string, but the length of the currently allocated buffer; as the string
1065grows, Perl automatically extends the available storage via a routine
1066called C<SvGROW>.
1067
1068You can get at any of these quantities from C very easily; just add
1069C<Sv> to the name of the field shown in the snippet, and you've got a
1070macro which will return the value: C<SvCUR(sv)> returns the current
1071length of the string, C<SvREFCOUNT(sv)> returns the reference count,
1072C<SvPV(sv, len)> returns the string itself with its length, and so on.
1073More macros to manipulate these properties can be found in L<perlguts>.
1074
1075Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
1076
1077 1 void
1078 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
1079 3 {
1080 4 STRLEN tlen;
1081 5 char *junk;
1082
1083 6 junk = SvPV_force(sv, tlen);
1084 7 SvGROW(sv, tlen + len + 1);
1085 8 if (ptr == junk)
1086 9 ptr = SvPVX(sv);
1087 10 Move(ptr,SvPVX(sv)+tlen,len,char);
1088 11 SvCUR(sv) += len;
1089 12 *SvEND(sv) = '\0';
1090 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
1091 14 SvTAINT(sv);
1092 15 }
1093
1094This is a function which adds a string, C<ptr>, of length C<len> onto
1095the end of the PV stored in C<sv>. The first thing we do in line 6 is
1096make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
1097macro to force a PV. As a side effect, C<tlen> gets set to the current
1098value of the PV, and the PV itself is returned to C<junk>.
1099
b1866b2d 1100In line 7, we make sure that the SV will have enough room to accommodate
a422fd2d
SC
1101the old string, the new string and the null terminator. If C<LEN> isn't
1102big enough, C<SvGROW> will reallocate space for us.
1103
1104Now, if C<junk> is the same as the string we're trying to add, we can
1105grab the string directly from the SV; C<SvPVX> is the address of the PV
1106in the SV.
1107
1108Line 10 does the actual catenation: the C<Move> macro moves a chunk of
1109memory around: we move the string C<ptr> to the end of the PV - that's
1110the start of the PV plus its current length. We're moving C<len> bytes
1111of type C<char>. After doing so, we need to tell Perl we've extended the
1112string, by altering C<CUR> to reflect the new length. C<SvEND> is a
1113macro which gives us the end of the string, so that needs to be a
1114C<"\0">.
1115
1116Line 13 manipulates the flags; since we've changed the PV, any IV or NV
1117values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
1e54db1a 1118want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF-8-aware
a422fd2d
SC
1119version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
1120and turns on POK. The final C<SvTAINT> is a macro which launders tainted
1121data if taint mode is turned on.
1122
1123AVs and HVs are more complicated, but SVs are by far the most common
1124variable type being thrown around. Having seen something of how we
1125manipulate these, let's go on and look at how the op tree is
1126constructed.
1127
1128=head2 Op Trees
1129
1130First, what is the op tree, anyway? The op tree is the parsed
1131representation of your program, as we saw in our section on parsing, and
1132it's the sequence of operations that Perl goes through to execute your
1133program, as we saw in L</Running>.
1134
1135An op is a fundamental operation that Perl can perform: all the built-in
1136functions and operators are ops, and there are a series of ops which
1137deal with concepts the interpreter needs internally - entering and
1138leaving a block, ending a statement, fetching a variable, and so on.
1139
1140The op tree is connected in two ways: you can imagine that there are two
1141"routes" through it, two orders in which you can traverse the tree.
1142First, parse order reflects how the parser understood the code, and
1143secondly, execution order tells perl what order to perform the
1144operations in.
1145
1146The easiest way to examine the op tree is to stop Perl after it has
1147finished parsing, and get it to dump out the tree. This is exactly what
7d7d5695
RGS
1148the compiler backends L<B::Terse|B::Terse>, L<B::Concise|B::Concise>
1149and L<B::Debug|B::Debug> do.
a422fd2d
SC
1150
1151Let's have a look at how Perl sees C<$a = $b + $c>:
1152
1153 % perl -MO=Terse -e '$a=$b+$c'
1154 1 LISTOP (0x8179888) leave
1155 2 OP (0x81798b0) enter
1156 3 COP (0x8179850) nextstate
1157 4 BINOP (0x8179828) sassign
1158 5 BINOP (0x8179800) add [1]
1159 6 UNOP (0x81796e0) null [15]
1160 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
1161 8 UNOP (0x81797e0) null [15]
1162 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
1163 10 UNOP (0x816b4f0) null [15]
1164 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
1165
1166Let's start in the middle, at line 4. This is a BINOP, a binary
1167operator, which is at location C<0x8179828>. The specific operator in
1168question is C<sassign> - scalar assignment - and you can find the code
1169which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
1170binary operator, it has two children: the add operator, providing the
1171result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
1172line 10.
1173
1174Line 10 is the null op: this does exactly nothing. What is that doing
1175there? If you see the null op, it's a sign that something has been
1176optimized away after parsing. As we mentioned in L</Optimization>,
1177the optimization stage sometimes converts two operations into one, for
1178example when fetching a scalar variable. When this happens, instead of
1179rewriting the op tree and cleaning up the dangling pointers, it's easier
1180just to replace the redundant operation with the null op. Originally,
1181the tree would have looked like this:
1182
1183 10 SVOP (0x816b4f0) rv2sv [15]
1184 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
1185
1186That is, fetch the C<a> entry from the main symbol table, and then look
1187at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
1188happens to do both these things.
1189
1190The right hand side, starting at line 5 is similar to what we've just
1191seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
1192two C<gvsv>s.
1193
1194Now, what's this about?
1195
1196 1 LISTOP (0x8179888) leave
1197 2 OP (0x81798b0) enter
1198 3 COP (0x8179850) nextstate
1199
1200C<enter> and C<leave> are scoping ops, and their job is to perform any
1201housekeeping every time you enter and leave a block: lexical variables
1202are tidied up, unreferenced variables are destroyed, and so on. Every
1203program will have those first three lines: C<leave> is a list, and its
1204children are all the statements in the block. Statements are delimited
1205by C<nextstate>, so a block is a collection of C<nextstate> ops, with
1206the ops to be performed for each statement being the children of
1207C<nextstate>. C<enter> is a single op which functions as a marker.
1208
1209That's how Perl parsed the program, from top to bottom:
1210
1211 Program
1212 |
1213 Statement
1214 |
1215 =
1216 / \
1217 / \
1218 $a +
1219 / \
1220 $b $c
1221
1222However, it's impossible to B<perform> the operations in this order:
1223you have to find the values of C<$b> and C<$c> before you add them
1224together, for instance. So, the other thread that runs through the op
1225tree is the execution order: each op has a field C<op_next> which points
1226to the next op to be run, so following these pointers tells us how perl
1227executes the code. We can traverse the tree in this order using
1228the C<exec> option to C<B::Terse>:
1229
1230 % perl -MO=Terse,exec -e '$a=$b+$c'
1231 1 OP (0x8179928) enter
1232 2 COP (0x81798c8) nextstate
1233 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
1234 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
1235 5 BINOP (0x8179878) add [1]
1236 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
1237 7 BINOP (0x81798a0) sassign
1238 8 LISTOP (0x8179900) leave
1239
1240This probably makes more sense for a human: enter a block, start a
1241statement. Get the values of C<$b> and C<$c>, and add them together.
1242Find C<$a>, and assign one to the other. Then leave.
1243
1244The way Perl builds up these op trees in the parsing process can be
1245unravelled by examining F<perly.y>, the YACC grammar. Let's take the
1246piece we need to construct the tree for C<$a = $b + $c>
1247
1248 1 term : term ASSIGNOP term
1249 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
1250 3 | term ADDOP term
1251 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1252
1253If you're not used to reading BNF grammars, this is how it works: You're
1254fed certain things by the tokeniser, which generally end up in upper
1255case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
1256code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
1257`terminal symbols', because you can't get any simpler than them.
1258
1259The grammar, lines one and three of the snippet above, tells you how to
1260build up more complex forms. These complex forms, `non-terminal symbols'
1261are generally placed in lower case. C<term> here is a non-terminal
1262symbol, representing a single expression.
