3 perlhack - How to hack at the Perl internals
7 This document attempts to explain how Perl development takes place,
8 and ends with some suggestions for people wanting to become bona fide
11 =head1 HOW PERL DEVELOPMENT HAPPENS
15 The perl5-porters mailing list is where the Perl standard distribution
16 is maintained and developed. The list can get anywhere from 10 to 150
17 messages a day, depending on the heatedness of the debate. Most days
18 there are two or three patches, extensions, features, or bugs being
21 A searchable archive of the list is at either:
23 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
27 http://archive.develooper.com/perl5-porters@perl.org/
29 List subscribers (the porters themselves) come in several flavours.
30 Some are quiet curious lurkers, who rarely pitch in and instead watch
31 the ongoing development to ensure they're forewarned of new changes or
32 features in Perl. Some are representatives of vendors, who are there
33 to make sure that Perl continues to compile and work on their
34 platforms. Some patch any reported bug that they know how to fix,
35 some are actively patching their pet area (threads, Win32, the regexp
36 engine), while others seem to do nothing but complain. In other
37 words, it's your usual mix of technical people.
39 Over this group of porters presides Larry Wall. He has the final word
40 in what does and does not change in the Perl language. Various
41 releases of Perl are shepherded by a "pumpking", a porter
42 responsible for gathering patches, deciding on a patch-by-patch,
43 feature-by-feature basis what will and will not go into the release.
44 For instance, Gurusamy Sarathy was the pumpking for the 5.6 release of
45 Perl, and Jarkko Hietaniemi was the pumpking for the 5.8 release, and
46 Rafael Garcia-Suarez holds the pumpking crown for the 5.10 release.
48 In addition, various people are pumpkings for different things. For
49 instance, Andy Dougherty and Jarkko Hietaniemi did a grand job as the
50 I<Configure> pumpkin up till the 5.8 release. For the 5.10 release
51 H.Merijn Brand took over.
53 Larry sees Perl development along the lines of the US government:
54 there's the Legislature (the porters), the Executive branch (the
55 pumpkings), and the Supreme Court (Larry). The legislature can
56 discuss and submit patches to the executive branch all they like, but
57 the executive branch is free to veto them. Rarely, the Supreme Court
58 will side with the executive branch over the legislature, or the
59 legislature over the executive branch. Mostly, however, the
60 legislature and the executive branch are supposed to get along and
61 work out their differences without impeachment or court cases.
63 You might sometimes see reference to Rule 1 and Rule 2. Larry's power
64 as Supreme Court is expressed in The Rules:
70 Larry is always by definition right about how Perl should behave.
71 This means he has final veto power on the core functionality.
75 Larry is allowed to change his mind about any matter at a later date,
76 regardless of whether he previously invoked Rule 1.
80 Got that? Larry is always right, even when he was wrong. It's rare
81 to see either Rule exercised, but they are often alluded to.
83 =head2 What makes for a good patch?
85 New features and extensions to the language are contentious, because
86 the criteria used by the pumpkings, Larry, and other porters to decide
87 which features should be implemented and incorporated are not codified
88 in a few small design goals as with some other languages. Instead,
89 the heuristics are flexible and often difficult to fathom. Here is
90 one person's list, roughly in decreasing order of importance, of
91 heuristics that new features have to be weighed against:
95 =item Does concept match the general goals of Perl?
97 These haven't been written anywhere in stone, but one approximation
100 1. Keep it fast, simple, and useful.
101 2. Keep features/concepts as orthogonal as possible.
102 3. No arbitrary limits (platforms, data sizes, cultures).
103 4. Keep it open and exciting to use/patch/advocate Perl everywhere.
104 5. Either assimilate new technologies, or build bridges to them.
106 =item Where is the implementation?
108 All the talk in the world is useless without an implementation. In
109 almost every case, the person or people who argue for a new feature
110 will be expected to be the ones who implement it. Porters capable
111 of coding new features have their own agendas, and are not available
112 to implement your (possibly good) idea.
114 =item Backwards compatibility
116 It's a cardinal sin to break existing Perl programs. New warnings are
117 contentious--some say that a program that emits warnings is not
118 broken, while others say it is. Adding keywords has the potential to
119 break programs, changing the meaning of existing token sequences or
120 functions might break programs.
122 =item Could it be a module instead?
124 Perl 5 has extension mechanisms, modules and XS, specifically to avoid
125 the need to keep changing the Perl interpreter. You can write modules
126 that export functions, you can give those functions prototypes so they
127 can be called like built-in functions, you can even write XS code to
128 mess with the runtime data structures of the Perl interpreter if you
129 want to implement really complicated things. If it can be done in a
130 module instead of in the core, it's highly unlikely to be added.
132 =item Is the feature generic enough?
134 Is this something that only the submitter wants added to the language,
135 or would it be broadly useful? Sometimes, instead of adding a feature
136 with a tight focus, the porters might decide to wait until someone
137 implements the more generalized feature. For instance, instead of
138 implementing a "delayed evaluation" feature, the porters are waiting
139 for a macro system that would permit delayed evaluation and much more.
141 =item Does it potentially introduce new bugs?
143 Radical rewrites of large chunks of the Perl interpreter have the
144 potential to introduce new bugs. The smaller and more localized the
147 =item Does it preclude other desirable features?
149 A patch is likely to be rejected if it closes off future avenues of
150 development. For instance, a patch that placed a true and final
151 interpretation on prototypes is likely to be rejected because there
152 are still options for the future of prototypes that haven't been
155 =item Is the implementation robust?
157 Good patches (tight code, complete, correct) stand more chance of
158 going in. Sloppy or incorrect patches might be placed on the back
159 burner until the pumpking has time to fix, or might be discarded
160 altogether without further notice.
162 =item Is the implementation generic enough to be portable?
164 The worst patches make use of a system-specific features. It's highly
165 unlikely that non-portable additions to the Perl language will be
168 =item Is the implementation tested?
170 Patches which change behaviour (fixing bugs or introducing new features)
171 must include regression tests to verify that everything works as expected.
172 Without tests provided by the original author, how can anyone else changing
173 perl in the future be sure that they haven't unwittingly broken the behaviour
174 the patch implements? And without tests, how can the patch's author be
175 confident that his/her hard work put into the patch won't be accidentally
176 thrown away by someone in the future?
178 =item Is there enough documentation?
180 Patches without documentation are probably ill-thought out or
181 incomplete. Nothing can be added without documentation, so submitting
182 a patch for the appropriate manpages as well as the source code is
185 =item Is there another way to do it?
187 Larry said "Although the Perl Slogan is I<There's More Than One Way
188 to Do It>, I hesitate to make 10 ways to do something". This is a
189 tricky heuristic to navigate, though--one man's essential addition is
190 another man's pointless cruft.
192 =item Does it create too much work?
194 Work for the pumpking, work for Perl programmers, work for module
195 authors, ... Perl is supposed to be easy.
197 =item Patches speak louder than words
199 Working code is always preferred to pie-in-the-sky ideas. A patch to
200 add a feature stands a much higher chance of making it to the language
201 than does a random feature request, no matter how fervently argued the
202 request might be. This ties into "Will it be useful?", as the fact
203 that someone took the time to make the patch demonstrates a strong
204 desire for the feature.
208 If you're on the list, you might hear the word "core" bandied
209 around. It refers to the standard distribution. "Hacking on the
210 core" means you're changing the C source code to the Perl
211 interpreter. "A core module" is one that ships with Perl.
213 =head2 Getting the Perl source
215 The source code to the Perl interpreter, in its different versions, is
216 kept in a repository managed by the git revision control system. The
217 pumpkings and a few others have write access to the repository to check in
220 How to clone and use the git perl repository is described in L<perlrepository>.
222 You can also choose to use rsync to get a copy of the current source tree
223 for the bleadperl branch and all maintenance branches:
225 $ rsync -avz rsync://perl5.git.perl.org/perl-current .
226 $ rsync -avz rsync://perl5.git.perl.org/perl-5.12.x .
227 $ rsync -avz rsync://perl5.git.perl.org/perl-5.10.x .
228 $ rsync -avz rsync://perl5.git.perl.org/perl-5.8.x .
229 $ rsync -avz rsync://perl5.git.perl.org/perl-5.6.x .
230 $ rsync -avz rsync://perl5.git.perl.org/perl-5.005xx .
232 (Add the C<--delete> option to remove leftover files)
234 To get a full list of the available sync points:
236 $ rsync perl5.git.perl.org::
238 You may also want to subscribe to the perl5-changes mailing list to
239 receive a copy of each patch that gets submitted to the maintenance
240 and development "branches" of the perl repository. See
241 http://lists.perl.org/ for subscription information.
243 If you are a member of the perl5-porters mailing list, it is a good
244 thing to keep in touch with the most recent changes. If not only to
245 verify if what you would have posted as a bug report isn't already
246 solved in the most recent available perl development branch, also
247 known as perl-current, bleading edge perl, bleedperl or bleadperl.
249 Needless to say, the source code in perl-current is usually in a perpetual
250 state of evolution. You should expect it to be very buggy. Do B<not> use
251 it for any purpose other than testing and development.
253 =head2 Bug tracking with Perlbug
255 There is a single remote administrative interface for modifying bug status,
256 category, open issues etc. using the B<RT> bugtracker system, maintained
257 by Robert Spier. Become an administrator, and close any bugs you can get
258 your sticky mitts on:
260 http://bugs.perl.org/
262 To email the bug system administrators:
264 "perlbug-admin" <perlbug-admin@perl.org>
266 =head2 Submitting patches
268 Always submit patches to I<perl5-porters@perl.org>. If you're
269 patching a core module and there's an author listed, send the author a
270 copy (see L<Patching a core module>). This lets other porters review
271 your patch, which catches a surprising number of errors in patches.
272 Please patch against the latest B<development> version. (e.g., even if
273 you're fixing a bug in the 5.8 track, patch against the C<blead> branch in
276 If changes are accepted, they are applied to the development branch. Then
277 the maintenance pumpking decides which of those patches is to be
278 backported to the maint branch. Only patches that survive the heat of the
279 development branch get applied to maintenance versions.
281 Your patch should update the documentation and test suite. See
282 L<TESTING>. If you have added or removed files in the distribution,
283 edit the MANIFEST file accordingly, sort the MANIFEST file using
284 C<make manisort>, and include those changes as part of your patch.
286 Patching documentation also follows the same order: if accepted, a patch
287 is first applied to B<development>, and if relevant then it's backported
288 to B<maintenance>. (With an exception for some patches that document
289 behaviour that only appears in the maintenance branch, but which has
290 changed in the development version.)
292 To report a bug in Perl, use the program L<perlbug> which comes with
293 Perl (if you can't get Perl to work, send mail to the address
294 I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through
295 I<perlbug> feeds into the automated bug-tracking system, access to
296 which is provided through the web at http://rt.perl.org/rt3/ . It
297 often pays to check the archives of the perl5-porters mailing list to
298 see whether the bug you're reporting has been reported before, and if
299 so whether it was considered a bug. See above for the location of
300 the searchable archives.
302 The CPAN testers ( http://testers.cpan.org/ ) are a group of
303 volunteers who test CPAN modules on a variety of platforms. Perl
304 Smokers ( http://www.nntp.perl.org/group/perl.daily-build and
305 http://www.nntp.perl.org/group/perl.daily-build.reports/ )
306 automatically test Perl source releases on platforms with various
307 configurations. Both efforts welcome volunteers. In order to get
308 involved in smoke testing of the perl itself visit
309 L<http://search.cpan.org/dist/Test-Smoke>. In order to start smoke
310 testing CPAN modules visit L<http://search.cpan.org/dist/CPANPLUS-YACSmoke/>
311 or L<http://search.cpan.org/dist/minismokebox/> or
312 L<http://search.cpan.org/dist/CPAN-Reporter/>.
314 It's a good idea to read and lurk for a while before chipping in.
315 That way you'll get to see the dynamic of the conversations, learn the
316 personalities of the players, and hopefully be better prepared to make
317 a useful contribution when do you speak up.
319 If after all this you still think you want to join the perl5-porters
320 mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To
321 unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>.
323 =head2 Patching a core module
325 This works just like patching anything else, with an extra
326 consideration. Many core modules also live on CPAN. If this is so,
327 patch the CPAN version instead of the core and send the patch off to
328 the module maintainer (with a copy to p5p). This will help the module
329 maintainer keep the CPAN version in sync with the core version without
330 constantly scanning p5p.
332 The list of maintainers of core modules is usefully documented in
333 F<Porting/Maintainers.pl>.
335 =head2 Adding a new function to the core
337 If, as part of a patch to fix a bug, or just because you have an
338 especially good idea, you decide to add a new function to the core,
339 discuss your ideas on p5p well before you start work. It may be that
340 someone else has already attempted to do what you are considering and
341 can give lots of good advice or even provide you with bits of code
342 that they already started (but never finished).
344 You have to follow all of the advice given above for patching. It is
345 extremely important to test any addition thoroughly and add new tests
346 to explore all boundary conditions that your new function is expected
347 to handle. If your new function is used only by one module (e.g. toke),
348 then it should probably be named S_your_function (for static); on the
349 other hand, if you expect it to accessible from other functions in
350 Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions>
353 The location of any new code is also an important consideration. Don't
354 just create a new top level .c file and put your code there; you would
355 have to make changes to Configure (so the Makefile is created properly),
356 as well as possibly lots of include files. This is strictly pumpking
359 It is better to add your function to one of the existing top level
360 source code files, but your choice is complicated by the nature of
361 the Perl distribution. Only the files that are marked as compiled
362 static are located in the perl executable. Everything else is located
363 in the shared library (or DLL if you are running under WIN32). So,
364 for example, if a function was only used by functions located in
365 toke.c, then your code can go in toke.c. If, however, you want to call
366 the function from universal.c, then you should put your code in another
367 location, for example util.c.
