NAME

perlhack - How to hack at the Perl internals

DESCRIPTION

This document attempts to explain how Perl development takes place, and ends with some suggestions for people wanting to become bona fide porters.

The perl5-porters mailing list is where the Perl standard distribution is maintained and developed. The list can get anywhere from 10 to 150 messages a day, depending on the heatedness of the debate. Most days there are two or three patches, extensions, features, or bugs being discussed at a time.

A searchable archive of the list is at either:

http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/

or

http://archive.develooper.com/perl5-porters@perl.org/

List subscribers (the porters themselves) come in several flavours. Some are quiet curious lurkers, who rarely pitch in and instead watch the ongoing development to ensure they're forewarned of new changes or features in Perl. Some are representatives of vendors, who are there to make sure that Perl continues to compile and work on their platforms. Some patch any reported bug that they know how to fix, some are actively patching their pet area (threads, Win32, the regexp engine), while others seem to do nothing but complain. In other words, it's your usual mix of technical people.

Over this group of porters presides Larry Wall. He has the final word in what does and does not change in the Perl language. Various releases of Perl are shepherded by a "pumpking", a porter responsible for gathering patches, deciding on a patch-by-patch, feature-by-feature basis what will and will not go into the release. For instance, Gurusamy Sarathy was the pumpking for the 5.6 release of Perl, and Jarkko Hietaniemi was the pumpking for the 5.8 release, and Rafael Garcia-Suarez holds the pumpking crown for the 5.10 release.

In addition, various people are pumpkings for different things. For instance, Andy Dougherty and Jarkko Hietaniemi did a grand job as the Configure pumpkin up till the 5.8 release. For the 5.10 release H.Merijn Brand took over.

Larry sees Perl development along the lines of the US government: there's the Legislature (the porters), the Executive branch (the pumpkings), and the Supreme Court (Larry). The legislature can discuss and submit patches to the executive branch all they like, but the executive branch is free to veto them. Rarely, the Supreme Court will side with the executive branch over the legislature, or the legislature over the executive branch. Mostly, however, the legislature and the executive branch are supposed to get along and work out their differences without impeachment or court cases.

You might sometimes see reference to Rule 1 and Rule 2. Larry's power as Supreme Court is expressed in The Rules:

1. Larry is always by definition right about how Perl should behave. This means he has final veto power on the core functionality.

2. Larry is allowed to change his mind about any matter at a later date, regardless of whether he previously invoked Rule 1.

Got that? Larry is always right, even when he was wrong. It's rare to see either Rule exercised, but they are often alluded to.

New features and extensions to the language are contentious, because the criteria used by the pumpkings, Larry, and other porters to decide which features should be implemented and incorporated are not codified in a few small design goals as with some other languages. Instead, the heuristics are flexible and often difficult to fathom. Here is one person's list, roughly in decreasing order of importance, of heuristics that new features have to be weighed against:

Does concept match the general goals of Perl?

These haven't been written anywhere in stone, but one approximation is:

1. Keep it fast, simple, and useful.
2. Keep features/concepts as orthogonal as possible.
3. No arbitrary limits (platforms, data sizes, cultures).
4. Keep it open and exciting to use/patch/advocate Perl everywhere.
5. Either assimilate new technologies, or build bridges to them.

Where is the implementation?

All the talk in the world is useless without an implementation. In almost every case, the person or people who argue for a new feature will be expected to be the ones who implement it. Porters capable of coding new features have their own agendas, and are not available to implement your (possibly good) idea.

Backwards compatibility

It's a cardinal sin to break existing Perl programs. New warnings are contentious--some say that a program that emits warnings is not broken, while others say it is. Adding keywords has the potential to break programs, changing the meaning of existing token sequences or functions might break programs.

Could it be a module instead?

Perl 5 has extension mechanisms, modules and XS, specifically to avoid the need to keep changing the Perl interpreter. You can write modules that export functions, you can give those functions prototypes so they can be called like built-in functions, you can even write XS code to mess with the runtime data structures of the Perl interpreter if you want to implement really complicated things. If it can be done in a module instead of in the core, it's highly unlikely to be added.

Is the feature generic enough?

Is this something that only the submitter wants added to the language, or would it be broadly useful? Sometimes, instead of adding a feature with a tight focus, the porters might decide to wait until someone implements the more generalized feature. For instance, instead of implementing a "delayed evaluation" feature, the porters are waiting for a macro system that would permit delayed evaluation and much more.

Does it potentially introduce new bugs?