1263
1264The grammar gives you the following rule: you can make the thing on the
1265left of the colon if you see all the things on the right in sequence.
1266This is called a "reduction", and the aim of parsing is to completely
1267reduce the input. There are several different ways you can perform a
1268reduction, separated by vertical bars: so, C<term> followed by C<=>
1269followed by C<term> makes a C<term>, and C<term> followed by C<+>
1270followed by C<term> can also make a C<term>.
1271
1272So, if you see two terms with an C<=> or C<+>, between them, you can
1273turn them into a single expression. When you do this, you execute the
1274code in the block on the next line: if you see C<=>, you'll do the code
1275in line 2. If you see C<+>, you'll do the code in line 4. It's this code
1276which contributes to the op tree.
1277
1278 | term ADDOP term
1279 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1280
1281What this does is creates a new binary op, and feeds it a number of
1282variables. The variables refer to the tokens: C<$1> is the first token in
1283the input, C<$2> the second, and so on - think regular expression
1284backreferences. C<$$> is the op returned from this reduction. So, we
1285call C<newBINOP> to create a new binary operator. The first parameter to
1286C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
1287operator, so we want the type to be C<ADDOP>. We could specify this
1288directly, but it's right there as the second token in the input, so we
1289use C<$2>. The second parameter is the op's flags: 0 means `nothing
1290special'. Then the things to add: the left and right hand side of our
1291expression, in scalar context.
1292
1293=head2 Stacks
1294
1295When perl executes something like C<addop>, how does it pass on its
1296results to the next op? The answer is, through the use of stacks. Perl
1297has a number of stacks to store things it's currently working on, and
1298we'll look at the three most important ones here.
1299
1300=over 3
1301
1302=item Argument stack
1303
1304Arguments are passed to PP code and returned from PP code using the
1305argument stack, C<ST>. The typical way to handle arguments is to pop
1306them off the stack, deal with them how you wish, and then push the result
1307back onto the stack. This is how, for instance, the cosine operator
1308works:
1309
1310 NV value;
1311 value = POPn;
1312 value = Perl_cos(value);
1313 XPUSHn(value);
1314
1315We'll see a more tricky example of this when we consider Perl's macros
1316below. C<POPn> gives you the NV (floating point value) of the top SV on
1317the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
1318the result back as an NV. The C<X> in C<XPUSHn> means that the stack
1319should be extended if necessary - it can't be necessary here, because we
1320know there's room for one more item on the stack, since we've just
1321removed one! The C<XPUSH*> macros at least guarantee safety.
1322
1323Alternatively, you can fiddle with the stack directly: C<SP> gives you
1324the first element in your portion of the stack, and C<TOP*> gives you
1325the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
1326negation of an integer:
1327
1328 SETi(-TOPi);
1329
1330Just set the integer value of the top stack entry to its negation.
1331
1332Argument stack manipulation in the core is exactly the same as it is in
1333XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
1334description of the macros used in stack manipulation.
1335
1336=item Mark stack
1337
1338I say `your portion of the stack' above because PP code doesn't
1339necessarily get the whole stack to itself: if your function calls
1340another function, you'll only want to expose the arguments aimed for the
1341called function, and not (necessarily) let it get at your own data. The
1342way we do this is to have a `virtual' bottom-of-stack, exposed to each
1343function. The mark stack keeps bookmarks to locations in the argument
1344stack usable by each function. For instance, when dealing with a tied
1345variable, (internally, something with `P' magic) Perl has to call
1346methods for accesses to the tied variables. However, we need to separate
1347the arguments exposed to the method to the argument exposed to the
1348original function - the store or fetch or whatever it may be. Here's how
1349the tied C<push> is implemented; see C<av_push> in F<av.c>:
1350
1351 1 PUSHMARK(SP);
1352 2 EXTEND(SP,2);
1353 3 PUSHs(SvTIED_obj((SV*)av, mg));
1354 4 PUSHs(val);
1355 5 PUTBACK;
1356 6 ENTER;
1357 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1358 8 LEAVE;
1359 9 POPSTACK;
13a2d996 1360
a422fd2d
SC
1361The lines which concern the mark stack are the first, fifth and last
1362lines: they save away, restore and remove the current position of the
1363argument stack.
1364
1365Let's examine the whole implementation, for practice:
1366
1367 1 PUSHMARK(SP);
1368
1369Push the current state of the stack pointer onto the mark stack. This is
1370so that when we've finished adding items to the argument stack, Perl
1371knows how many things we've added recently.
1372
1373 2 EXTEND(SP,2);
1374 3 PUSHs(SvTIED_obj((SV*)av, mg));
1375 4 PUSHs(val);
1376
1377We're going to add two more items onto the argument stack: when you have
1378a tied array, the C<PUSH> subroutine receives the object and the value
1379to be pushed, and that's exactly what we have here - the tied object,
1380retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
1381
1382 5 PUTBACK;
1383
1384Next we tell Perl to make the change to the global stack pointer: C<dSP>
1385only gave us a local copy, not a reference to the global.
1386
1387 6 ENTER;
1388 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1389 8 LEAVE;
1390
1391C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
1392variables are tidied up, everything that has been localised gets
1393its previous value returned, and so on. Think of them as the C<{> and
1394C<}> of a Perl block.
1395
1396To actually do the magic method call, we have to call a subroutine in
1397Perl space: C<call_method> takes care of that, and it's described in
1398L<perlcall>. We call the C<PUSH> method in scalar context, and we're
1399going to discard its return value.
1400
1401 9 POPSTACK;
1402
1403Finally, we remove the value we placed on the mark stack, since we
1404don't need it any more.
1405
1406=item Save stack
1407
1408C doesn't have a concept of local scope, so perl provides one. We've
1409seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
1410stack implements the C equivalent of, for example:
1411
1412 {
1413 local $foo = 42;
1414 ...
1415 }
1416
1417See L<perlguts/Localising Changes> for how to use the save stack.
1418
1419=back
1420
1421=head2 Millions of Macros
1422
1423One thing you'll notice about the Perl source is that it's full of
1424macros. Some have called the pervasive use of macros the hardest thing
1425to understand, others find it adds to clarity. Let's take an example,
1426the code which implements the addition operator:
1427
1428 1 PP(pp_add)
1429 2 {
39644a26 1430 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
a422fd2d
SC
1431 4 {
1432 5 dPOPTOPnnrl_ul;
1433 6 SETn( left + right );
1434 7 RETURN;
1435 8 }
1436 9 }
1437
1438Every line here (apart from the braces, of course) contains a macro. The
1439first line sets up the function declaration as Perl expects for PP code;
1440line 3 sets up variable declarations for the argument stack and the
1441target, the return value of the operation. Finally, it tries to see if
1442the addition operation is overloaded; if so, the appropriate subroutine
1443is called.
1444
1445Line 5 is another variable declaration - all variable declarations start
1446with C<d> - which pops from the top of the argument stack two NVs (hence
1447C<nn>) and puts them into the variables C<right> and C<left>, hence the
1448C<rl>. These are the two operands to the addition operator. Next, we
1449call C<SETn> to set the NV of the return value to the result of adding
1450the two values. This done, we return - the C<RETURN> macro makes sure
1451that our return value is properly handled, and we pass the next operator
1452to run back to the main run loop.
1453
1454Most of these macros are explained in L<perlapi>, and some of the more
1455important ones are explained in L<perlxs> as well. Pay special attention
1456to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
1457the C<[pad]THX_?> macros.
1458
52d59bef
JH
1459=head2 The .i Targets
1460
1461You can expand the macros in a F<foo.c> file by saying
1462
1463 make foo.i
1464
1465which will expand the macros using cpp. Don't be scared by the results.
1466
a422fd2d
SC
1467=head2 Poking at Perl
1468
1469To really poke around with Perl, you'll probably want to build Perl for
1470debugging, like this:
1471
1472 ./Configure -d -D optimize=-g
1473 make
1474
1475C<-g> is a flag to the C compiler to have it produce debugging
1476information which will allow us to step through a running program.
