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 17A searchable archive of the list is at either: 18 19 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/ 20 21or 22 23 http://archive.develooper.com/perl5-porters@perl.org/ 24 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 36in what does and does not change in the Perl language. Various 37releases of Perl are shepherded by a "pumpking", a porter 38responsible for gathering patches, deciding on a patch-by-patch, 39feature-by-feature basis what will and will not go into the release. 40For instance, Gurusamy Sarathy was the pumpking for the 5.6 release of 41Perl, and Jarkko Hietaniemi was the pumpking for the 5.8 release, and 42Rafael Garcia-Suarez holds the pumpking crown for the 5.10 release. 43 44In addition, various people are pumpkings for different things. For 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. 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 132implementing a "delayed evaluation" feature, the porters are waiting 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 159unlikely that non-portable additions to the Perl language will be 160accepted. 161 162=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 169confident that his/her hard work put into the patch won't be accidentally 170thrown away by someone in the future? 171 172=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 177always a good idea. 178 179=item Is there another way to do it? 180 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 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. 190 191=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 196request might be. This ties into "Will it be useful?", as the fact 197that someone took the time to make the patch demonstrates a strong 198desire for the feature. 199 200=back 201 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. 206 207=head2 Keeping in sync 208 209The source code to the Perl interpreter, in its different versions, is 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. 213 214How to clone and use the git perl repository is described in L<perlrepository>. 215 216You can also choose to use rsync to get a copy of the current source tree 217for the bleadperl branch and all maintenance branches: 218 219 $ 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 . 225 226(Add the C<--delete> option to remove leftover files) 227 228To get a full list of the available sync points: 229 230 $ rsync perl5.git.perl.org:: 231 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 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. 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. 246 247=head2 Perlbug administration 248 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 252your sticky mitts on: 253 254 http://bugs.perl.org/ 255 256To email the bug system administrators: 257 258 "perlbug-admin" <perlbug-admin@perl.org> 259 260=head2 Submitting patches 261 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. 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.) 269 270If changes are accepted, they are applied to the development branch. Then 271the maintenance pumpking decides which of those patches is to be 272backported to the maint branch. Only patches that survive the heat of the 273development branch get applied to maintenance versions. 274 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. 279 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 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 288I<perlbug@perl.org> or I<perlbug@perl.com>). Reporting bugs through 289I<perlbug> feeds into the automated bug-tracking system, access to 290which is provided through the web at http://rt.perl.org/rt3/ . It 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 296The CPAN testers ( http://testers.cpan.org/ ) are a group of 297volunteers who test CPAN modules on a variety of platforms. Perl 298Smokers ( http://www.nntp.perl.org/group/perl.daily-build and 299http://www.nntp.perl.org/group/perl.daily-build.reports/ ) 300automatically test Perl source releases on platforms with various 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 304testing CPAN modules visit L<http://search.cpan.org/dist/CPANPLUS-YACSmoke/> 305or L<http://search.cpan.org/dist/minismokebox/> or 306L<http://search.cpan.org/dist/CPAN-Reporter/>. 307 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 314mailing list, send mail to I<perl5-porters-subscribe@perl.org>. To 315unsubscribe, send mail to I<perl5-porters-unsubscribe@perl.org>. 316 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 329Gisle Aas's "illustrated perlguts", also known as I<illguts>, has very 330helpful pictures: 331 332L<http://search.cpan.org/dist/illguts/> 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 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. 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 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. 377 378 ext/ is for XS extensions, and modules with special Makefile.PL requirements, which exist in the core only. 379 380 cpan/ is for dual-life modules, where the CPAN module is canonical (should be patched first). 381 382 dist/ is for dual-life modules, where the blead source is canonical. 383 384For some dual-life modules it has not been discussed if the CPAN version or the 385blead source is canonical. Until that is done, those modules should be in 386F<cpan/>. 387 388=item Tests 389 390There are tests for nearly all the modules, built-ins and major bits 391of functionality. Test files all have a .t suffix. Module tests live 392in the F<lib/> and F<ext/> directories next to the module being 393tested. Others live in F<t/>. See L<Writing a test> 394 395=item Documentation 396 397Documentation maintenance includes looking after everything in the 398F<pod/> directory, (as well as contributing new documentation) and 399the documentation to the modules in core. 400 401=item Configure 402 403The Configure process is the way we make Perl portable across the 404myriad of operating systems it supports. Responsibility for the 405Configure, build and installation process, as well as the overall 406portability of the core code rests with the Configure pumpkin - 407others help out with individual operating systems. 408 409The three files that fall under his/her responsibility are Configure, 410config_h.SH, and Porting/Glossary (and a whole bunch of small related 411files that are less important here). The Configure pumpkin decides how 412patches to these are dealt with. Currently, the Configure pumpkin will 413accept patches in most common formats, even directly to these files. 414Other committers are allowed to commit to these files under the strict 415condition that they will inform the Configure pumpkin, either on IRC 416(if he/she happens to be around) or through (personal) e-mail. 417 418The files involved are the operating system directories, (F<win32/>, 419F<os2/>, F<vms/> and so on) the shell scripts which generate F<config.h> 420and F<Makefile>, as well as the metaconfig files which generate 421F<Configure>. (metaconfig isn't included in the core distribution.) 422 423See http://perl5.git.perl.org/metaconfig.git/blob/HEAD:/README for a 424description of the full process involved. 425 426=item Interpreter 427 428And of course, there's the core of the Perl interpreter itself. Let's 429have a look at that in a little more detail. 430 431=back 432 433Before we leave looking at the layout, though, don't forget that 434F<MANIFEST> contains not only the file names in the Perl distribution, 435but short descriptions of what's in them, too. For an overview of the 436important files, try this: 437 438 perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST 439 440=head2 Elements of the interpreter 441 442The work of the interpreter has two main stages: compiling the code 443into the internal representation, or bytecode, and then executing it. 444L<perlguts/Compiled code> explains exactly how the compilation stage 445happens. 446 447Here is a short breakdown of perl's operation: 448 449=over 3 450 451=item Startup 452 453The action begins in F<perlmain.c>. (or F<miniperlmain.c> for miniperl) 454This is very high-level code, enough to fit on a single screen, and it 455resembles the code found in L<perlembed>; most of the real action takes 456place in F<perl.c> 457 458F<perlmain.c> is generated by L<writemain> from F<miniperlmain.c> at 459make time, so you should make perl to follow this along. 460 461First, F<perlmain.c> allocates some memory and constructs a Perl 462interpreter, along these lines: 463 464 1 PERL_SYS_INIT3(&argc,&argv,&env); 465 2 466 3 if (!PL_do_undump) { 467 4 my_perl = perl_alloc(); 468 5 if (!my_perl) 469 6 exit(1); 470 7 perl_construct(my_perl); 471 8 PL_perl_destruct_level = 0; 472 9 } 473 474Line 1 is a macro, and its definition is dependent on your operating 475system. Line 3 references C<PL_do_undump>, a global variable - all 476global variables in Perl start with C<PL_>. This tells you whether the 477current running program was created with the C<-u> flag to perl and then 478F<undump>, which means it's going to be false in any sane context. 479 480Line 4 calls a function in F<perl.c> to allocate memory for a Perl 481interpreter. It's quite a simple function, and the guts of it looks like 482this: 483 484 my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter)); 485 486Here you see an example of Perl's system abstraction, which we'll see 487later: C<PerlMem_malloc> is either your system's C<malloc>, or Perl's 488own C<malloc> as defined in F<malloc.c> if you selected that option at 489configure time. 490 491Next, in line 7, we construct the interpreter using perl_construct, 492also in F<perl.c>; this sets up all the special variables that Perl 493needs, the stacks, and so on. 494 495Now we pass Perl the command line options, and tell it to go: 496 497 exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL); 498 if (!exitstatus) 499 perl_run(my_perl); 500 501 exitstatus = perl_destruct(my_perl); 502 503 perl_free(my_perl); 504 505C<perl_parse> is actually a wrapper around C<S_parse_body>, as defined 506in F<perl.c>, which processes the command line options, sets up any 507statically linked XS modules, opens the program and calls C<yyparse> to 508parse it. 509 510=item Parsing 511 512The aim of this stage is to take the Perl source, and turn it into an op 513tree. We'll see what one of those looks like later. Strictly speaking, 514there's three things going on here. 515 516C<yyparse>, the parser, lives in F<perly.c>, although you're better off 517reading the original YACC input in F<perly.y>. (Yes, Virginia, there 518B<is> a YACC grammar for Perl!) The job of the parser is to take your 519code and "understand" it, splitting it into sentences, deciding which 520operands go with which operators and so on. 521 522The parser is nobly assisted by the lexer, which chunks up your input 523into tokens, and decides what type of thing each token is: a variable 524name, an operator, a bareword, a subroutine, a core function, and so on. 525The main point of entry to the lexer is C<yylex>, and that and its 526associated routines can be found in F<toke.c>. Perl isn't much like 527other computer languages; it's highly context sensitive at times, it can 528be tricky to work out what sort of token something is, or where a token 529ends. As such, there's a lot of interplay between the tokeniser and the 530parser, which can get pretty frightening if you're not used to it. 531 532As the parser understands a Perl program, it builds up a tree of 533operations for the interpreter to perform during execution. The routines 534which construct and link together the various operations are to be found 535in F<op.c>, and will be examined later. 536 537=item Optimization 538 539Now the parsing stage is complete, and the finished tree represents 540the operations that the Perl interpreter needs to perform to execute our 541program. Next, Perl does a dry run over the tree looking for 542optimisations: constant expressions such as C<3 + 4> will be computed 543now, and the optimizer will also see if any multiple operations can be 544replaced with a single one. For instance, to fetch the variable C<$foo>, 545instead of grabbing the glob C<*foo> and looking at the scalar 546component, the optimizer fiddles the op tree to use a function which 547directly looks up the scalar in question. The main optimizer is C<peep> 548in F<op.c>, and many ops have their own optimizing functions. 549 550=item Running 551 552Now we're finally ready to go: we have compiled Perl byte code, and all 553that's left to do is run it. The actual execution is done by the 554C<runops_standard> function in F<run.c>; more specifically, it's done by 555these three innocent looking lines: 556 557 while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) { 558 PERL_ASYNC_CHECK(); 559 } 560 561You may be more comfortable with the Perl version of that: 562 563 PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}}; 564 565Well, maybe not. Anyway, each op contains a function pointer, which 566stipulates the function which will actually carry out the operation. 567This function will return the next op in the sequence - this allows for 568things like C<if> which choose the next op dynamically at run time. 569The C<PERL_ASYNC_CHECK> makes sure that things like signals interrupt 570execution if required. 571 572The actual functions called are known as PP code, and they're spread 573between four files: F<pp_hot.c> contains the "hot" code, which is most 574often used and highly optimized, F<pp_sys.c> contains all the 575system-specific functions, F<pp_ctl.c> contains the functions which 576implement control structures (C<if>, C<while> and the like) and F<pp.c> 577contains everything else. These are, if you like, the C code for Perl's 578built-in functions and operators. 579 580Note that each C<pp_> function is expected to return a pointer to the next 581op. Calls to perl subs (and eval blocks) are handled within the same 582runops loop, and do not consume extra space on the C stack. For example, 583C<pp_entersub> and C<pp_entertry> just push a C<CxSUB> or C<CxEVAL> block 584struct onto the context stack which contain the address of the op 585following the sub call or eval. They then return the first op of that sub 586or eval block, and so execution continues of that sub or block. Later, a 587C<pp_leavesub> or C<pp_leavetry> op pops the C<CxSUB> or C<CxEVAL>, 588retrieves the return op from it, and returns it. 589 590=item Exception handing 591 592Perl's exception handing (i.e. C<die> etc.) is built on top of the low-level 593C<setjmp()>/C<longjmp()> C-library functions. These basically provide a 594way to capture the current PC and SP registers and later restore them; i.e. 595a C<longjmp()> continues at the point in code where a previous C<setjmp()> 596was done, with anything further up on the C stack being lost. This is why 597code should always save values using C<SAVE_FOO> rather than in auto 598variables. 