369 In addition to writing your c-code, you will need to create an
370 appropriate entry in embed.pl describing your function, then run
371 'make regen_headers' to create the entries in the numerous header
372 files that perl needs to compile correctly. See L<perlguts/Internal Functions>
373 for information on the various options that you can set in embed.pl.
374 You will forget to do this a few (or many) times and you will get
375 warnings during the compilation phase. Make sure that you mention
376 this when you post your patch to P5P; the pumpking needs to know this.
378 When you write your new code, please be conscious of existing code
379 conventions used in the perl source files. See L<perlstyle> for
380 details. Although most of the guidelines discussed seem to focus on
381 Perl code, rather than c, they all apply (except when they don't ;).
382 Also see L<perlrepository> for lots of details about both formatting and
383 submitting patches of your changes.
385 Lastly, TEST TEST TEST TEST TEST any code before posting to p5p.
386 Test on as many platforms as you can find. Test as many perl
387 Configure options as you can (e.g. MULTIPLICITY). If you have
388 profiling or memory tools, see L<MEMORY DEBUGGERS> and L<PROFILING>
389 below for how to use them to further test your code. Remember that
390 most of the people on P5P are doing this on their own time and
391 don't have the time to debug your code.
393 =head2 Background reading
395 To hack on the Perl guts, you'll need to read the following things:
401 This is of paramount importance, since it's the documentation of what
402 goes where in the Perl source. Read it over a couple of times and it
403 might start to make sense - don't worry if it doesn't yet, because the
404 best way to study it is to read it in conjunction with poking at Perl
405 source, and we'll do that later on.
407 Gisle Aas's "illustrated perlguts", also known as I<illguts>, has very
410 L<http://search.cpan.org/dist/illguts/>
412 =item L<perlxstut> and L<perlxs>
414 A working knowledge of XSUB programming is incredibly useful for core
415 hacking; XSUBs use techniques drawn from the PP code, the portion of the
416 guts that actually executes a Perl program. It's a lot gentler to learn
417 those techniques from simple examples and explanation than from the core
422 The documentation for the Perl API explains what some of the internal
423 functions do, as well as the many macros used in the source.
425 =item F<Porting/pumpkin.pod>
427 This is a collection of words of wisdom for a Perl porter; some of it is
428 only useful to the pumpkin holder, but most of it applies to anyone
429 wanting to go about Perl development.
431 =item The perl5-porters FAQ
433 This should be available from http://dev.perl.org/perl5/docs/p5p-faq.html .
434 It contains hints on reading perl5-porters, information on how
435 perl5-porters works and how Perl development in general works.
439 =head1 UNDERSTANDING THE SOURCE
441 =head2 Finding your way around
443 Perl maintenance can be split into a number of areas, and certain people
444 (pumpkins) will have responsibility for each area. These areas sometimes
445 correspond to files or directories in the source kit. Among the areas are:
451 Modules shipped as part of the Perl core live in various subdirectories, where
452 two are dedicated to core-only modules, and two are for the dual-life modules
453 which live on CPAN and may be maintained separately with respect to the Perl
456 lib/ is for pure-Perl modules, which exist in the core only.
458 ext/ is for XS extensions, and modules with special Makefile.PL
459 requirements, which exist in the core only.
461 cpan/ is for dual-life modules, where the CPAN module is
462 canonical (should be patched first).
464 dist/ is for dual-life modules, where the blead source is
467 For some dual-life modules it has not been discussed if the CPAN version or the
468 blead source is canonical. Until that is done, those modules should be in
473 There are tests for nearly all the modules, built-ins and major bits
474 of functionality. Test files all have a .t suffix. Module tests live
475 in the F<lib/> and F<ext/> directories next to the module being
476 tested. Others live in F<t/>. See L<TESTING>
480 Documentation maintenance includes looking after everything in the
481 F<pod/> directory, (as well as contributing new documentation) and
482 the documentation to the modules in core.
486 The Configure process is the way we make Perl portable across the
487 myriad of operating systems it supports. Responsibility for the
488 Configure, build and installation process, as well as the overall
489 portability of the core code rests with the Configure pumpkin -
490 others help out with individual operating systems.
492 The three files that fall under his/her responsibility are Configure,
493 config_h.SH, and Porting/Glossary (and a whole bunch of small related
494 files that are less important here). The Configure pumpkin decides how
495 patches to these are dealt with. Currently, the Configure pumpkin will
496 accept patches in most common formats, even directly to these files.
497 Other committers are allowed to commit to these files under the strict
498 condition that they will inform the Configure pumpkin, either on IRC
499 (if he/she happens to be around) or through (personal) e-mail.
501 The files involved are the operating system directories, (F<win32/>,
502 F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
503 and F<Makefile>, as well as the metaconfig files which generate
504 F<Configure>. (metaconfig isn't included in the core distribution.)
506 See http://perl5.git.perl.org/metaconfig.git/blob/HEAD:/README for a
507 description of the full process involved.
511 And of course, there's the core of the Perl interpreter itself. Let's
512 have a look at that in a little more detail.
516 Before we leave looking at the layout, though, don't forget that
517 F<MANIFEST> contains not only the file names in the Perl distribution,
518 but short descriptions of what's in them, too. For an overview of the
519 important files, try this:
521 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
523 =head2 Elements of the interpreter
525 The work of the interpreter has two main stages: compiling the code
526 into the internal representation, or bytecode, and then executing it.
527 L<perlguts/Compiled code> explains exactly how the compilation stage
530 Here is a short breakdown of perl's operation:
536 The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
537 This is very high-level code, enough to fit on a single screen, and it
538 resembles the code found in L<perlembed>; most of the real action takes
541 F<perlmain.c> is generated by C<ExtUtils::Miniperl> from F<miniperlmain.c> at
542 make time, so you should make perl to follow this along.
544 First, F<perlmain.c> allocates some memory and constructs a Perl
545 interpreter, along these lines:
547 1 PERL_SYS_INIT3(&argc,&argv,&env);
549 3 if (!PL_do_undump) {
550 4 my_perl = perl_alloc();
553 7 perl_construct(my_perl);
554 8 PL_perl_destruct_level = 0;
557 Line 1 is a macro, and its definition is dependent on your operating
558 system. Line 3 references C<PL_do_undump>, a global variable - all
559 global variables in Perl start with C<PL_>. This tells you whether the
560 current running program was created with the C<-u> flag to perl and then
561 F<undump>, which means it's going to be false in any sane context.
563 Line 4 calls a function in F<perl.c> to allocate memory for a Perl
564 interpreter. It's quite a simple function, and the guts of it looks like
567 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
569 Here you see an example of Perl's system abstraction, which we'll see
570 later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
571 own C<malloc> as defined in F<malloc.c> if you selected that option at
574 Next, in line 7, we construct the interpreter using perl_construct,
575 also in F<perl.c>; this sets up all the special variables that Perl
576 needs, the stacks, and so on.
578 Now we pass Perl the command line options, and tell it to go:
580 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
584 exitstatus = perl_destruct(my_perl);
588 C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
589 in F<perl.c>, which processes the command line options, sets up any
590 statically linked XS modules, opens the program and calls C<yyparse> to
595 The aim of this stage is to take the Perl source, and turn it into an op
596 tree. We'll see what one of those looks like later. Strictly speaking,
597 there's three things going on here.
599 C<yyparse>, the parser, lives in F<perly.c>, although you're better off
600 reading the original YACC input in F<perly.y>. (Yes, Virginia, there
601 B<is> a YACC grammar for Perl!) The job of the parser is to take your
602 code and "understand" it, splitting it into sentences, deciding which
603 operands go with which operators and so on.
605 The parser is nobly assisted by the lexer, which chunks up your input
606 into tokens, and decides what type of thing each token is: a variable
607 name, an operator, a bareword, a subroutine, a core function, and so on.
608 The main point of entry to the lexer is C<yylex>, and that and its
609 associated routines can be found in F<toke.c>. Perl isn't much like
610 other computer languages; it's highly context sensitive at times, it can
611 be tricky to work out what sort of token something is, or where a token
612 ends. As such, there's a lot of interplay between the tokeniser and the
613 parser, which can get pretty frightening if you're not used to it.
615 As the parser understands a Perl program, it builds up a tree of
616 operations for the interpreter to perform during execution. The routines
617 which construct and link together the various operations are to be found
618 in F<op.c>, and will be examined later.
622 Now the parsing stage is complete, and the finished tree represents
623 the operations that the Perl interpreter needs to perform to execute our
624 program. Next, Perl does a dry run over the tree looking for
625 optimisations: constant expressions such as C<3 + 4> will be computed
626 now, and the optimizer will also see if any multiple operations can be
627 replaced with a single one. For instance, to fetch the variable C<$foo>,
628 instead of grabbing the glob C<*foo> and looking at the scalar
629 component, the optimizer fiddles the op tree to use a function which
630 directly looks up the scalar in question. The main optimizer is C<peep>
631 in F<op.c>, and many ops have their own optimizing functions.
635 Now we're finally ready to go: we have compiled Perl byte code, and all
636 that's left to do is run it. The actual execution is done by the
637 C<runops_standard> function in F<run.c>; more specifically, it's done by
638 these three innocent looking lines:
640 while ((PL_op = PL_op->op_ppaddr(aTHX))) {
644 You may be more comfortable with the Perl version of that:
646 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
648 Well, maybe not. Anyway, each op contains a function pointer, which
649 stipulates the function which will actually carry out the operation.
650 This function will return the next op in the sequence - this allows for
651 things like C<if> which choose the next op dynamically at run time.
652 The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
653 execution if required.
655 The actual functions called are known as PP code, and they're spread
656 between four files: F<pp_hot.c> contains the "hot" code, which is most
657 often used and highly optimized, F<pp_sys.c> contains all the
658 system-specific functions, F<pp_ctl.c> contains the functions which
659 implement control structures (C<if>, C<while> and the like) and F<pp.c>
660 contains everything else. These are, if you like, the C code for Perl's
661 built-in functions and operators.
663 Note that each C<pp_> function is expected to return a pointer to the next
664 op. Calls to perl subs (and eval blocks) are handled within the same
665 runops loop, and do not consume extra space on the C stack. For example,
666 C<pp_entersub> and C<pp_entertry> just push a C<CxSUB> or C<CxEVAL> block
667 struct onto the context stack which contain the address of the op
668 following the sub call or eval. They then return the first op of that sub
669 or eval block, and so execution continues of that sub or block. Later, a
670 C<pp_leavesub> or C<pp_leavetry> op pops the C<CxSUB> or C<CxEVAL>,
671 retrieves the return op from it, and returns it.
673 =item Exception handing
675 Perl's exception handing (i.e. C<die> etc.) is built on top of the low-level
676 C<setjmp()>/C<longjmp()> C-library functions. These basically provide a
677 way to capture the current PC and SP registers and later restore them; i.e.
678 a C<longjmp()> continues at the point in code where a previous C<setjmp()>
679 was done, with anything further up on the C stack being lost. This is why
680 code should always save values using C<SAVE_FOO> rather than in auto
683 The perl core wraps C<setjmp()> etc in the macros C<JMPENV_PUSH> and
684 C<JMPENV_JUMP>. The basic rule of perl exceptions is that C<exit>, and
685 C<die> (in the absence of C<eval>) perform a C<JMPENV_JUMP(2)>, while
686 C<die> within C<eval> does a C<JMPENV_JUMP(3)>.
688 At entry points to perl, such as C<perl_parse()>, C<perl_run()> and
689 C<call_sv(cv, G_EVAL)> each does a C<JMPENV_PUSH>, then enter a runops
690 loop or whatever, and handle possible exception returns. For a 2 return,
691 final cleanup is performed, such as popping stacks and calling C<CHECK> or
692 C<END> blocks. Amongst other things, this is how scope cleanup still
693 occurs during an C<exit>.
695 If a C<die> can find a C<CxEVAL> block on the context stack, then the
696 stack is popped to that level and the return op in that block is assigned
697 to C<PL_restartop>; then a C<JMPENV_JUMP(3)> is performed. This normally
698 passes control back to the guard. In the case of C<perl_run> and
699 C<call_sv>, a non-null C<PL_restartop> triggers re-entry to the runops
700 loop. The is the normal way that C<die> or C<croak> is handled within an
703 Sometimes ops are executed within an inner runops loop, such as tie, sort
704 or overload code. In this case, something like
706 sub FETCH { eval { die } }
708 would cause a longjmp right back to the guard in C<perl_run>, popping both
709 runops loops, which is clearly incorrect. One way to avoid this is for the
710 tie code to do a C<JMPENV_PUSH> before executing C<FETCH> in the inner
711 runops loop, but for efficiency reasons, perl in fact just sets a flag,
712 using C<CATCH_SET(TRUE)>. The C<pp_require>, C<pp_entereval> and
713 C<pp_entertry> ops check this flag, and if true, they call C<docatch>,
714 which does a C<JMPENV_PUSH> and starts a new runops level to execute the
715 code, rather than doing it on the current loop.