1477F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
1478enables all the internal debugging code in Perl. There are a whole bunch
1479of things you can debug with this: L<perlrun> lists them all, and the
1480best way to find out about them is to play about with them. The most
1481useful options are probably
1482
1483 l Context (loop) stack processing
1484 t Trace execution
1485 o Method and overloading resolution
1486 c String/numeric conversions
1487
1488Some of the functionality of the debugging code can be achieved using XS
1489modules.
13a2d996 1490
a422fd2d
SC
1491 -Dr => use re 'debug'
1492 -Dx => use O 'Debug'
1493
1494=head2 Using a source-level debugger
1495
1496If the debugging output of C<-D> doesn't help you, it's time to step
1497through perl's execution with a source-level debugger.
1498
1499=over 3
1500
1501=item *
1502
1503We'll use C<gdb> for our examples here; the principles will apply to any
1504debugger, but check the manual of the one you're using.
1505
1506=back
1507
1508To fire up the debugger, type
1509
1510 gdb ./perl
1511
1512You'll want to do that in your Perl source tree so the debugger can read
1513the source code. You should see the copyright message, followed by the
1514prompt.
1515
1516 (gdb)
1517
1518C<help> will get you into the documentation, but here are the most
1519useful commands:
1520
1521=over 3
1522
1523=item run [args]
1524
1525Run the program with the given arguments.
1526
1527=item break function_name
1528
1529=item break source.c:xxx
1530
1531Tells the debugger that we'll want to pause execution when we reach
cea6626f 1532either the named function (but see L<perlguts/Internal Functions>!) or the given
a422fd2d
SC
1533line in the named source file.
1534
1535=item step
1536
1537Steps through the program a line at a time.
1538
1539=item next
1540
1541Steps through the program a line at a time, without descending into
1542functions.
1543
1544=item continue
1545
1546Run until the next breakpoint.
1547
1548=item finish
1549
1550Run until the end of the current function, then stop again.
1551
13a2d996 1552=item 'enter'
a422fd2d
SC
1553
1554Just pressing Enter will do the most recent operation again - it's a
1555blessing when stepping through miles of source code.
1556
1557=item print
1558
1559Execute the given C code and print its results. B<WARNING>: Perl makes
52d59bef
JH
1560heavy use of macros, and F<gdb> does not necessarily support macros
1561(see later L</"gdb macro support">). You'll have to substitute them
1562yourself, or to invoke cpp on the source code files
1563(see L</"The .i Targets">)
1564So, for instance, you can't say
a422fd2d
SC
1565
1566 print SvPV_nolen(sv)
1567
1568but you have to say
1569
1570 print Perl_sv_2pv_nolen(sv)
1571
ffc145e8
RK
1572=back
1573
a422fd2d
SC
1574You may find it helpful to have a "macro dictionary", which you can
1575produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
52d59bef
JH
1576recursively apply those macros for you.
1577
1578=head2 gdb macro support
a422fd2d 1579
52d59bef 1580Recent versions of F<gdb> have fairly good macro support, but
ea031e66
RGS
1581in order to use it you'll need to compile perl with macro definitions
1582included in the debugging information. Using F<gcc> version 3.1, this
1583means configuring with C<-Doptimize=-g3>. Other compilers might use a
1584different switch (if they support debugging macros at all).
1585
a422fd2d
SC
1586=head2 Dumping Perl Data Structures
1587
1588One way to get around this macro hell is to use the dumping functions in
1589F<dump.c>; these work a little like an internal
1590L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
1591that you can't get at from Perl. Let's take an example. We'll use the
1592C<$a = $b + $c> we used before, but give it a bit of context:
1593C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
1594
1595What about C<pp_add>, the function we examined earlier to implement the
1596C<+> operator:
1597
1598 (gdb) break Perl_pp_add
1599 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
1600
cea6626f 1601Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>.
a422fd2d
SC
1602With the breakpoint in place, we can run our program:
1603
1604 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
1605
1606Lots of junk will go past as gdb reads in the relevant source files and
1607libraries, and then:
1608
1609 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
39644a26 1610 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
a422fd2d
SC
1611 (gdb) step
1612 311 dPOPTOPnnrl_ul;
1613 (gdb)
1614
1615We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
1616arranges for two C<NV>s to be placed into C<left> and C<right> - let's
1617slightly expand it:
1618
1619 #define dPOPTOPnnrl_ul NV right = POPn; \
1620 SV *leftsv = TOPs; \
1621 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
1622
1623C<POPn> takes the SV from the top of the stack and obtains its NV either
1624directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
1625C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
1626C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
1627C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
1628
1629Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
1630convert it. If we step again, we'll find ourselves there:
1631
1632 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1633 1669 if (!sv)
1634 (gdb)
1635
1636We can now use C<Perl_sv_dump> to investigate the SV:
1637
1638 SV = PV(0xa057cc0) at 0xa0675d0
1639 REFCNT = 1
1640 FLAGS = (POK,pPOK)
1641 PV = 0xa06a510 "6XXXX"\0
1642 CUR = 5
1643 LEN = 6
1644 $1 = void
1645
1646We know we're going to get C<6> from this, so let's finish the
1647subroutine:
1648
1649 (gdb) finish
1650 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
1651 0x462669 in Perl_pp_add () at pp_hot.c:311
1652 311 dPOPTOPnnrl_ul;
1653
1654We can also dump out this op: the current op is always stored in
1655C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
1656similar output to L<B::Debug|B::Debug>.
1657
1658 {
1659 13 TYPE = add ===> 14
1660 TARG = 1
1661 FLAGS = (SCALAR,KIDS)
1662 {
1663 TYPE = null ===> (12)
1664 (was rv2sv)
1665 FLAGS = (SCALAR,KIDS)
1666 {
1667 11 TYPE = gvsv ===> 12
1668 FLAGS = (SCALAR)
1669 GV = main::b
1670 }
1671 }
1672
10f58044 1673# finish this later #
a422fd2d
SC
1674
1675=head2 Patching
1676
1677All right, we've now had a look at how to navigate the Perl sources and
1678some things you'll need to know when fiddling with them. Let's now get
1679on and create a simple patch. Here's something Larry suggested: if a
1680C<U> is the first active format during a C<pack>, (for example,
1681C<pack "U3C8", @stuff>) then the resulting string should be treated as
1e54db1a 1682UTF-8 encoded.
a422fd2d
SC
1683
1684How do we prepare to fix this up? First we locate the code in question -
1685the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1686files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1687altering this file, let's copy it to F<pp.c~>.
1688
a6ec74c1
JH
1689[Well, it was in F<pp.c> when this tutorial was written. It has now been
1690split off with C<pp_unpack> to its own file, F<pp_pack.c>]
1691
a422fd2d
SC
1692Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1693loop over the pattern, taking each format character in turn into
1694C<datum_type>. Then for each possible format character, we swallow up
1695the other arguments in the pattern (a field width, an asterisk, and so
1696on) and convert the next chunk input into the specified format, adding
1697it onto the output SV C<cat>.