599 600The perl core wraps C<setjmp()> etc in the macros C<JMPENV_PUSH> and 601C<JMPENV_JUMP>. The basic rule of perl exceptions is that C<exit>, and 602C<die> (in the absence of C<eval>) perform a C<JMPENV_JUMP(2)>, while 603C<die> within C<eval> does a C<JMPENV_JUMP(3)>. 604 605At entry points to perl, such as C<perl_parse()>, C<perl_run()> and 606C<call_sv(cv, G_EVAL)> each does a C<JMPENV_PUSH>, then enter a runops 607loop or whatever, and handle possible exception returns. For a 2 return, 608final cleanup is performed, such as popping stacks and calling C<CHECK> or 609C<END> blocks. Amongst other things, this is how scope cleanup still 610occurs during an C<exit>. 611 612If a C<die> can find a C<CxEVAL> block on the context stack, then the 613stack is popped to that level and the return op in that block is assigned 614to C<PL_restartop>; then a C<JMPENV_JUMP(3)> is performed. This normally 615passes control back to the guard. In the case of C<perl_run> and 616C<call_sv>, a non-null C<PL_restartop> triggers re-entry to the runops 617loop. The is the normal way that C<die> or C<croak> is handled within an 618C<eval>. 619 620Sometimes ops are executed within an inner runops loop, such as tie, sort 621or overload code. In this case, something like 622 623 sub FETCH { eval { die } } 624 625would cause a longjmp right back to the guard in C<perl_run>, popping both 626runops loops, which is clearly incorrect. One way to avoid this is for the 627tie code to do a C<JMPENV_PUSH> before executing C<FETCH> in the inner 628runops loop, but for efficiency reasons, perl in fact just sets a flag, 629using C<CATCH_SET(TRUE)>. The C<pp_require>, C<pp_entereval> and 630C<pp_entertry> ops check this flag, and if true, they call C<docatch>, 631which does a C<JMPENV_PUSH> and starts a new runops level to execute the 632code, rather than doing it on the current loop. 633 634As a further optimisation, on exit from the eval block in the C<FETCH>, 635execution of the code following the block is still carried on in the inner 636loop. When an exception is raised, C<docatch> compares the C<JMPENV> 637level of the C<CxEVAL> with C<PL_top_env> and if they differ, just 638re-throws the exception. In this way any inner loops get popped. 639 640Here's an example. 641 642 1: eval { tie @a, 'A' }; 643 2: sub A::TIEARRAY { 644 3: eval { die }; 645 4: die; 646 5: } 647 648To run this code, C<perl_run> is called, which does a C<JMPENV_PUSH> then 649enters a runops loop. This loop executes the eval and tie ops on line 1, 650with the eval pushing a C<CxEVAL> onto the context stack. 651 652The C<pp_tie> does a C<CATCH_SET(TRUE)>, then starts a second runops loop 653to execute the body of C<TIEARRAY>. When it executes the entertry op on 654line 3, C<CATCH_GET> is true, so C<pp_entertry> calls C<docatch> which 655does a C<JMPENV_PUSH> and starts a third runops loop, which then executes 656the die op. At this point the C call stack looks like this: 657 658 Perl_pp_die 659 Perl_runops # third loop 660 S_docatch_body 661 S_docatch 662 Perl_pp_entertry 663 Perl_runops # second loop 664 S_call_body 665 Perl_call_sv 666 Perl_pp_tie 667 Perl_runops # first loop 668 S_run_body 669 perl_run 670 main 671 672and the context and data stacks, as shown by C<-Dstv>, look like: 673 674 STACK 0: MAIN 675 CX 0: BLOCK => 676 CX 1: EVAL => AV() PV("A"\0) 677 retop=leave 678 STACK 1: MAGIC 679 CX 0: SUB => 680 retop=(null) 681 CX 1: EVAL => * 682 retop=nextstate 683 684The die pops the first C<CxEVAL> off the context stack, sets 685C<PL_restartop> from it, does a C<JMPENV_JUMP(3)>, and control returns to 686the top C<docatch>. This then starts another third-level runops level, 687which executes the nextstate, pushmark and die ops on line 4. At the point 688that the second C<pp_die> is called, the C call stack looks exactly like 689that above, even though we are no longer within an inner eval; this is 690because of the optimization mentioned earlier. However, the context stack 691now looks like this, ie with the top CxEVAL popped: 692 693 STACK 0: MAIN 694 CX 0: BLOCK => 695 CX 1: EVAL => AV() PV("A"\0) 696 retop=leave 697 STACK 1: MAGIC 698 CX 0: SUB => 699 retop=(null) 700 701The die on line 4 pops the context stack back down to the CxEVAL, leaving 702it as: 703 704 STACK 0: MAIN 705 CX 0: BLOCK => 706 707As usual, C<PL_restartop> is extracted from the C<CxEVAL>, and a 708C<JMPENV_JUMP(3)> done, which pops the C stack back to the docatch: 709 710 S_docatch 711 Perl_pp_entertry 712 Perl_runops # second loop 713 S_call_body 714 Perl_call_sv 715 Perl_pp_tie 716 Perl_runops # first loop 717 S_run_body 718 perl_run 719 main 720 721In this case, because the C<JMPENV> level recorded in the C<CxEVAL> 722differs from the current one, C<docatch> just does a C<JMPENV_JUMP(3)> 723and the C stack unwinds to: 724 725 perl_run 726 main 727 728Because C<PL_restartop> is non-null, C<run_body> starts a new runops loop 729and execution continues. 730 731=back 732 733=head2 Internal Variable Types 734 735You should by now have had a look at L<perlguts>, which tells you about 736Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do 737that now. 738 739These variables are used not only to represent Perl-space variables, but 740also any constants in the code, as well as some structures completely 741internal to Perl. The symbol table, for instance, is an ordinary Perl 742hash. Your code is represented by an SV as it's read into the parser; 743any program files you call are opened via ordinary Perl filehandles, and 744so on. 745 746The core L<Devel::Peek|Devel::Peek> module lets us examine SVs from a 747Perl program. Let's see, for instance, how Perl treats the constant 748C<"hello">. 749 750 % perl -MDevel::Peek -e 'Dump("hello")' 751 1 SV = PV(0xa041450) at 0xa04ecbc 752 2 REFCNT = 1 753 3 FLAGS = (POK,READONLY,pPOK) 754 4 PV = 0xa0484e0 "hello"\0 755 5 CUR = 5 756 6 LEN = 6 757 758Reading C<Devel::Peek> output takes a bit of practise, so let's go 759through it line by line. 760 761Line 1 tells us we're looking at an SV which lives at C<0xa04ecbc> in 762memory. SVs themselves are very simple structures, but they contain a 763pointer to a more complex structure. In this case, it's a PV, a 764structure which holds a string value, at location C<0xa041450>. Line 2 765is the reference count; there are no other references to this data, so 766it's 1. 767 768Line 3 are the flags for this SV - it's OK to use it as a PV, it's a 769read-only SV (because it's a constant) and the data is a PV internally. 770Next we've got the contents of the string, starting at location 771C<0xa0484e0>. 772 773Line 5 gives us the current length of the string - note that this does 774B<not> include the null terminator. Line 6 is not the length of the 775string, but the length of the currently allocated buffer; as the string 776grows, Perl automatically extends the available storage via a routine 777called C<SvGROW>. 778 779You can get at any of these quantities from C very easily; just add 780C<Sv> to the name of the field shown in the snippet, and you've got a 781macro which will return the value: C<SvCUR(sv)> returns the current 782length of the string, C<SvREFCOUNT(sv)> returns the reference count, 783C<SvPV(sv, len)> returns the string itself with its length, and so on. 784More macros to manipulate these properties can be found in L<perlguts>. 785 786Let's take an example of manipulating a PV, from C<sv_catpvn>, in F<sv.c> 787 788 1 void 789 2 Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len) 790 3 { 791 4 STRLEN tlen; 792 5 char *junk; 793 794 6 junk = SvPV_force(sv, tlen); 795 7 SvGROW(sv, tlen + len + 1); 796 8 if (ptr == junk) 797 9 ptr = SvPVX(sv); 798 10 Move(ptr,SvPVX(sv)+tlen,len,char); 799 11 SvCUR(sv) += len; 800 12 *SvEND(sv) = '\0'; 801 13 (void)SvPOK_only_UTF8(sv); /* validate pointer */ 802 14 SvTAINT(sv); 803 15 } 804 805This is a function which adds a string, C<ptr>, of length C<len> onto 806the end of the PV stored in C<sv>. The first thing we do in line 6 is 807make sure that the SV B<has> a valid PV, by calling the C<SvPV_force> 808macro to force a PV. As a side effect, C<tlen> gets set to the current 809value of the PV, and the PV itself is returned to C<junk>. 810 811In line 7, we make sure that the SV will have enough room to accommodate 812the old string, the new string and the null terminator. If C<LEN> isn't 813big enough, C<SvGROW> will reallocate space for us. 814 815Now, if C<junk> is the same as the string we're trying to add, we can 816grab the string directly from the SV; C<SvPVX> is the address of the PV 817in the SV. 818 819Line 10 does the actual catenation: the C<Move> macro moves a chunk of 820memory around: we move the string C<ptr> to the end of the PV - that's 821the start of the PV plus its current length. We're moving C<len> bytes 822of type C<char>. After doing so, we need to tell Perl we've extended the 823string, by altering C<CUR> to reflect the new length. C<SvEND> is a 824macro which gives us the end of the string, so that needs to be a 825C<"\0">. 826 827Line 13 manipulates the flags; since we've changed the PV, any IV or NV 828values will no longer be valid: if we have C<$a=10; $a.="6";> we don't 829want to use the old IV of 10. C<SvPOK_only_utf8> is a special UTF-8-aware 830version of C<SvPOK_only>, a macro which turns off the IOK and NOK flags 831and turns on POK. The final C<SvTAINT> is a macro which launders tainted 832data if taint mode is turned on. 833 834AVs and HVs are more complicated, but SVs are by far the most common 835variable type being thrown around. Having seen something of how we 836manipulate these, let's go on and look at how the op tree is 837constructed. 838 839=head2 Op Trees 840 841First, what is the op tree, anyway? The op tree is the parsed 842representation of your program, as we saw in our section on parsing, and 843it's the sequence of operations that Perl goes through to execute your 844program, as we saw in L</Running>. 845 846An op is a fundamental operation that Perl can perform: all the built-in 847functions and operators are ops, and there are a series of ops which 848deal with concepts the interpreter needs internally - entering and 849leaving a block, ending a statement, fetching a variable, and so on. 850 851The op tree is connected in two ways: you can imagine that there are two 852"routes" through it, two orders in which you can traverse the tree. 853First, parse order reflects how the parser understood the code, and 854secondly, execution order tells perl what order to perform the 855operations in. 856 857The easiest way to examine the op tree is to stop Perl after it has 858finished parsing, and get it to dump out the tree. This is exactly what 859the compiler backends L<B::Terse|B::Terse>, L<B::Concise|B::Concise> 860and L<B::Debug|B::Debug> do. 861 862Let's have a look at how Perl sees C<$a = $b + $c>: 863 864 % perl -MO=Terse -e '$a=$b+$c' 865 1 LISTOP (0x8179888) leave 866 2 OP (0x81798b0) enter 867 3 COP (0x8179850) nextstate 868 4 BINOP (0x8179828) sassign 869 5 BINOP (0x8179800) add [1] 870 6 UNOP (0x81796e0) null [15] 871 7 SVOP (0x80fafe0) gvsv GV (0x80fa4cc) *b 872 8 UNOP (0x81797e0) null [15] 873 9 SVOP (0x8179700) gvsv GV (0x80efeb0) *c 874 10 UNOP (0x816b4f0) null [15] 875 11 SVOP (0x816dcf0) gvsv GV (0x80fa460) *a 876 877Let's start in the middle, at line 4. This is a BINOP, a binary 878operator, which is at location C<0x8179828>. The specific operator in 879question is C<sassign> - scalar assignment - and you can find the code 880which implements it in the function C<pp_sassign> in F<pp_hot.c>. As a 881binary operator, it has two children: the add operator, providing the 882result of C<$b+$c>, is uppermost on line 5, and the left hand side is on 883line 10. 884 885Line 10 is the null op: this does exactly nothing. What is that doing 886there? If you see the null op, it's a sign that something has been 887optimized away after parsing. As we mentioned in L</Optimization>, 888the optimization stage sometimes converts two operations into one, for 889example when fetching a scalar variable. When this happens, instead of 890rewriting the op tree and cleaning up the dangling pointers, it's easier 891just to replace the redundant operation with the null op. Originally, 892the tree would have looked like this: 893 894 10 SVOP (0x816b4f0) rv2sv [15] 895 11 SVOP (0x816dcf0) gv GV (0x80fa460) *a 896 897That is, fetch the C<a> entry from the main symbol table, and then look 898at the scalar component of it: C<gvsv> (C<pp_gvsv> into F<pp_hot.c>) 899happens to do both these things. 900 901The right hand side, starting at line 5 is similar to what we've just 902seen: we have the C<add> op (C<pp_add> also in F<pp_hot.c>) add together 903two C<gvsv>s. 904 905Now, what's this about? 906 907 1 LISTOP (0x8179888) leave 908 2 OP (0x81798b0) enter 909 3 COP (0x8179850) nextstate 910 911C<enter> and C<leave> are scoping ops, and their job is to perform any 912housekeeping every time you enter and leave a block: lexical variables 913are tidied up, unreferenced variables are destroyed, and so on. Every 914program will have those first three lines: C<leave> is a list, and its 915children are all the statements in the block. Statements are delimited 916by C<nextstate>, so a block is a collection of C<nextstate> ops, with 917the ops to be performed for each statement being the children of 918C<nextstate>. C<enter> is a single op which functions as a marker. 919 920That's how Perl parsed the program, from top to bottom: 921 922 Program 923 | 924 Statement 925 | 926 = 927 / \ 928 / \ 929 $a + 930 / \ 931 $b $c 932 933However, it's impossible to B<perform> the operations in this order: 934you have to find the values of C<$b> and C<$c> before you add them 935together, for instance. So, the other thread that runs through the op 936tree is the execution order: each op has a field C<op_next> which points 937to the next op to be run, so following these pointers tells us how perl 938executes the code. We can traverse the tree in this order using 939the C<exec> option to C<B::Terse>: 940 941 % perl -MO=Terse,exec -e '$a=$b+$c' 942 1 OP (0x8179928) enter 943 2 COP (0x81798c8) nextstate 944 3 SVOP (0x81796c8) gvsv GV (0x80fa4d4) *b 945 4 SVOP (0x8179798) gvsv GV (0x80efeb0) *c 946 5 BINOP (0x8179878) add [1] 947 6 SVOP (0x816dd38) gvsv GV (0x80fa468) *a 948 7 BINOP (0x81798a0) sassign 949 8 LISTOP (0x8179900) leave 950 951This probably makes more sense for a human: enter a block, start a 952statement. Get the values of C<$b> and C<$c>, and add them together. 953Find C<$a>, and assign one to the other. Then leave. 