717 As a further optimisation, on exit from the eval block in the C<FETCH>,
718 execution of the code following the block is still carried on in the inner
719 loop. When an exception is raised, C<docatch> compares the C<JMPENV>
720 level of the C<CxEVAL> with C<PL_top_env> and if they differ, just
721 re-throws the exception. In this way any inner loops get popped.
725 1: eval { tie @a, 'A' };
731 To run this code, C<perl_run> is called, which does a C<JMPENV_PUSH> then
732 enters a runops loop. This loop executes the eval and tie ops on line 1,
733 with the eval pushing a C<CxEVAL> onto the context stack.
735 The C<pp_tie> does a C<CATCH_SET(TRUE)>, then starts a second runops loop
736 to execute the body of C<TIEARRAY>. When it executes the entertry op on
737 line 3, C<CATCH_GET> is true, so C<pp_entertry> calls C<docatch> which
738 does a C<JMPENV_PUSH> and starts a third runops loop, which then executes
739 the die op. At this point the C call stack looks like this:
742 Perl_runops # third loop
746 Perl_runops # second loop
750 Perl_runops # first loop
755 and the context and data stacks, as shown by C<-Dstv>, look like:
759 CX 1: EVAL => AV() PV("A"\0)
767 The die pops the first C<CxEVAL> off the context stack, sets
768 C<PL_restartop> from it, does a C<JMPENV_JUMP(3)>, and control returns to
769 the top C<docatch>. This then starts another third-level runops level,
770 which executes the nextstate, pushmark and die ops on line 4. At the point
771 that the second C<pp_die> is called, the C call stack looks exactly like
772 that above, even though we are no longer within an inner eval; this is
773 because of the optimization mentioned earlier. However, the context stack
774 now looks like this, ie with the top CxEVAL popped:
778 CX 1: EVAL => AV() PV("A"\0)
784 The die on line 4 pops the context stack back down to the CxEVAL, leaving
790 As usual, C<PL_restartop> is extracted from the C<CxEVAL>, and a
791 C<JMPENV_JUMP(3)> done, which pops the C stack back to the docatch:
795 Perl_runops # second loop
799 Perl_runops # first loop
804 In this case, because the C<JMPENV> level recorded in the C<CxEVAL>
805 differs from the current one, C<docatch> just does a C<JMPENV_JUMP(3)>
806 and the C stack unwinds to:
811 Because C<PL_restartop> is non-null, C<run_body> starts a new runops loop
812 and execution continues.
816 =head2 Internal Variable Types
818 You should by now have had a look at L<perlguts>, which tells you about
819 Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
822 These variables are used not only to represent Perl-space variables, but
823 also any constants in the code, as well as some structures completely
824 internal to Perl. The symbol table, for instance, is an ordinary Perl
825 hash. Your code is represented by an SV as it's read into the parser;
826 any program files you call are opened via ordinary Perl filehandles, and
829 The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
830 Perl program. Let's see, for instance, how Perl treats the constant
833 % perl -MDevel::Peek -e 'Dump("hello")'
834 1 SV = PV(0xa041450) at 0xa04ecbc
836 3 FLAGS = (POK,READONLY,pPOK)
837 4 PV = 0xa0484e0 "hello"\0
841 Reading C<Devel::Peek> output takes a bit of practise, so let's go
842 through it line by line.
844 Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
845 memory. SVs themselves are very simple structures, but they contain a
846 pointer to a more complex structure. In this case, it's a PV, a
847 structure which holds a string value, at location C<0xa041450>. Line 2
848 is the reference count; there are no other references to this data, so
851 Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
852 read-only SV (because it's a constant) and the data is a PV internally.
853 Next we've got the contents of the string, starting at location
856 Line 5 gives us the current length of the string - note that this does
857 B<not> include the null terminator. Line 6 is not the length of the
858 string, but the length of the currently allocated buffer; as the string
859 grows, Perl automatically extends the available storage via a routine
862 You can get at any of these quantities from C very easily; just add
863 C<Sv> to the name of the field shown in the snippet, and you've got a
864 macro which will return the value: C<SvCUR(sv)> returns the current
865 length of the string, C<SvREFCOUNT(sv)> returns the reference count,
866 C<SvPV(sv, len)> returns the string itself with its length, and so on.
867 More macros to manipulate these properties can be found in L<perlguts>.
869 Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
872 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
877 6 junk = SvPV_force(sv, tlen);
878 7 SvGROW(sv, tlen + len + 1);
881 10 Move(ptr,SvPVX(sv)+tlen,len,char);
883 12 *SvEND(sv) = '\0';
884 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
888 This is a function which adds a string, C<ptr>, of length C<len> onto
889 the end of the PV stored in C<sv>. The first thing we do in line 6 is
890 make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
891 macro to force a PV. As a side effect, C<tlen> gets set to the current
892 value of the PV, and the PV itself is returned to C<junk>.
894 In line 7, we make sure that the SV will have enough room to accommodate
895 the old string, the new string and the null terminator. If C<LEN> isn't
896 big enough, C<SvGROW> will reallocate space for us.
898 Now, if C<junk> is the same as the string we're trying to add, we can
899 grab the string directly from the SV; C<SvPVX> is the address of the PV
902 Line 10 does the actual catenation: the C<Move> macro moves a chunk of
903 memory around: we move the string C<ptr> to the end of the PV - that's
904 the start of the PV plus its current length. We're moving C<len> bytes
905 of type C<char>. After doing so, we need to tell Perl we've extended the
906 string, by altering C<CUR> to reflect the new length. C<SvEND> is a
907 macro which gives us the end of the string, so that needs to be a
910 Line 13 manipulates the flags; since we've changed the PV, any IV or NV
911 values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
912 want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF-8-aware
913 version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
914 and turns on POK. The final C<SvTAINT> is a macro which launders tainted
915 data if taint mode is turned on.
917 AVs and HVs are more complicated, but SVs are by far the most common
918 variable type being thrown around. Having seen something of how we
919 manipulate these, let's go on and look at how the op tree is
924 First, what is the op tree, anyway? The op tree is the parsed
925 representation of your program, as we saw in our section on parsing, and
926 it's the sequence of operations that Perl goes through to execute your
927 program, as we saw in L</Running>.
929 An op is a fundamental operation that Perl can perform: all the built-in
930 functions and operators are ops, and there are a series of ops which
931 deal with concepts the interpreter needs internally - entering and
932 leaving a block, ending a statement, fetching a variable, and so on.
934 The op tree is connected in two ways: you can imagine that there are two
935 "routes" through it, two orders in which you can traverse the tree.
936 First, parse order reflects how the parser understood the code, and
937 secondly, execution order tells perl what order to perform the
940 The easiest way to examine the op tree is to stop Perl after it has
941 finished parsing, and get it to dump out the tree. This is exactly what
942 the compiler backends L<B::Terse|B::Terse>, L<B::Concise|B::Concise>
943 and L<B::Debug|B::Debug> do.
945 Let's have a look at how Perl sees C<$a = $b + $c>:
947 % perl -MO=Terse -e '$a=$b+$c'
948 1 LISTOP (0x8179888) leave
949 2 OP (0x81798b0) enter
950 3 COP (0x8179850) nextstate
951 4 BINOP (0x8179828) sassign
952 5 BINOP (0x8179800) add [1]
953 6 UNOP (0x81796e0) null [15]
954 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
955 8 UNOP (0x81797e0) null [15]
956 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
957 10 UNOP (0x816b4f0) null [15]
958 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
960 Let's start in the middle, at line 4. This is a BINOP, a binary
961 operator, which is at location C<0x8179828>. The specific operator in
962 question is C<sassign> - scalar assignment - and you can find the code
963 which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
964 binary operator, it has two children: the add operator, providing the
965 result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
968 Line 10 is the null op: this does exactly nothing. What is that doing
969 there? If you see the null op, it's a sign that something has been
970 optimized away after parsing. As we mentioned in L</Optimization>,
971 the optimization stage sometimes converts two operations into one, for
972 example when fetching a scalar variable. When this happens, instead of
973 rewriting the op tree and cleaning up the dangling pointers, it's easier
974 just to replace the redundant operation with the null op. Originally,
975 the tree would have looked like this:
977 10 SVOP (0x816b4f0) rv2sv [15]
978 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
980 That is, fetch the C<a> entry from the main symbol table, and then look
981 at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
982 happens to do both these things.
984 The right hand side, starting at line 5 is similar to what we've just
985 seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
988 Now, what's this about?
990 1 LISTOP (0x8179888) leave
991 2 OP (0x81798b0) enter
992 3 COP (0x8179850) nextstate
994 C<enter> and C<leave> are scoping ops, and their job is to perform any
995 housekeeping every time you enter and leave a block: lexical variables
996 are tidied up, unreferenced variables are destroyed, and so on. Every
997 program will have those first three lines: C<leave> is a list, and its
998 children are all the statements in the block. Statements are delimited
999 by C<nextstate>, so a block is a collection of C<nextstate> ops, with
1000 the ops to be performed for each statement being the children of
1001 C<nextstate>. C<enter> is a single op which functions as a marker.
1003 That's how Perl parsed the program, from top to bottom:
1016 However, it's impossible to B<perform> the operations in this order:
1017 you have to find the values of C<$b> and C<$c> before you add them
1018 together, for instance. So, the other thread that runs through the op
1019 tree is the execution order: each op has a field C<op_next> which points
1020 to the next op to be run, so following these pointers tells us how perl
1021 executes the code. We can traverse the tree in this order using
1022 the C<exec> option to C<B::Terse>:
1024 % perl -MO=Terse,exec -e '$a=$b+$c'
1025 1 OP (0x8179928) enter
1026 2 COP (0x81798c8) nextstate
1027 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
1028 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
1029 5 BINOP (0x8179878) add [1]
1030 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
1031 7 BINOP (0x81798a0) sassign
1032 8 LISTOP (0x8179900) leave
1034 This probably makes more sense for a human: enter a block, start a
1035 statement. Get the values of C<$b> and C<$c>, and add them together.
1036 Find C<$a>, and assign one to the other. Then leave.
1038 The way Perl builds up these op trees in the parsing process can be
1039 unravelled by examining F<perly.y>, the YACC grammar. Let's take the
1040 piece we need to construct the tree for C<$a = $b + $c>
1042 1 term : term ASSIGNOP term
1043 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
1045 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1047 If you're not used to reading BNF grammars, this is how it works: You're
1048 fed certain things by the tokeniser, which generally end up in upper
1049 case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
1050 code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
1051 "terminal symbols", because you can't get any simpler than them.
1053 The grammar, lines one and three of the snippet above, tells you how to
1054 build up more complex forms. These complex forms, "non-terminal symbols"
1055 are generally placed in lower case. C<term> here is a non-terminal
1056 symbol, representing a single expression.
1058 The grammar gives you the following rule: you can make the thing on the
1059 left of the colon if you see all the things on the right in sequence.
1060 This is called a "reduction", and the aim of parsing is to completely
1061 reduce the input. There are several different ways you can perform a
1062 reduction, separated by vertical bars: so, C<term> followed by C<=>
1063 followed by C<term> makes a C<term>, and C<term> followed by C<+>
1064 followed by C<term> can also make a C<term>.
1066 So, if you see two terms with an C<=> or C<+>, between them, you can
1067 turn them into a single expression. When you do this, you execute the
1068 code in the block on the next line: if you see C<=>, you'll do the code
1069 in line 2. If you see C<+>, you'll do the code in line 4. It's this code
1070 which contributes to the op tree.
1073 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
1075 What this does is creates a new binary op, and feeds it a number of
1076 variables. The variables refer to the tokens: C<$1> is the first token in
1077 the input, C<$2> the second, and so on - think regular expression
1078 backreferences. C<$$> is the op returned from this reduction. So, we
1079 call C<newBINOP> to create a new binary operator. The first parameter to
1080 C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
1081 operator, so we want the type to be C<ADDOP>. We could specify this
1082 directly, but it's right there as the second token in the input, so we
1083 use C<$2>. The second parameter is the op's flags: 0 means "nothing
1084 special". Then the things to add: the left and right hand side of our
1085 expression, in scalar context.
1089 When perl executes something like C<addop>, how does it pass on its
1090 results to the next op? The answer is, through the use of stacks. Perl
1091 has a number of stacks to store things it's currently working on, and
1092 we'll look at the three most important ones here.
1096 =item Argument stack
1098 Arguments are passed to PP code and returned from PP code using the
1099 argument stack, C<ST>. The typical way to handle arguments is to pop
1100 them off the stack, deal with them how you wish, and then push the result
1101 back onto the stack. This is how, for instance, the cosine operator
1106 value = Perl_cos(value);
1109 We'll see a more tricky example of this when we consider Perl's macros
1110 below. C<POPn> gives you the NV (floating point value) of the top SV on
1111 the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
1112 the result back as an NV. The C<X> in C<XPUSHn> means that the stack
1113 should be extended if necessary - it can't be necessary here, because we
1114 know there's room for one more item on the stack, since we've just
1115 removed one! The C<XPUSH*> macros at least guarantee safety.
1117 Alternatively, you can fiddle with the stack directly: C<SP> gives you
1118 the first element in your portion of the stack, and C<TOP*> gives you
1119 the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
1120 negation of an integer:
1124 Just set the integer value of the top stack entry to its negation.
1126 Argument stack manipulation in the core is exactly the same as it is in
1127 XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
1128 description of the macros used in stack manipulation.