1698
1699How do we know if the C<U> is the first format in the C<pat>? Well, if
1700we have a pointer to the start of C<pat> then, if we see a C<U> we can
1701test whether we're still at the start of the string. So, here's where
1702C<pat> is set up:
1703
1704 STRLEN fromlen;
1705 register char *pat = SvPVx(*++MARK, fromlen);
1706 register char *patend = pat + fromlen;
1707 register I32 len;
1708 I32 datumtype;
1709 SV *fromstr;
1710
1711We'll have another string pointer in there:
1712
1713 STRLEN fromlen;
1714 register char *pat = SvPVx(*++MARK, fromlen);
1715 register char *patend = pat + fromlen;
1716 + char *patcopy;
1717 register I32 len;
1718 I32 datumtype;
1719 SV *fromstr;
1720
1721And just before we start the loop, we'll set C<patcopy> to be the start
1722of C<pat>:
1723
1724 items = SP - MARK;
1725 MARK++;
1726 sv_setpvn(cat, "", 0);
1727 + patcopy = pat;
1728 while (pat < patend) {
1729
1730Now if we see a C<U> which was at the start of the string, we turn on
1e54db1a 1731the C<UTF8> flag for the output SV, C<cat>:
a422fd2d
SC
1732
1733 + if (datumtype == 'U' && pat==patcopy+1)
1734 + SvUTF8_on(cat);
1735 if (datumtype == '#') {
1736 while (pat < patend && *pat != '\n')
1737 pat++;
1738
1739Remember that it has to be C<patcopy+1> because the first character of
1740the string is the C<U> which has been swallowed into C<datumtype!>
1741
1742Oops, we forgot one thing: what if there are spaces at the start of the
1743pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1744character, even though it's not the first thing in the pattern. In this
1745case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1746
1747 if (isSPACE(datumtype))
1748 continue;
1749
1750needs to become
1751
1752 if (isSPACE(datumtype)) {
1753 patcopy++;
1754 continue;
1755 }
1756
1757OK. That's the C part done. Now we must do two additional things before
1758this patch is ready to go: we've changed the behaviour of Perl, and so
1759we must document that change. We must also provide some more regression
1760tests to make sure our patch works and doesn't create a bug somewhere
1761else along the line.
1762
b23b8711
MS
1763The regression tests for each operator live in F<t/op/>, and so we
1764make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our
1765tests to the end. First, we'll test that the C<U> does indeed create
1766Unicode strings.
1767
1768t/op/pack.t has a sensible ok() function, but if it didn't we could
35c336e6 1769use the one from t/test.pl.
b23b8711 1770
35c336e6
MS
1771 require './test.pl';
1772 plan( tests => 159 );
b23b8711
MS
1773
1774so instead of this:
a422fd2d
SC
1775
1776 print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000);
1777 print "ok $test\n"; $test++;
1778
35c336e6
MS
1779we can write the more sensible (see L<Test::More> for a full
1780explanation of is() and other testing functions).
b23b8711 1781
35c336e6 1782 is( "1.20.300.4000", sprintf "%vd", pack("U*",1,20,300,4000),
812f5127 1783 "U* produces unicode" );
b23b8711 1784
a422fd2d
SC
1785Now we'll test that we got that space-at-the-beginning business right:
1786
35c336e6 1787 is( "1.20.300.4000", sprintf "%vd", pack(" U*",1,20,300,4000),
812f5127 1788 " with spaces at the beginning" );
a422fd2d
SC
1789
1790And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1791the first active format:
1792
35c336e6 1793 isnt( v1.20.300.4000, sprintf "%vd", pack("C0U*",1,20,300,4000),
812f5127 1794 "U* not first isn't unicode" );
a422fd2d 1795
35c336e6
MS
1796Mustn't forget to change the number of tests which appears at the top,
1797or else the automated tester will get confused. This will either look
1798like this:
a422fd2d 1799
35c336e6
MS
1800 print "1..156\n";
1801
1802or this:
1803
1804 plan( tests => 156 );
a422fd2d
SC
1805
1806We now compile up Perl, and run it through the test suite. Our new
1807tests pass, hooray!
1808
1809Finally, the documentation. The job is never done until the paperwork is
1810over, so let's describe the change we've just made. The relevant place
1811is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1812this text in the description of C<pack>:
1813
1814 =item *
1815
1816 If the pattern begins with a C<U>, the resulting string will be treated
1e54db1a
JH
1817 as UTF-8-encoded Unicode. You can force UTF-8 encoding on in a string
1818 with an initial C<U0>, and the bytes that follow will be interpreted as
1819 Unicode characters. If you don't want this to happen, you can begin your
1820 pattern with C<C0> (or anything else) to force Perl not to UTF-8 encode your
a422fd2d
SC
1821 string, and then follow this with a C<U*> somewhere in your pattern.
1822
1823All done. Now let's create the patch. F<Porting/patching.pod> tells us
1824that if we're making major changes, we should copy the entire directory
1825to somewhere safe before we begin fiddling, and then do
13a2d996 1826
a422fd2d
SC
1827 diff -ruN old new > patch
1828
1829However, we know which files we've changed, and we can simply do this:
1830
1831 diff -u pp.c~ pp.c > patch
1832 diff -u t/op/pack.t~ t/op/pack.t >> patch
1833 diff -u pod/perlfunc.pod~ pod/perlfunc.pod >> patch
1834
1835We end up with a patch looking a little like this:
1836
1837 --- pp.c~ Fri Jun 02 04:34:10 2000
1838 +++ pp.c Fri Jun 16 11:37:25 2000
1839 @@ -4375,6 +4375,7 @@
1840 register I32 items;
1841 STRLEN fromlen;
1842 register char *pat = SvPVx(*++MARK, fromlen);
1843 + char *patcopy;
1844 register char *patend = pat + fromlen;
1845 register I32 len;
1846 I32 datumtype;
1847 @@ -4405,6 +4406,7 @@
1848 ...
1849
1850And finally, we submit it, with our rationale, to perl5-porters. Job
1851done!
1852
f7e1e956
MS
1853=head2 Patching a core module
1854
1855This works just like patching anything else, with an extra
1856consideration. Many core modules also live on CPAN. If this is so,
1857patch the CPAN version instead of the core and send the patch off to
1858the module maintainer (with a copy to p5p). This will help the module
1859maintainer keep the CPAN version in sync with the core version without
1860constantly scanning p5p.
1861
acbe17fc
JP
1862=head2 Adding a new function to the core
1863
1864If, as part of a patch to fix a bug, or just because you have an
1865especially good idea, you decide to add a new function to the core,
1866discuss your ideas on p5p well before you start work. It may be that
1867someone else has already attempted to do what you are considering and
1868can give lots of good advice or even provide you with bits of code
1869that they already started (but never finished).
1870
1871You have to follow all of the advice given above for patching. It is
1872extremely important to test any addition thoroughly and add new tests
1873to explore all boundary conditions that your new function is expected
1874to handle. If your new function is used only by one module (e.g. toke),
1875then it should probably be named S_your_function (for static); on the
210b36aa 1876other hand, if you expect it to accessible from other functions in
acbe17fc
JP
1877Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions>
1878for more details.
1879
1880The location of any new code is also an important consideration. Don't
1881just create a new top level .c file and put your code there; you would
1882have to make changes to Configure (so the Makefile is created properly),
1883as well as possibly lots of include files. This is strictly pumpking
1884business.
1885
1886It is better to add your function to one of the existing top level
1887source code files, but your choice is complicated by the nature of
1888the Perl distribution. Only the files that are marked as compiled
1889static are located in the perl executable. Everything else is located
1890in the shared library (or DLL if you are running under WIN32). So,
1891for example, if a function was only used by functions located in
1892toke.c, then your code can go in toke.c. If, however, you want to call
1893the function from universal.c, then you should put your code in another
1894location, for example util.c.
1895
1896In addition to writing your c-code, you will need to create an
1897appropriate entry in embed.pl describing your function, then run
1898'make regen_headers' to create the entries in the numerous header
1899files that perl needs to compile correctly. See L<perlguts/Internal Functions>
1900for information on the various options that you can set in embed.pl.
1901You will forget to do this a few (or many) times and you will get
1902warnings during the compilation phase. Make sure that you mention
1903this when you post your patch to P5P; the pumpking needs to know this.
1904
1905When you write your new code, please be conscious of existing code
884bad00 1906conventions used in the perl source files. See L<perlstyle> for
acbe17fc
JP
1907details. Although most of the guidelines discussed seem to focus on
1908Perl code, rather than c, they all apply (except when they don't ;).
1909See also I<Porting/patching.pod> file in the Perl source distribution
1910for lots of details about both formatting and submitting patches of
1911your changes.
1912
1913Lastly, TEST TEST TEST TEST TEST any code before posting to p5p.
