This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perlhack: Note existence of t/lib/warnings
[perl5.git] / pod / perlhack.pod
CommitLineData
e8cd7eae
GS
1=head1 NAME
2
3perlhack - How to hack at the Perl internals
4
5=head1 DESCRIPTION
6
7This document attempts to explain how Perl development takes place,
8and ends with some suggestions for people wanting to become bona fide
9porters.
10
11The perl5-porters mailing list is where the Perl standard distribution
12is maintained and developed. The list can get anywhere from 10 to 150
13messages a day, depending on the heatedness of the debate. Most days
14there are two or three patches, extensions, features, or bugs being
15discussed at a time.
16
f8e3975a 17A searchable archive of the list is at either:
e8cd7eae
GS
18
19 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
20
f8e3975a
IP
21or
22
23 http://archive.develooper.com/perl5-porters@perl.org/
24
e8cd7eae
GS
25List subscribers (the porters themselves) come in several flavours.
26Some are quiet curious lurkers, who rarely pitch in and instead watch
27the ongoing development to ensure they're forewarned of new changes or
28features in Perl. Some are representatives of vendors, who are there
29to make sure that Perl continues to compile and work on their
30platforms. Some patch any reported bug that they know how to fix,
31some are actively patching their pet area (threads, Win32, the regexp
32engine), while others seem to do nothing but complain. In other
33words, it's your usual mix of technical people.
34
35Over this group of porters presides Larry Wall. He has the final word
f6c51b38 36in what does and does not change in the Perl language. Various
b432a672
AL
37releases of Perl are shepherded by a "pumpking", a porter
38responsible for gathering patches, deciding on a patch-by-patch,
f6c51b38 39feature-by-feature basis what will and will not go into the release.
caf100c0 40For instance, Gurusamy Sarathy was the pumpking for the 5.6 release of
961f29c6 41Perl, and Jarkko Hietaniemi was the pumpking for the 5.8 release, and
1a88dbf8 42Rafael Garcia-Suarez holds the pumpking crown for the 5.10 release.
e8cd7eae
GS
43
44In addition, various people are pumpkings for different things. For
961f29c6
MB
45instance, Andy Dougherty and Jarkko Hietaniemi did a grand job as the
46I<Configure> pumpkin up till the 5.8 release. For the 5.10 release
47H.Merijn Brand took over.
e8cd7eae
GS
48
49Larry sees Perl development along the lines of the US government:
50there's the Legislature (the porters), the Executive branch (the
51pumpkings), and the Supreme Court (Larry). The legislature can
52discuss and submit patches to the executive branch all they like, but
53the executive branch is free to veto them. Rarely, the Supreme Court
54will side with the executive branch over the legislature, or the
55legislature over the executive branch. Mostly, however, the
56legislature and the executive branch are supposed to get along and
57work out their differences without impeachment or court cases.
58
59You might sometimes see reference to Rule 1 and Rule 2. Larry's power
60as Supreme Court is expressed in The Rules:
61
62=over 4
63
64=item 1
65
66Larry is always by definition right about how Perl should behave.
67This means he has final veto power on the core functionality.
68
69=item 2
70
71Larry is allowed to change his mind about any matter at a later date,
72regardless of whether he previously invoked Rule 1.
73
74=back
75
76Got that? Larry is always right, even when he was wrong. It's rare
77to see either Rule exercised, but they are often alluded to.
78
79New features and extensions to the language are contentious, because
80the criteria used by the pumpkings, Larry, and other porters to decide
81which features should be implemented and incorporated are not codified
82in a few small design goals as with some other languages. Instead,
83the heuristics are flexible and often difficult to fathom. Here is
84one person's list, roughly in decreasing order of importance, of
85heuristics that new features have to be weighed against:
86
87=over 4
88
89=item Does concept match the general goals of Perl?
90
91These haven't been written anywhere in stone, but one approximation
92is:
93
94 1. Keep it fast, simple, and useful.
95 2. Keep features/concepts as orthogonal as possible.
96 3. No arbitrary limits (platforms, data sizes, cultures).
97 4. Keep it open and exciting to use/patch/advocate Perl everywhere.
98 5. Either assimilate new technologies, or build bridges to them.
99
100=item Where is the implementation?
101
102All the talk in the world is useless without an implementation. In
103almost every case, the person or people who argue for a new feature
104will be expected to be the ones who implement it. Porters capable
105of coding new features have their own agendas, and are not available
106to implement your (possibly good) idea.
107
108=item Backwards compatibility
109
110It's a cardinal sin to break existing Perl programs. New warnings are
111contentious--some say that a program that emits warnings is not
112broken, while others say it is. Adding keywords has the potential to
113break programs, changing the meaning of existing token sequences or
114functions might break programs.
115
116=item Could it be a module instead?
117
118Perl 5 has extension mechanisms, modules and XS, specifically to avoid
119the need to keep changing the Perl interpreter. You can write modules
120that export functions, you can give those functions prototypes so they
121can be called like built-in functions, you can even write XS code to
122mess with the runtime data structures of the Perl interpreter if you
123want to implement really complicated things. If it can be done in a
124module instead of in the core, it's highly unlikely to be added.
125
126=item Is the feature generic enough?
127
128Is this something that only the submitter wants added to the language,
129or would it be broadly useful? Sometimes, instead of adding a feature
130with a tight focus, the porters might decide to wait until someone
131implements the more generalized feature. For instance, instead of
b432a672 132implementing a "delayed evaluation" feature, the porters are waiting
e8cd7eae
GS
133for a macro system that would permit delayed evaluation and much more.
134
135=item Does it potentially introduce new bugs?
136
137Radical rewrites of large chunks of the Perl interpreter have the
138potential to introduce new bugs. The smaller and more localized the
139change, the better.
140
141=item Does it preclude other desirable features?
142
143A patch is likely to be rejected if it closes off future avenues of
144development. For instance, a patch that placed a true and final
145interpretation on prototypes is likely to be rejected because there
146are still options for the future of prototypes that haven't been
147addressed.
148
149=item Is the implementation robust?
150
151Good patches (tight code, complete, correct) stand more chance of
152going in. Sloppy or incorrect patches might be placed on the back
153burner until the pumpking has time to fix, or might be discarded
154altogether without further notice.
155
156=item Is the implementation generic enough to be portable?
157
158The worst patches make use of a system-specific features. It's highly
353c6505 159unlikely that non-portable additions to the Perl language will be
e8cd7eae
GS
160accepted.
161
a936dd3c
NC
162=item Is the implementation tested?
163
164Patches which change behaviour (fixing bugs or introducing new features)
165must include regression tests to verify that everything works as expected.
166Without tests provided by the original author, how can anyone else changing
167perl in the future be sure that they haven't unwittingly broken the behaviour
168the patch implements? And without tests, how can the patch's author be
9d077eaa 169confident that his/her hard work put into the patch won't be accidentally
a936dd3c
NC
170thrown away by someone in the future?
171
e8cd7eae
GS
172=item Is there enough documentation?
173
174Patches without documentation are probably ill-thought out or
175incomplete. Nothing can be added without documentation, so submitting
176a patch for the appropriate manpages as well as the source code is
a936dd3c 177always a good idea.
e8cd7eae
GS
178
179=item Is there another way to do it?
180
b432a672
AL
181Larry said "Although the Perl Slogan is I<There's More Than One Way
182to Do It>, I hesitate to make 10 ways to do something". This is a
e8cd7eae
GS
183tricky heuristic to navigate, though--one man's essential addition is
184another man's pointless cruft.
185
186=item Does it create too much work?
187
188Work for the pumpking, work for Perl programmers, work for module
189authors, ... Perl is supposed to be easy.
190
f6c51b38
GS
191=item Patches speak louder than words
192
193Working code is always preferred to pie-in-the-sky ideas. A patch to
194add a feature stands a much higher chance of making it to the language
195than does a random feature request, no matter how fervently argued the
b432a672 196request might be. This ties into "Will it be useful?", as the fact
f6c51b38
GS
197that someone took the time to make the patch demonstrates a strong
198desire for the feature.
199
e8cd7eae
GS
200=back
201
b432a672
AL
202If you're on the list, you might hear the word "core" bandied
203around. It refers to the standard distribution. "Hacking on the
204core" means you're changing the C source code to the Perl
205interpreter. "A core module" is one that ships with Perl.
e8cd7eae 206
a1f349fd
MB
207=head2 Keeping in sync
208
e8cd7eae 209The source code to the Perl interpreter, in its different versions, is
b16c2e4a
RGS
210kept in a repository managed by the git revision control system. The
211pumpkings and a few others have write access to the repository to check in
212changes.
2be4c08b 213
b16c2e4a 214How to clone and use the git perl repository is described in L<perlrepository>.
2be4c08b 215
b16c2e4a 216You can also choose to use rsync to get a copy of the current source tree
a77cd7b8 217for the bleadperl branch and all maintenance branches:
0cfb3454 218
a77cd7b8
MB
219 $ rsync -avz rsync://perl5.git.perl.org/perl-current .
220 $ rsync -avz rsync://perl5.git.perl.org/perl-5.12.x .
221 $ rsync -avz rsync://perl5.git.perl.org/perl-5.10.x .
222 $ rsync -avz rsync://perl5.git.perl.org/perl-5.8.x .
223 $ rsync -avz rsync://perl5.git.perl.org/perl-5.6.x .
224 $ rsync -avz rsync://perl5.git.perl.org/perl-5.005xx .
b16c2e4a
RGS
225
226(Add the C<--delete> option to remove leftover files)
0cfb3454 227
a77cd7b8
MB
228To get a full list of the available sync points:
229
230 $ rsync perl5.git.perl.org::
231
0cfb3454
GS
232You may also want to subscribe to the perl5-changes mailing list to
233receive a copy of each patch that gets submitted to the maintenance
234and development "branches" of the perl repository. See
235http://lists.perl.org/ for subscription information.
236
a1f349fd
MB
237If you are a member of the perl5-porters mailing list, it is a good
238thing to keep in touch with the most recent changes. If not only to
239verify if what you would have posted as a bug report isn't already
240solved in the most recent available perl development branch, also
241known as perl-current, bleading edge perl, bleedperl or bleadperl.
2be4c08b
GS
242
243Needless to say, the source code in perl-current is usually in a perpetual
244state of evolution. You should expect it to be very buggy. Do B<not> use
245it for any purpose other than testing and development.
e8cd7eae 246
3fd28c4e 247=head2 Perlbug administration
52315700 248
902821cc
RGS
249There is a single remote administrative interface for modifying bug status,
250category, open issues etc. using the B<RT> bugtracker system, maintained
251by Robert Spier. Become an administrator, and close any bugs you can get
3fd28c4e 252your sticky mitts on:
52315700 253
39417508 254 http://bugs.perl.org/
52315700 255
3fd28c4e 256To email the bug system administrators:
52315700 257
3fd28c4e 258 "perlbug-admin" <perlbug-admin@perl.org>
52315700 259
a1f349fd
MB
260=head2 Submitting patches
261
f7e1e956
MS
262Always submit patches to I<perl5-porters@perl.org>. If you're
263patching a core module and there's an author listed, send the author a
264copy (see L<Patching a core module>). This lets other porters review
265your patch, which catches a surprising number of errors in patches.
b16c2e4a
RGS
266Please patch against the latest B<development> version. (e.g., even if
267you're fixing a bug in the 5.8 track, patch against the C<blead> branch in
268the git repository.)
824d470b
RGS
269
270If changes are accepted, they are applied to the development branch. Then
fe749c9a 271the maintenance pumpking decides which of those patches is to be
b16c2e4a
RGS
272backported to the maint branch. Only patches that survive the heat of the
273development branch get applied to maintenance versions.
f7e1e956 274
2c8694a7
JH
275Your patch should update the documentation and test suite. See
276L<Writing a test>. If you have added or removed files in the distribution,
277edit the MANIFEST file accordingly, sort the MANIFEST file using
278C<make manisort>, and include those changes as part of your patch.
e8cd7eae 279
824d470b
RGS
280Patching documentation also follows the same order: if accepted, a patch
281is first applied to B<development>, and if relevant then it's backported
282to B<maintenance>. (With an exception for some patches that document
283behaviour that only appears in the maintenance branch, but which has
284changed in the development version.)
285
e8cd7eae
GS
286To report a bug in Perl, use the program I<perlbug> which comes with
287Perl (if you can't get Perl to work, send mail to the address
f18956b7 288I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through
e8cd7eae 289I<perlbug> feeds into the automated bug-tracking system, access to
902821cc 290which is provided through the web at http://rt.perl.org/rt3/ . It
e8cd7eae
GS
291often pays to check the archives of the perl5-porters mailing list to
292see whether the bug you're reporting has been reported before, and if
293so whether it was considered a bug. See above for the location of
294the searchable archives.
295
f224927c 296The CPAN testers ( http://testers.cpan.org/ ) are a group of
ba139f7d 297volunteers who test CPAN modules on a variety of platforms. Perl
d3e8af89
GS
298Smokers ( http://www.nntp.perl.org/group/perl.daily-build and
299http://www.nntp.perl.org/group/perl.daily-build.reports/ )
902821cc 300automatically test Perl source releases on platforms with various
d3e8af89
GS
301configurations. Both efforts welcome volunteers. In order to get
302involved in smoke testing of the perl itself visit
303L<http://search.cpan.org/dist/Test-Smoke>. In order to start smoke
68cbce50
CBW
304testing CPAN modules visit L<http://search.cpan.org/dist/CPANPLUS-YACSmoke/>
305or L<http://search.cpan.org/dist/minismokebox/> or
d3e8af89 306L<http://search.cpan.org/dist/CPAN-Reporter/>.
e8cd7eae 307
e8cd7eae
GS
308It's a good idea to read and lurk for a while before chipping in.
309That way you'll get to see the dynamic of the conversations, learn the
310personalities of the players, and hopefully be better prepared to make
311a useful contribution when do you speak up.
312
313If after all this you still think you want to join the perl5-porters
f6c51b38
GS
314mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To
315unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>.
e8cd7eae 316
a422fd2d
SC
317To hack on the Perl guts, you'll need to read the following things:
318
319=over 3
320
321=item L<perlguts>
322
323This is of paramount importance, since it's the documentation of what
324goes where in the Perl source. Read it over a couple of times and it
325might start to make sense - don't worry if it doesn't yet, because the
326best way to study it is to read it in conjunction with poking at Perl
327source, and we'll do that later on.
328
0aa6d4a5
RU
329Gisle Aas's "illustrated perlguts", also known as I<illguts>, has very
330helpful pictures:
de10be12 331
0aa6d4a5 332L<http://search.cpan.org/dist/illguts/>
a422fd2d
SC
333
334=item L<perlxstut> and L<perlxs>
335
336A working knowledge of XSUB programming is incredibly useful for core
337hacking; XSUBs use techniques drawn from the PP code, the portion of the
338guts that actually executes a Perl program. It's a lot gentler to learn
339those techniques from simple examples and explanation than from the core
340itself.
341
342=item L<perlapi>
343
344The documentation for the Perl API explains what some of the internal
345functions do, as well as the many macros used in the source.
346
347=item F<Porting/pumpkin.pod>
348
349This is a collection of words of wisdom for a Perl porter; some of it is
350only useful to the pumpkin holder, but most of it applies to anyone
351wanting to go about Perl development.
352
353=item The perl5-porters FAQ
354
902821cc
RGS
355This should be available from http://dev.perl.org/perl5/docs/p5p-faq.html .
356It contains hints on reading perl5-porters, information on how
357perl5-porters works and how Perl development in general works.
a422fd2d
SC
358
359=back
360
361=head2 Finding Your Way Around
362
363Perl maintenance can be split into a number of areas, and certain people
364(pumpkins) will have responsibility for each area. These areas sometimes
365correspond to files or directories in the source kit. Among the areas are:
366
367=over 3
368
369=item Core modules
370
c53fdc5e
RF
371Modules shipped as part of the Perl core live in various subdirectories, where
372two are dedicated to core-only modules, and two are for the dual-life modules
373which live on CPAN and may be maintained separately with respect to the Perl
374core:
375
376 lib/ is for pure-Perl modules, which exist in the core only.
377
195c30ce
KW
378 ext/ is for XS extensions, and modules with special Makefile.PL
379 requirements, which exist in the core only.
c53fdc5e 380
195c30ce
KW
381 cpan/ is for dual-life modules, where the CPAN module is
382 canonical (should be patched first).
c53fdc5e 383
195c30ce
KW
384 dist/ is for dual-life modules, where the blead source is
385 canonical.
a422fd2d 386
6bdc4f2c
FR
387For some dual-life modules it has not been discussed if the CPAN version or the
388blead source is canonical. Until that is done, those modules should be in
389F<cpan/>.