Radical rewrites of large chunks of the Perl interpreter have the potential to introduce new bugs. The smaller and more localized the change, the better.

Does it preclude other desirable features?

A patch is likely to be rejected if it closes off future avenues of development. For instance, a patch that placed a true and final interpretation on prototypes is likely to be rejected because there are still options for the future of prototypes that haven't been addressed.

Is the implementation robust?

Good patches (tight code, complete, correct) stand more chance of going in. Sloppy or incorrect patches might be placed on the back burner until the pumpking has time to fix, or might be discarded altogether without further notice.

Is the implementation generic enough to be portable?

The worst patches make use of a system-specific features. It's highly unlikely that non-portable additions to the Perl language will be accepted.

Is the implementation tested?

Patches which change behaviour (fixing bugs or introducing new features) must include regression tests to verify that everything works as expected. Without tests provided by the original author, how can anyone else changing perl in the future be sure that they haven't unwittingly broken the behaviour the patch implements? And without tests, how can the patch's author be confident that his/her hard work put into the patch won't be accidentally thrown away by someone in the future?

Is there enough documentation?

Patches without documentation are probably ill-thought out or incomplete. Nothing can be added without documentation, so submitting a patch for the appropriate manpages as well as the source code is always a good idea.

Is there another way to do it?

Larry said "Although the Perl Slogan is There's More Than One Way to Do It, I hesitate to make 10 ways to do something". This is a tricky heuristic to navigate, though--one man's essential addition is another man's pointless cruft.

Does it create too much work?

Work for the pumpking, work for Perl programmers, work for module authors, ... Perl is supposed to be easy.

Patches speak louder than words

Working code is always preferred to pie-in-the-sky ideas. A patch to add a feature stands a much higher chance of making it to the language than does a random feature request, no matter how fervently argued the request might be. This ties into "Will it be useful?", as the fact that someone took the time to make the patch demonstrates a strong desire for the feature.

If you're on the list, you might hear the word "core" bandied around. It refers to the standard distribution. "Hacking on the core" means you're changing the C source code to the Perl interpreter. "A core module" is one that ships with Perl.

Keeping in sync

The source code to the Perl interpreter, in its different versions, is kept in a repository managed by the git revision control system. The pumpkings and a few others have write access to the repository to check in changes.

How to clone and use the git perl repository is described in perlrepository.

You can also choose to use rsync to get a copy of the current source tree for the bleadperl branch and all maintenance branches :

$ rsync -avz rsync://perl5.git.perl.org/APC/perl-current .
$ rsync -avz rsync://perl5.git.perl.org/APC/perl-5.10.x .
$ rsync -avz rsync://perl5.git.perl.org/APC/perl-5.8.x .
$ rsync -avz rsync://perl5.git.perl.org/APC/perl-5.6.x .
$ rsync -avz rsync://perl5.git.perl.org/APC/perl-5.005xx .

(Add the --delete option to remove leftover files)

You may also want to subscribe to the perl5-changes mailing list to receive a copy of each patch that gets submitted to the maintenance and development "branches" of the perl repository. See http://lists.perl.org/ for subscription information.

If you are a member of the perl5-porters mailing list, it is a good thing to keep in touch with the most recent changes. If not only to verify if what you would have posted as a bug report isn't already solved in the most recent available perl development branch, also known as perl-current, bleading edge perl, bleedperl or bleadperl.

Needless to say, the source code in perl-current is usually in a perpetual state of evolution. You should expect it to be very buggy. Do not use it for any purpose other than testing and development.

Perlbug administration

There is a single remote administrative interface for modifying bug status, category, open issues etc. using the RT bugtracker system, maintained by Robert Spier. Become an administrator, and close any bugs you can get your sticky mitts on:

http://bugs.perl.org/

To email the bug system administrators:

"perlbug-admin" <perlbug-admin@perl.org>

Submitting patches

Always submit patches to perl5-porters@perl.org. If you're patching a core module and there's an author listed, send the author a copy (see "Patching a core module"). This lets other porters review your patch, which catches a surprising number of errors in patches. Please patch against the latest development version. (e.g., even if you're fixing a bug in the 5.8 track, patch against the blead branch in the git repository.)

If changes are accepted, they are applied to the development branch. Then the maintenance pumpking decides which of those patches is to be backported to the maint branch. Only patches that survive the heat of the development branch get applied to maintenance versions.

Your patch should update the documentation and test suite. See "Writing a test". If you have added or removed files in the distribution, edit the MANIFEST file accordingly, sort the MANIFEST file using make manisort, and include those changes as part of your patch.