1914Test on as many platforms as you can find. Test as many perl
1915Configure options as you can (e.g. MULTIPLICITY). If you have
1916profiling or memory tools, see L<EXTERNAL TOOLS FOR DEBUGGING PERL>
210b36aa 1917below for how to use them to further test your code. Remember that
acbe17fc
JP
1918most of the people on P5P are doing this on their own time and
1919don't have the time to debug your code.
f7e1e956
MS
1920
1921=head2 Writing a test
1922
1923Every module and built-in function has an associated test file (or
1924should...). If you add or change functionality, you have to write a
1925test. If you fix a bug, you have to write a test so that bug never
1926comes back. If you alter the docs, it would be nice to test what the
1927new documentation says.
1928
1929In short, if you submit a patch you probably also have to patch the
1930tests.
1931
1932For modules, the test file is right next to the module itself.
1933F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation,
1934so there are some snags (and it would be wonderful for you to brush
1935them out), but it basically works that way. Everything else lives in
1936F<t/>.
1937
1938=over 3
1939
1940=item F<t/base/>
1941
1942Testing of the absolute basic functionality of Perl. Things like
1943C<if>, basic file reads and writes, simple regexes, etc. These are
1944run first in the test suite and if any of them fail, something is
1945I<really> broken.
1946
1947=item F<t/cmd/>
1948
1949These test the basic control structures, C<if/else>, C<while>,
35c336e6 1950subroutines, etc.
f7e1e956
MS
1951
1952=item F<t/comp/>
1953
1954Tests basic issues of how Perl parses and compiles itself.
1955
1956=item F<t/io/>
1957
1958Tests for built-in IO functions, including command line arguments.
1959
1960=item F<t/lib/>
1961
1962The old home for the module tests, you shouldn't put anything new in
1963here. There are still some bits and pieces hanging around in here
1964that need to be moved. Perhaps you could move them? Thanks!
1965
1966=item F<t/op/>
1967
1968Tests for perl's built in functions that don't fit into any of the
1969other directories.
1970
1971=item F<t/pod/>
1972
1973Tests for POD directives. There are still some tests for the Pod
1974modules hanging around in here that need to be moved out into F<lib/>.
1975
1976=item F<t/run/>
1977
1978Testing features of how perl actually runs, including exit codes and
1979handling of PERL* environment variables.
1980
244d9cb7
RGS
1981=item F<t/uni/>
1982
1983Tests for the core support of Unicode.
1984
1985=item F<t/win32/>
1986
1987Windows-specific tests.
1988
1989=item F<t/x2p>
1990
1991A test suite for the s2p converter.
1992
f7e1e956
MS
1993=back
1994
1995The core uses the same testing style as the rest of Perl, a simple
1996"ok/not ok" run through Test::Harness, but there are a few special
1997considerations.
1998
35c336e6
MS
1999There are three ways to write a test in the core. Test::More,
2000t/test.pl and ad hoc C<print $test ? "ok 42\n" : "not ok 42\n">. The
2001decision of which to use depends on what part of the test suite you're
2002working on. This is a measure to prevent a high-level failure (such
2003as Config.pm breaking) from causing basic functionality tests to fail.
2004
2005=over 4
2006
2007=item t/base t/comp
2008
2009Since we don't know if require works, or even subroutines, use ad hoc
2010tests for these two. Step carefully to avoid using the feature being
2011tested.
2012
2013=item t/cmd t/run t/io t/op
2014
2015Now that basic require() and subroutines are tested, you can use the
2016t/test.pl library which emulates the important features of Test::More
2017while using a minimum of core features.
2018
2019You can also conditionally use certain libraries like Config, but be
2020sure to skip the test gracefully if it's not there.
2021
2022=item t/lib ext lib
2023
2024Now that the core of Perl is tested, Test::More can be used. You can
2025also use the full suite of core modules in the tests.
2026
2027=back
f7e1e956
MS
2028
2029When you say "make test" Perl uses the F<t/TEST> program to run the
7205a85d
YO
2030test suite (except under Win32 where it uses F<t/harness> instead.)
2031All tests are run from the F<t/> directory, B<not> the directory
2032which contains the test. This causes some problems with the tests
2033in F<lib/>, so here's some opportunity for some patching.
f7e1e956
MS
2034
2035You must be triply conscious of cross-platform concerns. This usually
2036boils down to using File::Spec and avoiding things like C<fork()> and
2037C<system()> unless absolutely necessary.
2038
e018f8be
JH
2039=head2 Special Make Test Targets
2040
2041There are various special make targets that can be used to test Perl
2042slightly differently than the standard "test" target. Not all them
2043are expected to give a 100% success rate. Many of them have several
7205a85d
YO
2044aliases, and many of them are not available on certain operating
2045systems.
e018f8be
JH
2046
2047=over 4
2048
2049=item coretest
2050
7d7d5695 2051Run F<perl> on all core tests (F<t/*> and F<lib/[a-z]*> pragma tests).
e018f8be 2052
7205a85d
YO
2053(Not available on Win32)
2054
e018f8be
JH
2055=item test.deparse
2056
b26492ee
RGS
2057Run all the tests through B::Deparse. Not all tests will succeed.
2058
7205a85d
YO
2059(Not available on Win32)
2060
b26492ee
RGS
2061=item test.taintwarn
2062
2063Run all tests with the B<-t> command-line switch. Not all tests
2064are expected to succeed (until they're specifically fixed, of course).
e018f8be 2065
7205a85d
YO
2066(Not available on Win32)
2067
e018f8be
JH
2068=item minitest
2069
2070Run F<miniperl> on F<t/base>, F<t/comp>, F<t/cmd>, F<t/run>, F<t/io>,
2071F<t/op>, and F<t/uni> tests.
2072
7a834142
JH
2073=item test.valgrind check.valgrind utest.valgrind ucheck.valgrind
2074
2075(Only in Linux) Run all the tests using the memory leak + naughty
2076memory access tool "valgrind". The log files will be named
2077F<testname.valgrind>.
2078
e018f8be
JH
2079=item test.third check.third utest.third ucheck.third
2080
2081(Only in Tru64) Run all the tests using the memory leak + naughty
2082memory access tool "Third Degree". The log files will be named
2083F<perl3.log.testname>.
2084
2085=item test.torture torturetest
2086
2087Run all the usual tests and some extra tests. As of Perl 5.8.0 the
244d9cb7 2088only extra tests are Abigail's JAPHs, F<t/japh/abigail.t>.
e018f8be
JH
2089
2090You can also run the torture test with F<t/harness> by giving
2091C<-torture> argument to F<t/harness>.
2092
2093=item utest ucheck test.utf8 check.utf8
2094
2095Run all the tests with -Mutf8. Not all tests will succeed.
2096
7205a85d
YO
2097(Not available on Win32)
2098
cc0710ff
RGS
2099=item minitest.utf16 test.utf16
2100
2101Runs the tests with UTF-16 encoded scripts, encoded with different
2102versions of this encoding.
2103
2104C<make utest.utf16> runs the test suite with a combination of C<-utf8> and
2105C<-utf16> arguments to F<t/TEST>.
2106
7205a85d
YO
2107(Not available on Win32)
2108
244d9cb7
RGS
2109=item test_harness
2110
2111Run the test suite with the F<t/harness> controlling program, instead of
2112F<t/TEST>. F<t/harness> is more sophisticated, and uses the
2113L<Test::Harness> module, thus using this test target supposes that perl
2114mostly works. The main advantage for our purposes is that it prints a
00bf5cd9
RGS
2115detailed summary of failed tests at the end. Also, unlike F<t/TEST>, it
2116doesn't redirect stderr to stdout.
244d9cb7 2117
7205a85d
YO
2118Note that under Win32 F<t/harness> is always used instead of F<t/TEST>, so
2119there is no special "test_harness" target.
2120
2121Under Win32's "test" target you may use the TEST_SWITCHES and TEST_FILES
2122environment variables to control the behaviour of F<t/harness>. This means
2123you can say
2124
2125 nmake test TEST_FILES="op/*.t"
2126 nmake test TEST_SWITCHES="-torture" TEST_FILES="op/*.t"
2127
2128=item test-notty test_notty
2129
2130Sets PERL_SKIP_TTY_TEST to true before running normal test.