390
f7e1e956
MS
391=item Tests
392
393There are tests for nearly all the modules, built-ins and major bits
394of functionality. Test files all have a .t suffix. Module tests live
395in the F<lib/> and F<ext/> directories next to the module being
396tested. Others live in F<t/>. See L<Writing a test>
397
a422fd2d
SC
398=item Documentation
399
400Documentation maintenance includes looking after everything in the
401F<pod/> directory, (as well as contributing new documentation) and
402the documentation to the modules in core.
403
404=item Configure
405
99c47ece 406The Configure process is the way we make Perl portable across the
a422fd2d 407myriad of operating systems it supports. Responsibility for the
99c47ece
MB
408Configure, build and installation process, as well as the overall
409portability of the core code rests with the Configure pumpkin -
410others help out with individual operating systems.
411
e1020413 412The three files that fall under his/her responsibility are Configure,
99c47ece
MB
413config_h.SH, and Porting/Glossary (and a whole bunch of small related
414files that are less important here). The Configure pumpkin decides how
415patches to these are dealt with. Currently, the Configure pumpkin will
416accept patches in most common formats, even directly to these files.
417Other committers are allowed to commit to these files under the strict
418condition that they will inform the Configure pumpkin, either on IRC
419(if he/she happens to be around) or through (personal) e-mail.
a422fd2d
SC
420
421The files involved are the operating system directories, (F<win32/>,
422F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h>
423and F<Makefile>, as well as the metaconfig files which generate
424F<Configure>. (metaconfig isn't included in the core distribution.)
425
99c47ece
MB
426See http://perl5.git.perl.org/metaconfig.git/blob/HEAD:/README for a
427description of the full process involved.
428
a422fd2d
SC
429=item Interpreter
430
431And of course, there's the core of the Perl interpreter itself. Let's
432have a look at that in a little more detail.
433
434=back
435
436Before we leave looking at the layout, though, don't forget that
437F<MANIFEST> contains not only the file names in the Perl distribution,
438but short descriptions of what's in them, too. For an overview of the
439important files, try this:
440
441 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST
442
443=head2 Elements of the interpreter
444
445The work of the interpreter has two main stages: compiling the code
446into the internal representation, or bytecode, and then executing it.
447L<perlguts/Compiled code> explains exactly how the compilation stage
448happens.
449
450Here is a short breakdown of perl's operation:
451
452=over 3
453
454=item Startup
455
456The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl)
457This is very high-level code, enough to fit on a single screen, and it
458resembles the code found in L<perlembed>; most of the real action takes
459place in F<perl.c>
460
9df8f87f
LB
461F<perlmain.c> is generated by L<writemain> from F<miniperlmain.c> at
462make time, so you should make perl to follow this along.
463
a422fd2d 464First, F<perlmain.c> allocates some memory and constructs a Perl
9df8f87f 465interpreter, along these lines:
a422fd2d
SC
466
467 1 PERL_SYS_INIT3(&argc,&argv,&env);
468 2
469 3 if (!PL_do_undump) {
470 4 my_perl = perl_alloc();
471 5 if (!my_perl)
472 6 exit(1);
473 7 perl_construct(my_perl);
474 8 PL_perl_destruct_level = 0;
475 9 }
476
477Line 1 is a macro, and its definition is dependent on your operating
478system. Line 3 references C<PL_do_undump>, a global variable - all
479global variables in Perl start with C<PL_>. This tells you whether the
480current running program was created with the C<-u> flag to perl and then
481F<undump>, which means it's going to be false in any sane context.
482
483Line 4 calls a function in F<perl.c> to allocate memory for a Perl
484interpreter. It's quite a simple function, and the guts of it looks like
485this:
486
195c30ce 487 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
a422fd2d
SC
488
489Here you see an example of Perl's system abstraction, which we'll see
490later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's
491own C<malloc> as defined in F<malloc.c> if you selected that option at
492configure time.
493
9df8f87f
LB
494Next, in line 7, we construct the interpreter using perl_construct,
495also in F<perl.c>; this sets up all the special variables that Perl
496needs, the stacks, and so on.
a422fd2d
SC
497
498Now we pass Perl the command line options, and tell it to go:
499
195c30ce
KW
500 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
501 if (!exitstatus)
502 perl_run(my_perl);
9df8f87f 503
195c30ce 504 exitstatus = perl_destruct(my_perl);
a422fd2d 505
195c30ce 506 perl_free(my_perl);
a422fd2d
SC
507
508C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined
509in F<perl.c>, which processes the command line options, sets up any
510statically linked XS modules, opens the program and calls C<yyparse> to
511parse it.
512
513=item Parsing
514
515The aim of this stage is to take the Perl source, and turn it into an op
516tree. We'll see what one of those looks like later. Strictly speaking,
517there's three things going on here.
518
519C<yyparse>, the parser, lives in F<perly.c>, although you're better off
520reading the original YACC input in F<perly.y>. (Yes, Virginia, there
521B<is> a YACC grammar for Perl!) The job of the parser is to take your
b432a672 522code and "understand" it, splitting it into sentences, deciding which
a422fd2d
SC
523operands go with which operators and so on.
524
525The parser is nobly assisted by the lexer, which chunks up your input
526into tokens, and decides what type of thing each token is: a variable
527name, an operator, a bareword, a subroutine, a core function, and so on.
528The main point of entry to the lexer is C<yylex>, and that and its
529associated routines can be found in F<toke.c>. Perl isn't much like
530other computer languages; it's highly context sensitive at times, it can
531be tricky to work out what sort of token something is, or where a token
532ends. As such, there's a lot of interplay between the tokeniser and the
533parser, which can get pretty frightening if you're not used to it.
534
535As the parser understands a Perl program, it builds up a tree of
536operations for the interpreter to perform during execution. The routines
537which construct and link together the various operations are to be found
538in F<op.c>, and will be examined later.
539
540=item Optimization
541
542Now the parsing stage is complete, and the finished tree represents
543the operations that the Perl interpreter needs to perform to execute our
544program. Next, Perl does a dry run over the tree looking for
545optimisations: constant expressions such as C<3 + 4> will be computed
546now, and the optimizer will also see if any multiple operations can be
547replaced with a single one. For instance, to fetch the variable C<$foo>,
548instead of grabbing the glob C<*foo> and looking at the scalar
549component, the optimizer fiddles the op tree to use a function which
550directly looks up the scalar in question. The main optimizer is C<peep>
551in F<op.c>, and many ops have their own optimizing functions.
552
553=item Running
554
555Now we're finally ready to go: we have compiled Perl byte code, and all
556that's left to do is run it. The actual execution is done by the
557C<runops_standard> function in F<run.c>; more specifically, it's done by
558these three innocent looking lines:
559
560 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
561 PERL_ASYNC_CHECK();
562 }
563
564You may be more comfortable with the Perl version of that:
565
566 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};
567
568Well, maybe not. Anyway, each op contains a function pointer, which
569stipulates the function which will actually carry out the operation.
570This function will return the next op in the sequence - this allows for
571things like C<if> which choose the next op dynamically at run time.
572The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt
573execution if required.
574
575The actual functions called are known as PP code, and they're spread
b432a672 576between four files: F<pp_hot.c> contains the "hot" code, which is most
a422fd2d
SC
577often used and highly optimized, F<pp_sys.c> contains all the
578system-specific functions, F<pp_ctl.c> contains the functions which
579implement control structures (C<if>, C<while> and the like) and F<pp.c>
580contains everything else. These are, if you like, the C code for Perl's
581built-in functions and operators.
582
dfc98234
DM
583Note that each C<pp_> function is expected to return a pointer to the next
584op. Calls to perl subs (and eval blocks) are handled within the same
585runops loop, and do not consume extra space on the C stack. For example,
586C<pp_entersub> and C<pp_entertry> just push a C<CxSUB> or C<CxEVAL> block
587struct onto the context stack which contain the address of the op
588following the sub call or eval. They then return the first op of that sub
589or eval block, and so execution continues of that sub or block. Later, a
590C<pp_leavesub> or C<pp_leavetry> op pops the C<CxSUB> or C<CxEVAL>,
591retrieves the return op from it, and returns it.
592
593=item Exception handing
594
0503309d 595Perl's exception handing (i.e. C<die> etc.) is built on top of the low-level
dfc98234 596C<setjmp()>/C<longjmp()> C-library functions. These basically provide a
28a5cf3b 597way to capture the current PC and SP registers and later restore them; i.e.
dfc98234
DM
598a C<longjmp()> continues at the point in code where a previous C<setjmp()>
599was done, with anything further up on the C stack being lost. This is why
600code should always save values using C<SAVE_FOO> rather than in auto
601variables.
602
603The perl core wraps C<setjmp()> etc in the macros C<JMPENV_PUSH> and
604C<JMPENV_JUMP>. The basic rule of perl exceptions is that C<exit>, and
605C<die> (in the absence of C<eval>) perform a C<JMPENV_JUMP(2)>, while
606C<die> within C<eval> does a C<JMPENV_JUMP(3)>.
607
608At entry points to perl, such as C<perl_parse()>, C<perl_run()> and
609C<call_sv(cv, G_EVAL)> each does a C<JMPENV_PUSH>, then enter a runops
610loop or whatever, and handle possible exception returns. For a 2 return,
611final cleanup is performed, such as popping stacks and calling C<CHECK> or
612C<END> blocks. Amongst other things, this is how scope cleanup still
613occurs during an C<exit>.
614
615If a C<die> can find a C<CxEVAL> block on the context stack, then the
616stack is popped to that level and the return op in that block is assigned
617to C<PL_restartop>; then a C<JMPENV_JUMP(3)> is performed. This normally
618passes control back to the guard. In the case of C<perl_run> and
619C<call_sv>, a non-null C<PL_restartop> triggers re-entry to the runops
620loop. The is the normal way that C<die> or C<croak> is handled within an
621C<eval>.
622
623Sometimes ops are executed within an inner runops loop, such as tie, sort
624or overload code. In this case, something like
625
626 sub FETCH { eval { die } }
627
628would cause a longjmp right back to the guard in C<perl_run>, popping both
629runops loops, which is clearly incorrect. One way to avoid this is for the
630tie code to do a C<JMPENV_PUSH> before executing C<FETCH> in the inner
631runops loop, but for efficiency reasons, perl in fact just sets a flag,
632using C<CATCH_SET(TRUE)>. The C<pp_require>, C<pp_entereval> and
633C<pp_entertry> ops check this flag, and if true, they call C<docatch>,
634which does a C<JMPENV_PUSH> and starts a new runops level to execute the
635code, rather than doing it on the current loop.
636
637As a further optimisation, on exit from the eval block in the C<FETCH>,
638execution of the code following the block is still carried on in the inner
639loop. When an exception is raised, C<docatch> compares the C<JMPENV>
640level of the C<CxEVAL> with C<PL_top_env> and if they differ, just
641re-throws the exception. In this way any inner loops get popped.
642
643Here's an example.
644
645 1: eval { tie @a, 'A' };
646 2: sub A::TIEARRAY {
647 3: eval { die };
648 4: die;
649 5: }
650
651To run this code, C<perl_run> is called, which does a C<JMPENV_PUSH> then
652enters a runops loop. This loop executes the eval and tie ops on line 1,
653with the eval pushing a C<CxEVAL> onto the context stack.
654
655The C<pp_tie> does a C<CATCH_SET(TRUE)>, then starts a second runops loop
656to execute the body of C<TIEARRAY>. When it executes the entertry op on
657line 3, C<CATCH_GET> is true, so C<pp_entertry> calls C<docatch> which
658does a C<JMPENV_PUSH> and starts a third runops loop, which then executes
659the die op. At this point the C call stack looks like this:
660
661 Perl_pp_die
662 Perl_runops # third loop
663 S_docatch_body
664 S_docatch
665 Perl_pp_entertry
666 Perl_runops # second loop
667 S_call_body
668 Perl_call_sv
669 Perl_pp_tie
670 Perl_runops # first loop
671 S_run_body
672 perl_run
673 main
674
675and the context and data stacks, as shown by C<-Dstv>, look like:
676
677 STACK 0: MAIN
678 CX 0: BLOCK =>
679 CX 1: EVAL => AV() PV("A"\0)
680 retop=leave
681 STACK 1: MAGIC
682 CX 0: SUB =>
683 retop=(null)
684 CX 1: EVAL => *
685 retop=nextstate
686
687The die pops the first C<CxEVAL> off the context stack, sets
688C<PL_restartop> from it, does a C<JMPENV_JUMP(3)>, and control returns to
689the top C<docatch>. This then starts another third-level runops level,
690which executes the nextstate, pushmark and die ops on line 4. At the point
691that the second C<pp_die> is called, the C call stack looks exactly like
692that above, even though we are no longer within an inner eval; this is
693because of the optimization mentioned earlier. However, the context stack
694now looks like this, ie with the top CxEVAL popped:
695
696 STACK 0: MAIN
697 CX 0: BLOCK =>
698 CX 1: EVAL => AV() PV("A"\0)
699 retop=leave
700 STACK 1: MAGIC
701 CX 0: SUB =>
702 retop=(null)
703
704The die on line 4 pops the context stack back down to the CxEVAL, leaving
705it as:
706
707 STACK 0: MAIN
708 CX 0: BLOCK =>
709
710As usual, C<PL_restartop> is extracted from the C<CxEVAL>, and a
711C<JMPENV_JUMP(3)> done, which pops the C stack back to the docatch:
712
713 S_docatch
714 Perl_pp_entertry
715 Perl_runops # second loop
716 S_call_body
717 Perl_call_sv
718 Perl_pp_tie
719 Perl_runops # first loop
720 S_run_body
721 perl_run
722 main
723
724In this case, because the C<JMPENV> level recorded in the C<CxEVAL>
725differs from the current one, C<docatch> just does a C<JMPENV_JUMP(3)>
726and the C stack unwinds to:
727
728 perl_run
729 main
730
731Because C<PL_restartop> is non-null, C<run_body> starts a new runops loop
732and execution continues.
733
a422fd2d
SC
734=back
735
736=head2 Internal Variable Types
737
738You should by now have had a look at L<perlguts>, which tells you about
739Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do
740that now.
741
742These variables are used not only to represent Perl-space variables, but
743also any constants in the code, as well as some structures completely
744internal to Perl. The symbol table, for instance, is an ordinary Perl
745hash. Your code is represented by an SV as it's read into the parser;
746any program files you call are opened via ordinary Perl filehandles, and
747so on.
748
749The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a
750Perl program. Let's see, for instance, how Perl treats the constant
751C<"hello">.
752
753 % perl -MDevel::Peek -e 'Dump("hello")'
754 1 SV = PV(0xa041450) at 0xa04ecbc
755 2 REFCNT = 1
756 3 FLAGS = (POK,READONLY,pPOK)
757 4 PV = 0xa0484e0 "hello"\0
758 5 CUR = 5
759 6 LEN = 6
760
761Reading C<Devel::Peek> output takes a bit of practise, so let's go
762through it line by line.
763
764Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in
765memory. SVs themselves are very simple structures, but they contain a
766pointer to a more complex structure. In this case, it's a PV, a
767structure which holds a string value, at location C<0xa041450>. Line 2
768is the reference count; there are no other references to this data, so
769it's 1.
770
771Line 3 are the flags for this SV - it's OK to use it as a PV, it's a
772read-only SV (because it's a constant) and the data is a PV internally.
773Next we've got the contents of the string, starting at location
774C<0xa0484e0>.
775
776Line 5 gives us the current length of the string - note that this does
777B<not> include the null terminator. Line 6 is not the length of the
778string, but the length of the currently allocated buffer; as the string
779grows, Perl automatically extends the available storage via a routine
780called C<SvGROW>.
781
782You can get at any of these quantities from C very easily; just add
783C<Sv> to the name of the field shown in the snippet, and you've got a
784macro which will return the value: C<SvCUR(sv)> returns the current
785length of the string, C<SvREFCOUNT(sv)> returns the reference count,
786C<SvPV(sv, len)> returns the string itself with its length, and so on.
787More macros to manipulate these properties can be found in L<perlguts>.
788
789Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c>
790
791 1 void
792 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
793 3 {
794 4 STRLEN tlen;
795 5 char *junk;
796
797 6 junk = SvPV_force(sv, tlen);
798 7 SvGROW(sv, tlen + len + 1);
799 8 if (ptr == junk)
800 9 ptr = SvPVX(sv);
801 10 Move(ptr,SvPVX(sv)+tlen,len,char);
802 11 SvCUR(sv) += len;
803 12 *SvEND(sv) = '\0';
804 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */
805 14 SvTAINT(sv);
806 15 }
807
808This is a function which adds a string, C<ptr>, of length C<len> onto
809the end of the PV stored in C<sv>. The first thing we do in line 6 is
810make sure that the SV B<has> a valid PV, by calling the C<SvPV_force>
811macro to force a PV. As a side effect, C<tlen> gets set to the current
812value of the PV, and the PV itself is returned to C<junk>.