954 955The way Perl builds up these op trees in the parsing process can be 956unravelled by examining F<perly.y>, the YACC grammar. Let's take the 957piece we need to construct the tree for C<$a = $b + $c> 958 959 1 term : term ASSIGNOP term 960 2 { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); } 961 3 | term ADDOP term 962 4 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); } 963 964If you're not used to reading BNF grammars, this is how it works: You're 965fed certain things by the tokeniser, which generally end up in upper 966case. Here, C<ADDOP>, is provided when the tokeniser sees C<+> in your 967code. C<ASSIGNOP> is provided when C<=> is used for assigning. These are 968"terminal symbols", because you can't get any simpler than them. 969 970The grammar, lines one and three of the snippet above, tells you how to 971build up more complex forms. These complex forms, "non-terminal symbols" 972are generally placed in lower case. C<term> here is a non-terminal 973symbol, representing a single expression. 974 975The grammar gives you the following rule: you can make the thing on the 976left of the colon if you see all the things on the right in sequence. 977This is called a "reduction", and the aim of parsing is to completely 978reduce the input. There are several different ways you can perform a 979reduction, separated by vertical bars: so, C<term> followed by C<=> 980followed by C<term> makes a C<term>, and C<term> followed by C<+> 981followed by C<term> can also make a C<term>. 982 983So, if you see two terms with an C<=> or C<+>, between them, you can 984turn them into a single expression. When you do this, you execute the 985code in the block on the next line: if you see C<=>, you'll do the code 986in line 2. If you see C<+>, you'll do the code in line 4. It's this code 987which contributes to the op tree. 988 989 | term ADDOP term 990 { $$ = newBINOP($2, 0, scalar($1), scalar($3)); } 991 992What this does is creates a new binary op, and feeds it a number of 993variables. The variables refer to the tokens: C<$1> is the first token in 994the input, C<$2> the second, and so on - think regular expression 995backreferences. C<$$> is the op returned from this reduction. So, we 996call C<newBINOP> to create a new binary operator. The first parameter to 997C<newBINOP>, a function in F<op.c>, is the op type. It's an addition 998operator, so we want the type to be C<ADDOP>. We could specify this 999directly, but it's right there as the second token in the input, so we 1000use C<$2>. The second parameter is the op's flags: 0 means "nothing 1001special". Then the things to add: the left and right hand side of our 1002expression, in scalar context. 1003 1004=head2 Stacks 1005 1006When perl executes something like C<addop>, how does it pass on its 1007results to the next op? The answer is, through the use of stacks. Perl 1008has a number of stacks to store things it's currently working on, and 1009we'll look at the three most important ones here. 1010 1011=over 3 1012 1013=item Argument stack 1014 1015Arguments are passed to PP code and returned from PP code using the 1016argument stack, C<ST>. The typical way to handle arguments is to pop 1017them off the stack, deal with them how you wish, and then push the result 1018back onto the stack. This is how, for instance, the cosine operator 1019works: 1020 1021 NV value; 1022 value = POPn; 1023 value = Perl_cos(value); 1024 XPUSHn(value); 1025 1026We'll see a more tricky example of this when we consider Perl's macros 1027below. C<POPn> gives you the NV (floating point value) of the top SV on 1028the stack: the C<$x> in C<cos($x)>. Then we compute the cosine, and push 1029the result back as an NV. The C<X> in C<XPUSHn> means that the stack 1030should be extended if necessary - it can't be necessary here, because we 1031know there's room for one more item on the stack, since we've just 1032removed one! The C<XPUSH*> macros at least guarantee safety. 1033 1034Alternatively, you can fiddle with the stack directly: C<SP> gives you 1035the first element in your portion of the stack, and C<TOP*> gives you 1036the top SV/IV/NV/etc. on the stack. So, for instance, to do unary 1037negation of an integer: 1038 1039 SETi(-TOPi); 1040 1041Just set the integer value of the top stack entry to its negation. 1042 1043Argument stack manipulation in the core is exactly the same as it is in 1044XSUBs - see L<perlxstut>, L<perlxs> and L<perlguts> for a longer 1045description of the macros used in stack manipulation. 1046 1047=item Mark stack 1048 1049I say "your portion of the stack" above because PP code doesn't 1050necessarily get the whole stack to itself: if your function calls 1051another function, you'll only want to expose the arguments aimed for the 1052called function, and not (necessarily) let it get at your own data. The 1053way we do this is to have a "virtual" bottom-of-stack, exposed to each 1054function. The mark stack keeps bookmarks to locations in the argument 1055stack usable by each function. For instance, when dealing with a tied 1056variable, (internally, something with "P" magic) Perl has to call 1057methods for accesses to the tied variables. However, we need to separate 1058the arguments exposed to the method to the argument exposed to the 1059original function - the store or fetch or whatever it may be. Here's 1060roughly how the tied C<push> is implemented; see C<av_push> in F<av.c>: 1061 1062 1 PUSHMARK(SP); 1063 2 EXTEND(SP,2); 1064 3 PUSHs(SvTIED_obj((SV*)av, mg)); 1065 4 PUSHs(val); 1066 5 PUTBACK; 1067 6 ENTER; 1068 7 call_method("PUSH", G_SCALAR|G_DISCARD); 1069 8 LEAVE; 1070 1071Let's examine the whole implementation, for practice: 1072 1073 1 PUSHMARK(SP); 1074 1075Push the current state of the stack pointer onto the mark stack. This is 1076so that when we've finished adding items to the argument stack, Perl 1077knows how many things we've added recently. 1078 1079 2 EXTEND(SP,2); 1080 3 PUSHs(SvTIED_obj((SV*)av, mg)); 1081 4 PUSHs(val); 1082 1083We're going to add two more items onto the argument stack: when you have 1084a tied array, the C<PUSH> subroutine receives the object and the value 1085to be pushed, and that's exactly what we have here - the tied object, 1086retrieved with C<SvTIED_obj>, and the value, the SV C<val>. 1087 1088 5 PUTBACK; 1089 1090Next we tell Perl to update the global stack pointer from our internal 1091variable: C<dSP> only gave us a local copy, not a reference to the global. 1092 1093 6 ENTER; 1094 7 call_method("PUSH", G_SCALAR|G_DISCARD); 1095 8 LEAVE; 1096 1097C<ENTER> and C<LEAVE> localise a block of code - they make sure that all 1098variables are tidied up, everything that has been localised gets 1099its previous value returned, and so on. Think of them as the C<{> and 1100C<}> of a Perl block. 1101 1102To actually do the magic method call, we have to call a subroutine in 1103Perl space: C<call_method> takes care of that, and it's described in 1104L<perlcall>. We call the C<PUSH> method in scalar context, and we're 1105going to discard its return value. The call_method() function 1106removes the top element of the mark stack, so there is nothing for 1107the caller to clean up. 1108 1109=item Save stack 1110 1111C doesn't have a concept of local scope, so perl provides one. We've 1112seen that C<ENTER> and C<LEAVE> are used as scoping braces; the save 1113stack implements the C equivalent of, for example: 1114 1115 { 1116 local $foo = 42; 1117 ... 1118 } 1119 1120See L<perlguts/Localising Changes> for how to use the save stack. 1121 1122=back 1123 1124=head2 Millions of Macros 1125 1126One thing you'll notice about the Perl source is that it's full of 1127macros. Some have called the pervasive use of macros the hardest thing 1128to understand, others find it adds to clarity. Let's take an example, 1129the code which implements the addition operator: 1130 1131 1 PP(pp_add) 1132 2 { 1133 3 dSP; dATARGET; tryAMAGICbin(add,opASSIGN); 1134 4 { 1135 5 dPOPTOPnnrl_ul; 1136 6 SETn( left + right ); 1137 7 RETURN; 1138 8 } 1139 9 } 1140 1141Every line here (apart from the braces, of course) contains a macro. The 1142first line sets up the function declaration as Perl expects for PP code; 1143line 3 sets up variable declarations for the argument stack and the 1144target, the return value of the operation. Finally, it tries to see if 1145the addition operation is overloaded; if so, the appropriate subroutine 1146is called. 1147 1148Line 5 is another variable declaration - all variable declarations start 1149with C<d> - which pops from the top of the argument stack two NVs (hence 1150C<nn>) and puts them into the variables C<right> and C<left>, hence the 1151C<rl>. These are the two operands to the addition operator. Next, we 1152call C<SETn> to set the NV of the return value to the result of adding 1153the two values. This done, we return - the C<RETURN> macro makes sure 1154that our return value is properly handled, and we pass the next operator 1155to run back to the main run loop. 1156 1157Most of these macros are explained in L<perlapi>, and some of the more 1158important ones are explained in L<perlxs> as well. Pay special attention 1159to L<perlguts/Background and PERL_IMPLICIT_CONTEXT> for information on 1160the C<[pad]THX_?> macros. 1161 1162=head2 The .i Targets 1163 1164You can expand the macros in a F<foo.c> file by saying 1165 1166 make foo.i 1167 1168which will expand the macros using cpp. Don't be scared by the results. 1169 1170=head1 SOURCE CODE STATIC ANALYSIS 1171 1172Various tools exist for analysing C source code B<statically>, as 1173opposed to B<dynamically>, that is, without executing the code. 1174It is possible to detect resource leaks, undefined behaviour, type 1175mismatches, portability problems, code paths that would cause illegal 1176memory accesses, and other similar problems by just parsing the C code 1177and looking at the resulting graph, what does it tell about the 1178execution and data flows. As a matter of fact, this is exactly 1179how C compilers know to give warnings about dubious code. 1180 1181=head2 lint, splint 1182 1183The good old C code quality inspector, C<lint>, is available in 1184several platforms, but please be aware that there are several 1185different implementations of it by different vendors, which means that 1186the flags are not identical across different platforms. 1187 1188There is a lint variant called C<splint> (Secure Programming Lint) 1189available from http://www.splint.org/ that should compile on any 1190Unix-like platform. 1191 1192There are C<lint> and <splint> targets in Makefile, but you may have 1193to diddle with the flags (see above). 1194 1195=head2 Coverity 1196 1197Coverity (http://www.coverity.com/) is a product similar to lint and 1198as a testbed for their product they periodically check several open 1199source projects, and they give out accounts to open source developers 1200to the defect databases. 1201 1202=head2 cpd (cut-and-paste detector) 1203 1204The cpd tool detects cut-and-paste coding. If one instance of the 1205cut-and-pasted code changes, all the other spots should probably be 1206changed, too. Therefore such code should probably be turned into a 1207subroutine or a macro. 1208 1209cpd (http://pmd.sourceforge.net/cpd.html) is part of the pmd project 1210(http://pmd.sourceforge.net/). pmd was originally written for static 1211analysis of Java code, but later the cpd part of it was extended to 1212parse also C and C++. 1213 1214Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the 1215pmd-X.Y.jar from it, and then run that on source code thusly: 1216 1217 java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD --minimum-tokens 100 --files /some/where/src --language c > cpd.txt 1218 1219You may run into memory limits, in which case you should use the -Xmx option: 1220 1221 java -Xmx512M ... 1222 1223=head2 gcc warnings 1224 1225Though much can be written about the inconsistency and coverage 1226problems of gcc warnings (like C<-Wall> not meaning "all the 1227warnings", or some common portability problems not being covered by 1228C<-Wall>, or C<-ansi> and C<-pedantic> both being a poorly defined 1229collection of warnings, and so forth), gcc is still a useful tool in 1230keeping our coding nose clean. 1231 1232The C<-Wall> is by default on. 1233 1234The C<-ansi> (and its sidekick, C<-pedantic>) would be nice to be on 1235always, but unfortunately they are not safe on all platforms, they can 1236for example cause fatal conflicts with the system headers (Solaris 1237being a prime example). If Configure C<-Dgccansipedantic> is used, 1238the C<cflags> frontend selects C<-ansi -pedantic> for the platforms 1239where they are known to be safe. 1240 1241Starting from Perl 5.9.4 the following extra flags are added: 1242 1243=over 4 1244 1245=item * 1246 1247C<-Wendif-labels> 1248 1249=item * 1250 1251C<-Wextra> 1252 1253=item * 1254 1255C<-Wdeclaration-after-statement> 1256 1257=back 1258 1259The following flags would be nice to have but they would first need 1260their own Augean stablemaster: 1261 1262=over 4 1263 1264=item * 1265 1266C<-Wpointer-arith> 1267 1268=item * 1269 1270C<-Wshadow> 1271 1272=item * 1273 1274C<-Wstrict-prototypes> 1275 1276=back 1277 1278The C<-Wtraditional> is another example of the annoying tendency of 1279gcc to bundle a lot of warnings under one switch (it would be 1280impossible to deploy in practice because it would complain a lot) but 1281it does contain some warnings that would be beneficial to have available 1282on their own, such as the warning about string constants inside macros 1283containing the macro arguments: this behaved differently pre-ANSI 1284than it does in ANSI, and some C compilers are still in transition, 1285AIX being an example. 1286 1287=head2 Warnings of other C compilers 1288 1289Other C compilers (yes, there B<are> other C compilers than gcc) often 1290have their "strict ANSI" or "strict ANSI with some portability extensions" 1291modes on, like for example the Sun Workshop has its C<-Xa> mode on 1292(though implicitly), or the DEC (these days, HP...) has its C<-std1> 1293mode on. 1294 1295=head2 DEBUGGING 1296 1297You can compile a special debugging version of Perl, which allows you 1298to use the C<-D> option of Perl to tell more about what Perl is doing. 1299But sometimes there is no alternative than to dive in with a debugger, 1300either to see the stack trace of a core dump (very useful in a bug 1301report), or trying to figure out what went wrong before the core dump 1302happened, or how did we end up having wrong or unexpected results. 1303 1304=head2 Poking at Perl 1305 1306To really poke around with Perl, you'll probably want to build Perl for 1307debugging, like this: 1308 1309 ./Configure -d -D optimize=-g 1310 make 1311 1312C<-g> is a flag to the C compiler to have it produce debugging 1313information which will allow us to step through a running program, 1314and to see in which C function we are at (without the debugging 1315information we might see only the numerical addresses of the functions, 1316which is not very helpful). 1317 1318F<Configure> will also turn on the C<DEBUGGING> compilation symbol which 1319enables all the internal debugging code in Perl. There are a whole bunch 1320of things you can debug with this: L<perlrun> lists them all, and the 1321best way to find out about them is to play about with them. The most 1322useful options are probably 1323 1324 l Context (loop) stack processing 1325 t Trace execution 1326 o Method and overloading resolution 1327 c String/numeric conversions 1328 1329Some of the functionality of the debugging code can be achieved using XS 1330modules. 