1132 I say "your portion of the stack" above because PP code doesn't
1133 necessarily get the whole stack to itself: if your function calls
1134 another function, you'll only want to expose the arguments aimed for the
1135 called function, and not (necessarily) let it get at your own data. The
1136 way we do this is to have a "virtual" bottom-of-stack, exposed to each
1137 function. The mark stack keeps bookmarks to locations in the argument
1138 stack usable by each function. For instance, when dealing with a tied
1139 variable, (internally, something with "P" magic) Perl has to call
1140 methods for accesses to the tied variables. However, we need to separate
1141 the arguments exposed to the method to the argument exposed to the
1142 original function - the store or fetch or whatever it may be. Here's
1143 roughly how the tied C<push> is implemented; see C<av_push> in F<av.c>:
1147 3 PUSHs(SvTIED_obj((SV*)av, mg));
1151 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1154 Let's examine the whole implementation, for practice:
1158 Push the current state of the stack pointer onto the mark stack. This is
1159 so that when we've finished adding items to the argument stack, Perl
1160 knows how many things we've added recently.
1163 3 PUSHs(SvTIED_obj((SV*)av, mg));
1166 We're going to add two more items onto the argument stack: when you have
1167 a tied array, the C<PUSH> subroutine receives the object and the value
1168 to be pushed, and that's exactly what we have here - the tied object,
1169 retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
1173 Next we tell Perl to update the global stack pointer from our internal
1174 variable: C<dSP> only gave us a local copy, not a reference to the global.
1177 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1180 C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
1181 variables are tidied up, everything that has been localised gets
1182 its previous value returned, and so on. Think of them as the C<{> and
1183 C<}> of a Perl block.
1185 To actually do the magic method call, we have to call a subroutine in
1186 Perl space: C<call_method> takes care of that, and it's described in
1187 L<perlcall>. We call the C<PUSH> method in scalar context, and we're
1188 going to discard its return value. The call_method() function
1189 removes the top element of the mark stack, so there is nothing for
1190 the caller to clean up.
1194 C doesn't have a concept of local scope, so perl provides one. We've
1195 seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
1196 stack implements the C equivalent of, for example:
1203 See L<perlguts/Localising Changes> for how to use the save stack.
1207 =head2 Millions of Macros
1209 One thing you'll notice about the Perl source is that it's full of
1210 macros. Some have called the pervasive use of macros the hardest thing
1211 to understand, others find it adds to clarity. Let's take an example,
1212 the code which implements the addition operator:
1216 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
1219 6 SETn( left + right );
1224 Every line here (apart from the braces, of course) contains a macro. The
1225 first line sets up the function declaration as Perl expects for PP code;
1226 line 3 sets up variable declarations for the argument stack and the
1227 target, the return value of the operation. Finally, it tries to see if
1228 the addition operation is overloaded; if so, the appropriate subroutine
1231 Line 5 is another variable declaration - all variable declarations start
1232 with C<d> - which pops from the top of the argument stack two NVs (hence
1233 C<nn>) and puts them into the variables C<right> and C<left>, hence the
1234 C<rl>. These are the two operands to the addition operator. Next, we
1235 call C<SETn> to set the NV of the return value to the result of adding
1236 the two values. This done, we return - the C<RETURN> macro makes sure
1237 that our return value is properly handled, and we pass the next operator
1238 to run back to the main run loop.
1240 Most of these macros are explained in L<perlapi>, and some of the more
1241 important ones are explained in L<perlxs> as well. Pay special attention
1242 to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
1243 the C<[pad]THX_?> macros.
1245 =head2 The .i Targets
1247 You can expand the macros in a F<foo.c> file by saying
1251 which will expand the macros using cpp. Don't be scared by the results.
1255 Every module and built-in function has an associated test file (or
1256 should...). If you add or change functionality, you have to write a
1257 test. If you fix a bug, you have to write a test so that bug never
1258 comes back. If you alter the docs, it would be nice to test what the
1259 new documentation says.
1261 In short, if you submit a patch you probably also have to patch the
1264 =head2 Where to find test files
1266 For modules, the test file is right next to the module itself.
1267 F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation,
1268 so there are some snags (and it would be wonderful for you to brush
1269 them out), but it basically works that way. Everything else lives in
1272 Testing of warning messages is often separately done by using expect scripts in
1273 F<t/lib/warnings>. This is because much of the setup for them is already done
1276 If you add a new test directory under F<t/>, it is imperative that you
1277 add that directory to F<t/HARNESS> and F<t/TEST>.
1283 Testing of the absolute basic functionality of Perl. Things like
1284 C<if>, basic file reads and writes, simple regexes, etc. These are
1285 run first in the test suite and if any of them fail, something is
1290 These test the basic control structures, C<if/else>, C<while>,
1295 Tests basic issues of how Perl parses and compiles itself.
1299 Tests for built-in IO functions, including command line arguments.
1303 The old home for the module tests, you shouldn't put anything new in
1304 here. There are still some bits and pieces hanging around in here
1305 that need to be moved. Perhaps you could move them? Thanks!
1309 Tests for perl's method resolution order implementations
1314 Tests for perl's built in functions that don't fit into any of the
1319 Tests for regex related functions or behaviour. (These used to live
1324 Testing features of how perl actually runs, including exit codes and
1325 handling of PERL* environment variables.
1329 Tests for the core support of Unicode.
1333 Windows-specific tests.
1337 A test suite for the s2p converter.
1341 The core uses the same testing style as the rest of Perl, a simple
1342 "ok/not ok" run through Test::Harness, but there are a few special
1345 There are three ways to write a test in the core. Test::More,
1346 t/test.pl and ad hoc C<print $test ? "ok 42\n" : "not ok 42\n">. The
1347 decision of which to use depends on what part of the test suite you're
1348 working on. This is a measure to prevent a high-level failure (such
1349 as Config.pm breaking) from causing basic functionality tests to fail.
1350 If you write your own test, use the L<Test Anything Protocol|TAP>.
1356 Since we don't know if require works, or even subroutines, use ad hoc
1357 tests for these two. Step carefully to avoid using the feature being
1360 =item t/cmd t/run t/io t/op
1362 Now that basic require() and subroutines are tested, you can use the
1363 t/test.pl library which emulates the important features of Test::More
1364 while using a minimum of core features.
1366 You can also conditionally use certain libraries like Config, but be
1367 sure to skip the test gracefully if it's not there.
1371 Now that the core of Perl is tested, Test::More can be used. You can
1372 also use the full suite of core modules in the tests.
1376 When you say "make test" Perl uses the F<t/TEST> program to run the
1377 test suite (except under Win32 where it uses F<t/harness> instead.)
1378 All tests are run from the F<t/> directory, B<not> the directory
1379 which contains the test. This causes some problems with the tests
1380 in F<lib/>, so here's some opportunity for some patching.
1382 You must be triply conscious of cross-platform concerns. This usually
1383 boils down to using File::Spec and avoiding things like C<fork()> and
1384 C<system()> unless absolutely necessary.
1386 =head2 Special Make Test Targets
1388 There are various special make targets that can be used to test Perl
1389 slightly differently than the standard "test" target. Not all them
1390 are expected to give a 100% success rate. Many of them have several
1391 aliases, and many of them are not available on certain operating
1398 Run F<perl> on all core tests (F<t/*> and F<lib/[a-z]*> pragma tests).
1400 (Not available on Win32)
1404 Run all the tests through B::Deparse. Not all tests will succeed.
1406 (Not available on Win32)
1408 =item test.taintwarn
1410 Run all tests with the B<-t> command-line switch. Not all tests
1411 are expected to succeed (until they're specifically fixed, of course).
1413 (Not available on Win32)
1417 Run F<miniperl> on F<t/base>, F<t/comp>, F<t/cmd>, F<t/run>, F<t/io>,
1418 F<t/op>, F<t/uni> and F<t/mro> tests.
1420 =item test.valgrind check.valgrind utest.valgrind ucheck.valgrind
1422 (Only in Linux) Run all the tests using the memory leak + naughty
1423 memory access tool "valgrind". The log files will be named
1424 F<testname.valgrind>.
1426 =item test.third check.third utest.third ucheck.third
1428 (Only in Tru64) Run all the tests using the memory leak + naughty
1429 memory access tool "Third Degree". The log files will be named
1430 F<perl.3log.testname>.
1432 =item test.torture torturetest
1434 Run all the usual tests and some extra tests. As of Perl 5.8.0 the
1435 only extra tests are Abigail's JAPHs, F<t/japh/abigail.t>.
1437 You can also run the torture test with F<t/harness> by giving
1438 C<-torture> argument to F<t/harness>.
1440 =item utest ucheck test.utf8 check.utf8
1442 Run all the tests with -Mutf8. Not all tests will succeed.
1444 (Not available on Win32)
1446 =item minitest.utf16 test.utf16
1448 Runs the tests with UTF-16 encoded scripts, encoded with different
1449 versions of this encoding.
1451 C<make utest.utf16> runs the test suite with a combination of C<-utf8> and
1452 C<-utf16> arguments to F<t/TEST>.
1454 (Not available on Win32)
1458 Run the test suite with the F<t/harness> controlling program, instead of
1459 F<t/TEST>. F<t/harness> is more sophisticated, and uses the
1460 L<Test::Harness> module, thus using this test target supposes that perl
1461 mostly works. The main advantage for our purposes is that it prints a
1462 detailed summary of failed tests at the end. Also, unlike F<t/TEST>, it
1463 doesn't redirect stderr to stdout.
1465 Note that under Win32 F<t/harness> is always used instead of F<t/TEST>, so
1466 there is no special "test_harness" target.
1468 Under Win32's "test" target you may use the TEST_SWITCHES and TEST_FILES
1469 environment variables to control the behaviour of F<t/harness>. This means
1472 nmake test TEST_FILES="op/*.t"
1473 nmake test TEST_SWITCHES="-torture" TEST_FILES="op/*.t"
1475 =item Parallel tests
1477 The core distribution can now run its regression tests in parallel on
1478 Unix-like platforms. Instead of running C<make test>, set C<TEST_JOBS> in
1479 your environment to the number of tests to run in parallel, and run
1480 C<make test_harness>. On a Bourne-like shell, this can be done as
1482 TEST_JOBS=3 make test_harness # Run 3 tests in parallel
1484 An environment variable is used, rather than parallel make itself, because
1485 L<TAP::Harness> needs to be able to schedule individual non-conflicting test
1486 scripts itself, and there is no standard interface to C<make> utilities to
1487 interact with their job schedulers.
1489 Note that currently some test scripts may fail when run in parallel (most
1490 notably C<ext/IO/t/io_dir.t>). If necessary run just the failing scripts
1491 again sequentially and see if the failures go away.
1492 =item test-notty test_notty
1494 Sets PERL_SKIP_TTY_TEST to true before running normal test.
1498 =head2 Running tests by hand
1500 You can run part of the test suite by hand by using one the following
1501 commands from the F<t/> directory :
1503 ./perl -I../lib TEST list-of-.t-files
1507 ./perl -I../lib harness list-of-.t-files
1509 (if you don't specify test scripts, the whole test suite will be run.)
1511 =head3 Using t/harness for testing
1513 If you use C<harness> for testing you have several command line options
1514 available to you. The arguments are as follows, and are in the order
1515 that they must appear if used together.
1517 harness -v -torture -re=pattern LIST OF FILES TO TEST
1518 harness -v -torture -re LIST OF PATTERNS TO MATCH
1520 If C<LIST OF FILES TO TEST> is omitted the file list is obtained from
1521 the manifest. The file list may include shell wildcards which will be
1528 Run the tests under verbose mode so you can see what tests were run,
1533 Run the torture tests as well as the normal set.
1537 Filter the file list so that all the test files run match PATTERN.
1538 Note that this form is distinct from the B<-re LIST OF PATTERNS> form below
1539 in that it allows the file list to be provided as well.
1541 =item -re LIST OF PATTERNS
1543 Filter the file list so that all the test files run match
1544 /(LIST|OF|PATTERNS)/. Note that with this form the patterns
1545 are joined by '|' and you cannot supply a list of files, instead
1546 the test files are obtained from the MANIFEST.
1550 You can run an individual test by a command similar to
1552 ./perl -I../lib patho/to/foo.t
1554 except that the harnesses set up some environment variables that may
1555 affect the execution of the test :
1561 indicates that we're running this test part of the perl core test suite.
1562 This is useful for modules that have a dual life on CPAN.
1564 =item PERL_DESTRUCT_LEVEL=2
1566 is set to 2 if it isn't set already (see L</PERL_DESTRUCT_LEVEL>)
1570 (used only by F<t/TEST>) if set, overrides the path to the perl executable
1571 that should be used to run the tests (the default being F<./perl>).
1573 =item PERL_SKIP_TTY_TEST
1575 if set, tells to skip the tests that need a terminal. It's actually set
1576 automatically by the Makefile, but can also be forced artificially by
1577 running 'make test_notty'.
1581 =head3 Other environment variables that may influence tests
1585 =item PERL_TEST_Net_Ping
1587 Setting this variable runs all the Net::Ping modules tests,
1588 otherwise some tests that interact with the outside world are skipped.
1591 =item PERL_TEST_NOVREXX
1593 Setting this variable skips the vrexx.t tests for OS2::REXX.
1595 =item PERL_TEST_NUMCONVERTS
1597 This sets a variable in op/numconvert.t.
1601 See also the documentation for the Test and Test::Harness modules,
1602 for more environment variables that affect testing.