813
b1866b2d 814In line 7, we make sure that the SV will have enough room to accommodate
a422fd2d
SC
815the old string, the new string and the null terminator. If C<LEN> isn't
816big enough, C<SvGROW> will reallocate space for us.
817
818Now, if C<junk> is the same as the string we're trying to add, we can
819grab the string directly from the SV; C<SvPVX> is the address of the PV
820in the SV.
821
822Line 10 does the actual catenation: the C<Move> macro moves a chunk of
823memory around: we move the string C<ptr> to the end of the PV - that's
824the start of the PV plus its current length. We're moving C<len> bytes
825of type C<char>. After doing so, we need to tell Perl we've extended the
826string, by altering C<CUR> to reflect the new length. C<SvEND> is a
827macro which gives us the end of the string, so that needs to be a
828C<"\0">.
829
830Line 13 manipulates the flags; since we've changed the PV, any IV or NV
831values will no longer be valid: if we have C<$a=10; $a.="6";> we don't
1e54db1a 832want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF-8-aware
a422fd2d
SC
833version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags
834and turns on POK. The final C<SvTAINT> is a macro which launders tainted
835data if taint mode is turned on.
836
837AVs and HVs are more complicated, but SVs are by far the most common
838variable type being thrown around. Having seen something of how we
839manipulate these, let's go on and look at how the op tree is
840constructed.
841
842=head2 Op Trees
843
844First, what is the op tree, anyway? The op tree is the parsed
845representation of your program, as we saw in our section on parsing, and
846it's the sequence of operations that Perl goes through to execute your
847program, as we saw in L</Running>.
848
849An op is a fundamental operation that Perl can perform: all the built-in
850functions and operators are ops, and there are a series of ops which
851deal with concepts the interpreter needs internally - entering and
852leaving a block, ending a statement, fetching a variable, and so on.
853
854The op tree is connected in two ways: you can imagine that there are two
855"routes" through it, two orders in which you can traverse the tree.
856First, parse order reflects how the parser understood the code, and
857secondly, execution order tells perl what order to perform the
858operations in.
859
860The easiest way to examine the op tree is to stop Perl after it has
861finished parsing, and get it to dump out the tree. This is exactly what
7d7d5695
RGS
862the compiler backends L<B::Terse|B::Terse>, L<B::Concise|B::Concise>
863and L<B::Debug|B::Debug> do.
a422fd2d
SC
864
865Let's have a look at how Perl sees C<$a = $b + $c>:
866
867 % perl -MO=Terse -e '$a=$b+$c'
868 1 LISTOP (0x8179888) leave
869 2 OP (0x81798b0) enter
870 3 COP (0x8179850) nextstate
871 4 BINOP (0x8179828) sassign
872 5 BINOP (0x8179800) add [1]
873 6 UNOP (0x81796e0) null [15]
874 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b
875 8 UNOP (0x81797e0) null [15]
876 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c
877 10 UNOP (0x816b4f0) null [15]
878 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a
879
880Let's start in the middle, at line 4. This is a BINOP, a binary
881operator, which is at location C<0x8179828>. The specific operator in
882question is C<sassign> - scalar assignment - and you can find the code
883which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a
884binary operator, it has two children: the add operator, providing the
885result of C<$b+$c>, is uppermost on line 5, and the left hand side is on
886line 10.
887
888Line 10 is the null op: this does exactly nothing. What is that doing
889there? If you see the null op, it's a sign that something has been
890optimized away after parsing. As we mentioned in L</Optimization>,
891the optimization stage sometimes converts two operations into one, for
892example when fetching a scalar variable. When this happens, instead of
893rewriting the op tree and cleaning up the dangling pointers, it's easier
894just to replace the redundant operation with the null op. Originally,
895the tree would have looked like this:
896
897 10 SVOP (0x816b4f0) rv2sv [15]
898 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a
899
900That is, fetch the C<a> entry from the main symbol table, and then look
901at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>)
902happens to do both these things.
903
904The right hand side, starting at line 5 is similar to what we've just
905seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together
906two C<gvsv>s.
907
908Now, what's this about?
909
910 1 LISTOP (0x8179888) leave
911 2 OP (0x81798b0) enter
912 3 COP (0x8179850) nextstate
913
914C<enter> and C<leave> are scoping ops, and their job is to perform any
915housekeeping every time you enter and leave a block: lexical variables
916are tidied up, unreferenced variables are destroyed, and so on. Every
917program will have those first three lines: C<leave> is a list, and its
918children are all the statements in the block. Statements are delimited
919by C<nextstate>, so a block is a collection of C<nextstate> ops, with
920the ops to be performed for each statement being the children of
921C<nextstate>. C<enter> is a single op which functions as a marker.
922
923That's how Perl parsed the program, from top to bottom:
924
925 Program
926 |
927 Statement
928 |
929 =
930 / \
931 / \
932 $a +
933 / \
934 $b $c
935
936However, it's impossible to B<perform> the operations in this order:
937you have to find the values of C<$b> and C<$c> before you add them
938together, for instance. So, the other thread that runs through the op
939tree is the execution order: each op has a field C<op_next> which points
940to the next op to be run, so following these pointers tells us how perl
941executes the code. We can traverse the tree in this order using
942the C<exec> option to C<B::Terse>:
943
944 % perl -MO=Terse,exec -e '$a=$b+$c'
945 1 OP (0x8179928) enter
946 2 COP (0x81798c8) nextstate
947 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b
948 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c
949 5 BINOP (0x8179878) add [1]
950 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a
951 7 BINOP (0x81798a0) sassign
952 8 LISTOP (0x8179900) leave
953
954This probably makes more sense for a human: enter a block, start a
955statement. Get the values of C<$b> and C<$c>, and add them together.
956Find C<$a>, and assign one to the other. Then leave.
957
958The way Perl builds up these op trees in the parsing process can be
959unravelled by examining F<perly.y>, the YACC grammar. Let's take the
960piece we need to construct the tree for C<$a = $b + $c>
961
962 1 term : term ASSIGNOP term
963 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
964 3 | term ADDOP term
965 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
966
967If you're not used to reading BNF grammars, this is how it works: You're
968fed certain things by the tokeniser, which generally end up in upper
969case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your
970code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are
b432a672 971"terminal symbols", because you can't get any simpler than them.
a422fd2d
SC
972
973The grammar, lines one and three of the snippet above, tells you how to
b432a672 974build up more complex forms. These complex forms, "non-terminal symbols"
a422fd2d
SC
975are generally placed in lower case. C<term> here is a non-terminal
976symbol, representing a single expression.
977
978The grammar gives you the following rule: you can make the thing on the
979left of the colon if you see all the things on the right in sequence.
980This is called a "reduction", and the aim of parsing is to completely
981reduce the input. There are several different ways you can perform a
982reduction, separated by vertical bars: so, C<term> followed by C<=>
983followed by C<term> makes a C<term>, and C<term> followed by C<+>
984followed by C<term> can also make a C<term>.
985
986So, if you see two terms with an C<=> or C<+>, between them, you can
987turn them into a single expression. When you do this, you execute the
988code in the block on the next line: if you see C<=>, you'll do the code
989in line 2. If you see C<+>, you'll do the code in line 4. It's this code
990which contributes to the op tree.
991
992 | term ADDOP term
993 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }
994
995What this does is creates a new binary op, and feeds it a number of
996variables. The variables refer to the tokens: C<$1> is the first token in
997the input, C<$2> the second, and so on - think regular expression
998backreferences. C<$$> is the op returned from this reduction. So, we
999call C<newBINOP> to create a new binary operator. The first parameter to
1000C<newBINOP>, a function in F<op.c>, is the op type. It's an addition
1001operator, so we want the type to be C<ADDOP>. We could specify this
1002directly, but it's right there as the second token in the input, so we
b432a672
AL
1003use C<$2>. The second parameter is the op's flags: 0 means "nothing
1004special". Then the things to add: the left and right hand side of our
a422fd2d
SC
1005expression, in scalar context.
1006
1007=head2 Stacks
1008
1009When perl executes something like C<addop>, how does it pass on its
1010results to the next op? The answer is, through the use of stacks. Perl
1011has a number of stacks to store things it's currently working on, and
1012we'll look at the three most important ones here.
1013
1014=over 3
1015
1016=item Argument stack
1017
1018Arguments are passed to PP code and returned from PP code using the
1019argument stack, C<ST>. The typical way to handle arguments is to pop
1020them off the stack, deal with them how you wish, and then push the result
1021back onto the stack. This is how, for instance, the cosine operator
1022works:
1023
1024 NV value;
1025 value = POPn;
1026 value = Perl_cos(value);
1027 XPUSHn(value);
1028
1029We'll see a more tricky example of this when we consider Perl's macros
1030below. C<POPn> gives you the NV (floating point value) of the top SV on
1031the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push
1032the result back as an NV. The C<X> in C<XPUSHn> means that the stack
1033should be extended if necessary - it can't be necessary here, because we
1034know there's room for one more item on the stack, since we've just
1035removed one! The C<XPUSH*> macros at least guarantee safety.
1036
1037Alternatively, you can fiddle with the stack directly: C<SP> gives you
1038the first element in your portion of the stack, and C<TOP*> gives you
1039the top SV/IV/NV/etc. on the stack. So, for instance, to do unary
1040negation of an integer:
1041
1042 SETi(-TOPi);
1043
1044Just set the integer value of the top stack entry to its negation.
1045
1046Argument stack manipulation in the core is exactly the same as it is in
1047XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer
1048description of the macros used in stack manipulation.
1049
1050=item Mark stack
1051
b432a672 1052I say "your portion of the stack" above because PP code doesn't
a422fd2d
SC
1053necessarily get the whole stack to itself: if your function calls
1054another function, you'll only want to expose the arguments aimed for the
1055called function, and not (necessarily) let it get at your own data. The
b432a672 1056way we do this is to have a "virtual" bottom-of-stack, exposed to each
a422fd2d
SC
1057function. The mark stack keeps bookmarks to locations in the argument
1058stack usable by each function. For instance, when dealing with a tied
b432a672 1059variable, (internally, something with "P" magic) Perl has to call
a422fd2d
SC
1060methods for accesses to the tied variables. However, we need to separate
1061the arguments exposed to the method to the argument exposed to the
ed233832
DM
1062original function - the store or fetch or whatever it may be. Here's
1063roughly how the tied C<push> is implemented; see C<av_push> in F<av.c>:
a422fd2d
SC
1064
1065 1 PUSHMARK(SP);
1066 2 EXTEND(SP,2);
1067 3 PUSHs(SvTIED_obj((SV*)av, mg));
1068 4 PUSHs(val);
1069 5 PUTBACK;
1070 6 ENTER;
1071 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1072 8 LEAVE;
13a2d996 1073
a422fd2d
SC
1074Let's examine the whole implementation, for practice:
1075
1076 1 PUSHMARK(SP);
1077
1078Push the current state of the stack pointer onto the mark stack. This is
1079so that when we've finished adding items to the argument stack, Perl
1080knows how many things we've added recently.
1081
1082 2 EXTEND(SP,2);
1083 3 PUSHs(SvTIED_obj((SV*)av, mg));
1084 4 PUSHs(val);
1085
1086We're going to add two more items onto the argument stack: when you have
1087a tied array, the C<PUSH> subroutine receives the object and the value
1088to be pushed, and that's exactly what we have here - the tied object,
1089retrieved with C<SvTIED_obj>, and the value, the SV C<val>.
1090
1091 5 PUTBACK;
1092
e89a6d4e
JD
1093Next we tell Perl to update the global stack pointer from our internal
1094variable: C<dSP> only gave us a local copy, not a reference to the global.
a422fd2d
SC
1095
1096 6 ENTER;
1097 7 call_method("PUSH", G_SCALAR|G_DISCARD);
1098 8 LEAVE;
1099
1100C<ENTER> and C<LEAVE> localise a block of code - they make sure that all
1101variables are tidied up, everything that has been localised gets
1102its previous value returned, and so on. Think of them as the C<{> and
1103C<}> of a Perl block.
1104
1105To actually do the magic method call, we have to call a subroutine in
1106Perl space: C<call_method> takes care of that, and it's described in
1107L<perlcall>. We call the C<PUSH> method in scalar context, and we're
e89a6d4e
JD
1108going to discard its return value. The call_method() function
1109removes the top element of the mark stack, so there is nothing for
1110the caller to clean up.
a422fd2d 1111
a422fd2d
SC
1112=item Save stack
1113
1114C doesn't have a concept of local scope, so perl provides one. We've
1115seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save
1116stack implements the C equivalent of, for example:
1117
1118 {
1119 local $foo = 42;
1120 ...
1121 }
1122
1123See L<perlguts/Localising Changes> for how to use the save stack.
1124
1125=back
1126
1127=head2 Millions of Macros
1128
1129One thing you'll notice about the Perl source is that it's full of
1130macros. Some have called the pervasive use of macros the hardest thing
1131to understand, others find it adds to clarity. Let's take an example,
1132the code which implements the addition operator:
1133
1134 1 PP(pp_add)
1135 2 {
39644a26 1136 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
a422fd2d
SC
1137 4 {
1138 5 dPOPTOPnnrl_ul;
1139 6 SETn( left + right );
1140 7 RETURN;
1141 8 }
1142 9 }
1143
1144Every line here (apart from the braces, of course) contains a macro. The
1145first line sets up the function declaration as Perl expects for PP code;
1146line 3 sets up variable declarations for the argument stack and the
1147target, the return value of the operation. Finally, it tries to see if
1148the addition operation is overloaded; if so, the appropriate subroutine
1149is called.
1150
1151Line 5 is another variable declaration - all variable declarations start
1152with C<d> - which pops from the top of the argument stack two NVs (hence
1153C<nn>) and puts them into the variables C<right> and C<left>, hence the
1154C<rl>. These are the two operands to the addition operator. Next, we
1155call C<SETn> to set the NV of the return value to the result of adding
1156the two values. This done, we return - the C<RETURN> macro makes sure
1157that our return value is properly handled, and we pass the next operator
1158to run back to the main run loop.
1159
1160Most of these macros are explained in L<perlapi>, and some of the more
1161important ones are explained in L<perlxs> as well. Pay special attention
1162to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on
1163the C<[pad]THX_?> macros.
1164
52d59bef
JH
1165=head2 The .i Targets
1166
1167You can expand the macros in a F<foo.c> file by saying
1168
1169 make foo.i
1170
1171which will expand the macros using cpp. Don't be scared by the results.
1172
955fec6b
JH
1173=head1 SOURCE CODE STATIC ANALYSIS
1174
1175Various tools exist for analysing C source code B<statically>, as
1176opposed to B<dynamically>, that is, without executing the code.
1177It is possible to detect resource leaks, undefined behaviour, type
1178mismatches, portability problems, code paths that would cause illegal
1179memory accesses, and other similar problems by just parsing the C code
1180and looking at the resulting graph, what does it tell about the
1181execution and data flows. As a matter of fact, this is exactly
1182how C compilers know to give warnings about dubious code.
1183
1184=head2 lint, splint
1185
1186The good old C code quality inspector, C<lint>, is available in
1187several platforms, but please be aware that there are several
1188different implementations of it by different vendors, which means that
1189the flags are not identical across different platforms.
1190
1191There is a lint variant called C<splint> (Secure Programming Lint)
1192available from http://www.splint.org/ that should compile on any
1193Unix-like platform.
1194
1195There are C<lint> and <splint> targets in Makefile, but you may have
1196to diddle with the flags (see above).
1197
1198=head2 Coverity
1199
1200Coverity (http://www.coverity.com/) is a product similar to lint and
1201as a testbed for their product they periodically check several open
1202source projects, and they give out accounts to open source developers
1203to the defect databases.
1204
1205=head2 cpd (cut-and-paste detector)
1206
1207The cpd tool detects cut-and-paste coding. If one instance of the
1208cut-and-pasted code changes, all the other spots should probably be
1209changed, too. Therefore such code should probably be turned into a
1210subroutine or a macro.
1211
1212cpd (http://pmd.sourceforge.net/cpd.html) is part of the pmd project
1213(http://pmd.sourceforge.net/). pmd was originally written for static
1214analysis of Java code, but later the cpd part of it was extended to
1215parse also C and C++.
1216
a52aaefa
RGS
1217Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the
1218pmd-X.Y.jar from it, and then run that on source code thusly:
955fec6b
JH
1219
1220 java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD --minimum-tokens 100 --files /some/where/src --language c > cpd.txt
1221
1222You may run into memory limits, in which case you should use the -Xmx option:
1223
1224 java -Xmx512M ...