Patching documentation also follows the same order: if accepted, a patch is first applied to development, and if relevant then it's backported to maintenance. (With an exception for some patches that document behaviour that only appears in the maintenance branch, but which has changed in the development version.)

To report a bug in Perl, use the program perlbug which comes with Perl (if you can't get Perl to work, send mail to the address perlbug@perl.org or perlbug@perl.com). Reporting bugs through perlbug feeds into the automated bug-tracking system, access to which is provided through the web at http://rt.perl.org/rt3/ . It often pays to check the archives of the perl5-porters mailing list to see whether the bug you're reporting has been reported before, and if so whether it was considered a bug. See above for the location of the searchable archives.

The CPAN testers ( http://testers.cpan.org/ ) are a group of volunteers who test CPAN modules on a variety of platforms. Perl Smokers ( http://www.nntp.perl.org/group/perl.daily-build and http://www.nntp.perl.org/group/perl.daily-build.reports/ ) automatically test Perl source releases on platforms with various configurations. Both efforts welcome volunteers. In order to get involved in smoke testing of the perl itself visit http://search.cpan.org/dist/Test-Smoke. In order to start smoke testing CPAN modules visit http://search.cpan.org/dist/CPANPLUS-YACSmoke/ or http://search.cpan.org/dist/minismokebox/ or http://search.cpan.org/dist/CPAN-Reporter/.

It's a good idea to read and lurk for a while before chipping in. That way you'll get to see the dynamic of the conversations, learn the personalities of the players, and hopefully be better prepared to make a useful contribution when do you speak up.

If after all this you still think you want to join the perl5-porters mailing list, send mail to perl5-porters-subscribe@perl.org. To unsubscribe, send mail to perl5-porters-unsubscribe@perl.org.

To hack on the Perl guts, you'll need to read the following things:

perlguts

This is of paramount importance, since it's the documentation of what goes where in the Perl source. Read it over a couple of times and it might start to make sense - don't worry if it doesn't yet, because the best way to study it is to read it in conjunction with poking at Perl source, and we'll do that later on.

Gisle Aas's illustrated perlguts (also known as illguts) is wonderful, although a little out of date with regard to some size details; the various SV structures have since been reworked for smaller memory footprint. The fundamentals are right however, and the pictures are very helpful.

http://www.perl.org/tpc/1998/Perl_Language_and_Modules/Perl%20Illustrated/

perlxstut and perlxs

A working knowledge of XSUB programming is incredibly useful for core hacking; XSUBs use techniques drawn from the PP code, the portion of the guts that actually executes a Perl program. It's a lot gentler to learn those techniques from simple examples and explanation than from the core itself.

perlapi

The documentation for the Perl API explains what some of the internal functions do, as well as the many macros used in the source.

Porting/pumpkin.pod

This is a collection of words of wisdom for a Perl porter; some of it is only useful to the pumpkin holder, but most of it applies to anyone wanting to go about Perl development.

The perl5-porters FAQ

This should be available from http://dev.perl.org/perl5/docs/p5p-faq.html . It contains hints on reading perl5-porters, information on how perl5-porters works and how Perl development in general works.

Finding Your Way Around

Perl maintenance can be split into a number of areas, and certain people (pumpkins) will have responsibility for each area. These areas sometimes correspond to files or directories in the source kit. Among the areas are:

Core modules

Modules shipped as part of the Perl core live in various subdirectories, where two are dedicated to core-only modules, and two are for the dual-life modules which live on CPAN and may be maintained separately with respect to the Perl core:

lib/  is for pure-Perl modules, which exist in the core only.

ext/  is for XS extensions, and modules with special Makefile.PL requirements, which exist in the core only.

cpan/ is for dual-life modules, where the CPAN module is canonical (should be patched first).

dist/ is for dual-life modules, where the blead source is canonical.

Tests

There are tests for nearly all the modules, built-ins and major bits of functionality. Test files all have a .t suffix. Module tests live in the lib/ and ext/ directories next to the module being tested. Others live in t/. See "Writing a test"

Documentation

Documentation maintenance includes looking after everything in the pod/ directory, (as well as contributing new documentation) and the documentation to the modules in core.

Configure

The Configure process is the way we make Perl portable across the myriad of operating systems it supports. Responsibility for the Configure, build and installation process, as well as the overall portability of the core code rests with the Configure pumpkin - others help out with individual operating systems.