1331 1332 -Dr => use re 'debug' 1333 -Dx => use O 'Debug' 1334 1335=head2 Using a source-level debugger 1336 1337If the debugging output of C<-D> doesn't help you, it's time to step 1338through perl's execution with a source-level debugger. 1339 1340=over 3 1341 1342=item * 1343 1344We'll use C<gdb> for our examples here; the principles will apply to 1345any debugger (many vendors call their debugger C<dbx>), but check the 1346manual of the one you're using. 1347 1348=back 1349 1350To fire up the debugger, type 1351 1352 gdb ./perl 1353 1354Or if you have a core dump: 1355 1356 gdb ./perl core 1357 1358You'll want to do that in your Perl source tree so the debugger can read 1359the source code. You should see the copyright message, followed by the 1360prompt. 1361 1362 (gdb) 1363 1364C<help> will get you into the documentation, but here are the most 1365useful commands: 1366 1367=over 3 1368 1369=item run [args] 1370 1371Run the program with the given arguments. 1372 1373=item break function_name 1374 1375=item break source.c:xxx 1376 1377Tells the debugger that we'll want to pause execution when we reach 1378either the named function (but see L<perlguts/Internal Functions>!) or the given 1379line in the named source file. 1380 1381=item step 1382 1383Steps through the program a line at a time. 1384 1385=item next 1386 1387Steps through the program a line at a time, without descending into 1388functions. 1389 1390=item continue 1391 1392Run until the next breakpoint. 1393 1394=item finish 1395 1396Run until the end of the current function, then stop again. 1397 1398=item 'enter' 1399 1400Just pressing Enter will do the most recent operation again - it's a 1401blessing when stepping through miles of source code. 1402 1403=item print 1404 1405Execute the given C code and print its results. B<WARNING>: Perl makes 1406heavy use of macros, and F<gdb> does not necessarily support macros 1407(see later L</"gdb macro support">). You'll have to substitute them 1408yourself, or to invoke cpp on the source code files 1409(see L</"The .i Targets">) 1410So, for instance, you can't say 1411 1412 print SvPV_nolen(sv) 1413 1414but you have to say 1415 1416 print Perl_sv_2pv_nolen(sv) 1417 1418=back 1419 1420You may find it helpful to have a "macro dictionary", which you can 1421produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't 1422recursively apply those macros for you. 1423 1424=head2 gdb macro support 1425 1426Recent versions of F<gdb> have fairly good macro support, but 1427in order to use it you'll need to compile perl with macro definitions 1428included in the debugging information. Using F<gcc> version 3.1, this 1429means configuring with C<-Doptimize=-g3>. Other compilers might use a 1430different switch (if they support debugging macros at all). 1431 1432=head2 Dumping Perl Data Structures 1433 1434One way to get around this macro hell is to use the dumping functions in 1435F<dump.c>; these work a little like an internal 1436L<Devel::Peek|Devel::Peek>, but they also cover OPs and other structures 1437that you can't get at from Perl. Let's take an example. We'll use the 1438C<$a = $b + $c> we used before, but give it a bit of context: 1439C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and poke around? 1440 1441What about C<pp_add>, the function we examined earlier to implement the 1442C<+> operator: 1443 1444 (gdb) break Perl_pp_add 1445 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309. 1446 1447Notice we use C<Perl_pp_add> and not C<pp_add> - see L<perlguts/Internal Functions>. 1448With the breakpoint in place, we can run our program: 1449 1450 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c' 1451 1452Lots of junk will go past as gdb reads in the relevant source files and 1453libraries, and then: 1454 1455 Breakpoint 1, Perl_pp_add () at pp_hot.c:309 1456 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN); 1457 (gdb) step 1458 311 dPOPTOPnnrl_ul; 1459 (gdb) 1460 1461We looked at this bit of code before, and we said that C<dPOPTOPnnrl_ul> 1462arranges for two C<NV>s to be placed into C<left> and C<right> - let's 1463slightly expand it: 1464 1465 #define dPOPTOPnnrl_ul NV right = POPn; \ 1466 SV *leftsv = TOPs; \ 1467 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0 1468 1469C<POPn> takes the SV from the top of the stack and obtains its NV either 1470directly (if C<SvNOK> is set) or by calling the C<sv_2nv> function. 1471C<TOPs> takes the next SV from the top of the stack - yes, C<POPn> uses 1472C<TOPs> - but doesn't remove it. We then use C<SvNV> to get the NV from 1473C<leftsv> in the same way as before - yes, C<POPn> uses C<SvNV>. 1474 1475Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to 1476convert it. If we step again, we'll find ourselves there: 1477 1478 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669 1479 1669 if (!sv) 1480 (gdb) 1481 1482We can now use C<Perl_sv_dump> to investigate the SV: 1483 1484 SV = PV(0xa057cc0) at 0xa0675d0 1485 REFCNT = 1 1486 FLAGS = (POK,pPOK) 1487 PV = 0xa06a510 "6XXXX"\0 1488 CUR = 5 1489 LEN = 6 1490 $1 = void 1491 1492We know we're going to get C<6> from this, so let's finish the 1493subroutine: 1494 1495 (gdb) finish 1496 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671 1497 0x462669 in Perl_pp_add () at pp_hot.c:311 1498 311 dPOPTOPnnrl_ul; 1499 1500We can also dump out this op: the current op is always stored in 1501C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us 1502similar output to L<B::Debug|B::Debug>. 1503 1504 { 1505 13 TYPE = add ===> 14 1506 TARG = 1 1507 FLAGS = (SCALAR,KIDS) 1508 { 1509 TYPE = null ===> (12) 1510 (was rv2sv) 1511 FLAGS = (SCALAR,KIDS) 1512 { 1513 11 TYPE = gvsv ===> 12 1514 FLAGS = (SCALAR) 1515 GV = main::b 1516 } 1517 } 1518 1519# finish this later # 1520 1521=head2 Patching 1522 1523All right, we've now had a look at how to navigate the Perl sources and 1524some things you'll need to know when fiddling with them. Let's now get 1525on and create a simple patch. Here's something Larry suggested: if a 1526C<U> is the first active format during a C<pack>, (for example, 1527C<pack "U3C8", @stuff>) then the resulting string should be treated as 1528UTF-8 encoded. 1529 1530If you are working with a git clone of the Perl repository, you will want to 1531create a branch for your changes. This will make creating a proper patch much 1532simpler. See the L<perlrepository> for details on how to do this. 1533 1534How do we prepare to fix this up? First we locate the code in question - 1535the C<pack> happens at runtime, so it's going to be in one of the F<pp> 1536files. Sure enough, C<pp_pack> is in F<pp.c>. Since we're going to be 1537altering this file, let's copy it to F<pp.c~>. 1538 1539[Well, it was in F<pp.c> when this tutorial was written. It has now been 1540split off with C<pp_unpack> to its own file, F<pp_pack.c>] 1541 1542Now let's look over C<pp_pack>: we take a pattern into C<pat>, and then 1543loop over the pattern, taking each format character in turn into 1544C<datum_type>. Then for each possible format character, we swallow up 1545the other arguments in the pattern (a field width, an asterisk, and so 1546on) and convert the next chunk input into the specified format, adding 1547it onto the output SV C<cat>. 1548 1549How do we know if the C<U> is the first format in the C<pat>? Well, if 1550we have a pointer to the start of C<pat> then, if we see a C<U> we can 1551test whether we're still at the start of the string. So, here's where 1552C<pat> is set up: 1553 1554 STRLEN fromlen; 1555 register char *pat = SvPVx(*++MARK, fromlen); 1556 register char *patend = pat + fromlen; 1557 register I32 len; 1558 I32 datumtype; 1559 SV *fromstr; 1560 1561We'll have another string pointer in there: 1562 1563 STRLEN fromlen; 1564 register char *pat = SvPVx(*++MARK, fromlen); 1565 register char *patend = pat + fromlen; 1566 + char *patcopy; 1567 register I32 len; 1568 I32 datumtype; 1569 SV *fromstr; 1570 1571And just before we start the loop, we'll set C<patcopy> to be the start 1572of C<pat>: 1573 1574 items = SP - MARK; 1575 MARK++; 1576 sv_setpvn(cat, "", 0); 1577 + patcopy = pat; 1578 while (pat < patend) { 1579 1580Now if we see a C<U> which was at the start of the string, we turn on 1581the C<UTF8> flag for the output SV, C<cat>: 1582 1583 + if (datumtype == 'U' && pat==patcopy+1) 1584 + SvUTF8_on(cat); 1585 if (datumtype == '#') { 1586 while (pat < patend && *pat != '\n') 1587 pat++; 1588 1589Remember that it has to be C<patcopy+1> because the first character of 1590the string is the C<U> which has been swallowed into C<datumtype!> 1591 1592Oops, we forgot one thing: what if there are spaces at the start of the 1593pattern? C<pack(" U*", @stuff)> will have C<U> as the first active 1594character, even though it's not the first thing in the pattern. In this 1595case, we have to advance C<patcopy> along with C<pat> when we see spaces: 1596 1597 if (isSPACE(datumtype)) 1598 continue; 1599 1600needs to become 1601 1602 if (isSPACE(datumtype)) { 1603 patcopy++; 1604 continue; 1605 } 1606 1607OK. That's the C part done. Now we must do two additional things before 1608this patch is ready to go: we've changed the behaviour of Perl, and so 1609we must document that change. We must also provide some more regression 1610tests to make sure our patch works and doesn't create a bug somewhere 1611else along the line. 1612 1613The regression tests for each operator live in F<t/op/>, and so we 1614make a copy of F<t/op/pack.t> to F<t/op/pack.t~>. Now we can add our 1615tests to the end. First, we'll test that the C<U> does indeed create 1616Unicode strings. 1617 1618t/op/pack.t has a sensible ok() function, but if it didn't we could 1619use the one from t/test.pl. 1620 1621 require './test.pl'; 1622 plan( tests => 159 ); 1623 1624so instead of this: 1625 1626 print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000); 1627 print "ok $test\n"; $test++; 1628 1629we can write the more sensible (see L<Test::More> for a full 1630explanation of is() and other testing functions). 1631 1632 is( "1.20.300.4000", sprintf "%vd", pack("U*",1,20,300,4000), 1633 "U* produces Unicode" ); 1634 1635Now we'll test that we got that space-at-the-beginning business right: 1636 1637 is( "1.20.300.4000", sprintf "%vd", pack(" U*",1,20,300,4000), 1638 " with spaces at the beginning" ); 1639 1640And finally we'll test that we don't make Unicode strings if C<U> is B<not> 1641the first active format: 1642 1643 isnt( v1.20.300.4000, sprintf "%vd", pack("C0U*",1,20,300,4000), 1644 "U* not first isn't Unicode" ); 1645 1646Mustn't forget to change the number of tests which appears at the top, 1647or else the automated tester will get confused. This will either look 1648like this: 1649 1650 print "1..156\n"; 1651 1652or this: 1653 1654 plan( tests => 156 ); 1655 1656We now compile up Perl, and run it through the test suite. Our new 1657tests pass, hooray! 1658 1659Finally, the documentation. The job is never done until the paperwork is 1660over, so let's describe the change we've just made. The relevant place 1661is F<pod/perlfunc.pod>; again, we make a copy, and then we'll insert 1662this text in the description of C<pack>: 1663 1664 =item * 1665 1666 If the pattern begins with a C<U>, the resulting string will be treated 1667 as UTF-8-encoded Unicode. You can force UTF-8 encoding on in a string 1668 with an initial C<U0>, and the bytes that follow will be interpreted as 1669 Unicode characters. If you don't want this to happen, you can begin your 1670 pattern with C<C0> (or anything else) to force Perl not to UTF-8 encode your 1671 string, and then follow this with a C<U*> somewhere in your pattern. 1672 1673=head2 Patching a core module 1674 1675This works just like patching anything else, with an extra 1676consideration. Many core modules also live on CPAN. If this is so, 1677patch the CPAN version instead of the core and send the patch off to 1678the module maintainer (with a copy to p5p). This will help the module 1679maintainer keep the CPAN version in sync with the core version without 1680constantly scanning p5p. 1681 1682The list of maintainers of core modules is usefully documented in 1683F<Porting/Maintainers.pl>. 1684 1685=head2 Adding a new function to the core 1686 1687If, as part of a patch to fix a bug, or just because you have an 1688especially good idea, you decide to add a new function to the core, 1689discuss your ideas on p5p well before you start work. It may be that 1690someone else has already attempted to do what you are considering and 1691can give lots of good advice or even provide you with bits of code 1692that they already started (but never finished). 1693 1694You have to follow all of the advice given above for patching. It is 1695extremely important to test any addition thoroughly and add new tests 1696to explore all boundary conditions that your new function is expected 1697to handle. If your new function is used only by one module (e.g. toke), 1698then it should probably be named S_your_function (for static); on the 1699other hand, if you expect it to accessible from other functions in 1700Perl, you should name it Perl_your_function. See L<perlguts/Internal Functions> 1701for more details. 1702 1703The location of any new code is also an important consideration. Don't 1704just create a new top level .c file and put your code there; you would 1705have to make changes to Configure (so the Makefile is created properly), 1706as well as possibly lots of include files. This is strictly pumpking 1707business. 1708 1709It is better to add your function to one of the existing top level 1710source code files, but your choice is complicated by the nature of 1711the Perl distribution. Only the files that are marked as compiled 1712static are located in the perl executable. Everything else is located 1713in the shared library (or DLL if you are running under WIN32). So, 1714for example, if a function was only used by functions located in 1715toke.c, then your code can go in toke.c. If, however, you want to call 1716the function from universal.c, then you should put your code in another 1717location, for example util.c. 1718 1719In addition to writing your c-code, you will need to create an 1720appropriate entry in embed.pl describing your function, then run 1721'make regen_headers' to create the entries in the numerous header 1722files that perl needs to compile correctly. See L<perlguts/Internal Functions> 1723for information on the various options that you can set in embed.pl. 1724You will forget to do this a few (or many) times and you will get 1725warnings during the compilation phase. Make sure that you mention 1726this when you post your patch to P5P; the pumpking needs to know this. 1727 1728When you write your new code, please be conscious of existing code 1729conventions used in the perl source files. See L<perlstyle> for 1730details. Although most of the guidelines discussed seem to focus on 1731Perl code, rather than c, they all apply (except when they don't ;). 