1225
1226=head2 gcc warnings
1227
1228Though much can be written about the inconsistency and coverage
1229problems of gcc warnings (like C<-Wall> not meaning "all the
1230warnings", or some common portability problems not being covered by
1231C<-Wall>, or C<-ansi> and C<-pedantic> both being a poorly defined
1232collection of warnings, and so forth), gcc is still a useful tool in
1233keeping our coding nose clean.
1234
1235The C<-Wall> is by default on.
1236
a8e98a71
JH
1237The C<-ansi> (and its sidekick, C<-pedantic>) would be nice to be on
1238always, but unfortunately they are not safe on all platforms, they can
1239for example cause fatal conflicts with the system headers (Solaris
1240being a prime example). If Configure C<-Dgccansipedantic> is used,
1241the C<cflags> frontend selects C<-ansi -pedantic> for the platforms
1242where they are known to be safe.
955fec6b
JH
1243
1244Starting from Perl 5.9.4 the following extra flags are added:
1245
1246=over 4
1247
1248=item *
1249
1250C<-Wendif-labels>
1251
1252=item *
1253
1254C<-Wextra>
1255
1256=item *
1257
1258C<-Wdeclaration-after-statement>
1259
1260=back
1261
1262The following flags would be nice to have but they would first need
0503309d 1263their own Augean stablemaster:
955fec6b
JH
1264
1265=over 4
1266
1267=item *
1268
1269C<-Wpointer-arith>
1270
1271=item *
1272
1273C<-Wshadow>
1274
1275=item *
1276
1277C<-Wstrict-prototypes>
1278
955fec6b
JH
1279=back
1280
1281The C<-Wtraditional> is another example of the annoying tendency of
ac036724 1282gcc to bundle a lot of warnings under one switch (it would be
1283impossible to deploy in practice because it would complain a lot) but
955fec6b
JH
1284it does contain some warnings that would be beneficial to have available
1285on their own, such as the warning about string constants inside macros
1286containing the macro arguments: this behaved differently pre-ANSI
1287than it does in ANSI, and some C compilers are still in transition,
1288AIX being an example.
1289
1290=head2 Warnings of other C compilers
1291
1292Other C compilers (yes, there B<are> other C compilers than gcc) often
1293have their "strict ANSI" or "strict ANSI with some portability extensions"
1294modes on, like for example the Sun Workshop has its C<-Xa> mode on
1295(though implicitly), or the DEC (these days, HP...) has its C<-std1>
1296mode on.
1297
1298=head2 DEBUGGING
1299
1300You can compile a special debugging version of Perl, which allows you
1301to use the C<-D> option of Perl to tell more about what Perl is doing.
1302But sometimes there is no alternative than to dive in with a debugger,
1303either to see the stack trace of a core dump (very useful in a bug
1304report), or trying to figure out what went wrong before the core dump
1305happened, or how did we end up having wrong or unexpected results.
1306
a422fd2d
SC
1307=head2 Poking at Perl
1308
1309To really poke around with Perl, you'll probably want to build Perl for
1310debugging, like this:
1311
1312 ./Configure -d -D optimize=-g
1313 make
1314
1315C<-g> is a flag to the C compiler to have it produce debugging
955fec6b
JH
1316information which will allow us to step through a running program,
1317and to see in which C function we are at (without the debugging
1318information we might see only the numerical addresses of the functions,
1319which is not very helpful).
1320
a422fd2d
SC
1321F<Configure> will also turn on the C<DEBUGGING> compilation symbol which
1322enables all the internal debugging code in Perl. There are a whole bunch
1323of things you can debug with this: L<perlrun> lists them all, and the
1324best way to find out about them is to play about with them. The most
1325useful options are probably
1326
1327 l Context (loop) stack processing
1328 t Trace execution
1329 o Method and overloading resolution
1330 c String/numeric conversions
1331
1332Some of the functionality of the debugging code can be achieved using XS
1333modules.
13a2d996 1334
a422fd2d
SC
1335 -Dr => use re 'debug'
1336 -Dx => use O 'Debug'
1337
1338=head2 Using a source-level debugger
1339
1340If the debugging output of C<-D> doesn't help you, it's time to step
1341through perl's execution with a source-level debugger.
1342
1343=over 3
1344
1345=item *
1346
955fec6b
JH
1347We'll use C<gdb> for our examples here; the principles will apply to
1348any debugger (many vendors call their debugger C<dbx>), but check the
1349manual of the one you're using.
a422fd2d
SC
1350
1351=back
1352
1353To fire up the debugger, type
1354
1355 gdb ./perl
1356
955fec6b
JH
1357Or if you have a core dump:
1358
1359 gdb ./perl core
1360
a422fd2d
SC
1361You'll want to do that in your Perl source tree so the debugger can read
1362the source code. You should see the copyright message, followed by the
1363prompt.
1364
1365 (gdb)
1366
1367C<help> will get you into the documentation, but here are the most
1368useful commands:
1369
1370=over 3
1371
1372=item run [args]
1373
1374Run the program with the given arguments.
1375
1376=item break function_name
1377
1378=item break source.c:xxx
1379
1380Tells the debugger that we'll want to pause execution when we reach
cea6626f 1381either the named function (but see L<perlguts/Internal Functions>!) or the given
a422fd2d
SC
1382line in the named source file.
1383
1384=item step
1385
1386Steps through the program a line at a time.
1387
1388=item next
1389
1390Steps through the program a line at a time, without descending into
1391functions.
1392
1393=item continue
1394
1395Run until the next breakpoint.
1396
1397=item finish
1398
1399Run until the end of the current function, then stop again.
1400
13a2d996 1401=item 'enter'
a422fd2d
SC
1402
1403Just pressing Enter will do the most recent operation again - it's a
1404blessing when stepping through miles of source code.
1405
1406=item print
1407
1408Execute the given C code and print its results. B<WARNING>: Perl makes
52d59bef
JH
1409heavy use of macros, and F<gdb> does not necessarily support macros
1410(see later L</"gdb macro support">). You'll have to substitute them
1411yourself, or to invoke cpp on the source code files
1412(see L</"The .i Targets">)
1413So, for instance, you can't say
a422fd2d
SC
1414
1415 print SvPV_nolen(sv)
1416
1417but you have to say
1418
1419 print Perl_sv_2pv_nolen(sv)
1420
ffc145e8
RK
1421=back
1422
a422fd2d
SC
1423You may find it helpful to have a "macro dictionary", which you can
1424produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
07aa3531 1425recursively apply those macros for you.
52d59bef
JH
1426
1427=head2 gdb macro support
a422fd2d 1428
52d59bef 1429Recent versions of F<gdb> have fairly good macro support, but
ea031e66
RGS
1430in order to use it you'll need to compile perl with macro definitions
1431included in the debugging information. Using F<gcc> version 3.1, this
1432means configuring with C<-Doptimize=-g3>. Other compilers might use a
1433different switch (if they support debugging macros at all).
1434
a422fd2d
SC
1435=head2 Dumping Perl Data Structures
1436
1437One way to get around this macro hell is to use the dumping functions in
1438F<dump.c>; these work a little like an internal
1439L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures
1440that you can't get at from Perl. Let's take an example. We'll use the
07aa3531 1441C<$a = $b + $c> we used before, but give it a bit of context:
a422fd2d
SC
1442C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around?
1443
1444What about C<pp_add>, the function we examined earlier to implement the
1445C<+> operator:
1446
1447 (gdb) break Perl_pp_add
1448 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
1449
cea6626f 1450Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>.
a422fd2d
SC
1451With the breakpoint in place, we can run our program:
1452
1453 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
1454
1455Lots of junk will go past as gdb reads in the relevant source files and
1456libraries, and then:
1457
1458 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
39644a26 1459 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
a422fd2d
SC
1460 (gdb) step
1461 311 dPOPTOPnnrl_ul;
1462 (gdb)
1463
1464We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul>
1465arranges for two C<NV>s to be placed into C<left> and C<right> - let's
1466slightly expand it:
1467
195c30ce
KW
1468 #define dPOPTOPnnrl_ul NV right = POPn; \
1469 SV *leftsv = TOPs; \
1470 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
a422fd2d
SC
1471
1472C<POPn> takes the SV from the top of the stack and obtains its NV either
1473directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function.
1474C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses
1475C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from
07aa3531 1476C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>.
a422fd2d
SC
1477
1478Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
1479convert it. If we step again, we'll find ourselves there:
1480
1481 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1482 1669 if (!sv)
1483 (gdb)
1484
1485We can now use C<Perl_sv_dump> to investigate the SV:
1486
1487 SV = PV(0xa057cc0) at 0xa0675d0
1488 REFCNT = 1
1489 FLAGS = (POK,pPOK)
1490 PV = 0xa06a510 "6XXXX"\0
1491 CUR = 5
1492 LEN = 6
1493 $1 = void
1494
1495We know we're going to get C<6> from this, so let's finish the
1496subroutine:
1497
1498 (gdb) finish
1499 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
1500 0x462669 in Perl_pp_add () at pp_hot.c:311
1501 311 dPOPTOPnnrl_ul;
1502
1503We can also dump out this op: the current op is always stored in
1504C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
1505similar output to L<B::Debug|B::Debug>.
1506
1507 {
1508 13 TYPE = add ===> 14
1509 TARG = 1
1510 FLAGS = (SCALAR,KIDS)
1511 {
1512 TYPE = null ===> (12)
1513 (was rv2sv)
1514 FLAGS = (SCALAR,KIDS)
1515 {
1516 11 TYPE = gvsv ===> 12
1517 FLAGS = (SCALAR)
1518 GV = main::b
1519 }
1520 }
1521
10f58044 1522# finish this later #
a422fd2d
SC
1523
1524=head2 Patching
1525
1526All right, we've now had a look at how to navigate the Perl sources and
1527some things you'll need to know when fiddling with them. Let's now get
1528on and create a simple patch. Here's something Larry suggested: if a
07aa3531 1529C<U> is the first active format during a C<pack>, (for example,
a422fd2d 1530C<pack "U3C8", @stuff>) then the resulting string should be treated as
1e54db1a 1531UTF-8 encoded.
a422fd2d 1532
168a53cc
DR
1533If you are working with a git clone of the Perl repository, you will want to
1534create a branch for your changes. This will make creating a proper patch much
1535simpler. See the L<perlrepository> for details on how to do this.
1536
a422fd2d
SC
1537How do we prepare to fix this up? First we locate the code in question -
1538the C<pack> happens at runtime, so it's going to be in one of the F<pp>
1539files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be
1540altering this file, let's copy it to F<pp.c~>.
1541
a6ec74c1
JH
1542[Well, it was in F<pp.c> when this tutorial was written. It has now been
1543split off with C<pp_unpack> to its own file, F<pp_pack.c>]
1544
a422fd2d
SC
1545Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then
1546loop over the pattern, taking each format character in turn into
1547C<datum_type>. Then for each possible format character, we swallow up
1548the other arguments in the pattern (a field width, an asterisk, and so
1549on) and convert the next chunk input into the specified format, adding
1550it onto the output SV C<cat>.
1551
1552How do we know if the C<U> is the first format in the C<pat>? Well, if
1553we have a pointer to the start of C<pat> then, if we see a C<U> we can
1554test whether we're still at the start of the string. So, here's where
1555C<pat> is set up:
1556
1557 STRLEN fromlen;
1558 register char *pat = SvPVx(*++MARK, fromlen);
1559 register char *patend = pat + fromlen;
1560 register I32 len;
1561 I32 datumtype;
1562 SV *fromstr;
1563
1564We'll have another string pointer in there:
1565
1566 STRLEN fromlen;
1567 register char *pat = SvPVx(*++MARK, fromlen);
1568 register char *patend = pat + fromlen;
1569 + char *patcopy;
1570 register I32 len;
1571 I32 datumtype;
1572 SV *fromstr;
1573
1574And just before we start the loop, we'll set C<patcopy> to be the start
1575of C<pat>:
1576
1577 items = SP - MARK;
1578 MARK++;
1579 sv_setpvn(cat, "", 0);
1580 + patcopy = pat;
1581 while (pat < patend) {
1582
1583Now if we see a C<U> which was at the start of the string, we turn on
1e54db1a 1584the C<UTF8> flag for the output SV, C<cat>:
a422fd2d
SC
1585
1586 + if (datumtype == 'U' && pat==patcopy+1)
1587 + SvUTF8_on(cat);
1588 if (datumtype == '#') {
1589 while (pat < patend && *pat != '\n')
1590 pat++;
1591
1592Remember that it has to be C<patcopy+1> because the first character of
1593the string is the C<U> which has been swallowed into C<datumtype!>
1594
1595Oops, we forgot one thing: what if there are spaces at the start of the
1596pattern? C<pack(" U*", @stuff)> will have C<U> as the first active
1597character, even though it's not the first thing in the pattern. In this
1598case, we have to advance C<patcopy> along with C<pat> when we see spaces:
1599
1600 if (isSPACE(datumtype))
1601 continue;
1602
1603needs to become
1604
1605 if (isSPACE(datumtype)) {
1606 patcopy++;
1607 continue;
1608 }
1609
1610OK. That's the C part done. Now we must do two additional things before
1611this patch is ready to go: we've changed the behaviour of Perl, and so
1612we must document that change. We must also provide some more regression
1613tests to make sure our patch works and doesn't create a bug somewhere
1614else along the line.
1615
b23b8711
MS
1616The regression tests for each operator live in F<t/op/>, and so we
1617make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our
1618tests to the end. First, we'll test that the C<U> does indeed create
07aa3531 1619Unicode strings.
b23b8711
MS
1620
1621t/op/pack.t has a sensible ok() function, but if it didn't we could
35c336e6 1622use the one from t/test.pl.
b23b8711 1623
35c336e6
MS
1624 require './test.pl';
1625 plan( tests => 159 );
b23b8711
MS
1626
1627so instead of this:
a422fd2d 1628
195c30ce
KW
1629 print 'not ' unless "1.20.300.4000" eq sprintf "%vd",
1630 pack("U*",1,20,300,4000);
a422fd2d
SC
1631 print "ok $test\n"; $test++;
1632
35c336e6
MS
1633we can write the more sensible (see L<Test::More> for a full
1634explanation of is() and other testing functions).
b23b8711 1635
07aa3531 1636 is( "1.20.300.4000", sprintf "%vd", pack("U*",1,20,300,4000),
38a44b82 1637 "U* produces Unicode" );
b23b8711 1638
a422fd2d
SC
1639Now we'll test that we got that space-at-the-beginning business right:
1640
35c336e6 1641 is( "1.20.300.4000", sprintf "%vd", pack(" U*",1,20,300,4000),
195c30ce 1642 " with spaces at the beginning" );
a422fd2d
SC
1643
1644And finally we'll test that we don't make Unicode strings if C<U> is B<not>
1645the first active format:
1646
35c336e6 1647 isnt( v1.20.300.4000, sprintf "%vd", pack("C0U*",1,20,300,4000),
38a44b82 1648 "U* not first isn't Unicode" );
a422fd2d 1649
35c336e6
MS
1650Mustn't forget to change the number of tests which appears at the top,
1651or else the automated tester will get confused. This will either look
1652like this:
a422fd2d 1653
35c336e6
MS
1654 print "1..156\n";
1655
1656or this:
1657
1658 plan( tests => 156 );
a422fd2d
SC
1659
1660We now compile up Perl, and run it through the test suite. Our new
1661tests pass, hooray!
1662
1663Finally, the documentation. The job is never done until the paperwork is
1664over, so let's describe the change we've just made. The relevant place
1665is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert
1666this text in the description of C<pack>:
1667
1668 =item *
1669
1670 If the pattern begins with a C<U>, the resulting string will be treated
1e54db1a
JH
1671 as UTF-8-encoded Unicode. You can force UTF-8 encoding on in a string
1672 with an initial C<U0>, and the bytes that follow will be interpreted as
195c30ce
KW
1673 Unicode characters. If you don't want this to happen, you can begin
1674 your pattern with C<C0> (or anything else) to force Perl not to UTF-8
1675 encode your string, and then follow this with a C<U*> somewhere in your
1676 pattern.
a422fd2d 1677
f7e1e956
MS
1678=head2 Patching a core module
1679
1680This works just like patching anything else, with an extra
1681consideration. Many core modules also live on CPAN. If this is so,
1682patch the CPAN version instead of the core and send the patch off to
1683the module maintainer (with a copy to p5p). This will help the module
1684maintainer keep the CPAN version in sync with the core version without
1685constantly scanning p5p.
1686
db300100
RGS
1687The list of maintainers of core modules is usefully documented in
1688F<Porting/Maintainers.pl>.
1689
acbe17fc
JP
1690=head2 Adding a new function to the core
1691
1692If, as part of a patch to fix a bug, or just because you have an
1693especially good idea, you decide to add a new function to the core,
1694discuss your ideas on p5p well before you start work. It may be that
1695someone else has already attempted to do what you are considering and
1696can give lots of good advice or even provide you with bits of code
1697that they already started (but never finished).