2131
244d9cb7
RGS
2132=back
2133
2134=head2 Running tests by hand
2135
2136You can run part of the test suite by hand by using one the following
2137commands from the F<t/> directory :
2138
2139 ./perl -I../lib TEST list-of-.t-files
2140
2141or
2142
2143 ./perl -I../lib harness list-of-.t-files
2144
2145(if you don't specify test scripts, the whole test suite will be run.)
2146
7205a85d
YO
2147=head3 Using t/harness for testing
2148
2149If you use C<harness> for testing you have several command line options
2150available to you. The arguments are as follows, and are in the order
2151that they must appear if used together.
2152
2153 harness -v -torture -re=pattern LIST OF FILES TO TEST
2154 harness -v -torture -re LIST OF PATTERNS TO MATCH
2155
2156If C<LIST OF FILES TO TEST> is omitted the file list is obtained from
2157the manifest. The file list may include shell wildcards which will be
2158expanded out.
2159
2160=over 4
2161
2162=item -v
2163
2164Run the tests under verbose mode so you can see what tests were run,
2165and debug outbut.
2166
2167=item -torture
2168
2169Run the torture tests as well as the normal set.
2170
2171=item -re=PATTERN
2172
2173Filter the file list so that all the test files run match PATTERN.
2174Note that this form is distinct from the B<-re LIST OF PATTERNS> form below
2175in that it allows the file list to be provided as well.
2176
2177=item -re LIST OF PATTERNS
2178
2179Filter the file list so that all the test files run match
2180/(LIST|OF|PATTERNS)/. Note that with this form the patterns
2181are joined by '|' and you cannot supply a list of files, instead
2182the test files are obtained from the MANIFEST.
2183
2184=back
2185
244d9cb7
RGS
2186You can run an individual test by a command similar to
2187
2188 ./perl -I../lib patho/to/foo.t
2189
2190except that the harnesses set up some environment variables that may
2191affect the execution of the test :
2192
2193=over 4
2194
2195=item PERL_CORE=1
2196
2197indicates that we're running this test part of the perl core test suite.
2198This is useful for modules that have a dual life on CPAN.
2199
2200=item PERL_DESTRUCT_LEVEL=2
2201
2202is set to 2 if it isn't set already (see L</PERL_DESTRUCT_LEVEL>)
2203
2204=item PERL
2205
2206(used only by F<t/TEST>) if set, overrides the path to the perl executable
2207that should be used to run the tests (the default being F<./perl>).
2208
2209=item PERL_SKIP_TTY_TEST
2210
2211if set, tells to skip the tests that need a terminal. It's actually set
2212automatically by the Makefile, but can also be forced artificially by
2213running 'make test_notty'.
2214
e018f8be 2215=back
f7e1e956 2216
902b9dbf
MF
2217=head1 EXTERNAL TOOLS FOR DEBUGGING PERL
2218
2219Sometimes it helps to use external tools while debugging and
2220testing Perl. This section tries to guide you through using
2221some common testing and debugging tools with Perl. This is
2222meant as a guide to interfacing these tools with Perl, not
2223as any kind of guide to the use of the tools themselves.
2224
a958818a
JH
2225B<NOTE 1>: Running under memory debuggers such as Purify, valgrind, or
2226Third Degree greatly slows down the execution: seconds become minutes,
2227minutes become hours. For example as of Perl 5.8.1, the
2228ext/Encode/t/Unicode.t takes extraordinarily long to complete under
2229e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more
2230than six hours, even on a snappy computer-- the said test must be
2231doing something that is quite unfriendly for memory debuggers. If you
2232don't feel like waiting, that you can simply kill away the perl
2233process.
2234
2235B<NOTE 2>: To minimize the number of memory leak false alarms (see
2236L</PERL_DESTRUCT_LEVEL> for more information), you have to have
2237environment variable PERL_DESTRUCT_LEVEL set to 2. The F<TEST>
2238and harness scripts do that automatically. But if you are running
2239some of the tests manually-- for csh-like shells:
2240
2241 setenv PERL_DESTRUCT_LEVEL 2
2242
2243and for Bourne-type shells:
2244
2245 PERL_DESTRUCT_LEVEL=2
2246 export PERL_DESTRUCT_LEVEL
2247
2248or in UNIXy environments you can also use the C<env> command:
2249
2250 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...
a1b65709 2251
37c0adeb
JH
2252B<NOTE 3>: There are known memory leaks when there are compile-time
2253errors within eval or require, seeing C<S_doeval> in the call stack
2254is a good sign of these. Fixing these leaks is non-trivial,
2255unfortunately, but they must be fixed eventually.
2256
902b9dbf
MF
2257=head2 Rational Software's Purify
2258
2259Purify is a commercial tool that is helpful in identifying
2260memory overruns, wild pointers, memory leaks and other such
2261badness. Perl must be compiled in a specific way for
2262optimal testing with Purify. Purify is available under
2263Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.
2264
902b9dbf
MF
2265=head2 Purify on Unix
2266
2267On Unix, Purify creates a new Perl binary. To get the most
2268benefit out of Purify, you should create the perl to Purify
2269using:
2270
2271 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
2272 -Uusemymalloc -Dusemultiplicity
2273
2274where these arguments mean:
2275
2276=over 4
2277
2278=item -Accflags=-DPURIFY
2279
2280Disables Perl's arena memory allocation functions, as well as
2281forcing use of memory allocation functions derived from the
2282system malloc.
2283
2284=item -Doptimize='-g'
2285
2286Adds debugging information so that you see the exact source
2287statements where the problem occurs. Without this flag, all
2288you will see is the source filename of where the error occurred.
2289
2290=item -Uusemymalloc
2291
2292Disable Perl's malloc so that Purify can more closely monitor
2293allocations and leaks. Using Perl's malloc will make Purify
2294report most leaks in the "potential" leaks category.
2295
2296=item -Dusemultiplicity
2297
2298Enabling the multiplicity option allows perl to clean up
2299thoroughly when the interpreter shuts down, which reduces the
2300number of bogus leak reports from Purify.
2301
2302=back
2303
2304Once you've compiled a perl suitable for Purify'ing, then you
2305can just:
2306
2307 make pureperl
2308
2309which creates a binary named 'pureperl' that has been Purify'ed.
2310This binary is used in place of the standard 'perl' binary
2311when you want to debug Perl memory problems.
2312
2313As an example, to show any memory leaks produced during the
2314standard Perl testset you would create and run the Purify'ed
2315perl as:
2316
2317 make pureperl
2318 cd t
2319 ../pureperl -I../lib harness
2320
2321which would run Perl on test.pl and report any memory problems.
2322
2323Purify outputs messages in "Viewer" windows by default. If
2324you don't have a windowing environment or if you simply
2325want the Purify output to unobtrusively go to a log file
2326instead of to the interactive window, use these following
2327options to output to the log file "perl.log":
2328
2329 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
2330 -log-file=perl.log -append-logfile=yes"
2331
2332If you plan to use the "Viewer" windows, then you only need this option:
2333
2334 setenv PURIFYOPTIONS "-chain-length=25"
2335
c406981e
JH
2336In Bourne-type shells:
2337
98631ff8
JL
2338 PURIFYOPTIONS="..."
2339 export PURIFYOPTIONS
c406981e
JH
2340
2341or if you have the "env" utility:
2342
98631ff8 2343 env PURIFYOPTIONS="..." ../pureperl ...
c406981e 2344
902b9dbf
MF
2345=head2 Purify on NT
2346
2347Purify on Windows NT instruments the Perl binary 'perl.exe'
2348on the fly. There are several options in the makefile you
2349should change to get the most use out of Purify:
2350
2351=over 4
2352
2353=item DEFINES
2354
2355You should add -DPURIFY to the DEFINES line so the DEFINES
2356line looks something like:
2357
2358 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1
2359
2360to disable Perl's arena memory allocation functions, as
2361well as to force use of memory allocation functions derived
2362from the system malloc.