1732Also see I<perlrepository> for lots of details about both formatting and 1733submitting patches of your changes. 1734 1735Lastly, TEST TEST TEST TEST TEST any code before posting to p5p. 1736Test on as many platforms as you can find. Test as many perl 1737Configure options as you can (e.g. MULTIPLICITY). If you have 1738profiling or memory tools, see L<EXTERNAL TOOLS FOR DEBUGGING PERL> 1739below for how to use them to further test your code. Remember that 1740most of the people on P5P are doing this on their own time and 1741don't have the time to debug your code. 1742 1743=head2 Writing a test 1744 1745Every module and built-in function has an associated test file (or 1746should...). If you add or change functionality, you have to write a 1747test. If you fix a bug, you have to write a test so that bug never 1748comes back. If you alter the docs, it would be nice to test what the 1749new documentation says. 1750 1751In short, if you submit a patch you probably also have to patch the 1752tests. 1753 1754For modules, the test file is right next to the module itself. 1755F<lib/strict.t> tests F<lib/strict.pm>. This is a recent innovation, 1756so there are some snags (and it would be wonderful for you to brush 1757them out), but it basically works that way. Everything else lives in 1758F<t/>. 1759 1760If you add a new test directory under F<t/>, it is imperative that you 1761add that directory to F<t/HARNESS> and F<t/TEST>. 1762 1763=over 3 1764 1765=item F<t/base/> 1766 1767Testing of the absolute basic functionality of Perl. Things like 1768C<if>, basic file reads and writes, simple regexes, etc. These are 1769run first in the test suite and if any of them fail, something is 1770I<really> broken. 1771 1772=item F<t/cmd/> 1773 1774These test the basic control structures, C<if/else>, C<while>, 1775subroutines, etc. 1776 1777=item F<t/comp/> 1778 1779Tests basic issues of how Perl parses and compiles itself. 1780 1781=item F<t/io/> 1782 1783Tests for built-in IO functions, including command line arguments. 1784 1785=item F<t/lib/> 1786 1787The old home for the module tests, you shouldn't put anything new in 1788here. There are still some bits and pieces hanging around in here 1789that need to be moved. Perhaps you could move them? Thanks! 1790 1791=item F<t/mro/> 1792 1793Tests for perl's method resolution order implementations 1794(see L<mro>). 1795 1796=item F<t/op/> 1797 1798Tests for perl's built in functions that don't fit into any of the 1799other directories. 1800 1801=item F<t/re/> 1802 1803Tests for regex related functions or behaviour. (These used to live 1804in t/op). 1805 1806=item F<t/run/> 1807 1808Testing features of how perl actually runs, including exit codes and 1809handling of PERL* environment variables. 1810 1811=item F<t/uni/> 1812 1813Tests for the core support of Unicode. 1814 1815=item F<t/win32/> 1816 1817Windows-specific tests. 1818 1819=item F<t/x2p> 1820 1821A test suite for the s2p converter. 1822 1823=back 1824 1825The core uses the same testing style as the rest of Perl, a simple 1826"ok/not ok" run through Test::Harness, but there are a few special 1827considerations. 1828 1829There are three ways to write a test in the core. Test::More, 1830t/test.pl and ad hoc C<print $test ? "ok 42\n" : "not ok 42\n">. The 1831decision of which to use depends on what part of the test suite you're 1832working on. This is a measure to prevent a high-level failure (such 1833as Config.pm breaking) from causing basic functionality tests to fail. 1834If you write your own test, use the L<Test Anything Protocol|TAP>. 1835 1836=over 4 1837 1838=item t/base t/comp 1839 1840Since we don't know if require works, or even subroutines, use ad hoc 1841tests for these two. Step carefully to avoid using the feature being 1842tested. 1843 1844=item t/cmd t/run t/io t/op 1845 1846Now that basic require() and subroutines are tested, you can use the 1847t/test.pl library which emulates the important features of Test::More 1848while using a minimum of core features. 1849 1850You can also conditionally use certain libraries like Config, but be 1851sure to skip the test gracefully if it's not there. 1852 1853=item t/lib ext lib 1854 1855Now that the core of Perl is tested, Test::More can be used. You can 1856also use the full suite of core modules in the tests. 1857 1858=back 1859 1860When you say "make test" Perl uses the F<t/TEST> program to run the 1861test suite (except under Win32 where it uses F<t/harness> instead.) 1862All tests are run from the F<t/> directory, B<not> the directory 1863which contains the test. This causes some problems with the tests 1864in F<lib/>, so here's some opportunity for some patching. 1865 1866You must be triply conscious of cross-platform concerns. This usually 1867boils down to using File::Spec and avoiding things like C<fork()> and 1868C<system()> unless absolutely necessary. 1869 1870=head2 Special Make Test Targets 1871 1872There are various special make targets that can be used to test Perl 1873slightly differently than the standard "test" target. Not all them 1874are expected to give a 100% success rate. Many of them have several 1875aliases, and many of them are not available on certain operating 1876systems. 1877 1878=over 4 1879 1880=item coretest 1881 1882Run F<perl> on all core tests (F<t/*> and F<lib/[a-z]*> pragma tests). 1883 1884(Not available on Win32) 1885 1886=item test.deparse 1887 1888Run all the tests through B::Deparse. Not all tests will succeed. 1889 1890(Not available on Win32) 1891 1892=item test.taintwarn 1893 1894Run all tests with the B<-t> command-line switch. Not all tests 1895are expected to succeed (until they're specifically fixed, of course). 1896 1897(Not available on Win32) 1898 1899=item minitest 1900 1901Run F<miniperl> on F<t/base>, F<t/comp>, F<t/cmd>, F<t/run>, F<t/io>, 1902F<t/op>, F<t/uni> and F<t/mro> tests. 1903 1904=item test.valgrind check.valgrind utest.valgrind ucheck.valgrind 1905 1906(Only in Linux) Run all the tests using the memory leak + naughty 1907memory access tool "valgrind". The log files will be named 1908F<testname.valgrind>. 1909 1910=item test.third check.third utest.third ucheck.third 1911 1912(Only in Tru64) Run all the tests using the memory leak + naughty 1913memory access tool "Third Degree". The log files will be named 1914F<perl.3log.testname>. 1915 1916=item test.torture torturetest 1917 1918Run all the usual tests and some extra tests. As of Perl 5.8.0 the 1919only extra tests are Abigail's JAPHs, F<t/japh/abigail.t>. 1920 1921You can also run the torture test with F<t/harness> by giving 1922C<-torture> argument to F<t/harness>. 1923 1924=item utest ucheck test.utf8 check.utf8 1925 1926Run all the tests with -Mutf8. Not all tests will succeed. 1927 1928(Not available on Win32) 1929 1930=item minitest.utf16 test.utf16 1931 1932Runs the tests with UTF-16 encoded scripts, encoded with different 1933versions of this encoding. 1934 1935C<make utest.utf16> runs the test suite with a combination of C<-utf8> and 1936C<-utf16> arguments to F<t/TEST>. 1937 1938(Not available on Win32) 1939 1940=item test_harness 1941 1942Run the test suite with the F<t/harness> controlling program, instead of 1943F<t/TEST>. F<t/harness> is more sophisticated, and uses the 1944L<Test::Harness> module, thus using this test target supposes that perl 1945mostly works. The main advantage for our purposes is that it prints a 1946detailed summary of failed tests at the end. Also, unlike F<t/TEST>, it 1947doesn't redirect stderr to stdout. 1948 1949Note that under Win32 F<t/harness> is always used instead of F<t/TEST>, so 1950there is no special "test_harness" target. 1951 1952Under Win32's "test" target you may use the TEST_SWITCHES and TEST_FILES 1953environment variables to control the behaviour of F<t/harness>. This means 1954you can say 1955 1956 nmake test TEST_FILES="op/*.t" 1957 nmake test TEST_SWITCHES="-torture" TEST_FILES="op/*.t" 1958 1959=item Parallel tests 1960 1961The core distribution can now run its regression tests in parallel on 1962Unix-like platforms. Instead of running C<make test>, set C<TEST_JOBS> in 1963your environment to the number of tests to run in parallel, and run 1964C<make test_harness>. On a Bourne-like shell, this can be done as 1965 1966 TEST_JOBS=3 make test_harness # Run 3 tests in parallel 1967 1968An environment variable is used, rather than parallel make itself, because 1969L<TAP::Harness> needs to be able to schedule individual non-conflicting test 1970scripts itself, and there is no standard interface to C<make> utilities to 1971interact with their job schedulers. 1972 1973Note that currently some test scripts may fail when run in parallel (most 1974notably C<ext/IO/t/io_dir.t>). If necessary run just the failing scripts 1975again sequentially and see if the failures go away. 1976=item test-notty test_notty 1977 1978Sets PERL_SKIP_TTY_TEST to true before running normal test. 1979 1980=back 1981 1982=head2 Running tests by hand 1983 1984You can run part of the test suite by hand by using one the following 1985commands from the F<t/> directory : 1986 1987 ./perl -I../lib TEST list-of-.t-files 1988 1989or 1990 1991 ./perl -I../lib harness list-of-.t-files 1992 1993(if you don't specify test scripts, the whole test suite will be run.) 1994 1995=head3 Using t/harness for testing 1996 1997If you use C<harness> for testing you have several command line options 1998available to you. The arguments are as follows, and are in the order 1999that they must appear if used together. 2000 2001 harness -v -torture -re=pattern LIST OF FILES TO TEST 2002 harness -v -torture -re LIST OF PATTERNS TO MATCH 2003 2004If C<LIST OF FILES TO TEST> is omitted the file list is obtained from 2005the manifest. The file list may include shell wildcards which will be 2006expanded out. 2007 2008=over 4 2009 2010=item -v 2011 2012Run the tests under verbose mode so you can see what tests were run, 2013and debug output. 2014 2015=item -torture 2016 2017Run the torture tests as well as the normal set. 2018 2019=item -re=PATTERN 2020 2021Filter the file list so that all the test files run match PATTERN. 2022Note that this form is distinct from the B<-re LIST OF PATTERNS> form below 2023in that it allows the file list to be provided as well. 2024 2025=item -re LIST OF PATTERNS 2026 2027Filter the file list so that all the test files run match 2028/(LIST|OF|PATTERNS)/. Note that with this form the patterns 2029are joined by '|' and you cannot supply a list of files, instead 2030the test files are obtained from the MANIFEST. 2031 2032=back 2033 2034You can run an individual test by a command similar to 2035 2036 ./perl -I../lib patho/to/foo.t 2037 2038except that the harnesses set up some environment variables that may 2039affect the execution of the test : 2040 2041=over 4 2042 2043=item PERL_CORE=1 2044 2045indicates that we're running this test part of the perl core test suite. 2046This is useful for modules that have a dual life on CPAN. 2047 2048=item PERL_DESTRUCT_LEVEL=2 2049 2050is set to 2 if it isn't set already (see L</PERL_DESTRUCT_LEVEL>) 2051 2052=item PERL 2053 2054(used only by F<t/TEST>) if set, overrides the path to the perl executable 2055that should be used to run the tests (the default being F<./perl>). 2056 2057=item PERL_SKIP_TTY_TEST 2058 2059if set, tells to skip the tests that need a terminal. It's actually set 2060automatically by the Makefile, but can also be forced artificially by 2061running 'make test_notty'. 2062 2063=back 2064 2065=head3 Other environment variables that may influence tests 2066 2067=over 4 2068 2069=item PERL_TEST_Net_Ping 2070 2071Setting this variable runs all the Net::Ping modules tests, 2072otherwise some tests that interact with the outside world are skipped. 2073See L<perl58delta>. 2074 2075=item PERL_TEST_NOVREXX 2076 2077Setting this variable skips the vrexx.t tests for OS2::REXX. 2078 2079=item PERL_TEST_NUMCONVERTS 2080 2081This sets a variable in op/numconvert.t. 2082 2083=back 2084 2085See also the documentation for the Test and Test::Harness modules, 2086for more environment variables that affect testing. 2087 2088=head2 Common problems when patching Perl source code 2089 2090Perl source plays by ANSI C89 rules: no C99 (or C++) extensions. In 2091some cases we have to take pre-ANSI requirements into consideration. 2092You don't care about some particular platform having broken Perl? 2093I hear there is still a strong demand for J2EE programmers. 2094 2095=head2 Perl environment problems 2096 2097=over 4 2098 2099=item * 2100 2101Not compiling with threading 2102 2103Compiling with threading (-Duseithreads) completely rewrites 2104the function prototypes of Perl. You better try your changes 2105with that. Related to this is the difference between "Perl_-less" 2106and "Perl_-ly" APIs, for example: 2107 2108 Perl_sv_setiv(aTHX_ ...); 2109 sv_setiv(...); 2110 2111The first one explicitly passes in the context, which is needed for e.g. 2112threaded builds. The second one does that implicitly; do not get them 2113mixed. If you are not passing in a aTHX_, you will need to do a dTHX 2114(or a dVAR) as the first thing in the function. 2115 2116See L<perlguts/"How multiple interpreters and concurrency are supported"> 2117for further discussion about context. 2118 2119=item * 2120 2121Not compiling with -DDEBUGGING 2122 2123The DEBUGGING define exposes more code to the compiler, 2124therefore more ways for things to go wrong. You should try it. 2125 2126=item * 2127 2128Introducing (non-read-only) globals 2129 2130Do not introduce any modifiable globals, truly global or file static. 2131They are bad form and complicate multithreading and other forms of 2132concurrency. The right way is to introduce them as new interpreter 2133variables, see F<intrpvar.h> (at the very end for binary compatibility). 2134 2135Introducing read-only (const) globals is okay, as long as you verify 2136with e.g. C<nm libperl.a|egrep -v ' [TURtr] '> (if your C<nm> has 2137BSD-style output) that the data you added really is read-only. 2138(If it is, it shouldn't show up in the output of that command.) 2139 2140If you want to have static strings, make them constant: 2141 2142 static const char etc[] = "..."; 2143 2144If you want to have arrays of constant strings, note carefully 2145the right combination of C<const>s: 2146 2147 static const char * const yippee[] = 2148 {"hi", "ho", "silver"}; 2149 2150There is a way to completely hide any modifiable globals (they are all 2151moved to heap), the compilation setting C<-DPERL_GLOBAL_STRUCT_PRIVATE>. 2152It is not normally used, but can be used for testing, read more 2153about it in L<perlguts/"Background and PERL_IMPLICIT_CONTEXT">. 