1698
1699You have to follow all of the advice given above for patching. It is
1700extremely important to test any addition thoroughly and add new tests
1701to explore all boundary conditions that your new function is expected
1702to handle. If your new function is used only by one module (e.g. toke),
1703then it should probably be named S_your_function (for static); on the
210b36aa 1704other hand, if you expect it to accessible from other functions in
acbe17fc
JP
1705Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions>
1706for more details.
1707
1708The location of any new code is also an important consideration. Don't
1709just create a new top level .c file and put your code there; you would
1710have to make changes to Configure (so the Makefile is created properly),
1711as well as possibly lots of include files. This is strictly pumpking
1712business.
1713
1714It is better to add your function to one of the existing top level
1715source code files, but your choice is complicated by the nature of
1716the Perl distribution. Only the files that are marked as compiled
1717static are located in the perl executable. Everything else is located
1718in the shared library (or DLL if you are running under WIN32). So,
1719for example, if a function was only used by functions located in
1720toke.c, then your code can go in toke.c. If, however, you want to call
1721the function from universal.c, then you should put your code in another
1722location, for example util.c.
1723
1724In addition to writing your c-code, you will need to create an
1725appropriate entry in embed.pl describing your function, then run
1726'make regen_headers' to create the entries in the numerous header
1727files that perl needs to compile correctly. See L<perlguts/Internal Functions>
1728for information on the various options that you can set in embed.pl.
1729You will forget to do this a few (or many) times and you will get
1730warnings during the compilation phase. Make sure that you mention
1731this when you post your patch to P5P; the pumpking needs to know this.
1732
1733When you write your new code, please be conscious of existing code
884bad00 1734conventions used in the perl source files. See L<perlstyle> for
acbe17fc
JP
1735details. Although most of the guidelines discussed seem to focus on
1736Perl code, rather than c, they all apply (except when they don't ;).
195c30ce 1737Also see L<perlrepository> for lots of details about both formatting and
552d2b87 1738submitting patches of your changes.
acbe17fc
JP
1739
1740Lastly, TEST TEST TEST TEST TEST any code before posting to p5p.
1741Test on as many platforms as you can find. Test as many perl
1742Configure options as you can (e.g. MULTIPLICITY). If you have
1743profiling or memory tools, see L<EXTERNAL TOOLS FOR DEBUGGING PERL>
210b36aa 1744below for how to use them to further test your code. Remember that
acbe17fc
JP
1745most of the people on P5P are doing this on their own time and
1746don't have the time to debug your code.
f7e1e956
MS
1747
1748=head2 Writing a test
1749
1750Every module and built-in function has an associated test file (or
1751should...). If you add or change functionality, you have to write a
1752test. If you fix a bug, you have to write a test so that bug never
1753comes back. If you alter the docs, it would be nice to test what the
1754new documentation says.
1755
1756In short, if you submit a patch you probably also have to patch the
1757tests.
1758
1759For modules, the test file is right next to the module itself.
1760F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation,
1761so there are some snags (and it would be wonderful for you to brush
1762them out), but it basically works that way. Everything else lives in
1763F<t/>.
1764
628f0a0a
KW
1765Testing of warning messages is often separately done by using expect scripts in
1766F<t/lib/warnings>. This is because much of the setup for them is already done
1767for you.
1768
d5f28025
JV
1769If you add a new test directory under F<t/>, it is imperative that you
1770add that directory to F<t/HARNESS> and F<t/TEST>.
1771
f7e1e956
MS
1772=over 3
1773
1774=item F<t/base/>
1775
1776Testing of the absolute basic functionality of Perl. Things like
1777C<if>, basic file reads and writes, simple regexes, etc. These are
1778run first in the test suite and if any of them fail, something is
1779I<really> broken.
1780
1781=item F<t/cmd/>
1782
1783These test the basic control structures, C<if/else>, C<while>,
35c336e6 1784subroutines, etc.
f7e1e956
MS
1785
1786=item F<t/comp/>
1787
1788Tests basic issues of how Perl parses and compiles itself.
1789
1790=item F<t/io/>
1791
1792Tests for built-in IO functions, including command line arguments.
1793
1794=item F<t/lib/>
1795
1796The old home for the module tests, you shouldn't put anything new in
1797here. There are still some bits and pieces hanging around in here
1798that need to be moved. Perhaps you could move them? Thanks!
1799
3c295041
RGS
1800=item F<t/mro/>
1801
0503309d 1802Tests for perl's method resolution order implementations
3c295041
RGS
1803(see L<mro>).
1804
f7e1e956
MS
1805=item F<t/op/>
1806
1807Tests for perl's built in functions that don't fit into any of the
1808other directories.
1809
a4499558
YO
1810=item F<t/re/>
1811
1812Tests for regex related functions or behaviour. (These used to live
1813in t/op).
1814
f7e1e956
MS
1815=item F<t/run/>
1816
1817Testing features of how perl actually runs, including exit codes and
1818handling of PERL* environment variables.
1819
244d9cb7
RGS
1820=item F<t/uni/>
1821
1822Tests for the core support of Unicode.
1823
1824=item F<t/win32/>
1825
1826Windows-specific tests.
1827
1828=item F<t/x2p>
1829
1830A test suite for the s2p converter.
1831
f7e1e956
MS
1832=back
1833
1834The core uses the same testing style as the rest of Perl, a simple
1835"ok/not ok" run through Test::Harness, but there are a few special
1836considerations.
1837
35c336e6
MS
1838There are three ways to write a test in the core. Test::More,
1839t/test.pl and ad hoc C<print $test ? "ok 42\n" : "not ok 42\n">. The
1840decision of which to use depends on what part of the test suite you're
1841working on. This is a measure to prevent a high-level failure (such
1842as Config.pm breaking) from causing basic functionality tests to fail.
b3fe9f3f 1843If you write your own test, use the L<Test Anything Protocol|TAP>.
35c336e6 1844
07aa3531 1845=over 4
35c336e6
MS
1846
1847=item t/base t/comp
1848
1849Since we don't know if require works, or even subroutines, use ad hoc
1850tests for these two. Step carefully to avoid using the feature being
1851tested.
1852
1853=item t/cmd t/run t/io t/op
1854
1855Now that basic require() and subroutines are tested, you can use the
1856t/test.pl library which emulates the important features of Test::More
1857while using a minimum of core features.
1858
1859You can also conditionally use certain libraries like Config, but be
1860sure to skip the test gracefully if it's not there.
1861
1862=item t/lib ext lib
1863
1864Now that the core of Perl is tested, Test::More can be used. You can
1865also use the full suite of core modules in the tests.
1866
1867=back
f7e1e956
MS
1868
1869When you say "make test" Perl uses the F<t/TEST> program to run the
07aa3531
JC
1870test suite (except under Win32 where it uses F<t/harness> instead.)
1871All tests are run from the F<t/> directory, B<not> the directory
1872which contains the test. This causes some problems with the tests
7205a85d 1873in F<lib/>, so here's some opportunity for some patching.
f7e1e956
MS
1874
1875You must be triply conscious of cross-platform concerns. This usually
1876boils down to using File::Spec and avoiding things like C<fork()> and
1877C<system()> unless absolutely necessary.
1878
e018f8be
JH
1879=head2 Special Make Test Targets
1880
1881There are various special make targets that can be used to test Perl
1882slightly differently than the standard "test" target. Not all them
1883are expected to give a 100% success rate. Many of them have several
7205a85d
YO
1884aliases, and many of them are not available on certain operating
1885systems.
e018f8be
JH
1886
1887=over 4
1888
1889=item coretest
1890
7d7d5695 1891Run F<perl> on all core tests (F<t/*> and F<lib/[a-z]*> pragma tests).
e018f8be 1892
7205a85d
YO
1893(Not available on Win32)
1894
e018f8be
JH
1895=item test.deparse
1896
b26492ee
RGS
1897Run all the tests through B::Deparse. Not all tests will succeed.
1898
7205a85d
YO
1899(Not available on Win32)
1900
b26492ee
RGS
1901=item test.taintwarn
1902
1903Run all tests with the B<-t> command-line switch. Not all tests
1904are expected to succeed (until they're specifically fixed, of course).
e018f8be 1905
7205a85d
YO
1906(Not available on Win32)
1907
e018f8be
JH
1908=item minitest
1909
1910Run F<miniperl> on F<t/base>, F<t/comp>, F<t/cmd>, F<t/run>, F<t/io>,
8cebccf4 1911F<t/op>, F<t/uni> and F<t/mro> tests.
e018f8be 1912
7a834142
JH
1913=item test.valgrind check.valgrind utest.valgrind ucheck.valgrind
1914
1915(Only in Linux) Run all the tests using the memory leak + naughty
1916memory access tool "valgrind". The log files will be named
1917F<testname.valgrind>.
1918
e018f8be
JH
1919=item test.third check.third utest.third ucheck.third
1920
1921(Only in Tru64) Run all the tests using the memory leak + naughty
1922memory access tool "Third Degree". The log files will be named
60a57c1c 1923F<perl.3log.testname>.
e018f8be
JH
1924
1925=item test.torture torturetest
1926
1927Run all the usual tests and some extra tests. As of Perl 5.8.0 the
244d9cb7 1928only extra tests are Abigail's JAPHs, F<t/japh/abigail.t>.
e018f8be
JH
1929
1930You can also run the torture test with F<t/harness> by giving
1931C<-torture> argument to F<t/harness>.
1932
1933=item utest ucheck test.utf8 check.utf8
1934
1935Run all the tests with -Mutf8. Not all tests will succeed.
1936
7205a85d
YO
1937(Not available on Win32)
1938
cc0710ff
RGS
1939=item minitest.utf16 test.utf16
1940
1941Runs the tests with UTF-16 encoded scripts, encoded with different
1942versions of this encoding.
1943
1944C<make utest.utf16> runs the test suite with a combination of C<-utf8> and
1945C<-utf16> arguments to F<t/TEST>.
1946
7205a85d
YO
1947(Not available on Win32)
1948
244d9cb7
RGS
1949=item test_harness
1950
1951Run the test suite with the F<t/harness> controlling program, instead of
1952F<t/TEST>. F<t/harness> is more sophisticated, and uses the
1953L<Test::Harness> module, thus using this test target supposes that perl
1954mostly works. The main advantage for our purposes is that it prints a
00bf5cd9
RGS
1955detailed summary of failed tests at the end. Also, unlike F<t/TEST>, it
1956doesn't redirect stderr to stdout.
244d9cb7 1957
7205a85d
YO
1958Note that under Win32 F<t/harness> is always used instead of F<t/TEST>, so
1959there is no special "test_harness" target.
1960
1961Under Win32's "test" target you may use the TEST_SWITCHES and TEST_FILES
1962environment variables to control the behaviour of F<t/harness>. This means
1963you can say
1964
1965 nmake test TEST_FILES="op/*.t"
1966 nmake test TEST_SWITCHES="-torture" TEST_FILES="op/*.t"
1967
a75f557c
JV
1968=item Parallel tests
1969
1970The core distribution can now run its regression tests in parallel on
1971Unix-like platforms. Instead of running C<make test>, set C<TEST_JOBS> in
1972your environment to the number of tests to run in parallel, and run
1973C<make test_harness>. On a Bourne-like shell, this can be done as
1974
1975 TEST_JOBS=3 make test_harness # Run 3 tests in parallel
1976
1977An environment variable is used, rather than parallel make itself, because
1978L<TAP::Harness> needs to be able to schedule individual non-conflicting test
1979scripts itself, and there is no standard interface to C<make> utilities to
1980interact with their job schedulers.
1981
1982Note that currently some test scripts may fail when run in parallel (most
1983notably C<ext/IO/t/io_dir.t>). If necessary run just the failing scripts
1984again sequentially and see if the failures go away.
7205a85d
YO
1985=item test-notty test_notty
1986
1987Sets PERL_SKIP_TTY_TEST to true before running normal test.
1988
244d9cb7
RGS
1989=back
1990
1991=head2 Running tests by hand
1992
1993You can run part of the test suite by hand by using one the following
1994commands from the F<t/> directory :
1995
1996 ./perl -I../lib TEST list-of-.t-files
1997
1998or
1999
2000 ./perl -I../lib harness list-of-.t-files
2001
2002(if you don't specify test scripts, the whole test suite will be run.)
2003
7205a85d
YO
2004=head3 Using t/harness for testing
2005
2006If you use C<harness> for testing you have several command line options
2007available to you. The arguments are as follows, and are in the order
2008that they must appear if used together.
2009
2010 harness -v -torture -re=pattern LIST OF FILES TO TEST
2011 harness -v -torture -re LIST OF PATTERNS TO MATCH
2012
2013If C<LIST OF FILES TO TEST> is omitted the file list is obtained from
07aa3531 2014the manifest. The file list may include shell wildcards which will be
7205a85d
YO
2015expanded out.
2016
2017=over 4
2018
2019=item -v
2020
07aa3531 2021Run the tests under verbose mode so you can see what tests were run,
8550bf48 2022and debug output.
7205a85d
YO
2023
2024=item -torture
2025
2026Run the torture tests as well as the normal set.
2027
2028=item -re=PATTERN
2029
2030Filter the file list so that all the test files run match PATTERN.
2031Note that this form is distinct from the B<-re LIST OF PATTERNS> form below
2032in that it allows the file list to be provided as well.
2033
2034=item -re LIST OF PATTERNS
2035
07aa3531 2036Filter the file list so that all the test files run match
7205a85d
YO
2037/(LIST|OF|PATTERNS)/. Note that with this form the patterns
2038are joined by '|' and you cannot supply a list of files, instead
2039the test files are obtained from the MANIFEST.
2040
2041=back
2042
244d9cb7
RGS
2043You can run an individual test by a command similar to
2044
2045 ./perl -I../lib patho/to/foo.t
2046
2047except that the harnesses set up some environment variables that may
2048affect the execution of the test :
2049
07aa3531 2050=over 4
244d9cb7
RGS
2051
2052=item PERL_CORE=1
2053
2054indicates that we're running this test part of the perl core test suite.
2055This is useful for modules that have a dual life on CPAN.
2056
2057=item PERL_DESTRUCT_LEVEL=2
2058
2059is set to 2 if it isn't set already (see L</PERL_DESTRUCT_LEVEL>)
2060
2061=item PERL
2062
2063(used only by F<t/TEST>) if set, overrides the path to the perl executable
2064that should be used to run the tests (the default being F<./perl>).
2065
2066=item PERL_SKIP_TTY_TEST
2067
2068if set, tells to skip the tests that need a terminal. It's actually set
2069automatically by the Makefile, but can also be forced artificially by
2070running 'make test_notty'.
2071
e018f8be 2072=back
f7e1e956 2073
7cd58830
RGS
2074=head3 Other environment variables that may influence tests
2075
2076=over 4
2077
2078=item PERL_TEST_Net_Ping
2079
2080Setting this variable runs all the Net::Ping modules tests,
2081otherwise some tests that interact with the outside world are skipped.
2082See L<perl58delta>.
2083
2084=item PERL_TEST_NOVREXX
2085
2086Setting this variable skips the vrexx.t tests for OS2::REXX.
2087
2088=item PERL_TEST_NUMCONVERTS
2089
2090This sets a variable in op/numconvert.t.
2091
2092=back
2093
2094See also the documentation for the Test and Test::Harness modules,
2095for more environment variables that affect testing.
2096
d7889f52
JH
2097=head2 Common problems when patching Perl source code
2098
2099Perl source plays by ANSI C89 rules: no C99 (or C++) extensions. In
2100some cases we have to take pre-ANSI requirements into consideration.
2101You don't care about some particular platform having broken Perl?
2102I hear there is still a strong demand for J2EE programmers.
2103
2104=head2 Perl environment problems
2105
2106=over 4
2107
2108=item *
2109
2110Not compiling with threading
2111
2112Compiling with threading (-Duseithreads) completely rewrites
2113the function prototypes of Perl. You better try your changes
0bec6c03 2114with that. Related to this is the difference between "Perl_-less"
d7889f52
JH
2115and "Perl_-ly" APIs, for example:
2116
2117 Perl_sv_setiv(aTHX_ ...);
2118 sv_setiv(...);
2119
ee9468a2
RGS
2120The first one explicitly passes in the context, which is needed for e.g.
2121threaded builds. The second one does that implicitly; do not get them
def4ed7d
JH
2122mixed. If you are not passing in a aTHX_, you will need to do a dTHX
2123(or a dVAR) as the first thing in the function.
d7889f52
JH
2124
2125See L<perlguts/"How multiple interpreters and concurrency are supported">
2126for further discussion about context.