1604 =head1 EXAMPLE OF A SIMPLE PATCH
1606 All right, we've now had a look at how to navigate the Perl sources and
1607 some things you'll need to know when fiddling with them. Let's now get
1608 on and create a simple patch. Here's something Larry suggested: if a
1609 C<U> is the first active format during a C<pack>, (for example,
1610 C<pack "U3C8", @stuff>) then the resulting string should be treated as
1613 If you are working with a git clone of the Perl repository, you will want to
1614 create a branch for your changes. This will make creating a proper patch much
1615 simpler. See the L<perlrepository> for details on how to do this.
1617 =head2 Writing the patch
1619 How do we prepare to fix this up? First we locate the code in question -
1620 the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1621 files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1622 altering this file, let's copy it to F<pp.c~>.
1624 [Well, it was in F<pp.c> when this tutorial was written. It has now been
1625 split off with C<pp_unpack> to its own file, F<pp_pack.c>]
1627 Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1628 loop over the pattern, taking each format character in turn into
1629 C<datum_type>. Then for each possible format character, we swallow up
1630 the other arguments in the pattern (a field width, an asterisk, and so
1631 on) and convert the next chunk input into the specified format, adding
1632 it onto the output SV C<cat>.
1634 How do we know if the C<U> is the first format in the C<pat>? Well, if
1635 we have a pointer to the start of C<pat> then, if we see a C<U> we can
1636 test whether we're still at the start of the string. So, here's where
1640 register char *pat = SvPVx(*++MARK, fromlen);
1641 register char *patend = pat + fromlen;
1646 We'll have another string pointer in there:
1649 register char *pat = SvPVx(*++MARK, fromlen);
1650 register char *patend = pat + fromlen;
1656 And just before we start the loop, we'll set C<patcopy> to be the start
1661 sv_setpvn(cat, "", 0);
1663 while (pat < patend) {
1665 Now if we see a C<U> which was at the start of the string, we turn on
1666 the C<UTF8> flag for the output SV, C<cat>:
1668 + if (datumtype == 'U' && pat==patcopy+1)
1670 if (datumtype == '#') {
1671 while (pat < patend && *pat != '\n')
1674 Remember that it has to be C<patcopy+1> because the first character of
1675 the string is the C<U> which has been swallowed into C<datumtype!>
1677 Oops, we forgot one thing: what if there are spaces at the start of the
1678 pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1679 character, even though it's not the first thing in the pattern. In this
1680 case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1682 if (isSPACE(datumtype))
1687 if (isSPACE(datumtype)) {
1692 OK. That's the C part done. Now we must do two additional things before
1693 this patch is ready to go: we've changed the behaviour of Perl, and so
1694 we must document that change. We must also provide some more regression
1695 tests to make sure our patch works and doesn't create a bug somewhere
1696 else along the line.
1698 =head2 Testing the patch
1700 The regression tests for each operator live in F<t/op/>, and so we
1701 make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our
1702 tests to the end. First, we'll test that the C<U> does indeed create
1705 t/op/pack.t has a sensible ok() function, but if it didn't we could
1706 use the one from t/test.pl.
1708 require './test.pl';
1709 plan( tests => 159 );
1713 print 'not ' unless "1.20.300.4000" eq sprintf "%vd",
1714 pack("U*",1,20,300,4000);
1715 print "ok $test\n"; $test++;
1717 we can write the more sensible (see L<Test::More> for a full
1718 explanation of is() and other testing functions).
1720 is( "1.20.300.4000", sprintf "%vd", pack("U*",1,20,300,4000),
1721 "U* produces Unicode" );
1723 Now we'll test that we got that space-at-the-beginning business right:
1725 is( "1.20.300.4000", sprintf "%vd", pack(" U*",1,20,300,4000),
1726 " with spaces at the beginning" );
1728 And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1729 the first active format:
1731 isnt( v1.20.300.4000, sprintf "%vd", pack("C0U*",1,20,300,4000),
1732 "U* not first isn't Unicode" );
1734 Mustn't forget to change the number of tests which appears at the top,
1735 or else the automated tester will get confused. This will either look
1742 plan( tests => 156 );
1744 We now compile up Perl, and run it through the test suite. Our new
1747 =head2 Documenting the patch
1749 Finally, the documentation. The job is never done until the paperwork is
1750 over, so let's describe the change we've just made. The relevant place
1751 is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1752 this text in the description of C<pack>:
1756 If the pattern begins with a C<U>, the resulting string will be treated
1757 as UTF-8-encoded Unicode. You can force UTF-8 encoding on in a string
1758 with an initial C<U0>, and the bytes that follow will be interpreted as
1759 Unicode characters. If you don't want this to happen, you can begin
1760 your pattern with C<C0> (or anything else) to force Perl not to UTF-8
1761 encode your string, and then follow this with a C<U*> somewhere in your
1764 =head1 COMMON PROBLEMS
1766 Perl source plays by ANSI C89 rules: no C99 (or C++) extensions. In
1767 some cases we have to take pre-ANSI requirements into consideration.
1768 You don't care about some particular platform having broken Perl?
1769 I hear there is still a strong demand for J2EE programmers.
1771 =head2 Perl environment problems
1777 Not compiling with threading
1779 Compiling with threading (-Duseithreads) completely rewrites
1780 the function prototypes of Perl. You better try your changes
1781 with that. Related to this is the difference between "Perl_-less"
1782 and "Perl_-ly" APIs, for example:
1784 Perl_sv_setiv(aTHX_ ...);
1787 The first one explicitly passes in the context, which is needed for e.g.
1788 threaded builds. The second one does that implicitly; do not get them
1789 mixed. If you are not passing in a aTHX_, you will need to do a dTHX
1790 (or a dVAR) as the first thing in the function.
1792 See L<perlguts/"How multiple interpreters and concurrency are supported">
1793 for further discussion about context.
1797 Not compiling with -DDEBUGGING
1799 The DEBUGGING define exposes more code to the compiler,
1800 therefore more ways for things to go wrong. You should try it.
1804 Introducing (non-read-only) globals
1806 Do not introduce any modifiable globals, truly global or file static.
1807 They are bad form and complicate multithreading and other forms of
1808 concurrency. The right way is to introduce them as new interpreter
1809 variables, see F<intrpvar.h> (at the very end for binary compatibility).
1811 Introducing read-only (const) globals is okay, as long as you verify
1812 with e.g. C<nm libperl.a|egrep -v ' [TURtr] '> (if your C<nm> has
1813 BSD-style output) that the data you added really is read-only.
1814 (If it is, it shouldn't show up in the output of that command.)
1816 If you want to have static strings, make them constant:
1818 static const char etc[] = "...";
1820 If you want to have arrays of constant strings, note carefully
1821 the right combination of C<const>s:
1823 static const char * const yippee[] =
1824 {"hi", "ho", "silver"};
1826 There is a way to completely hide any modifiable globals (they are all
1827 moved to heap), the compilation setting C<-DPERL_GLOBAL_STRUCT_PRIVATE>.
1828 It is not normally used, but can be used for testing, read more
1829 about it in L<perlguts/"Background and PERL_IMPLICIT_CONTEXT">.
1833 Not exporting your new function
1835 Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any
1836 function that is part of the public API (the shared Perl library)
1837 to be explicitly marked as exported. See the discussion about
1838 F<embed.pl> in L<perlguts>.
1842 Exporting your new function
1844 The new shiny result of either genuine new functionality or your
1845 arduous refactoring is now ready and correctly exported. So what
1846 could possibly go wrong?
1848 Maybe simply that your function did not need to be exported in the
1849 first place. Perl has a long and not so glorious history of exporting
1850 functions that it should not have.
1852 If the function is used only inside one source code file, make it
1853 static. See the discussion about F<embed.pl> in L<perlguts>.
1855 If the function is used across several files, but intended only for
1856 Perl's internal use (and this should be the common case), do not
1857 export it to the public API. See the discussion about F<embed.pl>
1862 =head2 Portability problems
1864 The following are common causes of compilation and/or execution
1865 failures, not common to Perl as such. The C FAQ is good bedtime
1866 reading. Please test your changes with as many C compilers and
1867 platforms as possible; we will, anyway, and it's nice to save
1868 oneself from public embarrassment.
1870 If using gcc, you can add the C<-std=c89> option which will hopefully
1871 catch most of these unportabilities. (However it might also catch
1872 incompatibilities in your system's header files.)
1874 Use the Configure C<-Dgccansipedantic> flag to enable the gcc
1875 C<-ansi -pedantic> flags which enforce stricter ANSI rules.
1877 If using the C<gcc -Wall> note that not all the possible warnings
1878 (like C<-Wunitialized>) are given unless you also compile with C<-O>.
1880 Note that if using gcc, starting from Perl 5.9.5 the Perl core source
1881 code files (the ones at the top level of the source code distribution,
1882 but not e.g. the extensions under ext/) are automatically compiled
1883 with as many as possible of the C<-std=c89>, C<-ansi>, C<-pedantic>,
1884 and a selection of C<-W> flags (see cflags.SH).
1886 Also study L<perlport> carefully to avoid any bad assumptions
1887 about the operating system, filesystems, and so forth.
1889 You may once in a while try a "make microperl" to see whether we
1890 can still compile Perl with just the bare minimum of interfaces.
1893 Do not assume an operating system indicates a certain compiler.
1899 Casting pointers to integers or casting integers to pointers
1901 void castaway(U8* p)
1907 void castaway(U8* p)
1911 Both are bad, and broken, and unportable. Use the PTR2IV()
1912 macro that does it right. (Likewise, there are PTR2UV(), PTR2NV(),
1913 INT2PTR(), and NUM2PTR().)
1917 Casting between data function pointers and data pointers
1919 Technically speaking casting between function pointers and data
1920 pointers is unportable and undefined, but practically speaking
1921 it seems to work, but you should use the FPTR2DPTR() and DPTR2FPTR()
1922 macros. Sometimes you can also play games with unions.
1926 Assuming sizeof(int) == sizeof(long)
1928 There are platforms where longs are 64 bits, and platforms where ints
1929 are 64 bits, and while we are out to shock you, even platforms where
1930 shorts are 64 bits. This is all legal according to the C standard.
1931 (In other words, "long long" is not a portable way to specify 64 bits,
1932 and "long long" is not even guaranteed to be any wider than "long".)
1934 Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth.
1935 Avoid things like I32 because they are B<not> guaranteed to be
1936 I<exactly> 32 bits, they are I<at least> 32 bits, nor are they
1937 guaranteed to be B<int> or B<long>. If you really explicitly need
1938 64-bit variables, use I64 and U64, but only if guarded by HAS_QUAD.
1942 Assuming one can dereference any type of pointer for any type of data
1945 long pony = *p; /* BAD */
1947 Many platforms, quite rightly so, will give you a core dump instead
1948 of a pony if the p happens not be correctly aligned.
1954 (int)*p = ...; /* BAD */
1956 Simply not portable. Get your lvalue to be of the right type,
1957 or maybe use temporary variables, or dirty tricks with unions.
1961 Assume B<anything> about structs (especially the ones you
1962 don't control, like the ones coming from the system headers)
1968 That a certain field exists in a struct
1972 That no other fields exist besides the ones you know of
1976 That a field is of certain signedness, sizeof, or type
1980 That the fields are in a certain order
1986 While C guarantees the ordering specified in the struct definition,
1987 between different platforms the definitions might differ
1993 That the sizeof(struct) or the alignments are the same everywhere
1999 There might be padding bytes between the fields to align the fields -
2000 the bytes can be anything
2004 Structs are required to be aligned to the maximum alignment required
2005 by the fields - which for native types is for usually equivalent to
2006 sizeof() of the field
2014 Assuming the character set is ASCIIish
2016 Perl can compile and run under EBCDIC platforms. See L<perlebcdic>.
2017 This is transparent for the most part, but because the character sets
2018 differ, you shouldn't use numeric (decimal, octal, nor hex) constants
2019 to refer to characters. You can safely say 'A', but not 0x41.
2020 You can safely say '\n', but not \012.
2021 If a character doesn't have a trivial input form, you can
2022 create a #define for it in both C<utfebcdic.h> and C<utf8.h>, so that
2023 it resolves to different values depending on the character set being used.
2024 (There are three different EBCDIC character sets defined in C<utfebcdic.h>,
2025 so it might be best to insert the #define three times in that file.)
2027 Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper case
2028 alphabetic characters. That is not true in EBCDIC. Nor for 'a' to 'z'.
2029 But '0' - '9' is an unbroken range in both systems. Don't assume anything
2032 Many of the comments in the existing code ignore the possibility of EBCDIC,
2033 and may be wrong therefore, even if the code works.
2034 This is actually a tribute to the successful transparent insertion of being
2035 able to handle EBCDIC without having to change pre-existing code.
2037 UTF-8 and UTF-EBCDIC are two different encodings used to represent Unicode
2038 code points as sequences of bytes. Macros
2039 with the same names (but different definitions)
2040 in C<utf8.h> and C<utfebcdic.h>
2041 are used to allow the calling code to think that there is only one such
2043 This is almost always referred to as C<utf8>, but it means the EBCDIC version
2044 as well. Again, comments in the code may well be wrong even if the code itself
2046 For example, the concept of C<invariant characters> differs between ASCII and
2048 On ASCII platforms, only characters that do not have the high-order
2049 bit set (i.e. whose ordinals are strict ASCII, 0 - 127)
2050 are invariant, and the documentation and comments in the code
2052 often referring to something like, say, C<hibit>.