2154 2155=item * 2156 2157Not exporting your new function 2158 2159Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any 2160function that is part of the public API (the shared Perl library) 2161to be explicitly marked as exported. See the discussion about 2162F<embed.pl> in L<perlguts>. 2163 2164=item * 2165 2166Exporting your new function 2167 2168The new shiny result of either genuine new functionality or your 2169arduous refactoring is now ready and correctly exported. So what 2170could possibly go wrong? 2171 2172Maybe simply that your function did not need to be exported in the 2173first place. Perl has a long and not so glorious history of exporting 2174functions that it should not have. 2175 2176If the function is used only inside one source code file, make it 2177static. See the discussion about F<embed.pl> in L<perlguts>. 2178 2179If the function is used across several files, but intended only for 2180Perl's internal use (and this should be the common case), do not 2181export it to the public API. See the discussion about F<embed.pl> 2182in L<perlguts>. 2183 2184=back 2185 2186=head2 Portability problems 2187 2188The following are common causes of compilation and/or execution 2189failures, not common to Perl as such. The C FAQ is good bedtime 2190reading. Please test your changes with as many C compilers and 2191platforms as possible; we will, anyway, and it's nice to save 2192oneself from public embarrassment. 2193 2194If using gcc, you can add the C<-std=c89> option which will hopefully 2195catch most of these unportabilities. (However it might also catch 2196incompatibilities in your system's header files.) 2197 2198Use the Configure C<-Dgccansipedantic> flag to enable the gcc 2199C<-ansi -pedantic> flags which enforce stricter ANSI rules. 2200 2201If using the C<gcc -Wall> note that not all the possible warnings 2202(like C<-Wunitialized>) are given unless you also compile with C<-O>. 2203 2204Note that if using gcc, starting from Perl 5.9.5 the Perl core source 2205code files (the ones at the top level of the source code distribution, 2206but not e.g. the extensions under ext/) are automatically compiled 2207with as many as possible of the C<-std=c89>, C<-ansi>, C<-pedantic>, 2208and a selection of C<-W> flags (see cflags.SH). 2209 2210Also study L<perlport> carefully to avoid any bad assumptions 2211about the operating system, filesystems, and so forth. 2212 2213You may once in a while try a "make microperl" to see whether we 2214can still compile Perl with just the bare minimum of interfaces. 2215(See README.micro.) 2216 2217Do not assume an operating system indicates a certain compiler. 2218 2219=over 4 2220 2221=item * 2222 2223Casting pointers to integers or casting integers to pointers 2224 2225 void castaway(U8* p) 2226 { 2227 IV i = p; 2228 2229or 2230 2231 void castaway(U8* p) 2232 { 2233 IV i = (IV)p; 2234 2235Both are bad, and broken, and unportable. Use the PTR2IV() 2236macro that does it right. (Likewise, there are PTR2UV(), PTR2NV(), 2237INT2PTR(), and NUM2PTR().) 2238 2239=item * 2240 2241Casting between data function pointers and data pointers 2242 2243Technically speaking casting between function pointers and data 2244pointers is unportable and undefined, but practically speaking 2245it seems to work, but you should use the FPTR2DPTR() and DPTR2FPTR() 2246macros. Sometimes you can also play games with unions. 2247 2248=item * 2249 2250Assuming sizeof(int) == sizeof(long) 2251 2252There are platforms where longs are 64 bits, and platforms where ints 2253are 64 bits, and while we are out to shock you, even platforms where 2254shorts are 64 bits. This is all legal according to the C standard. 2255(In other words, "long long" is not a portable way to specify 64 bits, 2256and "long long" is not even guaranteed to be any wider than "long".) 2257 2258Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth. 2259Avoid things like I32 because they are B<not> guaranteed to be 2260I<exactly> 32 bits, they are I<at least> 32 bits, nor are they 2261guaranteed to be B<int> or B<long>. If you really explicitly need 226264-bit variables, use I64 and U64, but only if guarded by HAS_QUAD. 2263 2264=item * 2265 2266Assuming one can dereference any type of pointer for any type of data 2267 2268 char *p = ...; 2269 long pony = *p; /* BAD */ 2270 2271Many platforms, quite rightly so, will give you a core dump instead 2272of a pony if the p happens not be correctly aligned. 2273 2274=item * 2275 2276Lvalue casts 2277 2278 (int)*p = ...; /* BAD */ 2279 2280Simply not portable. Get your lvalue to be of the right type, 2281or maybe use temporary variables, or dirty tricks with unions. 2282 2283=item * 2284 2285Assume B<anything> about structs (especially the ones you 2286don't control, like the ones coming from the system headers) 2287 2288=over 8 2289 2290=item * 2291 2292That a certain field exists in a struct 2293 2294=item * 2295 2296That no other fields exist besides the ones you know of 2297 2298=item * 2299 2300That a field is of certain signedness, sizeof, or type 2301 2302=item * 2303 2304That the fields are in a certain order 2305 2306=over 8 2307 2308=item * 2309 2310While C guarantees the ordering specified in the struct definition, 2311between different platforms the definitions might differ 2312 2313=back 2314 2315=item * 2316 2317That the sizeof(struct) or the alignments are the same everywhere 2318 2319=over 8 2320 2321=item * 2322 2323There might be padding bytes between the fields to align the fields - 2324the bytes can be anything 2325 2326=item * 2327 2328Structs are required to be aligned to the maximum alignment required 2329by the fields - which for native types is for usually equivalent to 2330sizeof() of the field 2331 2332=back 2333 2334=back 2335 2336=item * 2337 2338Assuming the character set is ASCIIish 2339 2340Perl can compile and run under EBCDIC platforms. See L<perlebcdic>. 2341This is transparent for the most part, but because the character sets 2342differ, you shouldn't use numeric (decimal, octal, nor hex) constants 2343to refer to characters. You can safely say 'A', but not 0x41. 2344You can safely say '\n', but not \012. 2345If a character doesn't have a trivial input form, you can 2346create a #define for it in both C<utfebcdic.h> and C<utf8.h>, so that 2347it resolves to different values depending on the character set being used. 2348(There are three different EBCDIC character sets defined in C<utfebcdic.h>, 2349so it might be best to insert the #define three times in that file.) 2350 2351Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper case 2352alphabetic characters. That is not true in EBCDIC. Nor for 'a' to 'z'. 2353But '0' - '9' is an unbroken range in both systems. Don't assume anything 2354about other ranges. 2355 2356Many of the comments in the existing code ignore the possibility of EBCDIC, 2357and may be wrong therefore, even if the code works. 2358This is actually a tribute to the successful transparent insertion of being 2359able to handle EBCDIC without having to change pre-existing code. 2360 2361UTF-8 and UTF-EBCDIC are two different encodings used to represent Unicode 2362code points as sequences of bytes. Macros 2363with the same names (but different definitions) 2364in C<utf8.h> and C<utfebcdic.h> 2365are used to allow the calling code to think that there is only one such 2366encoding. 2367This is almost always referred to as C<utf8>, but it means the EBCDIC version 2368as well. Again, comments in the code may well be wrong even if the code itself 2369is right. 2370For example, the concept of C<invariant characters> differs between ASCII and 2371EBCDIC. 2372On ASCII platforms, only characters that do not have the high-order 2373bit set (i.e. whose ordinals are strict ASCII, 0 - 127) 2374are invariant, and the documentation and comments in the code 2375may assume that, 2376often referring to something like, say, C<hibit>. 2377The situation differs and is not so simple on EBCDIC machines, but as long as 2378the code itself uses the C<NATIVE_IS_INVARIANT()> macro appropriately, it 2379works, even if the comments are wrong. 2380 2381=item * 2382 2383Assuming the character set is just ASCII 2384 2385ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra 2386characters have different meanings depending on the locale. Absent a locale, 2387currently these extra characters are generally considered to be unassigned, 2388and this has presented some problems. 2389This is being changed starting in 5.12 so that these characters will 2390be considered to be Latin-1 (ISO-8859-1). 2391 2392=item * 2393 2394Mixing #define and #ifdef 2395 2396 #define BURGLE(x) ... \ 2397 #ifdef BURGLE_OLD_STYLE /* BAD */ 2398 ... do it the old way ... \ 2399 #else 2400 ... do it the new way ... \ 2401 #endif 2402 2403You cannot portably "stack" cpp directives. For example in the above 2404you need two separate BURGLE() #defines, one for each #ifdef branch. 2405 2406=item * 2407 2408Adding non-comment stuff after #endif or #else 2409 2410 #ifdef SNOSH 2411 ... 2412 #else !SNOSH /* BAD */ 2413 ... 2414 #endif SNOSH /* BAD */ 2415 2416The #endif and #else cannot portably have anything non-comment after 2417them. If you want to document what is going (which is a good idea 2418especially if the branches are long), use (C) comments: 2419 2420 #ifdef SNOSH 2421 ... 2422 #else /* !SNOSH */ 2423 ... 2424 #endif /* SNOSH */ 2425 2426The gcc option C<-Wendif-labels> warns about the bad variant 2427(by default on starting from Perl 5.9.4). 2428 2429=item * 2430 2431Having a comma after the last element of an enum list 2432 2433 enum color { 2434 CERULEAN, 2435 CHARTREUSE, 2436 CINNABAR, /* BAD */ 2437 }; 2438 2439is not portable. Leave out the last comma. 2440 2441Also note that whether enums are implicitly morphable to ints 2442varies between compilers, you might need to (int). 2443 2444=item * 2445 2446Using //-comments 2447 2448 // This function bamfoodles the zorklator. /* BAD */ 2449 2450That is C99 or C++. Perl is C89. Using the //-comments is silently 2451allowed by many C compilers but cranking up the ANSI C89 strictness 2452(which we like to do) causes the compilation to fail. 2453 2454=item * 2455 2456Mixing declarations and code 2457 2458 void zorklator() 2459 { 2460 int n = 3; 2461 set_zorkmids(n); /* BAD */ 2462 int q = 4; 2463 2464That is C99 or C++. Some C compilers allow that, but you shouldn't. 2465 2466The gcc option C<-Wdeclaration-after-statements> scans for such problems 2467(by default on starting from Perl 5.9.4). 2468 2469=item * 2470 2471Introducing variables inside for() 2472 2473 for(int i = ...; ...; ...) { /* BAD */ 2474 2475That is C99 or C++. While it would indeed be awfully nice to have that 2476also in C89, to limit the scope of the loop variable, alas, we cannot. 2477 2478=item * 2479 2480Mixing signed char pointers with unsigned char pointers 2481 2482 int foo(char *s) { ... } 2483 ... 2484 unsigned char *t = ...; /* Or U8* t = ... */ 2485 foo(t); /* BAD */ 2486 2487While this is legal practice, it is certainly dubious, and downright 2488fatal in at least one platform: for example VMS cc considers this a 2489fatal error. One cause for people often making this mistake is that a 2490"naked char" and therefore dereferencing a "naked char pointer" have 2491an undefined signedness: it depends on the compiler and the flags of 2492the compiler and the underlying platform whether the result is signed 2493or unsigned. For this very same reason using a 'char' as an array 2494index is bad. 2495 2496=item * 2497 2498Macros that have string constants and their arguments as substrings of 2499the string constants 2500 2501 #define FOO(n) printf("number = %d\n", n) /* BAD */ 2502 FOO(10); 2503 2504Pre-ANSI semantics for that was equivalent to 2505 2506 printf("10umber = %d\10"); 2507 2508which is probably not what you were expecting. Unfortunately at least 2509one reasonably common and modern C compiler does "real backward 2510compatibility" here, in AIX that is what still happens even though the 2511rest of the AIX compiler is very happily C89. 2512 2513=item * 2514 2515Using printf formats for non-basic C types 2516 2517 IV i = ...; 2518 printf("i = %d\n", i); /* BAD */ 2519 2520While this might by accident work in some platform (where IV happens 2521to be an C<int>), in general it cannot. IV might be something larger. 2522Even worse the situation is with more specific types (defined by Perl's 2523configuration step in F<config.h>): 2524 2525 Uid_t who = ...; 2526 printf("who = %d\n", who); /* BAD */ 2527 2528The problem here is that Uid_t might be not only not C<int>-wide 2529but it might also be unsigned, in which case large uids would be 2530printed as negative values. 2531 2532There is no simple solution to this because of printf()'s limited 2533intelligence, but for many types the right format is available as 2534with either 'f' or '_f' suffix, for example: 2535 2536 IVdf /* IV in decimal */ 2537 UVxf /* UV is hexadecimal */ 2538 2539 printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */ 2540 2541 Uid_t_f /* Uid_t in decimal */ 2542 2543 printf("who = %"Uid_t_f"\n", who); 2544 2545Or you can try casting to a "wide enough" type: 2546 2547 printf("i = %"IVdf"\n", (IV)something_very_small_and_signed); 2548 2549Also remember that the C<%p> format really does require a void pointer: 2550 2551 U8* p = ...; 2552 printf("p = %p\n", (void*)p); 2553 2554The gcc option C<-Wformat> scans for such problems. 2555 2556=item * 2557 2558Blindly using variadic macros 2559 2560gcc has had them for a while with its own syntax, and C99 brought 2561them with a standardized syntax. Don't use the former, and use 2562the latter only if the HAS_C99_VARIADIC_MACROS is defined. 2563 2564=item * 2565 2566Blindly passing va_list 2567 2568Not all platforms support passing va_list to further varargs (stdarg) 2569functions. The right thing to do is to copy the va_list using the 2570Perl_va_copy() if the NEED_VA_COPY is defined. 2571 2572=item * 2573 2574Using gcc statement expressions 2575 2576 val = ({...;...;...}); /* BAD */ 2577 2578While a nice extension, it's not portable. The Perl code does 2579admittedly use them if available to gain some extra speed 2580(essentially as a funky form of inlining), but you shouldn't. 2581 2582=item * 2583 2584Binding together several statements in a macro 2585 2586Use the macros STMT_START and STMT_END. 2587 2588 STMT_START { 2589 ... 