2127
2128=item *
2129
2130Not compiling with -DDEBUGGING
2131
2132The DEBUGGING define exposes more code to the compiler,
0bec6c03 2133therefore more ways for things to go wrong. You should try it.
d7889f52
JH
2134
2135=item *
2136
ee9468a2
RGS
2137Introducing (non-read-only) globals
2138
2139Do not introduce any modifiable globals, truly global or file static.
bc028b6b
JH
2140They are bad form and complicate multithreading and other forms of
2141concurrency. The right way is to introduce them as new interpreter
2142variables, see F<intrpvar.h> (at the very end for binary compatibility).
ee9468a2
RGS
2143
2144Introducing read-only (const) globals is okay, as long as you verify
2145with e.g. C<nm libperl.a|egrep -v ' [TURtr] '> (if your C<nm> has
2146BSD-style output) that the data you added really is read-only.
2147(If it is, it shouldn't show up in the output of that command.)
2148
2149If you want to have static strings, make them constant:
2150
2151 static const char etc[] = "...";
2152
bc028b6b 2153If you want to have arrays of constant strings, note carefully
ee9468a2
RGS
2154the right combination of C<const>s:
2155
2156 static const char * const yippee[] =
2157 {"hi", "ho", "silver"};
2158
bc028b6b
JH
2159There is a way to completely hide any modifiable globals (they are all
2160moved to heap), the compilation setting C<-DPERL_GLOBAL_STRUCT_PRIVATE>.
2161It is not normally used, but can be used for testing, read more
def4ed7d 2162about it in L<perlguts/"Background and PERL_IMPLICIT_CONTEXT">.
bc028b6b 2163
ee9468a2
RGS
2164=item *
2165
d7889f52
JH
2166Not exporting your new function
2167
2168Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any
2169function that is part of the public API (the shared Perl library)
2170to be explicitly marked as exported. See the discussion about
2171F<embed.pl> in L<perlguts>.
2172
2173=item *
2174
2175Exporting your new function
2176
2177The new shiny result of either genuine new functionality or your
2178arduous refactoring is now ready and correctly exported. So what
def4ed7d 2179could possibly go wrong?
d7889f52
JH
2180
2181Maybe simply that your function did not need to be exported in the
2182first place. Perl has a long and not so glorious history of exporting
2183functions that it should not have.
2184
2185If the function is used only inside one source code file, make it
2186static. See the discussion about F<embed.pl> in L<perlguts>.
2187
2188If the function is used across several files, but intended only for
2189Perl's internal use (and this should be the common case), do not
2190export it to the public API. See the discussion about F<embed.pl>
2191in L<perlguts>.
2192
2193=back
2194
5b38b9cd 2195=head2 Portability problems
d7889f52
JH
2196
2197The following are common causes of compilation and/or execution
2198failures, not common to Perl as such. The C FAQ is good bedtime
0bec6c03 2199reading. Please test your changes with as many C compilers and
ac036724 2200platforms as possible; we will, anyway, and it's nice to save
0bec6c03
JH
2201oneself from public embarrassment.
2202
9aaf14db
RGS
2203If using gcc, you can add the C<-std=c89> option which will hopefully
2204catch most of these unportabilities. (However it might also catch
2205incompatibilities in your system's header files.)
d1307786 2206
a8e98a71
JH
2207Use the Configure C<-Dgccansipedantic> flag to enable the gcc
2208C<-ansi -pedantic> flags which enforce stricter ANSI rules.
2209
def4ed7d
JH
2210If using the C<gcc -Wall> note that not all the possible warnings
2211(like C<-Wunitialized>) are given unless you also compile with C<-O>.
2212
2213Note that if using gcc, starting from Perl 5.9.5 the Perl core source
2214code files (the ones at the top level of the source code distribution,
2215but not e.g. the extensions under ext/) are automatically compiled
2216with as many as possible of the C<-std=c89>, C<-ansi>, C<-pedantic>,
2217and a selection of C<-W> flags (see cflags.SH).
27565cb6 2218
0bec6c03 2219Also study L<perlport> carefully to avoid any bad assumptions
def4ed7d 2220about the operating system, filesystems, and so forth.
0bec6c03 2221
606fd33d 2222You may once in a while try a "make microperl" to see whether we
63796a85 2223can still compile Perl with just the bare minimum of interfaces.
606fd33d 2224(See README.micro.)
ee9468a2 2225
0bec6c03 2226Do not assume an operating system indicates a certain compiler.
d7889f52
JH
2227
2228=over 4
2229
2230=item *
2231
2232Casting pointers to integers or casting integers to pointers
2233
2234 void castaway(U8* p)
2235 {
2236 IV i = p;
2237
2238or
2239
2240 void castaway(U8* p)
2241 {
2242 IV i = (IV)p;
2243
ee9468a2 2244Both are bad, and broken, and unportable. Use the PTR2IV()
d7889f52
JH
2245macro that does it right. (Likewise, there are PTR2UV(), PTR2NV(),
2246INT2PTR(), and NUM2PTR().)
2247
2248=item *
2249
0bec6c03
JH
2250Casting between data function pointers and data pointers
2251
d7889f52
JH
2252Technically speaking casting between function pointers and data
2253pointers is unportable and undefined, but practically speaking
2254it seems to work, but you should use the FPTR2DPTR() and DPTR2FPTR()
0bec6c03 2255macros. Sometimes you can also play games with unions.
d7889f52
JH
2256
2257=item *
2258
2259Assuming sizeof(int) == sizeof(long)
2260
2261There are platforms where longs are 64 bits, and platforms where ints
2262are 64 bits, and while we are out to shock you, even platforms where
2263shorts are 64 bits. This is all legal according to the C standard.
2264(In other words, "long long" is not a portable way to specify 64 bits,
2265and "long long" is not even guaranteed to be any wider than "long".)
63796a85
JH
2266
2267Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth.
2268Avoid things like I32 because they are B<not> guaranteed to be
2269I<exactly> 32 bits, they are I<at least> 32 bits, nor are they
2270guaranteed to be B<int> or B<long>. If you really explicitly need
227164-bit variables, use I64 and U64, but only if guarded by HAS_QUAD.
d7889f52
JH
2272
2273=item *
2274
2275Assuming one can dereference any type of pointer for any type of data
2276
2277 char *p = ...;
def4ed7d 2278 long pony = *p; /* BAD */
d7889f52
JH
2279
2280Many platforms, quite rightly so, will give you a core dump instead
2281of a pony if the p happens not be correctly aligned.
2282
2283=item *
2284
2285Lvalue casts
2286
def4ed7d 2287 (int)*p = ...; /* BAD */
d7889f52
JH
2288
2289Simply not portable. Get your lvalue to be of the right type,
27565cb6
JH
2290or maybe use temporary variables, or dirty tricks with unions.
2291
2292=item *
2293
606fd33d
JH
2294Assume B<anything> about structs (especially the ones you
2295don't control, like the ones coming from the system headers)
27565cb6
JH
2296
2297=over 8
2298
2299=item *
2300
2301That a certain field exists in a struct
2302
2303=item *
2304
902821cc 2305That no other fields exist besides the ones you know of
27565cb6
JH
2306
2307=item *
2308
606fd33d 2309That a field is of certain signedness, sizeof, or type
27565cb6
JH
2310
2311=item *
2312
2313That the fields are in a certain order
2314
606fd33d
JH
2315=over 8
2316
27565cb6
JH
2317=item *
2318
606fd33d
JH
2319While C guarantees the ordering specified in the struct definition,
2320between different platforms the definitions might differ
2321
2322=back
27565cb6
JH
2323
2324=item *
2325
606fd33d
JH
2326That the sizeof(struct) or the alignments are the same everywhere
2327
2328=over 8
27565cb6
JH
2329
2330=item *
2331
606fd33d
JH
2332There might be padding bytes between the fields to align the fields -
2333the bytes can be anything
2334
2335=item *
2336
2337Structs are required to be aligned to the maximum alignment required
2338by the fields - which for native types is for usually equivalent to
2339sizeof() of the field
2340
2341=back
27565cb6
JH
2342
2343=back
d7889f52
JH
2344
2345=item *
2346
2bbc8d55
SP
2347Assuming the character set is ASCIIish
2348
2349Perl can compile and run under EBCDIC platforms. See L<perlebcdic>.
2350This is transparent for the most part, but because the character sets
2351differ, you shouldn't use numeric (decimal, octal, nor hex) constants
2352to refer to characters. You can safely say 'A', but not 0x41.
2353You can safely say '\n', but not \012.
2354If a character doesn't have a trivial input form, you can
2355create a #define for it in both C<utfebcdic.h> and C<utf8.h>, so that
2356it resolves to different values depending on the character set being used.
2357(There are three different EBCDIC character sets defined in C<utfebcdic.h>,
2358so it might be best to insert the #define three times in that file.)
2359
2360Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper case
2361alphabetic characters. That is not true in EBCDIC. Nor for 'a' to 'z'.
2362But '0' - '9' is an unbroken range in both systems. Don't assume anything
2363about other ranges.
2364
2365Many of the comments in the existing code ignore the possibility of EBCDIC,
2366and may be wrong therefore, even if the code works.
2367This is actually a tribute to the successful transparent insertion of being
fe749c9a 2368able to handle EBCDIC without having to change pre-existing code.
2bbc8d55
SP
2369
2370UTF-8 and UTF-EBCDIC are two different encodings used to represent Unicode
2371code points as sequences of bytes. Macros
2372with the same names (but different definitions)
2373in C<utf8.h> and C<utfebcdic.h>
fe749c9a
KW
2374are used to allow the calling code to think that there is only one such
2375encoding.
2376This is almost always referred to as C<utf8>, but it means the EBCDIC version
2377as well. Again, comments in the code may well be wrong even if the code itself
2378is right.
2bbc8d55
SP
2379For example, the concept of C<invariant characters> differs between ASCII and
2380EBCDIC.
2381On ASCII platforms, only characters that do not have the high-order
2382bit set (i.e. whose ordinals are strict ASCII, 0 - 127)
2383are invariant, and the documentation and comments in the code
2384may assume that,
2385often referring to something like, say, C<hibit>.
2386The situation differs and is not so simple on EBCDIC machines, but as long as
2387the code itself uses the C<NATIVE_IS_INVARIANT()> macro appropriately, it
2388works, even if the comments are wrong.
2389
2390=item *
2391
2392Assuming the character set is just ASCII
2393
2394ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra
2395characters have different meanings depending on the locale. Absent a locale,
2396currently these extra characters are generally considered to be unassigned,
2397and this has presented some problems.
e1b711da 2398This is being changed starting in 5.12 so that these characters will
2bbc8d55
SP
2399be considered to be Latin-1 (ISO-8859-1).
2400
2401=item *
2402
0bec6c03
JH
2403Mixing #define and #ifdef
2404
2405 #define BURGLE(x) ... \
def4ed7d 2406 #ifdef BURGLE_OLD_STYLE /* BAD */
0bec6c03
JH
2407 ... do it the old way ... \
2408 #else
2409 ... do it the new way ... \
2410 #endif
2411
ee9468a2
RGS
2412You cannot portably "stack" cpp directives. For example in the above
2413you need two separate BURGLE() #defines, one for each #ifdef branch.
2414
2415=item *
2416
2bbc8d55 2417Adding non-comment stuff after #endif or #else
ee9468a2
RGS
2418
2419 #ifdef SNOSH
2420 ...
def4ed7d 2421 #else !SNOSH /* BAD */
ee9468a2 2422 ...
def4ed7d 2423 #endif SNOSH /* BAD */
ee9468a2 2424
def4ed7d
JH
2425The #endif and #else cannot portably have anything non-comment after
2426them. If you want to document what is going (which is a good idea
2427especially if the branches are long), use (C) comments:
ee9468a2
RGS
2428
2429 #ifdef SNOSH
2430 ...
2431 #else /* !SNOSH */
2432 ...
2433 #endif /* SNOSH */
2434
2435The gcc option C<-Wendif-labels> warns about the bad variant
2436(by default on starting from Perl 5.9.4).
0bec6c03
JH
2437
2438=item *
2439
27565cb6
JH
2440Having a comma after the last element of an enum list
2441
2442 enum color {
2443 CERULEAN,
2444 CHARTREUSE,
def4ed7d 2445 CINNABAR, /* BAD */
27565cb6
JH
2446 };
2447
2448is not portable. Leave out the last comma.
2449
2450Also note that whether enums are implicitly morphable to ints
2451varies between compilers, you might need to (int).
2452
2453=item *
2454
d7889f52
JH
2455Using //-comments
2456
def4ed7d 2457 // This function bamfoodles the zorklator. /* BAD */
d7889f52
JH
2458
2459That is C99 or C++. Perl is C89. Using the //-comments is silently
0bec6c03
JH
2460allowed by many C compilers but cranking up the ANSI C89 strictness
2461(which we like to do) causes the compilation to fail.
d7889f52
JH
2462
2463=item *
2464
2465Mixing declarations and code
2466
2467 void zorklator()
2468 {
2469 int n = 3;
def4ed7d 2470 set_zorkmids(n); /* BAD */
d7889f52
JH
2471 int q = 4;
2472
0bec6c03
JH
2473That is C99 or C++. Some C compilers allow that, but you shouldn't.
2474
63796a85
JH
2475The gcc option C<-Wdeclaration-after-statements> scans for such problems
2476(by default on starting from Perl 5.9.4).
2477
0bec6c03
JH
2478=item *
2479
2480Introducing variables inside for()
2481
def4ed7d 2482 for(int i = ...; ...; ...) { /* BAD */
0bec6c03
JH
2483
2484That is C99 or C++. While it would indeed be awfully nice to have that
2485also in C89, to limit the scope of the loop variable, alas, we cannot.
d7889f52
JH
2486
2487=item *
2488
2489Mixing signed char pointers with unsigned char pointers
2490
2491 int foo(char *s) { ... }
2492 ...
2493 unsigned char *t = ...; /* Or U8* t = ... */
def4ed7d 2494 foo(t); /* BAD */
d7889f52
JH
2495
2496While this is legal practice, it is certainly dubious, and downright
2497fatal in at least one platform: for example VMS cc considers this a
def4ed7d
JH
2498fatal error. One cause for people often making this mistake is that a
2499"naked char" and therefore dereferencing a "naked char pointer" have
2500an undefined signedness: it depends on the compiler and the flags of
2501the compiler and the underlying platform whether the result is signed
2502or unsigned. For this very same reason using a 'char' as an array
2503index is bad.
d7889f52
JH
2504
2505=item *
2506
2507Macros that have string constants and their arguments as substrings of
2508the string constants
2509
def4ed7d 2510 #define FOO(n) printf("number = %d\n", n) /* BAD */
d7889f52
JH
2511 FOO(10);
2512
2513Pre-ANSI semantics for that was equivalent to
2514
2515 printf("10umber = %d\10");
2516
0bec6c03
JH
2517which is probably not what you were expecting. Unfortunately at least
2518one reasonably common and modern C compiler does "real backward
63796a85 2519compatibility" here, in AIX that is what still happens even though the
0bec6c03
JH
2520rest of the AIX compiler is very happily C89.
2521
2522=item *
2523
ee9468a2
RGS
2524Using printf formats for non-basic C types
2525
2526 IV i = ...;
def4ed7d 2527 printf("i = %d\n", i); /* BAD */
ee9468a2
RGS
2528
2529While this might by accident work in some platform (where IV happens
2530to be an C<int>), in general it cannot. IV might be something larger.
2531Even worse the situation is with more specific types (defined by Perl's
2532configuration step in F<config.h>):
2533
2534 Uid_t who = ...;
def4ed7d 2535 printf("who = %d\n", who); /* BAD */
ee9468a2
RGS
2536
2537The problem here is that Uid_t might be not only not C<int>-wide
2538but it might also be unsigned, in which case large uids would be
2539printed as negative values.
2540
2541There is no simple solution to this because of printf()'s limited
2542intelligence, but for many types the right format is available as
2543with either 'f' or '_f' suffix, for example:
2544
2545 IVdf /* IV in decimal */
2546 UVxf /* UV is hexadecimal */
2547
2548 printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
2549
2550 Uid_t_f /* Uid_t in decimal */
2551
2552 printf("who = %"Uid_t_f"\n", who);
2553
63796a85
JH
2554Or you can try casting to a "wide enough" type:
2555
2556 printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);
2557
2558Also remember that the C<%p> format really does require a void pointer:
2559
2560 U8* p = ...;
2561 printf("p = %p\n", (void*)p);
2562
ee9468a2
RGS
2563The gcc option C<-Wformat> scans for such problems.
2564
2565=item *
2566
0bec6c03
JH
2567Blindly using variadic macros
2568
63796a85
JH
2569gcc has had them for a while with its own syntax, and C99 brought
2570them with a standardized syntax. Don't use the former, and use
2571the latter only if the HAS_C99_VARIADIC_MACROS is defined.