2053 The situation differs and is not so simple on EBCDIC machines, but as long as
2054 the code itself uses the C<NATIVE_IS_INVARIANT()> macro appropriately, it
2055 works, even if the comments are wrong.
2059 Assuming the character set is just ASCII
2061 ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra
2062 characters have different meanings depending on the locale. Absent a locale,
2063 currently these extra characters are generally considered to be unassigned,
2064 and this has presented some problems.
2065 This is being changed starting in 5.12 so that these characters will
2066 be considered to be Latin-1 (ISO-8859-1).
2070 Mixing #define and #ifdef
2072 #define BURGLE(x) ... \
2073 #ifdef BURGLE_OLD_STYLE /* BAD */
2074 ... do it the old way ... \
2076 ... do it the new way ... \
2079 You cannot portably "stack" cpp directives. For example in the above
2080 you need two separate BURGLE() #defines, one for each #ifdef branch.
2084 Adding non-comment stuff after #endif or #else
2088 #else !SNOSH /* BAD */
2090 #endif SNOSH /* BAD */
2092 The #endif and #else cannot portably have anything non-comment after
2093 them. If you want to document what is going (which is a good idea
2094 especially if the branches are long), use (C) comments:
2102 The gcc option C<-Wendif-labels> warns about the bad variant
2103 (by default on starting from Perl 5.9.4).
2107 Having a comma after the last element of an enum list
2115 is not portable. Leave out the last comma.
2117 Also note that whether enums are implicitly morphable to ints
2118 varies between compilers, you might need to (int).
2124 // This function bamfoodles the zorklator. /* BAD */
2126 That is C99 or C++. Perl is C89. Using the //-comments is silently
2127 allowed by many C compilers but cranking up the ANSI C89 strictness
2128 (which we like to do) causes the compilation to fail.
2132 Mixing declarations and code
2137 set_zorkmids(n); /* BAD */
2140 That is C99 or C++. Some C compilers allow that, but you shouldn't.
2142 The gcc option C<-Wdeclaration-after-statements> scans for such problems
2143 (by default on starting from Perl 5.9.4).
2147 Introducing variables inside for()
2149 for(int i = ...; ...; ...) { /* BAD */
2151 That is C99 or C++. While it would indeed be awfully nice to have that
2152 also in C89, to limit the scope of the loop variable, alas, we cannot.
2156 Mixing signed char pointers with unsigned char pointers
2158 int foo(char *s) { ... }
2160 unsigned char *t = ...; /* Or U8* t = ... */
2163 While this is legal practice, it is certainly dubious, and downright
2164 fatal in at least one platform: for example VMS cc considers this a
2165 fatal error. One cause for people often making this mistake is that a
2166 "naked char" and therefore dereferencing a "naked char pointer" have
2167 an undefined signedness: it depends on the compiler and the flags of
2168 the compiler and the underlying platform whether the result is signed
2169 or unsigned. For this very same reason using a 'char' as an array
2174 Macros that have string constants and their arguments as substrings of
2175 the string constants
2177 #define FOO(n) printf("number = %d\n", n) /* BAD */
2180 Pre-ANSI semantics for that was equivalent to
2182 printf("10umber = %d\10");
2184 which is probably not what you were expecting. Unfortunately at least
2185 one reasonably common and modern C compiler does "real backward
2186 compatibility" here, in AIX that is what still happens even though the
2187 rest of the AIX compiler is very happily C89.
2191 Using printf formats for non-basic C types
2194 printf("i = %d\n", i); /* BAD */
2196 While this might by accident work in some platform (where IV happens
2197 to be an C<int>), in general it cannot. IV might be something larger.
2198 Even worse the situation is with more specific types (defined by Perl's
2199 configuration step in F<config.h>):
2202 printf("who = %d\n", who); /* BAD */
2204 The problem here is that Uid_t might be not only not C<int>-wide
2205 but it might also be unsigned, in which case large uids would be
2206 printed as negative values.
2208 There is no simple solution to this because of printf()'s limited
2209 intelligence, but for many types the right format is available as
2210 with either 'f' or '_f' suffix, for example:
2212 IVdf /* IV in decimal */
2213 UVxf /* UV is hexadecimal */
2215 printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
2217 Uid_t_f /* Uid_t in decimal */
2219 printf("who = %"Uid_t_f"\n", who);
2221 Or you can try casting to a "wide enough" type:
2223 printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);
2225 Also remember that the C<%p> format really does require a void pointer:
2228 printf("p = %p\n", (void*)p);
2230 The gcc option C<-Wformat> scans for such problems.
2234 Blindly using variadic macros
2236 gcc has had them for a while with its own syntax, and C99 brought
2237 them with a standardized syntax. Don't use the former, and use
2238 the latter only if the HAS_C99_VARIADIC_MACROS is defined.
2242 Blindly passing va_list
2244 Not all platforms support passing va_list to further varargs (stdarg)
2245 functions. The right thing to do is to copy the va_list using the
2246 Perl_va_copy() if the NEED_VA_COPY is defined.
2250 Using gcc statement expressions
2252 val = ({...;...;...}); /* BAD */
2254 While a nice extension, it's not portable. The Perl code does
2255 admittedly use them if available to gain some extra speed
2256 (essentially as a funky form of inlining), but you shouldn't.
2260 Binding together several statements in a macro
2262 Use the macros STMT_START and STMT_END.
2270 Testing for operating systems or versions when should be testing for features
2272 #ifdef __FOONIX__ /* BAD */
2276 Unless you know with 100% certainty that quux() is only ever available
2277 for the "Foonix" operating system B<and> that is available B<and>
2278 correctly working for B<all> past, present, B<and> future versions of
2279 "Foonix", the above is very wrong. This is more correct (though still
2280 not perfect, because the below is a compile-time check):
2286 How does the HAS_QUUX become defined where it needs to be? Well, if
2287 Foonix happens to be Unixy enough to be able to run the Configure
2288 script, and Configure has been taught about detecting and testing
2289 quux(), the HAS_QUUX will be correctly defined. In other platforms,
2290 the corresponding configuration step will hopefully do the same.
2292 In a pinch, if you cannot wait for Configure to be educated,
2293 or if you have a good hunch of where quux() might be available,
2294 you can temporarily try the following:
2296 #if (defined(__FOONIX__) || defined(__BARNIX__))
2306 But in any case, try to keep the features and operating systems separate.
2310 =head2 Problematic System Interfaces
2316 malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable
2317 allocate at least one byte. (In general you should rarely need to
2318 work at this low level, but instead use the various malloc wrappers.)
2322 snprintf() - the return type is unportable. Use my_snprintf() instead.
2326 =head2 Security problems
2328 Last but not least, here are various tips for safer coding.
2336 Or we will publicly ridicule you. Seriously.
2340 Do not use strcpy() or strcat() or strncpy() or strncat()
2342 Use my_strlcpy() and my_strlcat() instead: they either use the native
2343 implementation, or Perl's own implementation (borrowed from the public
2344 domain implementation of INN).
2348 Do not use sprintf() or vsprintf()
2350 If you really want just plain byte strings, use my_snprintf()
2351 and my_vsnprintf() instead, which will try to use snprintf() and
2352 vsnprintf() if those safer APIs are available. If you want something
2353 fancier than a plain byte string, use SVs and Perl_sv_catpvf().
2360 You can compile a special debugging version of Perl, which allows you
2361 to use the C<-D> option of Perl to tell more about what Perl is doing.
2362 But sometimes there is no alternative than to dive in with a debugger,
2363 either to see the stack trace of a core dump (very useful in a bug
2364 report), or trying to figure out what went wrong before the core dump
2365 happened, or how did we end up having wrong or unexpected results.
2367 =head2 Poking at Perl
2369 To really poke around with Perl, you'll probably want to build Perl for
2370 debugging, like this:
2372 ./Configure -d -D optimize=-g
2375 C<-g> is a flag to the C compiler to have it produce debugging
2376 information which will allow us to step through a running program,
2377 and to see in which C function we are at (without the debugging
2378 information we might see only the numerical addresses of the functions,
2379 which is not very helpful).
2381 F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
2382 enables all the internal debugging code in Perl. There are a whole bunch
2383 of things you can debug with this: L<perlrun> lists them all, and the
2384 best way to find out about them is to play about with them. The most
2385 useful options are probably
2387 l Context (loop) stack processing
2389 o Method and overloading resolution
2390 c String/numeric conversions
2392 Some of the functionality of the debugging code can be achieved using XS
2395 -Dr => use re 'debug'
2396 -Dx => use O 'Debug'
2398 =head2 Using a source-level debugger
2400 If the debugging output of C<-D> doesn't help you, it's time to step
2401 through perl's execution with a source-level debugger.
2407 We'll use C<gdb> for our examples here; the principles will apply to
2408 any debugger (many vendors call their debugger C<dbx>), but check the
2409 manual of the one you're using.
2413 To fire up the debugger, type
2417 Or if you have a core dump:
2421 You'll want to do that in your Perl source tree so the debugger can read
2422 the source code. You should see the copyright message, followed by the
2427 C<help> will get you into the documentation, but here are the most
2434 Run the program with the given arguments.
2436 =item break function_name
2438 =item break source.c:xxx
2440 Tells the debugger that we'll want to pause execution when we reach
2441 either the named function (but see L<perlguts/Internal Functions>!) or the given
2442 line in the named source file.
2446 Steps through the program a line at a time.
2450 Steps through the program a line at a time, without descending into
2455 Run until the next breakpoint.
2459 Run until the end of the current function, then stop again.
2463 Just pressing Enter will do the most recent operation again - it's a
2464 blessing when stepping through miles of source code.
2468 Execute the given C code and print its results. B<WARNING>: Perl makes
2469 heavy use of macros, and F<gdb> does not necessarily support macros
2470 (see later L</"gdb macro support">). You'll have to substitute them
2471 yourself, or to invoke cpp on the source code files
2472 (see L</"The .i Targets">)
2473 So, for instance, you can't say
2475 print SvPV_nolen(sv)
2479 print Perl_sv_2pv_nolen(sv)
2483 You may find it helpful to have a "macro dictionary", which you can
2484 produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
2485 recursively apply those macros for you.
2487 =head2 gdb macro support
2489 Recent versions of F<gdb> have fairly good macro support, but
2490 in order to use it you'll need to compile perl with macro definitions
2491 included in the debugging information. Using F<gcc> version 3.1, this
2492 means configuring with C<-Doptimize=-g3>. Other compilers might use a
2493 different switch (if they support debugging macros at all).
2495 =head2 Dumping Perl Data Structures
2497 One way to get around this macro hell is to use the dumping functions in
2498 F<dump.c>; these work a little like an internal
2499 L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
2500 that you can't get at from Perl. Let's take an example. We'll use the
2501 C<$a = $b + $c> we used before, but give it a bit of context:
2502 C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
2504 What about C<pp_add>, the function we examined earlier to implement the
2507 (gdb) break Perl_pp_add
2508 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
2510 Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>.
2511 With the breakpoint in place, we can run our program:
2513 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
2515 Lots of junk will go past as gdb reads in the relevant source files and
2516 libraries, and then:
2518 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
2519 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
2524 We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
2525 arranges for two C<NV>s to be placed into C<left> and C<right> - let's
2528 #define dPOPTOPnnrl_ul NV right = POPn; \
2529 SV *leftsv = TOPs; \
2530 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
2532 C<POPn> takes the SV from the top of the stack and obtains its NV either
2533 directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
2534 C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
2535 C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
2536 C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
2538 Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
2539 convert it. If we step again, we'll find ourselves there:
2541 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
2545 We can now use C<Perl_sv_dump> to investigate the SV:
2547 SV = PV(0xa057cc0) at 0xa0675d0
2550 PV = 0xa06a510 "6XXXX"\0
2555 We know we're going to get C<6> from this, so let's finish the
2559 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
2560 0x462669 in Perl_pp_add () at pp_hot.c:311
2563 We can also dump out this op: the current op is always stored in
2564 C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
2565 similar output to L<B::Debug|B::Debug>.
2568 13 TYPE = add ===> 14
2570 FLAGS = (SCALAR,KIDS)
2572 TYPE = null ===> (12)
2574 FLAGS = (SCALAR,KIDS)
2576 11 TYPE = gvsv ===> 12
2582 # finish this later #
2584 =head1 SOURCE CODE STATIC ANALYSIS
2586 Various tools exist for analysing C source code B<statically>, as
2587 opposed to B<dynamically>, that is, without executing the code.
2588 It is possible to detect resource leaks, undefined behaviour, type
2589 mismatches, portability problems, code paths that would cause illegal
2590 memory accesses, and other similar problems by just parsing the C code
2591 and looking at the resulting graph, what does it tell about the
2592 execution and data flows. As a matter of fact, this is exactly
2593 how C compilers know to give warnings about dubious code.
2597 The good old C code quality inspector, C<lint>, is available in
2598 several platforms, but please be aware that there are several
2599 different implementations of it by different vendors, which means that
2600 the flags are not identical across different platforms.
2602 There is a lint variant called C<splint> (Secure Programming Lint)
2603 available from http://www.splint.org/ that should compile on any
2606 There are C<lint> and <splint> targets in Makefile, but you may have
2607 to diddle with the flags (see above).