2590 } STMT_END 2591 2592=item * 2593 2594Testing for operating systems or versions when should be testing for features 2595 2596 #ifdef __FOONIX__ /* BAD */ 2597 foo = quux(); 2598 #endif 2599 2600Unless you know with 100% certainty that quux() is only ever available 2601for the "Foonix" operating system B<and> that is available B<and> 2602correctly working for B<all> past, present, B<and> future versions of 2603"Foonix", the above is very wrong. This is more correct (though still 2604not perfect, because the below is a compile-time check): 2605 2606 #ifdef HAS_QUUX 2607 foo = quux(); 2608 #endif 2609 2610How does the HAS_QUUX become defined where it needs to be? Well, if 2611Foonix happens to be Unixy enough to be able to run the Configure 2612script, and Configure has been taught about detecting and testing 2613quux(), the HAS_QUUX will be correctly defined. In other platforms, 2614the corresponding configuration step will hopefully do the same. 2615 2616In a pinch, if you cannot wait for Configure to be educated, 2617or if you have a good hunch of where quux() might be available, 2618you can temporarily try the following: 2619 2620 #if (defined(__FOONIX__) || defined(__BARNIX__)) 2621 # define HAS_QUUX 2622 #endif 2623 2624 ... 2625 2626 #ifdef HAS_QUUX 2627 foo = quux(); 2628 #endif 2629 2630But in any case, try to keep the features and operating systems separate. 2631 2632=back 2633 2634=head2 Problematic System Interfaces 2635 2636=over 4 2637 2638=item * 2639 2640malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable 2641allocate at least one byte. (In general you should rarely need to 2642work at this low level, but instead use the various malloc wrappers.) 2643 2644=item * 2645 2646snprintf() - the return type is unportable. Use my_snprintf() instead. 2647 2648=back 2649 2650=head2 Security problems 2651 2652Last but not least, here are various tips for safer coding. 2653 2654=over 4 2655 2656=item * 2657 2658Do not use gets() 2659 2660Or we will publicly ridicule you. Seriously. 2661 2662=item * 2663 2664Do not use strcpy() or strcat() or strncpy() or strncat() 2665 2666Use my_strlcpy() and my_strlcat() instead: they either use the native 2667implementation, or Perl's own implementation (borrowed from the public 2668domain implementation of INN). 2669 2670=item * 2671 2672Do not use sprintf() or vsprintf() 2673 2674If you really want just plain byte strings, use my_snprintf() 2675and my_vsnprintf() instead, which will try to use snprintf() and 2676vsnprintf() if those safer APIs are available. If you want something 2677fancier than a plain byte string, use SVs and Perl_sv_catpvf(). 2678 2679=back 2680 2681=head1 EXTERNAL TOOLS FOR DEBUGGING PERL 2682 2683Sometimes it helps to use external tools while debugging and 2684testing Perl. This section tries to guide you through using 2685some common testing and debugging tools with Perl. This is 2686meant as a guide to interfacing these tools with Perl, not 2687as any kind of guide to the use of the tools themselves. 2688 2689B<NOTE 1>: Running under memory debuggers such as Purify, valgrind, or 2690Third Degree greatly slows down the execution: seconds become minutes, 2691minutes become hours. For example as of Perl 5.8.1, the 2692ext/Encode/t/Unicode.t takes extraordinarily long to complete under 2693e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more 2694than six hours, even on a snappy computer. The said test must be 2695doing something that is quite unfriendly for memory debuggers. If you 2696don't feel like waiting, that you can simply kill away the perl 2697process. 2698 2699B<NOTE 2>: To minimize the number of memory leak false alarms (see 2700L</PERL_DESTRUCT_LEVEL> for more information), you have to set the 2701environment variable PERL_DESTRUCT_LEVEL to 2. 2702 2703For csh-like shells: 2704 2705 setenv PERL_DESTRUCT_LEVEL 2 2706 2707For Bourne-type shells: 2708 2709 PERL_DESTRUCT_LEVEL=2 2710 export PERL_DESTRUCT_LEVEL 2711 2712In Unixy environments you can also use the C<env> command: 2713 2714 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ... 2715 2716B<NOTE 3>: There are known memory leaks when there are compile-time 2717errors within eval or require, seeing C<S_doeval> in the call stack 2718is a good sign of these. Fixing these leaks is non-trivial, 2719unfortunately, but they must be fixed eventually. 2720 2721B<NOTE 4>: L<DynaLoader> will not clean up after itself completely 2722unless Perl is built with the Configure option 2723C<-Accflags=-DDL_UNLOAD_ALL_AT_EXIT>. 2724 2725=head2 Rational Software's Purify 2726 2727Purify is a commercial tool that is helpful in identifying 2728memory overruns, wild pointers, memory leaks and other such 2729badness. Perl must be compiled in a specific way for 2730optimal testing with Purify. Purify is available under 2731Windows NT, Solaris, HP-UX, SGI, and Siemens Unix. 2732 2733=head2 Purify on Unix 2734 2735On Unix, Purify creates a new Perl binary. To get the most 2736benefit out of Purify, you should create the perl to Purify 2737using: 2738 2739 sh Configure -Accflags=-DPURIFY -Doptimize='-g' \ 2740 -Uusemymalloc -Dusemultiplicity 2741 2742where these arguments mean: 2743 2744=over 4 2745 2746=item -Accflags=-DPURIFY 2747 2748Disables Perl's arena memory allocation functions, as well as 2749forcing use of memory allocation functions derived from the 2750system malloc. 2751 2752=item -Doptimize='-g' 2753 2754Adds debugging information so that you see the exact source 2755statements where the problem occurs. Without this flag, all 2756you will see is the source filename of where the error occurred. 2757 2758=item -Uusemymalloc 2759 2760Disable Perl's malloc so that Purify can more closely monitor 2761allocations and leaks. Using Perl's malloc will make Purify 2762report most leaks in the "potential" leaks category. 2763 2764=item -Dusemultiplicity 2765 2766Enabling the multiplicity option allows perl to clean up 2767thoroughly when the interpreter shuts down, which reduces the 2768number of bogus leak reports from Purify. 2769 2770=back 2771 2772Once you've compiled a perl suitable for Purify'ing, then you 2773can just: 2774 2775 make pureperl 2776 2777which creates a binary named 'pureperl' that has been Purify'ed. 2778This binary is used in place of the standard 'perl' binary 2779when you want to debug Perl memory problems. 2780 2781As an example, to show any memory leaks produced during the 2782standard Perl testset you would create and run the Purify'ed 2783perl as: 2784 2785 make pureperl 2786 cd t 2787 ../pureperl -I../lib harness 2788 2789which would run Perl on test.pl and report any memory problems. 2790 2791Purify outputs messages in "Viewer" windows by default. If 2792you don't have a windowing environment or if you simply 2793want the Purify output to unobtrusively go to a log file 2794instead of to the interactive window, use these following 2795options to output to the log file "perl.log": 2796 2797 setenv PURIFYOPTIONS "-chain-length=25 -windows=no \ 2798 -log-file=perl.log -append-logfile=yes" 2799 2800If you plan to use the "Viewer" windows, then you only need this option: 2801 2802 setenv PURIFYOPTIONS "-chain-length=25" 2803 2804In Bourne-type shells: 2805 2806 PURIFYOPTIONS="..." 2807 export PURIFYOPTIONS 2808 2809or if you have the "env" utility: 2810 2811 env PURIFYOPTIONS="..." ../pureperl ... 2812 2813=head2 Purify on NT 2814 2815Purify on Windows NT instruments the Perl binary 'perl.exe' 2816on the fly. There are several options in the makefile you 2817should change to get the most use out of Purify: 2818 2819=over 4 2820 2821=item DEFINES 2822 2823You should add -DPURIFY to the DEFINES line so the DEFINES 2824line looks something like: 2825 2826 DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1 2827 2828to disable Perl's arena memory allocation functions, as 2829well as to force use of memory allocation functions derived 2830from the system malloc. 2831 2832=item USE_MULTI = define 2833 2834Enabling the multiplicity option allows perl to clean up 2835thoroughly when the interpreter shuts down, which reduces the 2836number of bogus leak reports from Purify. 2837 2838=item #PERL_MALLOC = define 2839 2840Disable Perl's malloc so that Purify can more closely monitor 2841allocations and leaks. Using Perl's malloc will make Purify 2842report most leaks in the "potential" leaks category. 2843 2844=item CFG = Debug 2845 2846Adds debugging information so that you see the exact source 2847statements where the problem occurs. Without this flag, all 2848you will see is the source filename of where the error occurred. 2849 2850=back 2851 2852As an example, to show any memory leaks produced during the 2853standard Perl testset you would create and run Purify as: 2854 2855 cd win32 2856 make 2857 cd ../t 2858 purify ../perl -I../lib harness 2859 2860which would instrument Perl in memory, run Perl on test.pl, 2861then finally report any memory problems. 2862 2863=head2 valgrind 2864 2865The excellent valgrind tool can be used to find out both memory leaks 2866and illegal memory accesses. As of version 3.3.0, Valgrind only 2867supports Linux on x86, x86-64 and PowerPC. The special "test.valgrind" 2868target can be used to run the tests under valgrind. Found errors 2869and memory leaks are logged in files named F<testfile.valgrind>. 2870 2871Valgrind also provides a cachegrind tool, invoked on perl as: 2872 2873 VG_OPTS=--tool=cachegrind make test.valgrind 2874 2875As system libraries (most notably glibc) are also triggering errors, 2876valgrind allows to suppress such errors using suppression files. The 2877default suppression file that comes with valgrind already catches a lot 2878of them. Some additional suppressions are defined in F<t/perl.supp>. 2879 2880To get valgrind and for more information see 2881 2882 http://developer.kde.org/~sewardj/ 2883 2884=head2 Compaq's/Digital's/HP's Third Degree 2885 2886Third Degree is a tool for memory leak detection and memory access checks. 2887It is one of the many tools in the ATOM toolkit. The toolkit is only 2888available on Tru64 (formerly known as Digital UNIX formerly known as 2889DEC OSF/1). 2890 2891When building Perl, you must first run Configure with -Doptimize=-g 2892and -Uusemymalloc flags, after that you can use the make targets 2893"perl.third" and "test.third". (What is required is that Perl must be 2894compiled using the C<-g> flag, you may need to re-Configure.) 2895 2896The short story is that with "atom" you can instrument the Perl 2897executable to create a new executable called F<perl.third>. When the 2898instrumented executable is run, it creates a log of dubious memory 2899traffic in file called F<perl.3log>. See the manual pages of atom and 2900third for more information. The most extensive Third Degree 2901documentation is available in the Compaq "Tru64 UNIX Programmer's 2902Guide", chapter "Debugging Programs with Third Degree". 2903 2904The "test.third" leaves a lot of files named F<foo_bar.3log> in the t/ 2905subdirectory. There is a problem with these files: Third Degree is so 2906effective that it finds problems also in the system libraries. 2907Therefore you should used the Porting/thirdclean script to cleanup 2908the F<*.3log> files. 2909 2910There are also leaks that for given certain definition of a leak, 2911aren't. See L</PERL_DESTRUCT_LEVEL> for more information. 2912 2913=head2 PERL_DESTRUCT_LEVEL 2914 2915If you want to run any of the tests yourself manually using e.g. 2916valgrind, or the pureperl or perl.third executables, please note that 2917by default perl B<does not> explicitly cleanup all the memory it has 2918allocated (such as global memory arenas) but instead lets the exit() 2919of the whole program "take care" of such allocations, also known as 2920"global destruction of objects". 2921 2922There is a way to tell perl to do complete cleanup: set the 2923environment variable PERL_DESTRUCT_LEVEL to a non-zero value. 2924The t/TEST wrapper does set this to 2, and this is what you 2925need to do too, if you don't want to see the "global leaks": 2926For example, for "third-degreed" Perl: 2927 2928 env PERL_DESTRUCT_LEVEL=2 ./perl.third -Ilib t/foo/bar.t 2929 2930(Note: the mod_perl apache module uses also this environment variable 2931for its own purposes and extended its semantics. Refer to the mod_perl 2932documentation for more information. Also, spawned threads do the 2933equivalent of setting this variable to the value 1.) 2934 2935If, at the end of a run you get the message I<N scalars leaked>, you can 2936recompile with C<-DDEBUG_LEAKING_SCALARS>, which will cause the addresses 2937of all those leaked SVs to be dumped along with details as to where each 2938SV was originally allocated. This information is also displayed by 2939Devel::Peek. Note that the extra details recorded with each SV increases 2940memory usage, so it shouldn't be used in production environments. It also 2941converts C<new_SV()> from a macro into a real function, so you can use 2942your favourite debugger to discover where those pesky SVs were allocated. 2943 2944If you see that you're leaking memory at runtime, but neither valgrind 2945nor C<-DDEBUG_LEAKING_SCALARS> will find anything, you're probably 2946leaking SVs that are still reachable and will be properly cleaned up 2947during destruction of the interpreter. In such cases, using the C<-Dm> 2948switch can point you to the source of the leak. If the executable was 2949built with C<-DDEBUG_LEAKING_SCALARS>, C<-Dm> will output SV allocations 2950in addition to memory allocations. Each SV allocation has a distinct 2951serial number that will be written on creation and destruction of the SV. 2952So if you're executing the leaking code in a loop, you need to look for 2953SVs that are created, but never destroyed between each cycle. If such an 2954SV is found, set a conditional breakpoint within C<new_SV()> and make it 2955break only when C<PL_sv_serial> is equal to the serial number of the 2956leaking SV. Then you will catch the interpreter in exactly the state 2957where the leaking SV is allocated, which is sufficient in many cases to 2958find the source of the leak. 2959 2960As C<-Dm> is using the PerlIO layer for output, it will by itself 2961allocate quite a bunch of SVs, which are hidden to avoid recursion. 2962You can bypass the PerlIO layer if you use the SV logging provided 2963by C<-DPERL_MEM_LOG> instead. 2964 2965=head2 PERL_MEM_LOG 2966 2967If compiled with C<-DPERL_MEM_LOG>, both memory and SV allocations go 2968through logging functions, which is handy for breakpoint setting. 2969 2970Unless C<-DPERL_MEM_LOG_NOIMPL> is also compiled, the logging 2971functions read $ENV{PERL_MEM_LOG} to determine whether to log the 2972event, and if so how: 2973 2974 $ENV{PERL_MEM_LOG} =~ /m/ Log all memory ops 2975 $ENV{PERL_MEM_LOG} =~ /s/ Log all SV ops 2976 $ENV{PERL_MEM_LOG} =~ /t/ include timestamp in Log 2977 $ENV{PERL_MEM_LOG} =~ /^(\d+)/ write to FD given (default is 2) 2978 2979Memory logging is somewhat similar to C<-Dm> but is independent of 2980C<-DDEBUGGING>, and at a higher level; all uses of Newx(), Renew(), 2981and Safefree() are logged with the caller's source code file and line 2982number (and C function name, if supported by the C compiler). In 2983contrast, C<-Dm> is directly at the point of C<malloc()>. SV logging 2984is similar. 2985 2986Since the logging doesn't use PerlIO, all SV allocations are logged 2987and no extra SV allocations are introduced by enabling the logging. 2988If compiled with C<-DDEBUG_LEAKING_SCALARS>, the serial number for 2989each SV allocation is also logged. 