2363
2364=item USE_MULTI = define
2365
2366Enabling the multiplicity option allows perl to clean up
2367thoroughly when the interpreter shuts down, which reduces the
2368number of bogus leak reports from Purify.
2369
2370=item #PERL_MALLOC = define
2371
2372Disable Perl's malloc so that Purify can more closely monitor
2373allocations and leaks. Using Perl's malloc will make Purify
2374report most leaks in the "potential" leaks category.
2375
2376=item CFG = Debug
2377
2378Adds debugging information so that you see the exact source
2379statements where the problem occurs. Without this flag, all
2380you will see is the source filename of where the error occurred.
2381
2382=back
2383
2384As an example, to show any memory leaks produced during the
2385standard Perl testset you would create and run Purify as:
2386
2387 cd win32
2388 make
2389 cd ../t
2390 purify ../perl -I../lib harness
2391
2392which would instrument Perl in memory, run Perl on test.pl,
2393then finally report any memory problems.
2394
7a834142
JH
2395=head2 valgrind
2396
2397The excellent valgrind tool can be used to find out both memory leaks
2398and illegal memory accesses. As of August 2003 it unfortunately works
2399only on x86 (ELF) Linux. The special "test.valgrind" target can be used
d44161bf
MHM
2400to run the tests under valgrind. Found errors and memory leaks are
2401logged in files named F<test.valgrind>.
2402
2403As system libraries (most notably glibc) are also triggering errors,
2404valgrind allows to suppress such errors using suppression files. The
2405default suppression file that comes with valgrind already catches a lot
2406of them. Some additional suppressions are defined in F<t/perl.supp>.
7a834142
JH
2407
2408To get valgrind and for more information see
2409
2410 http://developer.kde.org/~sewardj/
2411
f134cc4e 2412=head2 Compaq's/Digital's/HP's Third Degree
09187cb1
JH
2413
2414Third Degree is a tool for memory leak detection and memory access checks.
2415It is one of the many tools in the ATOM toolkit. The toolkit is only
2416available on Tru64 (formerly known as Digital UNIX formerly known as
2417DEC OSF/1).
2418
2419When building Perl, you must first run Configure with -Doptimize=-g
2420and -Uusemymalloc flags, after that you can use the make targets
51a35ef1
JH
2421"perl.third" and "test.third". (What is required is that Perl must be
2422compiled using the C<-g> flag, you may need to re-Configure.)
09187cb1 2423
64cea5fd 2424The short story is that with "atom" you can instrument the Perl
83f0ef60 2425executable to create a new executable called F<perl.third>. When the
4ae3d70a 2426instrumented executable is run, it creates a log of dubious memory
83f0ef60 2427traffic in file called F<perl.3log>. See the manual pages of atom and
4ae3d70a
JH
2428third for more information. The most extensive Third Degree
2429documentation is available in the Compaq "Tru64 UNIX Programmer's
2430Guide", chapter "Debugging Programs with Third Degree".
64cea5fd 2431
9c54ecba 2432The "test.third" leaves a lot of files named F<foo_bar.3log> in the t/
64cea5fd
JH
2433subdirectory. There is a problem with these files: Third Degree is so
2434effective that it finds problems also in the system libraries.
9c54ecba
JH
2435Therefore you should used the Porting/thirdclean script to cleanup
2436the F<*.3log> files.
64cea5fd
JH
2437
2438There are also leaks that for given certain definition of a leak,
2439aren't. See L</PERL_DESTRUCT_LEVEL> for more information.
2440
2441=head2 PERL_DESTRUCT_LEVEL
2442
a958818a
JH
2443If you want to run any of the tests yourself manually using e.g.
2444valgrind, or the pureperl or perl.third executables, please note that
2445by default perl B<does not> explicitly cleanup all the memory it has
2446allocated (such as global memory arenas) but instead lets the exit()
2447of the whole program "take care" of such allocations, also known as
2448"global destruction of objects".
64cea5fd
JH
2449
2450There is a way to tell perl to do complete cleanup: set the
2451environment variable PERL_DESTRUCT_LEVEL to a non-zero value.
2452The t/TEST wrapper does set this to 2, and this is what you
2453need to do too, if you don't want to see the "global leaks":
1f56d61a 2454For example, for "third-degreed" Perl:
64cea5fd 2455
1f56d61a 2456 env PERL_DESTRUCT_LEVEL=2 ./perl.third -Ilib t/foo/bar.t
09187cb1 2457
414f2397
RGS
2458(Note: the mod_perl apache module uses also this environment variable
2459for its own purposes and extended its semantics. Refer to the mod_perl
287a822c
RGS
2460documentation for more information. Also, spawned threads do the
2461equivalent of setting this variable to the value 1.)
5a6c59ef
DM
2462
2463If, at the end of a run you get the message I<N scalars leaked>, you can
fd0854ff
DM
2464recompile with C<-DDEBUG_LEAKING_SCALARS>, which will cause the addresses
2465of all those leaked SVs to be dumped along with details as to where each
2466SV was originally allocated. This information is also displayed by
2467Devel::Peek. Note that the extra details recorded with each SV increases
2468memory usage, so it shouldn't be used in production environments. It also
2469converts C<new_SV()> from a macro into a real function, so you can use
2470your favourite debugger to discover where those pesky SVs were allocated.
414f2397 2471
51a35ef1
JH
2472=head2 Profiling
2473
2474Depending on your platform there are various of profiling Perl.
2475
2476There are two commonly used techniques of profiling executables:
10f58044 2477I<statistical time-sampling> and I<basic-block counting>.
51a35ef1
JH
2478
2479The first method takes periodically samples of the CPU program
2480counter, and since the program counter can be correlated with the code
2481generated for functions, we get a statistical view of in which
2482functions the program is spending its time. The caveats are that very
2483small/fast functions have lower probability of showing up in the
2484profile, and that periodically interrupting the program (this is
2485usually done rather frequently, in the scale of milliseconds) imposes
2486an additional overhead that may skew the results. The first problem
2487can be alleviated by running the code for longer (in general this is a
2488good idea for profiling), the second problem is usually kept in guard
2489by the profiling tools themselves.
2490
10f58044 2491The second method divides up the generated code into I<basic blocks>.
51a35ef1
JH
2492Basic blocks are sections of code that are entered only in the
2493beginning and exited only at the end. For example, a conditional jump
2494starts a basic block. Basic block profiling usually works by
10f58044 2495I<instrumenting> the code by adding I<enter basic block #nnnn>
51a35ef1
JH
2496book-keeping code to the generated code. During the execution of the
2497code the basic block counters are then updated appropriately. The
2498caveat is that the added extra code can skew the results: again, the
2499profiling tools usually try to factor their own effects out of the
2500results.
2501
83f0ef60
JH
2502=head2 Gprof Profiling
2503
51a35ef1
JH
2504gprof is a profiling tool available in many UNIX platforms,
2505it uses F<statistical time-sampling>.
83f0ef60
JH
2506
2507You can build a profiled version of perl called "perl.gprof" by
51a35ef1
JH
2508invoking the make target "perl.gprof" (What is required is that Perl
2509must be compiled using the C<-pg> flag, you may need to re-Configure).
2510Running the profiled version of Perl will create an output file called
2511F<gmon.out> is created which contains the profiling data collected
2512during the execution.
83f0ef60
JH
2513
2514The gprof tool can then display the collected data in various ways.
2515Usually gprof understands the following options:
2516
2517=over 4
2518
2519=item -a
2520
2521Suppress statically defined functions from the profile.
2522
2523=item -b
2524
2525Suppress the verbose descriptions in the profile.
2526
2527=item -e routine
2528
2529Exclude the given routine and its descendants from the profile.
2530
2531=item -f routine
2532
2533Display only the given routine and its descendants in the profile.
2534
2535=item -s
2536
2537Generate a summary file called F<gmon.sum> which then may be given
2538to subsequent gprof runs to accumulate data over several runs.
2539
2540=item -z
2541
2542Display routines that have zero usage.
2543
2544=back
2545
2546For more detailed explanation of the available commands and output
2547formats, see your own local documentation of gprof.