2611 Coverity (http://www.coverity.com/) is a product similar to lint and
2612 as a testbed for their product they periodically check several open
2613 source projects, and they give out accounts to open source developers
2614 to the defect databases.
2616 =head2 cpd (cut-and-paste detector)
2618 The cpd tool detects cut-and-paste coding. If one instance of the
2619 cut-and-pasted code changes, all the other spots should probably be
2620 changed, too. Therefore such code should probably be turned into a
2621 subroutine or a macro.
2623 cpd (http://pmd.sourceforge.net/cpd.html) is part of the pmd project
2624 (http://pmd.sourceforge.net/). pmd was originally written for static
2625 analysis of Java code, but later the cpd part of it was extended to
2626 parse also C and C++.
2628 Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the
2629 pmd-X.Y.jar from it, and then run that on source code thusly:
2631 java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD --minimum-tokens 100 --files /some/where/src --language c > cpd.txt
2633 You may run into memory limits, in which case you should use the -Xmx option:
2639 Though much can be written about the inconsistency and coverage
2640 problems of gcc warnings (like C<-Wall> not meaning "all the
2641 warnings", or some common portability problems not being covered by
2642 C<-Wall>, or C<-ansi> and C<-pedantic> both being a poorly defined
2643 collection of warnings, and so forth), gcc is still a useful tool in
2644 keeping our coding nose clean.
2646 The C<-Wall> is by default on.
2648 The C<-ansi> (and its sidekick, C<-pedantic>) would be nice to be on
2649 always, but unfortunately they are not safe on all platforms, they can
2650 for example cause fatal conflicts with the system headers (Solaris
2651 being a prime example). If Configure C<-Dgccansipedantic> is used,
2652 the C<cflags> frontend selects C<-ansi -pedantic> for the platforms
2653 where they are known to be safe.
2655 Starting from Perl 5.9.4 the following extra flags are added:
2669 C<-Wdeclaration-after-statement>
2673 The following flags would be nice to have but they would first need
2674 their own Augean stablemaster:
2688 C<-Wstrict-prototypes>
2692 The C<-Wtraditional> is another example of the annoying tendency of
2693 gcc to bundle a lot of warnings under one switch (it would be
2694 impossible to deploy in practice because it would complain a lot) but
2695 it does contain some warnings that would be beneficial to have available
2696 on their own, such as the warning about string constants inside macros
2697 containing the macro arguments: this behaved differently pre-ANSI
2698 than it does in ANSI, and some C compilers are still in transition,
2699 AIX being an example.
2701 =head2 Warnings of other C compilers
2703 Other C compilers (yes, there B<are> other C compilers than gcc) often
2704 have their "strict ANSI" or "strict ANSI with some portability extensions"
2705 modes on, like for example the Sun Workshop has its C<-Xa> mode on
2706 (though implicitly), or the DEC (these days, HP...) has its C<-std1>
2709 =head1 MEMORY DEBUGGERS
2711 B<NOTE 1>: Running under memory debuggers such as Purify, valgrind, or
2712 Third Degree greatly slows down the execution: seconds become minutes,
2713 minutes become hours. For example as of Perl 5.8.1, the
2714 ext/Encode/t/Unicode.t takes extraordinarily long to complete under
2715 e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more
2716 than six hours, even on a snappy computer. The said test must be
2717 doing something that is quite unfriendly for memory debuggers. If you
2718 don't feel like waiting, that you can simply kill away the perl
2721 B<NOTE 2>: To minimize the number of memory leak false alarms (see
2722 L</PERL_DESTRUCT_LEVEL> for more information), you have to set the
2723 environment variable PERL_DESTRUCT_LEVEL to 2.
2725 For csh-like shells:
2727 setenv PERL_DESTRUCT_LEVEL 2
2729 For Bourne-type shells:
2731 PERL_DESTRUCT_LEVEL=2
2732 export PERL_DESTRUCT_LEVEL
2734 In Unixy environments you can also use the C<env> command:
2736 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...
2738 B<NOTE 3>: There are known memory leaks when there are compile-time
2739 errors within eval or require, seeing C<S_doeval> in the call stack
2740 is a good sign of these. Fixing these leaks is non-trivial,
2741 unfortunately, but they must be fixed eventually.
2743 B<NOTE 4>: L<DynaLoader> will not clean up after itself completely
2744 unless Perl is built with the Configure option
2745 C<-Accflags=-DDL_UNLOAD_ALL_AT_EXIT>.
2747 =head2 Rational Software's Purify
2749 Purify is a commercial tool that is helpful in identifying
2750 memory overruns, wild pointers, memory leaks and other such
2751 badness. Perl must be compiled in a specific way for
2752 optimal testing with Purify. Purify is available under
2753 Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.
2755 =head3 Purify on Unix
2757 On Unix, Purify creates a new Perl binary. To get the most
2758 benefit out of Purify, you should create the perl to Purify
2761 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
2762 -Uusemymalloc -Dusemultiplicity
2764 where these arguments mean:
2768 =item -Accflags=-DPURIFY
2770 Disables Perl's arena memory allocation functions, as well as
2771 forcing use of memory allocation functions derived from the
2774 =item -Doptimize='-g'
2776 Adds debugging information so that you see the exact source
2777 statements where the problem occurs. Without this flag, all
2778 you will see is the source filename of where the error occurred.
2782 Disable Perl's malloc so that Purify can more closely monitor
2783 allocations and leaks. Using Perl's malloc will make Purify
2784 report most leaks in the "potential" leaks category.
2786 =item -Dusemultiplicity
2788 Enabling the multiplicity option allows perl to clean up
2789 thoroughly when the interpreter shuts down, which reduces the
2790 number of bogus leak reports from Purify.
2794 Once you've compiled a perl suitable for Purify'ing, then you
2799 which creates a binary named 'pureperl' that has been Purify'ed.
2800 This binary is used in place of the standard 'perl' binary
2801 when you want to debug Perl memory problems.
2803 As an example, to show any memory leaks produced during the
2804 standard Perl testset you would create and run the Purify'ed
2809 ../pureperl -I../lib harness
2811 which would run Perl on test.pl and report any memory problems.
2813 Purify outputs messages in "Viewer" windows by default. If
2814 you don't have a windowing environment or if you simply
2815 want the Purify output to unobtrusively go to a log file
2816 instead of to the interactive window, use these following
2817 options to output to the log file "perl.log":
2819 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
2820 -log-file=perl.log -append-logfile=yes"
2822 If you plan to use the "Viewer" windows, then you only need this option:
2824 setenv PURIFYOPTIONS "-chain-length=25"
2826 In Bourne-type shells:
2829 export PURIFYOPTIONS
2831 or if you have the "env" utility:
2833 env PURIFYOPTIONS="..." ../pureperl ...
2837 Purify on Windows NT instruments the Perl binary 'perl.exe'
2838 on the fly. There are several options in the makefile you
2839 should change to get the most use out of Purify:
2845 You should add -DPURIFY to the DEFINES line so the DEFINES
2846 line looks something like:
2848 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1
2850 to disable Perl's arena memory allocation functions, as
2851 well as to force use of memory allocation functions derived
2852 from the system malloc.
2854 =item USE_MULTI = define
2856 Enabling the multiplicity option allows perl to clean up
2857 thoroughly when the interpreter shuts down, which reduces the
2858 number of bogus leak reports from Purify.
2860 =item #PERL_MALLOC = define
2862 Disable Perl's malloc so that Purify can more closely monitor
2863 allocations and leaks. Using Perl's malloc will make Purify
2864 report most leaks in the "potential" leaks category.
2868 Adds debugging information so that you see the exact source
2869 statements where the problem occurs. Without this flag, all
2870 you will see is the source filename of where the error occurred.
2874 As an example, to show any memory leaks produced during the
2875 standard Perl testset you would create and run Purify as:
2880 purify ../perl -I../lib harness
2882 which would instrument Perl in memory, run Perl on test.pl,
2883 then finally report any memory problems.
2887 The excellent valgrind tool can be used to find out both memory leaks
2888 and illegal memory accesses. As of version 3.3.0, Valgrind only
2889 supports Linux on x86, x86-64 and PowerPC. The special "test.valgrind"
2890 target can be used to run the tests under valgrind. Found errors
2891 and memory leaks are logged in files named F<testfile.valgrind>.
2893 Valgrind also provides a cachegrind tool, invoked on perl as:
2895 VG_OPTS=--tool=cachegrind make test.valgrind
2897 As system libraries (most notably glibc) are also triggering errors,
2898 valgrind allows to suppress such errors using suppression files. The
2899 default suppression file that comes with valgrind already catches a lot
2900 of them. Some additional suppressions are defined in F<t/perl.supp>.
2902 To get valgrind and for more information see
2904 http://developer.kde.org/~sewardj/
2906 =head2 Compaq's/Digital's/HP's Third Degree
2908 Third Degree is a tool for memory leak detection and memory access checks.
2909 It is one of the many tools in the ATOM toolkit. The toolkit is only
2910 available on Tru64 (formerly known as Digital UNIX formerly known as
2913 When building Perl, you must first run Configure with -Doptimize=-g
2914 and -Uusemymalloc flags, after that you can use the make targets
2915 "perl.third" and "test.third". (What is required is that Perl must be
2916 compiled using the C<-g> flag, you may need to re-Configure.)
2918 The short story is that with "atom" you can instrument the Perl
2919 executable to create a new executable called F<perl.third>. When the
2920 instrumented executable is run, it creates a log of dubious memory
2921 traffic in file called F<perl.3log>. See the manual pages of atom and
2922 third for more information. The most extensive Third Degree
2923 documentation is available in the Compaq "Tru64 UNIX Programmer's
2924 Guide", chapter "Debugging Programs with Third Degree".
2926 The "test.third" leaves a lot of files named F<foo_bar.3log> in the t/
2927 subdirectory. There is a problem with these files: Third Degree is so
2928 effective that it finds problems also in the system libraries.
2929 Therefore you should used the Porting/thirdclean script to cleanup
2930 the F<*.3log> files.
2932 There are also leaks that for given certain definition of a leak,
2933 aren't. See L</PERL_DESTRUCT_LEVEL> for more information.
2937 Depending on your platform there are various ways of profiling Perl.
2939 There are two commonly used techniques of profiling executables:
2940 I<statistical time-sampling> and I<basic-block counting>.
2942 The first method takes periodically samples of the CPU program
2943 counter, and since the program counter can be correlated with the code
2944 generated for functions, we get a statistical view of in which
2945 functions the program is spending its time. The caveats are that very
2946 small/fast functions have lower probability of showing up in the
2947 profile, and that periodically interrupting the program (this is
2948 usually done rather frequently, in the scale of milliseconds) imposes
2949 an additional overhead that may skew the results. The first problem
2950 can be alleviated by running the code for longer (in general this is a
2951 good idea for profiling), the second problem is usually kept in guard
2952 by the profiling tools themselves.
2954 The second method divides up the generated code into I<basic blocks>.
2955 Basic blocks are sections of code that are entered only in the
2956 beginning and exited only at the end. For example, a conditional jump
2957 starts a basic block. Basic block profiling usually works by
2958 I<instrumenting> the code by adding I<enter basic block #nnnn>
2959 book-keeping code to the generated code. During the execution of the
2960 code the basic block counters are then updated appropriately. The
2961 caveat is that the added extra code can skew the results: again, the
2962 profiling tools usually try to factor their own effects out of the
2965 =head2 Gprof Profiling
2967 gprof is a profiling tool available in many Unix platforms,
2968 it uses F<statistical time-sampling>.
2970 You can build a profiled version of perl called "perl.gprof" by
2971 invoking the make target "perl.gprof" (What is required is that Perl
2972 must be compiled using the C<-pg> flag, you may need to re-Configure).
2973 Running the profiled version of Perl will create an output file called
2974 F<gmon.out> is created which contains the profiling data collected
2975 during the execution.
2977 The gprof tool can then display the collected data in various ways.
2978 Usually gprof understands the following options:
2984 Suppress statically defined functions from the profile.
2988 Suppress the verbose descriptions in the profile.
2992 Exclude the given routine and its descendants from the profile.
2996 Display only the given routine and its descendants in the profile.
3000 Generate a summary file called F<gmon.sum> which then may be given
3001 to subsequent gprof runs to accumulate data over several runs.
3005 Display routines that have zero usage.
3009 For more detailed explanation of the available commands and output
3010 formats, see your own local documentation of gprof.
3014 $ sh Configure -des -Dusedevel -Doptimize='-pg' && make perl.gprof
3015 $ ./perl.gprof someprog # creates gmon.out in current directory
3016 $ gprof ./perl.gprof > out
3019 =head2 GCC gcov Profiling
3021 Starting from GCC 3.0 I<basic block profiling> is officially available
3024 You can build a profiled version of perl called F<perl.gcov> by
3025 invoking the make target "perl.gcov" (what is required that Perl must
3026 be compiled using gcc with the flags C<-fprofile-arcs
3027 -ftest-coverage>, you may need to re-Configure).
3029 Running the profiled version of Perl will cause profile output to be
3030 generated. For each source file an accompanying ".da" file will be
3033 To display the results you use the "gcov" utility (which should
3034 be installed if you have gcc 3.0 or newer installed). F<gcov> is
3035 run on source code files, like this
3039 which will cause F<sv.c.gcov> to be created. The F<.gcov> files
3040 contain the source code annotated with relative frequencies of
3041 execution indicated by "#" markers.