2990 2991=head2 Profiling 2992 2993Depending on your platform there are various of profiling Perl. 2994 2995There are two commonly used techniques of profiling executables: 2996I<statistical time-sampling> and I<basic-block counting>. 2997 2998The first method takes periodically samples of the CPU program 2999counter, and since the program counter can be correlated with the code 3000generated for functions, we get a statistical view of in which 3001functions the program is spending its time. The caveats are that very 3002small/fast functions have lower probability of showing up in the 3003profile, and that periodically interrupting the program (this is 3004usually done rather frequently, in the scale of milliseconds) imposes 3005an additional overhead that may skew the results. The first problem 3006can be alleviated by running the code for longer (in general this is a 3007good idea for profiling), the second problem is usually kept in guard 3008by the profiling tools themselves. 3009 3010The second method divides up the generated code into I<basic blocks>. 3011Basic blocks are sections of code that are entered only in the 3012beginning and exited only at the end. For example, a conditional jump 3013starts a basic block. Basic block profiling usually works by 3014I<instrumenting> the code by adding I<enter basic block #nnnn> 3015book-keeping code to the generated code. During the execution of the 3016code the basic block counters are then updated appropriately. The 3017caveat is that the added extra code can skew the results: again, the 3018profiling tools usually try to factor their own effects out of the 3019results. 3020 3021=head2 Gprof Profiling 3022 3023gprof is a profiling tool available in many Unix platforms, 3024it uses F<statistical time-sampling>. 3025 3026You can build a profiled version of perl called "perl.gprof" by 3027invoking the make target "perl.gprof" (What is required is that Perl 3028must be compiled using the C<-pg> flag, you may need to re-Configure). 3029Running the profiled version of Perl will create an output file called 3030F<gmon.out> is created which contains the profiling data collected 3031during the execution. 3032 3033The gprof tool can then display the collected data in various ways. 3034Usually gprof understands the following options: 3035 3036=over 4 3037 3038=item -a 3039 3040Suppress statically defined functions from the profile. 3041 3042=item -b 3043 3044Suppress the verbose descriptions in the profile. 3045 3046=item -e routine 3047 3048Exclude the given routine and its descendants from the profile. 3049 3050=item -f routine 3051 3052Display only the given routine and its descendants in the profile. 3053 3054=item -s 3055 3056Generate a summary file called F<gmon.sum> which then may be given 3057to subsequent gprof runs to accumulate data over several runs. 3058 3059=item -z 3060 3061Display routines that have zero usage. 3062 3063=back 3064 3065For more detailed explanation of the available commands and output 3066formats, see your own local documentation of gprof. 3067 3068quick hint: 3069 3070 $ sh Configure -des -Dusedevel -Doptimize='-pg' && make perl.gprof 3071 $ ./perl.gprof someprog # creates gmon.out in current directory 3072 $ gprof ./perl.gprof > out 3073 $ view out 3074 3075=head2 GCC gcov Profiling 3076 3077Starting from GCC 3.0 I<basic block profiling> is officially available 3078for the GNU CC. 3079 3080You can build a profiled version of perl called F<perl.gcov> by 3081invoking the make target "perl.gcov" (what is required that Perl must 3082be compiled using gcc with the flags C<-fprofile-arcs 3083-ftest-coverage>, you may need to re-Configure). 3084 3085Running the profiled version of Perl will cause profile output to be 3086generated. For each source file an accompanying ".da" file will be 3087created. 3088 3089To display the results you use the "gcov" utility (which should 3090be installed if you have gcc 3.0 or newer installed). F<gcov> is 3091run on source code files, like this 3092 3093 gcov sv.c 3094 3095which will cause F<sv.c.gcov> to be created. The F<.gcov> files 3096contain the source code annotated with relative frequencies of 3097execution indicated by "#" markers. 3098 3099Useful options of F<gcov> include C<-b> which will summarise the 3100basic block, branch, and function call coverage, and C<-c> which 3101instead of relative frequencies will use the actual counts. For 3102more information on the use of F<gcov> and basic block profiling 3103with gcc, see the latest GNU CC manual, as of GCC 3.0 see 3104 3105 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html 3106 3107and its section titled "8. gcov: a Test Coverage Program" 3108 3109 http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132 3110 3111quick hint: 3112 3113 $ sh Configure -des -Doptimize='-g' -Accflags='-fprofile-arcs -ftest-coverage' \ 3114 -Aldflags='-fprofile-arcs -ftest-coverage' && make perl.gcov 3115 $ rm -f regexec.c.gcov regexec.gcda 3116 $ ./perl.gcov 3117 $ gcov regexec.c 3118 $ view regexec.c.gcov 3119 3120=head2 Pixie Profiling 3121 3122Pixie is a profiling tool available on IRIX and Tru64 (aka Digital 3123UNIX aka DEC OSF/1) platforms. Pixie does its profiling using 3124I<basic-block counting>. 3125 3126You can build a profiled version of perl called F<perl.pixie> by 3127invoking the make target "perl.pixie" (what is required is that Perl 3128must be compiled using the C<-g> flag, you may need to re-Configure). 3129 3130In Tru64 a file called F<perl.Addrs> will also be silently created, 3131this file contains the addresses of the basic blocks. Running the 3132profiled version of Perl will create a new file called "perl.Counts" 3133which contains the counts for the basic block for that particular 3134program execution. 3135 3136To display the results you use the F<prof> utility. The exact 3137incantation depends on your operating system, "prof perl.Counts" in 3138IRIX, and "prof -pixie -all -L. perl" in Tru64. 3139 3140In IRIX the following prof options are available: 3141 3142=over 4 3143 3144=item -h 3145 3146Reports the most heavily used lines in descending order of use. 3147Useful for finding the hotspot lines. 3148 3149=item -l 3150 3151Groups lines by procedure, with procedures sorted in descending order of use. 3152Within a procedure, lines are listed in source order. 3153Useful for finding the hotspots of procedures. 3154 3155=back 3156 3157In Tru64 the following options are available: 3158 3159=over 4 3160 3161=item -p[rocedures] 3162 3163Procedures sorted in descending order by the number of cycles executed 3164in each procedure. Useful for finding the hotspot procedures. 3165(This is the default option.) 3166 3167=item -h[eavy] 3168 3169Lines sorted in descending order by the number of cycles executed in 3170each line. Useful for finding the hotspot lines. 3171 3172=item -i[nvocations] 3173 3174The called procedures are sorted in descending order by number of calls 3175made to the procedures. Useful for finding the most used procedures. 3176 3177=item -l[ines] 3178 3179Grouped by procedure, sorted by cycles executed per procedure. 3180Useful for finding the hotspots of procedures. 3181 3182=item -testcoverage 3183 3184The compiler emitted code for these lines, but the code was unexecuted. 3185 3186=item -z[ero] 3187 3188Unexecuted procedures. 3189 3190=back 3191 3192For further information, see your system's manual pages for pixie and prof. 3193 3194=head2 Miscellaneous tricks 3195 3196=over 4 3197 3198=item * 3199 3200Those debugging perl with the DDD frontend over gdb may find the 3201following useful: 3202 3203You can extend the data conversion shortcuts menu, so for example you 3204can display an SV's IV value with one click, without doing any typing. 3205To do that simply edit ~/.ddd/init file and add after: 3206 3207 ! Display shortcuts. 3208 Ddd*gdbDisplayShortcuts: \ 3209 /t () // Convert to Bin\n\ 3210 /d () // Convert to Dec\n\ 3211 /x () // Convert to Hex\n\ 3212 /o () // Convert to Oct(\n\ 3213 3214the following two lines: 3215 3216 ((XPV*) (())->sv_any )->xpv_pv // 2pvx\n\ 3217 ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx 3218 3219so now you can do ivx and pvx lookups or you can plug there the 3220sv_peek "conversion": 3221 3222 Perl_sv_peek(my_perl, (SV*)()) // sv_peek 3223 3224(The my_perl is for threaded builds.) 3225Just remember that every line, but the last one, should end with \n\ 3226 3227Alternatively edit the init file interactively via: 32283rd mouse button -> New Display -> Edit Menu 3229 3230Note: you can define up to 20 conversion shortcuts in the gdb 3231section. 3232 3233=item * 3234 3235If you see in a debugger a memory area mysteriously full of 0xABABABAB 3236or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros, 3237see L<perlclib>. 3238 3239=item * 3240 3241Under ithreads the optree is read only. If you want to enforce this, to check 3242for write accesses from buggy code, compile with C<-DPL_OP_SLAB_ALLOC> to 3243enable the OP slab allocator and C<-DPERL_DEBUG_READONLY_OPS> to enable code 3244that allocates op memory via C<mmap>, and sets it read-only at run time. 3245Any write access to an op results in a C<SIGBUS> and abort. 3246 3247This code is intended for development only, and may not be portable even to 3248all Unix variants. Also, it is an 80% solution, in that it isn't able to make 3249all ops read only. Specifically it 3250 3251=over 3252 3253=item 1 3254 3255Only sets read-only on all slabs of ops at C<CHECK> time, hence ops allocated 3256later via C<require> or C<eval> will be re-write 3257 3258=item 2 3259 3260Turns an entire slab of ops read-write if the refcount of any op in the slab 3261needs to be decreased. 3262 3263=item 3 3264 3265Turns an entire slab of ops read-write if any op from the slab is freed. 3266 3267=back 3268 3269It's not possible to turn the slabs to read-only after an action requiring 3270read-write access, as either can happen during op tree building time, so 3271there may still be legitimate write access. 3272 3273However, as an 80% solution it is still effective, as currently it catches 3274a write access during the generation of F<Config.pm>, which means that we 3275can't yet build F<perl> with this enabled. 3276 3277=back 3278 3279 3280=head1 CONCLUSION 3281 3282We've had a brief look around the Perl source, how to maintain quality 3283of the source code, an overview of the stages F<perl> goes through 3284when it's running your code, how to use debuggers to poke at the Perl 3285guts, and finally how to analyse the execution of Perl. We took a very 3286simple problem and demonstrated how to solve it fully - with 3287documentation, regression tests, and finally a patch for submission to 3288p5p. Finally, we talked about how to use external tools to debug and 3289test Perl. 3290 3291I'd now suggest you read over those references again, and then, as soon 3292as possible, get your hands dirty. The best way to learn is by doing, 3293so: 3294 3295=over 3 3296 3297=item * 3298 3299Subscribe to perl5-porters, follow the patches and try and understand 3300them; don't be afraid to ask if there's a portion you're not clear on - 3301who knows, you may unearth a bug in the patch... 3302 3303=item * 3304 3305Keep up to date with the bleeding edge Perl distributions and get 3306familiar with the changes. Try and get an idea of what areas people are 3307working on and the changes they're making. 3308 3309=item * 3310 3311Do read the README associated with your operating system, e.g. README.aix 3312on the IBM AIX OS. Don't hesitate to supply patches to that README if 3313you find anything missing or changed over a new OS release. 3314 3315=item * 3316 3317Find an area of Perl that seems interesting to you, and see if you can 3318work out how it works. Scan through the source, and step over it in the 3319debugger. Play, poke, investigate, fiddle! You'll probably get to 3320understand not just your chosen area but a much wider range of F<perl>'s 3321activity as well, and probably sooner than you'd think. 3322 3323=back 3324 3325=over 3 3326 3327=item I<The Road goes ever on and on, down from the door where it began.> 3328 3329=back 3330 3331If you can do these things, you've started on the long road to Perl porting. 3332Thanks for wanting to help make Perl better - and happy hacking! 3333 3334=head2 Metaphoric Quotations 3335 3336If you recognized the quote about the Road above, you're in luck. 3337 3338Most software projects begin each file with a literal description of each 3339file's purpose. Perl instead begins each with a literary allusion to that 3340file's purpose. 3341 3342Like chapters in many books, all top-level Perl source files (along with a 3343few others here and there) begin with an epigramic inscription that alludes, 3344indirectly and metaphorically, to the material you're about to read. 3345 3346Quotations are taken from writings of J.R.R Tolkien pertaining to his 3347Legendarium, almost always from I<The Lord of the Rings>. Chapters and 3348page numbers are given using the following editions: 3349 3350=over 4 3351 3352=item * 3353 3354I<The Hobbit>, by J.R.R. Tolkien. The hardcover, 70th-anniversary 3355edition of 2007 was used, published in the UK by Harper Collins Publishers 3356and in the US by the Houghton Mifflin Company. 3357 3358=item * 3359 3360I<The Lord of the Rings>, by J.R.R. Tolkien. The hardcover, 336150th-anniversary edition of 2004 was used, published in the UK by Harper 3362Collins Publishers and in the US by the Houghton Mifflin Company. 3363 3364=item * 3365 3366I<The Lays of Beleriand>, by J.R.R. Tolkien and published posthumously by his 3367son and literary executor, C.J.R. Tolkien, being the 3rd of the 12 volumes 3368in Christopher's mammoth I<History of Middle Earth>. Page numbers derive 3369from the hardcover edition, first published in 1983 by George Allen & 3370Unwin; no page numbers changed for the special 3-volume omnibus edition of 33712002 or the various trade-paper editions, all again now by Harper Collins 3372or Houghton Mifflin. 3373 3374=back 3375 3376Other JRRT books fair game for quotes would thus include I<The Adventures of 3377Tom Bombadil>, I<The Silmarillion>, I<Unfinished Tales>, and I<The Tale of 3378the Children of Hurin>, all but the first posthumously assembled by CJRT. 3379But I<The Lord of the Rings> itself is perfectly fine and probably best to 3380quote from, provided you can find a suitable quote there. 3381 3382So if you were to supply a new, complete, top-level source file to add to 3383Perl, you should conform to this peculiar practice by yourself selecting an 3384appropriate quotation from Tolkien, retaining the original spelling and 3385punctuation and using the same format the rest of the quotes are in. 3386Indirect and oblique is just fine; remember, it's a metaphor, so being meta 3387is, after all, what it's for. 3388 3389=head1 AUTHOR 3390 3391This document was written by Nathan Torkington, and is maintained by 3392the perl5-porters mailing list. 3393 3394=head1 SEE ALSO 3395 3396L<perlrepository> 3397