0bec6c03
JH
2572
2573=item *
2574
2575Blindly passing va_list
2576
2577Not all platforms support passing va_list to further varargs (stdarg)
2578functions. The right thing to do is to copy the va_list using the
2579Perl_va_copy() if the NEED_VA_COPY is defined.
d7889f52 2580
ee9468a2
RGS
2581=item *
2582
e5afc1ae 2583Using gcc statement expressions
63796a85 2584
def4ed7d 2585 val = ({...;...;...}); /* BAD */
63796a85 2586
def4ed7d
JH
2587While a nice extension, it's not portable. The Perl code does
2588admittedly use them if available to gain some extra speed
2589(essentially as a funky form of inlining), but you shouldn't.
63796a85
JH
2590
2591=item *
2592
2bbc8d55 2593Binding together several statements in a macro
63796a85
JH
2594
2595Use the macros STMT_START and STMT_END.
2596
2597 STMT_START {
2598 ...
2599 } STMT_END
2600
2601=item *
2602
ee9468a2
RGS
2603Testing for operating systems or versions when should be testing for features
2604
def4ed7d 2605 #ifdef __FOONIX__ /* BAD */
ee9468a2
RGS
2606 foo = quux();
2607 #endif
2608
2609Unless you know with 100% certainty that quux() is only ever available
2610for the "Foonix" operating system B<and> that is available B<and>
2611correctly working for B<all> past, present, B<and> future versions of
2612"Foonix", the above is very wrong. This is more correct (though still
2613not perfect, because the below is a compile-time check):
2614
2615 #ifdef HAS_QUUX
2616 foo = quux();
2617 #endif
2618
def4ed7d 2619How does the HAS_QUUX become defined where it needs to be? Well, if
e1020413 2620Foonix happens to be Unixy enough to be able to run the Configure
ee9468a2
RGS
2621script, and Configure has been taught about detecting and testing
2622quux(), the HAS_QUUX will be correctly defined. In other platforms,
2623the corresponding configuration step will hopefully do the same.
2624
2625In a pinch, if you cannot wait for Configure to be educated,
2626or if you have a good hunch of where quux() might be available,
2627you can temporarily try the following:
2628
2629 #if (defined(__FOONIX__) || defined(__BARNIX__))
2630 # define HAS_QUUX
2631 #endif
2632
2633 ...
2634
2635 #ifdef HAS_QUUX
2636 foo = quux();
2637 #endif
2638
2639But in any case, try to keep the features and operating systems separate.
2640
d7889f52
JH
2641=back
2642
ad7244db
JH
2643=head2 Problematic System Interfaces
2644
2645=over 4
2646
2647=item *
2648
353c6505 2649malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable
ad7244db
JH
2650allocate at least one byte. (In general you should rarely need to
2651work at this low level, but instead use the various malloc wrappers.)
2652
2653=item *
2654
2655snprintf() - the return type is unportable. Use my_snprintf() instead.
2656
2657=back
2658
d7889f52
JH
2659=head2 Security problems
2660
2661Last but not least, here are various tips for safer coding.
2662
2663=over 4
2664
2665=item *
2666
2667Do not use gets()
2668
2669Or we will publicly ridicule you. Seriously.
2670
2671=item *
2672
d1307786 2673Do not use strcpy() or strcat() or strncpy() or strncat()
d7889f52 2674
d1307786
JH
2675Use my_strlcpy() and my_strlcat() instead: they either use the native
2676implementation, or Perl's own implementation (borrowed from the public
2677domain implementation of INN).
d7889f52
JH
2678
2679=item *
2680
2681Do not use sprintf() or vsprintf()
2682
0bec6c03 2683If you really want just plain byte strings, use my_snprintf()
64d9b66b 2684and my_vsnprintf() instead, which will try to use snprintf() and
0bec6c03
JH
2685vsnprintf() if those safer APIs are available. If you want something
2686fancier than a plain byte string, use SVs and Perl_sv_catpvf().
d7889f52
JH
2687
2688=back
2689
902b9dbf
MF
2690=head1 EXTERNAL TOOLS FOR DEBUGGING PERL
2691
2692Sometimes it helps to use external tools while debugging and
2693testing Perl. This section tries to guide you through using
2694some common testing and debugging tools with Perl. This is
2695meant as a guide to interfacing these tools with Perl, not
2696as any kind of guide to the use of the tools themselves.
2697
a958818a
JH
2698B<NOTE 1>: Running under memory debuggers such as Purify, valgrind, or
2699Third Degree greatly slows down the execution: seconds become minutes,
2700minutes become hours. For example as of Perl 5.8.1, the
2701ext/Encode/t/Unicode.t takes extraordinarily long to complete under
2702e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more
ac036724 2703than six hours, even on a snappy computer. The said test must be
a958818a
JH
2704doing something that is quite unfriendly for memory debuggers. If you
2705don't feel like waiting, that you can simply kill away the perl
2706process.
2707
2708B<NOTE 2>: To minimize the number of memory leak false alarms (see
ac036724 2709L</PERL_DESTRUCT_LEVEL> for more information), you have to set the
2710environment variable PERL_DESTRUCT_LEVEL to 2.
2711
2712For csh-like shells:
a958818a
JH
2713
2714 setenv PERL_DESTRUCT_LEVEL 2
2715
ac036724 2716For Bourne-type shells:
a958818a
JH
2717
2718 PERL_DESTRUCT_LEVEL=2
2719 export PERL_DESTRUCT_LEVEL
2720
ac036724 2721In Unixy environments you can also use the C<env> command:
a958818a
JH
2722
2723 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...
a1b65709 2724
37c0adeb
JH
2725B<NOTE 3>: There are known memory leaks when there are compile-time
2726errors within eval or require, seeing C<S_doeval> in the call stack
2727is a good sign of these. Fixing these leaks is non-trivial,
2728unfortunately, but they must be fixed eventually.
2729
f50e5b73
MH
2730B<NOTE 4>: L<DynaLoader> will not clean up after itself completely
2731unless Perl is built with the Configure option
2732C<-Accflags=-DDL_UNLOAD_ALL_AT_EXIT>.
2733
902b9dbf
MF
2734=head2 Rational Software's Purify
2735
2736Purify is a commercial tool that is helpful in identifying
2737memory overruns, wild pointers, memory leaks and other such
2738badness. Perl must be compiled in a specific way for
2739optimal testing with Purify. Purify is available under
2740Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.
2741
902b9dbf
MF
2742=head2 Purify on Unix
2743
2744On Unix, Purify creates a new Perl binary. To get the most
2745benefit out of Purify, you should create the perl to Purify
2746using:
2747
2748 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
2749 -Uusemymalloc -Dusemultiplicity
2750
2751where these arguments mean:
2752
2753=over 4
2754
2755=item -Accflags=-DPURIFY
2756
2757Disables Perl's arena memory allocation functions, as well as
2758forcing use of memory allocation functions derived from the
2759system malloc.
2760
2761=item -Doptimize='-g'
2762
2763Adds debugging information so that you see the exact source
2764statements where the problem occurs. Without this flag, all
2765you will see is the source filename of where the error occurred.
2766
2767=item -Uusemymalloc
2768
2769Disable Perl's malloc so that Purify can more closely monitor
2770allocations and leaks. Using Perl's malloc will make Purify
2771report most leaks in the "potential" leaks category.
2772
2773=item -Dusemultiplicity
2774
2775Enabling the multiplicity option allows perl to clean up
2776thoroughly when the interpreter shuts down, which reduces the
2777number of bogus leak reports from Purify.
2778
2779=back
2780
2781Once you've compiled a perl suitable for Purify'ing, then you
2782can just:
2783
07aa3531 2784 make pureperl
902b9dbf
MF
2785
2786which creates a binary named 'pureperl' that has been Purify'ed.
2787This binary is used in place of the standard 'perl' binary
2788when you want to debug Perl memory problems.
2789
2790As an example, to show any memory leaks produced during the
2791standard Perl testset you would create and run the Purify'ed
2792perl as:
2793
2794 make pureperl
2795 cd t
07aa3531 2796 ../pureperl -I../lib harness
902b9dbf
MF
2797
2798which would run Perl on test.pl and report any memory problems.
2799
2800Purify outputs messages in "Viewer" windows by default. If
2801you don't have a windowing environment or if you simply
2802want the Purify output to unobtrusively go to a log file
2803instead of to the interactive window, use these following
2804options to output to the log file "perl.log":
2805
2806 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
2807 -log-file=perl.log -append-logfile=yes"
2808
2809If you plan to use the "Viewer" windows, then you only need this option:
2810
2811 setenv PURIFYOPTIONS "-chain-length=25"
2812
c406981e
JH
2813In Bourne-type shells:
2814
98631ff8
JL
2815 PURIFYOPTIONS="..."
2816 export PURIFYOPTIONS
c406981e
JH
2817
2818or if you have the "env" utility:
2819
98631ff8 2820 env PURIFYOPTIONS="..." ../pureperl ...
c406981e 2821
902b9dbf
MF
2822=head2 Purify on NT
2823
2824Purify on Windows NT instruments the Perl binary 'perl.exe'
2825on the fly. There are several options in the makefile you
2826should change to get the most use out of Purify:
2827
2828=over 4
2829
2830=item DEFINES
2831
2832You should add -DPURIFY to the DEFINES line so the DEFINES
2833line looks something like:
2834
195c30ce 2835 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1
902b9dbf
MF
2836
2837to disable Perl's arena memory allocation functions, as
2838well as to force use of memory allocation functions derived
2839from the system malloc.
2840
2841=item USE_MULTI = define
2842
2843Enabling the multiplicity option allows perl to clean up
2844thoroughly when the interpreter shuts down, which reduces the
2845number of bogus leak reports from Purify.
2846
2847=item #PERL_MALLOC = define
2848
2849Disable Perl's malloc so that Purify can more closely monitor
2850allocations and leaks. Using Perl's malloc will make Purify
2851report most leaks in the "potential" leaks category.
2852
2853=item CFG = Debug
2854
2855Adds debugging information so that you see the exact source
2856statements where the problem occurs. Without this flag, all
2857you will see is the source filename of where the error occurred.
2858
2859=back
2860
2861As an example, to show any memory leaks produced during the
2862standard Perl testset you would create and run Purify as:
2863
2864 cd win32
2865 make
2866 cd ../t
07aa3531 2867 purify ../perl -I../lib harness
902b9dbf
MF
2868
2869which would instrument Perl in memory, run Perl on test.pl,
2870then finally report any memory problems.
2871
7a834142
JH
2872=head2 valgrind
2873
2874The excellent valgrind tool can be used to find out both memory leaks
9df8f87f
LB
2875and illegal memory accesses. As of version 3.3.0, Valgrind only
2876supports Linux on x86, x86-64 and PowerPC. The special "test.valgrind"
2877target can be used to run the tests under valgrind. Found errors
2878and memory leaks are logged in files named F<testfile.valgrind>.
07aa3531
JC
2879
2880Valgrind also provides a cachegrind tool, invoked on perl as:
2881
038c294a 2882 VG_OPTS=--tool=cachegrind make test.valgrind
d44161bf
MHM
2883
2884As system libraries (most notably glibc) are also triggering errors,
2885valgrind allows to suppress such errors using suppression files. The
2886default suppression file that comes with valgrind already catches a lot
2887of them. Some additional suppressions are defined in F<t/perl.supp>.
7a834142
JH
2888
2889To get valgrind and for more information see
2890
2891 http://developer.kde.org/~sewardj/
2892
f134cc4e 2893=head2 Compaq's/Digital's/HP's Third Degree
09187cb1
JH
2894
2895Third Degree is a tool for memory leak detection and memory access checks.
2896It is one of the many tools in the ATOM toolkit. The toolkit is only
2897available on Tru64 (formerly known as Digital UNIX formerly known as
2898DEC OSF/1).
2899
2900When building Perl, you must first run Configure with -Doptimize=-g
2901and -Uusemymalloc flags, after that you can use the make targets
51a35ef1
JH
2902"perl.third" and "test.third". (What is required is that Perl must be
2903compiled using the C<-g> flag, you may need to re-Configure.)
09187cb1 2904
64cea5fd 2905The short story is that with "atom" you can instrument the Perl
83f0ef60 2906executable to create a new executable called F<perl.third>. When the
4ae3d70a 2907instrumented executable is run, it creates a log of dubious memory
83f0ef60 2908traffic in file called F<perl.3log>. See the manual pages of atom and
4ae3d70a
JH
2909third for more information. The most extensive Third Degree
2910documentation is available in the Compaq "Tru64 UNIX Programmer's
2911Guide", chapter "Debugging Programs with Third Degree".
64cea5fd 2912
9c54ecba 2913The "test.third" leaves a lot of files named F<foo_bar.3log> in the t/
64cea5fd
JH
2914subdirectory. There is a problem with these files: Third Degree is so
2915effective that it finds problems also in the system libraries.
9c54ecba
JH
2916Therefore you should used the Porting/thirdclean script to cleanup
2917the F<*.3log> files.
64cea5fd
JH
2918
2919There are also leaks that for given certain definition of a leak,
2920aren't. See L</PERL_DESTRUCT_LEVEL> for more information.
2921
2922=head2 PERL_DESTRUCT_LEVEL
2923
a958818a
JH
2924If you want to run any of the tests yourself manually using e.g.
2925valgrind, or the pureperl or perl.third executables, please note that
2926by default perl B<does not> explicitly cleanup all the memory it has
2927allocated (such as global memory arenas) but instead lets the exit()
2928of the whole program "take care" of such allocations, also known as
2929"global destruction of objects".
64cea5fd
JH
2930
2931There is a way to tell perl to do complete cleanup: set the
2932environment variable PERL_DESTRUCT_LEVEL to a non-zero value.
2933The t/TEST wrapper does set this to 2, and this is what you
2934need to do too, if you don't want to see the "global leaks":
1f56d61a 2935For example, for "third-degreed" Perl:
64cea5fd 2936
1f56d61a 2937 env PERL_DESTRUCT_LEVEL=2 ./perl.third -Ilib t/foo/bar.t
09187cb1 2938
414f2397
RGS
2939(Note: the mod_perl apache module uses also this environment variable
2940for its own purposes and extended its semantics. Refer to the mod_perl
287a822c
RGS
2941documentation for more information. Also, spawned threads do the
2942equivalent of setting this variable to the value 1.)
5a6c59ef
DM
2943
2944If, at the end of a run you get the message I<N scalars leaked>, you can
fd0854ff
DM
2945recompile with C<-DDEBUG_LEAKING_SCALARS>, which will cause the addresses
2946of all those leaked SVs to be dumped along with details as to where each
2947SV was originally allocated. This information is also displayed by
2948Devel::Peek. Note that the extra details recorded with each SV increases
2949memory usage, so it shouldn't be used in production environments. It also
2950converts C<new_SV()> from a macro into a real function, so you can use
2951your favourite debugger to discover where those pesky SVs were allocated.
414f2397 2952
d7a2c63c
MHM
2953If you see that you're leaking memory at runtime, but neither valgrind
2954nor C<-DDEBUG_LEAKING_SCALARS> will find anything, you're probably
2955leaking SVs that are still reachable and will be properly cleaned up
2956during destruction of the interpreter. In such cases, using the C<-Dm>
2957switch can point you to the source of the leak. If the executable was
2958built with C<-DDEBUG_LEAKING_SCALARS>, C<-Dm> will output SV allocations
2959in addition to memory allocations. Each SV allocation has a distinct
2960serial number that will be written on creation and destruction of the SV.
2961So if you're executing the leaking code in a loop, you need to look for
2962SVs that are created, but never destroyed between each cycle. If such an
2963SV is found, set a conditional breakpoint within C<new_SV()> and make it
2964break only when C<PL_sv_serial> is equal to the serial number of the
2965leaking SV. Then you will catch the interpreter in exactly the state
2966where the leaking SV is allocated, which is sufficient in many cases to
2967find the source of the leak.
2968
2969As C<-Dm> is using the PerlIO layer for output, it will by itself
2970allocate quite a bunch of SVs, which are hidden to avoid recursion.
2971You can bypass the PerlIO layer if you use the SV logging provided
2972by C<-DPERL_MEM_LOG> instead.
2973
46c6c7e2
JH
2974=head2 PERL_MEM_LOG
2975
10a879f5
JC
2976If compiled with C<-DPERL_MEM_LOG>, both memory and SV allocations go
2977through logging functions, which is handy for breakpoint setting.