3043 Useful options of F<gcov> include C<-b> which will summarise the
3044 basic block, branch, and function call coverage, and C<-c> which
3045 instead of relative frequencies will use the actual counts. For
3046 more information on the use of F<gcov> and basic block profiling
3047 with gcc, see the latest GNU CC manual, as of GCC 3.0 see
3049 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html
3051 and its section titled "8. gcov: a Test Coverage Program"
3053 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132
3057 $ sh Configure -des -Dusedevel -Doptimize='-g' \
3058 -Accflags='-fprofile-arcs -ftest-coverage' \
3059 -Aldflags='-fprofile-arcs -ftest-coverage' && make perl.gcov
3060 $ rm -f regexec.c.gcov regexec.gcda
3063 $ view regexec.c.gcov
3065 =head2 Pixie Profiling
3067 Pixie is a profiling tool available on IRIX and Tru64 (aka Digital
3068 UNIX aka DEC OSF/1) platforms. Pixie does its profiling using
3069 I<basic-block counting>.
3071 You can build a profiled version of perl called F<perl.pixie> by
3072 invoking the make target "perl.pixie" (what is required is that Perl
3073 must be compiled using the C<-g> flag, you may need to re-Configure).
3075 In Tru64 a file called F<perl.Addrs> will also be silently created,
3076 this file contains the addresses of the basic blocks. Running the
3077 profiled version of Perl will create a new file called "perl.Counts"
3078 which contains the counts for the basic block for that particular
3081 To display the results you use the F<prof> utility. The exact
3082 incantation depends on your operating system, "prof perl.Counts" in
3083 IRIX, and "prof -pixie -all -L. perl" in Tru64.
3085 In IRIX the following prof options are available:
3091 Reports the most heavily used lines in descending order of use.
3092 Useful for finding the hotspot lines.
3096 Groups lines by procedure, with procedures sorted in descending order of use.
3097 Within a procedure, lines are listed in source order.
3098 Useful for finding the hotspots of procedures.
3102 In Tru64 the following options are available:
3108 Procedures sorted in descending order by the number of cycles executed
3109 in each procedure. Useful for finding the hotspot procedures.
3110 (This is the default option.)
3114 Lines sorted in descending order by the number of cycles executed in
3115 each line. Useful for finding the hotspot lines.
3117 =item -i[nvocations]
3119 The called procedures are sorted in descending order by number of calls
3120 made to the procedures. Useful for finding the most used procedures.
3124 Grouped by procedure, sorted by cycles executed per procedure.
3125 Useful for finding the hotspots of procedures.
3129 The compiler emitted code for these lines, but the code was unexecuted.
3133 Unexecuted procedures.
3137 For further information, see your system's manual pages for pixie and prof.
3139 =head1 MISCELLANEOUS TRICKS
3141 =head2 PERL_DESTRUCT_LEVEL
3143 If you want to run any of the tests yourself manually using e.g.
3144 valgrind, or the pureperl or perl.third executables, please note that
3145 by default perl B<does not> explicitly cleanup all the memory it has
3146 allocated (such as global memory arenas) but instead lets the exit()
3147 of the whole program "take care" of such allocations, also known as
3148 "global destruction of objects".
3150 There is a way to tell perl to do complete cleanup: set the
3151 environment variable PERL_DESTRUCT_LEVEL to a non-zero value.
3152 The t/TEST wrapper does set this to 2, and this is what you
3153 need to do too, if you don't want to see the "global leaks":
3154 For example, for "third-degreed" Perl:
3156 env PERL_DESTRUCT_LEVEL=2 ./perl.third -Ilib t/foo/bar.t
3158 (Note: the mod_perl apache module uses also this environment variable
3159 for its own purposes and extended its semantics. Refer to the mod_perl
3160 documentation for more information. Also, spawned threads do the
3161 equivalent of setting this variable to the value 1.)
3163 If, at the end of a run you get the message I<N scalars leaked>, you can
3164 recompile with C<-DDEBUG_LEAKING_SCALARS>, which will cause the addresses
3165 of all those leaked SVs to be dumped along with details as to where each
3166 SV was originally allocated. This information is also displayed by
3167 Devel::Peek. Note that the extra details recorded with each SV increases
3168 memory usage, so it shouldn't be used in production environments. It also
3169 converts C<new_SV()> from a macro into a real function, so you can use
3170 your favourite debugger to discover where those pesky SVs were allocated.
3172 If you see that you're leaking memory at runtime, but neither valgrind
3173 nor C<-DDEBUG_LEAKING_SCALARS> will find anything, you're probably
3174 leaking SVs that are still reachable and will be properly cleaned up
3175 during destruction of the interpreter. In such cases, using the C<-Dm>
3176 switch can point you to the source of the leak. If the executable was
3177 built with C<-DDEBUG_LEAKING_SCALARS>, C<-Dm> will output SV allocations
3178 in addition to memory allocations. Each SV allocation has a distinct
3179 serial number that will be written on creation and destruction of the SV.
3180 So if you're executing the leaking code in a loop, you need to look for
3181 SVs that are created, but never destroyed between each cycle. If such an
3182 SV is found, set a conditional breakpoint within C<new_SV()> and make it
3183 break only when C<PL_sv_serial> is equal to the serial number of the
3184 leaking SV. Then you will catch the interpreter in exactly the state
3185 where the leaking SV is allocated, which is sufficient in many cases to
3186 find the source of the leak.
3188 As C<-Dm> is using the PerlIO layer for output, it will by itself
3189 allocate quite a bunch of SVs, which are hidden to avoid recursion.
3190 You can bypass the PerlIO layer if you use the SV logging provided
3191 by C<-DPERL_MEM_LOG> instead.
3195 If compiled with C<-DPERL_MEM_LOG>, both memory and SV allocations go
3196 through logging functions, which is handy for breakpoint setting.
3198 Unless C<-DPERL_MEM_LOG_NOIMPL> is also compiled, the logging
3199 functions read $ENV{PERL_MEM_LOG} to determine whether to log the
3200 event, and if so how:
3202 $ENV{PERL_MEM_LOG} =~ /m/ Log all memory ops
3203 $ENV{PERL_MEM_LOG} =~ /s/ Log all SV ops
3204 $ENV{PERL_MEM_LOG} =~ /t/ include timestamp in Log
3205 $ENV{PERL_MEM_LOG} =~ /^(\d+)/ write to FD given (default is 2)
3207 Memory logging is somewhat similar to C<-Dm> but is independent of
3208 C<-DDEBUGGING>, and at a higher level; all uses of Newx(), Renew(),
3209 and Safefree() are logged with the caller's source code file and line
3210 number (and C function name, if supported by the C compiler). In
3211 contrast, C<-Dm> is directly at the point of C<malloc()>. SV logging
3214 Since the logging doesn't use PerlIO, all SV allocations are logged
3215 and no extra SV allocations are introduced by enabling the logging.
3216 If compiled with C<-DDEBUG_LEAKING_SCALARS>, the serial number for
3217 each SV allocation is also logged.
3221 Those debugging perl with the DDD frontend over gdb may find the
3224 You can extend the data conversion shortcuts menu, so for example you
3225 can display an SV's IV value with one click, without doing any typing.
3226 To do that simply edit ~/.ddd/init file and add after:
3228 ! Display shortcuts.
3229 Ddd*gdbDisplayShortcuts: \
3230 /t () // Convert to Bin\n\
3231 /d () // Convert to Dec\n\
3232 /x () // Convert to Hex\n\
3233 /o () // Convert to Oct(\n\
3235 the following two lines:
3237 ((XPV*) (())->sv_any )->xpv_pv // 2pvx\n\
3238 ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx
3240 so now you can do ivx and pvx lookups or you can plug there the
3241 sv_peek "conversion":
3243 Perl_sv_peek(my_perl, (SV*)()) // sv_peek
3245 (The my_perl is for threaded builds.)
3246 Just remember that every line, but the last one, should end with \n\
3248 Alternatively edit the init file interactively via:
3249 3rd mouse button -> New Display -> Edit Menu
3251 Note: you can define up to 20 conversion shortcuts in the gdb
3256 If you see in a debugger a memory area mysteriously full of 0xABABABAB
3257 or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros,
3260 =head2 Read-only optrees
3262 Under ithreads the optree is read only. If you want to enforce this, to check
3263 for write accesses from buggy code, compile with C<-DPL_OP_SLAB_ALLOC> to
3264 enable the OP slab allocator and C<-DPERL_DEBUG_READONLY_OPS> to enable code
3265 that allocates op memory via C<mmap>, and sets it read-only at run time.
3266 Any write access to an op results in a C<SIGBUS> and abort.
3268 This code is intended for development only, and may not be portable even to
3269 all Unix variants. Also, it is an 80% solution, in that it isn't able to make
3270 all ops read only. Specifically it
3276 Only sets read-only on all slabs of ops at C<CHECK> time, hence ops allocated
3277 later via C<require> or C<eval> will be re-write
3281 Turns an entire slab of ops read-write if the refcount of any op in the slab
3282 needs to be decreased.
3286 Turns an entire slab of ops read-write if any op from the slab is freed.
3290 It's not possible to turn the slabs to read-only after an action requiring
3291 read-write access, as either can happen during op tree building time, so
3292 there may still be legitimate write access.
3294 However, as an 80% solution it is still effective, as currently it catches
3295 a write access during the generation of F<Config.pm>, which means that we
3296 can't yet build F<perl> with this enabled.
3300 We've had a brief look around the Perl source, how to maintain quality
3301 of the source code, an overview of the stages F<perl> goes through
3302 when it's running your code, how to use debuggers to poke at the Perl
3303 guts, and finally how to analyse the execution of Perl. We took a very
3304 simple problem and demonstrated how to solve it fully - with
3305 documentation, regression tests, and finally a patch for submission to
3306 p5p. Finally, we talked about how to use external tools to debug and
3309 I'd now suggest you read over those references again, and then, as soon
3310 as possible, get your hands dirty. The best way to learn is by doing,
3317 Subscribe to perl5-porters, follow the patches and try and understand
3318 them; don't be afraid to ask if there's a portion you're not clear on -
3319 who knows, you may unearth a bug in the patch...
3323 Keep up to date with the bleeding edge Perl distributions and get
3324 familiar with the changes. Try and get an idea of what areas people are
3325 working on and the changes they're making.
3329 Do read the README associated with your operating system, e.g. README.aix
3330 on the IBM AIX OS. Don't hesitate to supply patches to that README if
3331 you find anything missing or changed over a new OS release.
3335 Find an area of Perl that seems interesting to you, and see if you can
3336 work out how it works. Scan through the source, and step over it in the
3337 debugger. Play, poke, investigate, fiddle! You'll probably get to
3338 understand not just your chosen area but a much wider range of F<perl>'s
3339 activity as well, and probably sooner than you'd think.
3345 =item I<The Road goes ever on and on, down from the door where it began.>
3349 If you can do these things, you've started on the long road to Perl porting.
3350 Thanks for wanting to help make Perl better - and happy hacking!
3352 =head2 Metaphoric Quotations
3354 If you recognized the quote about the Road above, you're in luck.
3356 Most software projects begin each file with a literal description of each
3357 file's purpose. Perl instead begins each with a literary allusion to that
3360 Like chapters in many books, all top-level Perl source files (along with a
3361 few others here and there) begin with an epigramic inscription that alludes,
3362 indirectly and metaphorically, to the material you're about to read.
3364 Quotations are taken from writings of J.R.R Tolkien pertaining to his
3365 Legendarium, almost always from I<The Lord of the Rings>. Chapters and
3366 page numbers are given using the following editions:
3372 I<The Hobbit>, by J.R.R. Tolkien. The hardcover, 70th-anniversary
3373 edition of 2007 was used, published in the UK by Harper Collins Publishers
3374 and in the US by the Houghton Mifflin Company.
3378 I<The Lord of the Rings>, by J.R.R. Tolkien. The hardcover,
3379 50th-anniversary edition of 2004 was used, published in the UK by Harper
3380 Collins Publishers and in the US by the Houghton Mifflin Company.
3384 I<The Lays of Beleriand>, by J.R.R. Tolkien and published posthumously by his
3385 son and literary executor, C.J.R. Tolkien, being the 3rd of the 12 volumes
3386 in Christopher's mammoth I<History of Middle Earth>. Page numbers derive
3387 from the hardcover edition, first published in 1983 by George Allen &
3388 Unwin; no page numbers changed for the special 3-volume omnibus edition of
3389 2002 or the various trade-paper editions, all again now by Harper Collins
3390 or Houghton Mifflin.
3394 Other JRRT books fair game for quotes would thus include I<The Adventures of
3395 Tom Bombadil>, I<The Silmarillion>, I<Unfinished Tales>, and I<The Tale of
3396 the Children of Hurin>, all but the first posthumously assembled by CJRT.
3397 But I<The Lord of the Rings> itself is perfectly fine and probably best to
3398 quote from, provided you can find a suitable quote there.
3400 So if you were to supply a new, complete, top-level source file to add to
3401 Perl, you should conform to this peculiar practice by yourself selecting an
3402 appropriate quotation from Tolkien, retaining the original spelling and
3403 punctuation and using the same format the rest of the quotes are in.
3404 Indirect and oblique is just fine; remember, it's a metaphor, so being meta
3405 is, after all, what it's for.
3409 This document was written by Nathan Torkington, and is maintained by
3410 the perl5-porters mailing list.