2548
51a35ef1
JH
2549=head2 GCC gcov Profiling
2550
10f58044 2551Starting from GCC 3.0 I<basic block profiling> is officially available
51a35ef1
JH
2552for the GNU CC.
2553
2554You can build a profiled version of perl called F<perl.gcov> by
2555invoking the make target "perl.gcov" (what is required that Perl must
2556be compiled using gcc with the flags C<-fprofile-arcs
2557-ftest-coverage>, you may need to re-Configure).
2558
2559Running the profiled version of Perl will cause profile output to be
2560generated. For each source file an accompanying ".da" file will be
2561created.
2562
2563To display the results you use the "gcov" utility (which should
2564be installed if you have gcc 3.0 or newer installed). F<gcov> is
2565run on source code files, like this
2566
2567 gcov sv.c
2568
2569which will cause F<sv.c.gcov> to be created. The F<.gcov> files
2570contain the source code annotated with relative frequencies of
2571execution indicated by "#" markers.
2572
2573Useful options of F<gcov> include C<-b> which will summarise the
2574basic block, branch, and function call coverage, and C<-c> which
2575instead of relative frequencies will use the actual counts. For
2576more information on the use of F<gcov> and basic block profiling
2577with gcc, see the latest GNU CC manual, as of GCC 3.0 see
2578
2579 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html
2580
2581and its section titled "8. gcov: a Test Coverage Program"
2582
2583 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132
2584
4ae3d70a
JH
2585=head2 Pixie Profiling
2586
51a35ef1
JH
2587Pixie is a profiling tool available on IRIX and Tru64 (aka Digital
2588UNIX aka DEC OSF/1) platforms. Pixie does its profiling using
10f58044 2589I<basic-block counting>.
4ae3d70a 2590
83f0ef60 2591You can build a profiled version of perl called F<perl.pixie> by
51a35ef1
JH
2592invoking the make target "perl.pixie" (what is required is that Perl
2593must be compiled using the C<-g> flag, you may need to re-Configure).
2594
2595In Tru64 a file called F<perl.Addrs> will also be silently created,
2596this file contains the addresses of the basic blocks. Running the
2597profiled version of Perl will create a new file called "perl.Counts"
2598which contains the counts for the basic block for that particular
2599program execution.
4ae3d70a 2600
51a35ef1 2601To display the results you use the F<prof> utility. The exact
4ae3d70a
JH
2602incantation depends on your operating system, "prof perl.Counts" in
2603IRIX, and "prof -pixie -all -L. perl" in Tru64.
2604
6c41479b
JH
2605In IRIX the following prof options are available:
2606
2607=over 4
2608
2609=item -h
2610
2611Reports the most heavily used lines in descending order of use.
6e36760b 2612Useful for finding the hotspot lines.
6c41479b
JH
2613
2614=item -l
2615
2616Groups lines by procedure, with procedures sorted in descending order of use.
2617Within a procedure, lines are listed in source order.
6e36760b 2618Useful for finding the hotspots of procedures.
6c41479b
JH
2619
2620=back
2621
2622In Tru64 the following options are available:
2623
2624=over 4
2625
3958b146 2626=item -p[rocedures]
6c41479b 2627
3958b146 2628Procedures sorted in descending order by the number of cycles executed
6e36760b 2629in each procedure. Useful for finding the hotspot procedures.
6c41479b
JH
2630(This is the default option.)
2631
24000d2f 2632=item -h[eavy]
6c41479b 2633
6e36760b
JH
2634Lines sorted in descending order by the number of cycles executed in
2635each line. Useful for finding the hotspot lines.
6c41479b 2636
24000d2f 2637=item -i[nvocations]
6c41479b 2638
6e36760b
JH
2639The called procedures are sorted in descending order by number of calls
2640made to the procedures. Useful for finding the most used procedures.
6c41479b 2641
24000d2f 2642=item -l[ines]
6c41479b
JH
2643
2644Grouped by procedure, sorted by cycles executed per procedure.
6e36760b 2645Useful for finding the hotspots of procedures.
6c41479b
JH
2646
2647=item -testcoverage
2648
2649The compiler emitted code for these lines, but the code was unexecuted.
2650
24000d2f 2651=item -z[ero]
6c41479b
JH
2652
2653Unexecuted procedures.
2654
aa500c9e 2655=back
6c41479b
JH
2656
2657For further information, see your system's manual pages for pixie and prof.
4ae3d70a 2658
b8ddf6b3
SB
2659=head2 Miscellaneous tricks
2660
2661=over 4
2662
2663=item *
2664
cc177e1a 2665Those debugging perl with the DDD frontend over gdb may find the
b8ddf6b3
SB
2666following useful:
2667
2668You can extend the data conversion shortcuts menu, so for example you
2669can display an SV's IV value with one click, without doing any typing.
2670To do that simply edit ~/.ddd/init file and add after:
2671
2672 ! Display shortcuts.
2673 Ddd*gdbDisplayShortcuts: \
2674 /t () // Convert to Bin\n\
2675 /d () // Convert to Dec\n\
2676 /x () // Convert to Hex\n\
2677 /o () // Convert to Oct(\n\
2678
2679the following two lines:
2680
2681 ((XPV*) (())->sv_any )->xpv_pv // 2pvx\n\
2682 ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx
2683
2684so now you can do ivx and pvx lookups or you can plug there the
2685sv_peek "conversion":
2686
2687 Perl_sv_peek(my_perl, (SV*)()) // sv_peek
2688
2689(The my_perl is for threaded builds.)
2690Just remember that every line, but the last one, should end with \n\
2691
2692Alternatively edit the init file interactively via:
26933rd mouse button -> New Display -> Edit Menu
2694
2695Note: you can define up to 20 conversion shortcuts in the gdb
2696section.
2697
9965345d
JH
2698=item *
2699
2700If you see in a debugger a memory area mysteriously full of 0xabababab,
2701you may be seeing the effect of the Poison() macro, see L<perlclib>.
2702
b8ddf6b3
SB
2703=back
2704
a422fd2d
SC
2705=head2 CONCLUSION
2706
2707We've had a brief look around the Perl source, an overview of the stages
2708F<perl> goes through when it's running your code, and how to use a
902b9dbf
MF
2709debugger to poke at the Perl guts. We took a very simple problem and
2710demonstrated how to solve it fully - with documentation, regression
2711tests, and finally a patch for submission to p5p. Finally, we talked
2712about how to use external tools to debug and test Perl.
a422fd2d
SC
2713
2714I'd now suggest you read over those references again, and then, as soon
2715as possible, get your hands dirty. The best way to learn is by doing,
2716so:
2717
2718=over 3
2719
2720=item *
2721
2722Subscribe to perl5-porters, follow the patches and try and understand
2723them; don't be afraid to ask if there's a portion you're not clear on -
2724who knows, you may unearth a bug in the patch...
2725
2726=item *
2727
2728Keep up to date with the bleeding edge Perl distributions and get
2729familiar with the changes. Try and get an idea of what areas people are
2730working on and the changes they're making.
2731
2732=item *
2733
3e148164 2734Do read the README associated with your operating system, e.g. README.aix
a1f349fd
MB
2735on the IBM AIX OS. Don't hesitate to supply patches to that README if
2736you find anything missing or changed over a new OS release.
2737
2738=item *
2739
a422fd2d
SC
2740Find an area of Perl that seems interesting to you, and see if you can
2741work out how it works. Scan through the source, and step over it in the
2742debugger. Play, poke, investigate, fiddle! You'll probably get to
2743understand not just your chosen area but a much wider range of F<perl>'s
2744activity as well, and probably sooner than you'd think.
2745
2746=back
2747
2748=over 3
2749
2750=item I<The Road goes ever on and on, down from the door where it began.>
2751
2752=back
2753
2754If you can do these things, you've started on the long road to Perl porting.
2755Thanks for wanting to help make Perl better - and happy hacking!
2756
e8cd7eae
GS
2757=head1 AUTHOR
2758
2759This document was written by Nathan Torkington, and is maintained by
2760the perl5-porters mailing list.
2761