2978
2979Unless C<-DPERL_MEM_LOG_NOIMPL> is also compiled, the logging
2e5b5004 2980functions read $ENV{PERL_MEM_LOG} to determine whether to log the
10a879f5
JC
2981event, and if so how:
2982
2e5b5004
RGS
2983 $ENV{PERL_MEM_LOG} =~ /m/ Log all memory ops
2984 $ENV{PERL_MEM_LOG} =~ /s/ Log all SV ops
2985 $ENV{PERL_MEM_LOG} =~ /t/ include timestamp in Log
2986 $ENV{PERL_MEM_LOG} =~ /^(\d+)/ write to FD given (default is 2)
10a879f5
JC
2987
2988Memory logging is somewhat similar to C<-Dm> but is independent of
2989C<-DDEBUGGING>, and at a higher level; all uses of Newx(), Renew(),
2e5b5004 2990and Safefree() are logged with the caller's source code file and line
10a879f5
JC
2991number (and C function name, if supported by the C compiler). In
2992contrast, C<-Dm> is directly at the point of C<malloc()>. SV logging
2993is similar.
2994
2995Since the logging doesn't use PerlIO, all SV allocations are logged
2996and no extra SV allocations are introduced by enabling the logging.
2997If compiled with C<-DDEBUG_LEAKING_SCALARS>, the serial number for
2998each SV allocation is also logged.
d7a2c63c 2999
51a35ef1
JH
3000=head2 Profiling
3001
3002Depending on your platform there are various of profiling Perl.
3003
3004There are two commonly used techniques of profiling executables:
10f58044 3005I<statistical time-sampling> and I<basic-block counting>.
51a35ef1
JH
3006
3007The first method takes periodically samples of the CPU program
3008counter, and since the program counter can be correlated with the code
3009generated for functions, we get a statistical view of in which
3010functions the program is spending its time. The caveats are that very
3011small/fast functions have lower probability of showing up in the
3012profile, and that periodically interrupting the program (this is
3013usually done rather frequently, in the scale of milliseconds) imposes
3014an additional overhead that may skew the results. The first problem
3015can be alleviated by running the code for longer (in general this is a
3016good idea for profiling), the second problem is usually kept in guard
3017by the profiling tools themselves.
3018
10f58044 3019The second method divides up the generated code into I<basic blocks>.
51a35ef1
JH
3020Basic blocks are sections of code that are entered only in the
3021beginning and exited only at the end. For example, a conditional jump
3022starts a basic block. Basic block profiling usually works by
10f58044 3023I<instrumenting> the code by adding I<enter basic block #nnnn>
51a35ef1
JH
3024book-keeping code to the generated code. During the execution of the
3025code the basic block counters are then updated appropriately. The
3026caveat is that the added extra code can skew the results: again, the
3027profiling tools usually try to factor their own effects out of the
3028results.
3029
83f0ef60
JH
3030=head2 Gprof Profiling
3031
e1020413 3032gprof is a profiling tool available in many Unix platforms,
51a35ef1 3033it uses F<statistical time-sampling>.
83f0ef60
JH
3034
3035You can build a profiled version of perl called "perl.gprof" by
51a35ef1
JH
3036invoking the make target "perl.gprof" (What is required is that Perl
3037must be compiled using the C<-pg> flag, you may need to re-Configure).
3038Running the profiled version of Perl will create an output file called
3039F<gmon.out> is created which contains the profiling data collected
3040during the execution.
83f0ef60
JH
3041
3042The gprof tool can then display the collected data in various ways.
3043Usually gprof understands the following options:
3044
3045=over 4
3046
3047=item -a
3048
3049Suppress statically defined functions from the profile.
3050
3051=item -b
3052
3053Suppress the verbose descriptions in the profile.
3054
3055=item -e routine
3056
3057Exclude the given routine and its descendants from the profile.
3058
3059=item -f routine
3060
3061Display only the given routine and its descendants in the profile.
3062
3063=item -s
3064
3065Generate a summary file called F<gmon.sum> which then may be given
3066to subsequent gprof runs to accumulate data over several runs.
3067
3068=item -z
3069
3070Display routines that have zero usage.
3071
3072=back
3073
3074For more detailed explanation of the available commands and output
3075formats, see your own local documentation of gprof.
3076
038c294a 3077quick hint:
07aa3531 3078
289d61c2
JL
3079 $ sh Configure -des -Dusedevel -Doptimize='-pg' && make perl.gprof
3080 $ ./perl.gprof someprog # creates gmon.out in current directory
3081 $ gprof ./perl.gprof > out
07aa3531
JC
3082 $ view out
3083
51a35ef1
JH
3084=head2 GCC gcov Profiling
3085
10f58044 3086Starting from GCC 3.0 I<basic block profiling> is officially available
51a35ef1
JH
3087for the GNU CC.
3088
3089You can build a profiled version of perl called F<perl.gcov> by
3090invoking the make target "perl.gcov" (what is required that Perl must
3091be compiled using gcc with the flags C<-fprofile-arcs
3092-ftest-coverage>, you may need to re-Configure).
3093
3094Running the profiled version of Perl will cause profile output to be
3095generated. For each source file an accompanying ".da" file will be
3096created.
3097
3098To display the results you use the "gcov" utility (which should
3099be installed if you have gcc 3.0 or newer installed). F<gcov> is
3100run on source code files, like this
3101
3102 gcov sv.c
3103
3104which will cause F<sv.c.gcov> to be created. The F<.gcov> files
3105contain the source code annotated with relative frequencies of
3106execution indicated by "#" markers.
3107
3108Useful options of F<gcov> include C<-b> which will summarise the
3109basic block, branch, and function call coverage, and C<-c> which
3110instead of relative frequencies will use the actual counts. For
3111more information on the use of F<gcov> and basic block profiling
3112with gcc, see the latest GNU CC manual, as of GCC 3.0 see
3113
3114 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html
3115
3116and its section titled "8. gcov: a Test Coverage Program"
3117
3118 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132
3119
07aa3531
JC
3120quick hint:
3121
27837272
ABR
3122 $ sh Configure -des -Dusedevel -Doptimize='-g' \
3123 -Accflags='-fprofile-arcs -ftest-coverage' \
07aa3531
JC
3124 -Aldflags='-fprofile-arcs -ftest-coverage' && make perl.gcov
3125 $ rm -f regexec.c.gcov regexec.gcda
3126 $ ./perl.gcov
3127 $ gcov regexec.c
3128 $ view regexec.c.gcov
3129
4ae3d70a
JH
3130=head2 Pixie Profiling
3131
51a35ef1
JH
3132Pixie is a profiling tool available on IRIX and Tru64 (aka Digital
3133UNIX aka DEC OSF/1) platforms. Pixie does its profiling using
10f58044 3134I<basic-block counting>.
4ae3d70a 3135
83f0ef60 3136You can build a profiled version of perl called F<perl.pixie> by
51a35ef1
JH
3137invoking the make target "perl.pixie" (what is required is that Perl
3138must be compiled using the C<-g> flag, you may need to re-Configure).
3139
3140In Tru64 a file called F<perl.Addrs> will also be silently created,
3141this file contains the addresses of the basic blocks. Running the
3142profiled version of Perl will create a new file called "perl.Counts"
3143which contains the counts for the basic block for that particular
3144program execution.
4ae3d70a 3145
51a35ef1 3146To display the results you use the F<prof> utility. The exact
4ae3d70a
JH
3147incantation depends on your operating system, "prof perl.Counts" in
3148IRIX, and "prof -pixie -all -L. perl" in Tru64.
3149
6c41479b
JH
3150In IRIX the following prof options are available:
3151
3152=over 4
3153
3154=item -h
3155
3156Reports the most heavily used lines in descending order of use.
6e36760b 3157Useful for finding the hotspot lines.
6c41479b
JH
3158
3159=item -l
3160
3161Groups lines by procedure, with procedures sorted in descending order of use.
3162Within a procedure, lines are listed in source order.
6e36760b 3163Useful for finding the hotspots of procedures.
6c41479b
JH
3164
3165=back
3166
3167In Tru64 the following options are available:
3168
3169=over 4
3170
3958b146 3171=item -p[rocedures]
6c41479b 3172
3958b146 3173Procedures sorted in descending order by the number of cycles executed
6e36760b 3174in each procedure. Useful for finding the hotspot procedures.
6c41479b
JH
3175(This is the default option.)
3176
24000d2f 3177=item -h[eavy]
6c41479b 3178
6e36760b
JH
3179Lines sorted in descending order by the number of cycles executed in
3180each line. Useful for finding the hotspot lines.
6c41479b 3181
24000d2f 3182=item -i[nvocations]
6c41479b 3183
6e36760b
JH
3184The called procedures are sorted in descending order by number of calls
3185made to the procedures. Useful for finding the most used procedures.
6c41479b 3186
24000d2f 3187=item -l[ines]
6c41479b
JH
3188
3189Grouped by procedure, sorted by cycles executed per procedure.
6e36760b 3190Useful for finding the hotspots of procedures.
6c41479b
JH
3191
3192=item -testcoverage
3193
3194The compiler emitted code for these lines, but the code was unexecuted.
3195
24000d2f 3196=item -z[ero]
6c41479b
JH
3197
3198Unexecuted procedures.
3199
aa500c9e 3200=back
6c41479b
JH
3201
3202For further information, see your system's manual pages for pixie and prof.
4ae3d70a 3203
b8ddf6b3
SB
3204=head2 Miscellaneous tricks
3205
3206=over 4
3207
3208=item *
3209
cc177e1a 3210Those debugging perl with the DDD frontend over gdb may find the
b8ddf6b3
SB
3211following useful:
3212
3213You can extend the data conversion shortcuts menu, so for example you
3214can display an SV's IV value with one click, without doing any typing.
3215To do that simply edit ~/.ddd/init file and add after:
3216
3217 ! Display shortcuts.
3218 Ddd*gdbDisplayShortcuts: \
3219 /t () // Convert to Bin\n\
3220 /d () // Convert to Dec\n\
3221 /x () // Convert to Hex\n\
3222 /o () // Convert to Oct(\n\
3223
3224the following two lines:
3225
3226 ((XPV*) (())->sv_any )->xpv_pv // 2pvx\n\
3227 ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx
3228
3229so now you can do ivx and pvx lookups or you can plug there the
3230sv_peek "conversion":
3231
3232 Perl_sv_peek(my_perl, (SV*)()) // sv_peek
3233
3234(The my_perl is for threaded builds.)
3235Just remember that every line, but the last one, should end with \n\
3236
3237Alternatively edit the init file interactively via:
32383rd mouse button -> New Display -> Edit Menu
3239
3240Note: you can define up to 20 conversion shortcuts in the gdb
3241section.
3242
9965345d
JH
3243=item *
3244
7e337ee0
JH
3245If you see in a debugger a memory area mysteriously full of 0xABABABAB
3246or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros,
3247see L<perlclib>.
9965345d 3248
f1fac472
NC
3249=item *
3250
3251Under ithreads the optree is read only. If you want to enforce this, to check
3252for write accesses from buggy code, compile with C<-DPL_OP_SLAB_ALLOC> to
3253enable the OP slab allocator and C<-DPERL_DEBUG_READONLY_OPS> to enable code
3254that allocates op memory via C<mmap>, and sets it read-only at run time.
3255Any write access to an op results in a C<SIGBUS> and abort.
3256
3257This code is intended for development only, and may not be portable even to
3258all Unix variants. Also, it is an 80% solution, in that it isn't able to make
3259all ops read only. Specifically it
3260
3261=over
3262
3263=item 1
3264
3265Only sets read-only on all slabs of ops at C<CHECK> time, hence ops allocated
3266later via C<require> or C<eval> will be re-write
3267
3268=item 2
3269
3270Turns an entire slab of ops read-write if the refcount of any op in the slab
3271needs to be decreased.
3272
3273=item 3
3274
3275Turns an entire slab of ops read-write if any op from the slab is freed.
3276
b8ddf6b3
SB
3277=back
3278
f1fac472
NC
3279It's not possible to turn the slabs to read-only after an action requiring
3280read-write access, as either can happen during op tree building time, so
3281there may still be legitimate write access.
3282
3283However, as an 80% solution it is still effective, as currently it catches
3284a write access during the generation of F<Config.pm>, which means that we
3285can't yet build F<perl> with this enabled.
3286
3287=back
3288
3289
955fec6b 3290=head1 CONCLUSION
a422fd2d 3291
955fec6b
JH
3292We've had a brief look around the Perl source, how to maintain quality
3293of the source code, an overview of the stages F<perl> goes through
3294when it's running your code, how to use debuggers to poke at the Perl
3295guts, and finally how to analyse the execution of Perl. We took a very
3296simple problem and demonstrated how to solve it fully - with
3297documentation, regression tests, and finally a patch for submission to
3298p5p. Finally, we talked about how to use external tools to debug and
3299test Perl.
a422fd2d
SC
3300
3301I'd now suggest you read over those references again, and then, as soon
3302as possible, get your hands dirty. The best way to learn is by doing,
07aa3531 3303so:
a422fd2d
SC
3304
3305=over 3
3306
3307=item *
3308
3309Subscribe to perl5-porters, follow the patches and try and understand
3310them; don't be afraid to ask if there's a portion you're not clear on -
3311who knows, you may unearth a bug in the patch...
3312
3313=item *
3314
3315Keep up to date with the bleeding edge Perl distributions and get
3316familiar with the changes. Try and get an idea of what areas people are
3317working on and the changes they're making.
3318
3319=item *
3320
3e148164 3321Do read the README associated with your operating system, e.g. README.aix
a1f349fd
MB
3322on the IBM AIX OS. Don't hesitate to supply patches to that README if
3323you find anything missing or changed over a new OS release.
3324
3325=item *
3326
a422fd2d
SC
3327Find an area of Perl that seems interesting to you, and see if you can
3328work out how it works. Scan through the source, and step over it in the
3329debugger. Play, poke, investigate, fiddle! You'll probably get to
3330understand not just your chosen area but a much wider range of F<perl>'s
3331activity as well, and probably sooner than you'd think.
3332
3333=back
3334
3335=over 3
3336
3337=item I<The Road goes ever on and on, down from the door where it began.>
3338
3339=back
3340
64d9b66b 3341If you can do these things, you've started on the long road to Perl porting.
a422fd2d
SC
3342Thanks for wanting to help make Perl better - and happy hacking!
3343
4ac71550
TC
3344=head2 Metaphoric Quotations
3345
3346If you recognized the quote about the Road above, you're in luck.
3347
3348Most software projects begin each file with a literal description of each
3349file's purpose. Perl instead begins each with a literary allusion to that
3350file's purpose.
3351
3352Like chapters in many books, all top-level Perl source files (along with a
3353few others here and there) begin with an epigramic inscription that alludes,
3354indirectly and metaphorically, to the material you're about to read.
3355
3356Quotations are taken from writings of J.R.R Tolkien pertaining to his
3357Legendarium, almost always from I<The Lord of the Rings>. Chapters and
3358page numbers are given using the following editions:
3359
3360=over 4
3361
3362=item *
3363
3364I<The Hobbit>, by J.R.R. Tolkien. The hardcover, 70th-anniversary
3365edition of 2007 was used, published in the UK by Harper Collins Publishers
3366and in the US by the Houghton Mifflin Company.
3367
3368=item *
3369
3370I<The Lord of the Rings>, by J.R.R. Tolkien. The hardcover,
337150th-anniversary edition of 2004 was used, published in the UK by Harper
3372Collins Publishers and in the US by the Houghton Mifflin Company.
3373
3374=item *
3375
3376I<The Lays of Beleriand>, by J.R.R. Tolkien and published posthumously by his
3377son and literary executor, C.J.R. Tolkien, being the 3rd of the 12 volumes
3378in Christopher's mammoth I<History of Middle Earth>. Page numbers derive
3379from the hardcover edition, first published in 1983 by George Allen &
3380Unwin; no page numbers changed for the special 3-volume omnibus edition of
33812002 or the various trade-paper editions, all again now by Harper Collins
3382or Houghton Mifflin.
3383
3384=back
3385
3386Other JRRT books fair game for quotes would thus include I<The Adventures of
3387Tom Bombadil>, I<The Silmarillion>, I<Unfinished Tales>, and I<The Tale of
3388the Children of Hurin>, all but the first posthumously assembled by CJRT.
3389But I<The Lord of the Rings> itself is perfectly fine and probably best to
3390quote from, provided you can find a suitable quote there.
3391
3392So if you were to supply a new, complete, top-level source file to add to
3393Perl, you should conform to this peculiar practice by yourself selecting an
3394appropriate quotation from Tolkien, retaining the original spelling and
3395punctuation and using the same format the rest of the quotes are in.
3396Indirect and oblique is just fine; remember, it's a metaphor, so being meta
3397is, after all, what it's for.
3398
e8cd7eae
GS
3399=head1 AUTHOR
3400
3401This document was written by Nathan Torkington, and is maintained by
3402the perl5-porters mailing list.
4ac71550 3403
b16c2e4a
RGS
3404=head1 SEE ALSO
3405
3406L<perlrepository>