The three files that fall under his/her resposibility are Configure, config_h.SH, and Porting/Glossary (and a whole bunch of small related files that are less important here). The Configure pumpkin decides how patches to these are dealt with. Currently, the Configure pumpkin will accept patches in most common formats, even directly to these files. Other committers are allowed to commit to these files under the strict condition that they will inform the Configure pumpkin, either on IRC (if he/she happens to be around) or through (personal) e-mail.

The files involved are the operating system directories, (win32/, os2/, vms/ and so on) the shell scripts which generate config.h and Makefile, as well as the metaconfig files which generate Configure. (metaconfig isn't included in the core distribution.)

See http://perl5.git.perl.org/metaconfig.git/blob/HEAD:/README for a description of the full process involved.

Interpreter

And of course, there's the core of the Perl interpreter itself. Let's have a look at that in a little more detail.

Before we leave looking at the layout, though, don't forget that MANIFEST contains not only the file names in the Perl distribution, but short descriptions of what's in them, too. For an overview of the important files, try this:

perl -lne 'print if /^[^\/]+\.[ch]\s+/' MANIFEST

Elements of the interpreter

The work of the interpreter has two main stages: compiling the code into the internal representation, or bytecode, and then executing it. "Compiled code" in perlguts explains exactly how the compilation stage happens.

Here is a short breakdown of perl's operation:

Startup

The action begins in perlmain.c. (or miniperlmain.c for miniperl) This is very high-level code, enough to fit on a single screen, and it resembles the code found in perlembed; most of the real action takes place in perl.c

perlmain.c is generated by writemain from miniperlmain.c at make time, so you should make perl to follow this along.

First, perlmain.c allocates some memory and constructs a Perl interpreter, along these lines:

1 PERL_SYS_INIT3(&argc,&argv,&env);
2
3 if (!PL_do_undump) {
4     my_perl = perl_alloc();
5     if (!my_perl)
6         exit(1);
7     perl_construct(my_perl);
8     PL_perl_destruct_level = 0;
9 }

Line 1 is a macro, and its definition is dependent on your operating system. Line 3 references PL_do_undump, a global variable - all global variables in Perl start with PL_. This tells you whether the current running program was created with the -u flag to perl and then undump, which means it's going to be false in any sane context.

Line 4 calls a function in perl.c to allocate memory for a Perl interpreter. It's quite a simple function, and the guts of it looks like this:

my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));

Here you see an example of Perl's system abstraction, which we'll see later: PerlMem_malloc is either your system's malloc, or Perl's own malloc as defined in malloc.c if you selected that option at configure time.

Next, in line 7, we construct the interpreter using perl_construct, also in perl.c; this sets up all the special variables that Perl needs, the stacks, and so on.

Now we pass Perl the command line options, and tell it to go:

exitstatus = perl_parse(my_perl, xs_init, argc, argv, (char **)NULL);
if (!exitstatus)
    perl_run(my_perl);

exitstatus = perl_destruct(my_perl);

perl_free(my_perl);

perl_parse is actually a wrapper around S_parse_body, as defined in perl.c, which processes the command line options, sets up any statically linked XS modules, opens the program and calls yyparse to parse it.

Parsing

The aim of this stage is to take the Perl source, and turn it into an op tree. We'll see what one of those looks like later. Strictly speaking, there's three things going on here.

yyparse, the parser, lives in perly.c, although you're better off reading the original YACC input in perly.y. (Yes, Virginia, there is a YACC grammar for Perl!) The job of the parser is to take your code and "understand" it, splitting it into sentences, deciding which operands go with which operators and so on.

The parser is nobly assisted by the lexer, which chunks up your input into tokens, and decides what type of thing each token is: a variable name, an operator, a bareword, a subroutine, a core function, and so on. The main point of entry to the lexer is yylex, and that and its associated routines can be found in toke.c. Perl isn't much like other computer languages; it's highly context sensitive at times, it can be tricky to work out what sort of token something is, or where a token ends. As such, there's a lot of interplay between the tokeniser and the parser, which can get pretty frightening if you're not used to it.

As the parser understands a Perl program, it builds up a tree of operations for the interpreter to perform during execution. The routines which construct and link together the various operations are to be found in op.c, and will be examined later.

Optimization

Now the parsing stage is complete, and the finished tree represents the operations that the Perl interpreter needs to perform to execute our program. Next, Perl does a dry run over the tree looking for optimisations: constant expressions such as 3 + 4 will be computed now, and the optimizer will also see if any multiple operations can be replaced with a single one. For instance, to fetch the variable $foo, instead of grabbing the glob *foo and looking at the scalar component, the optimizer fiddles the op tree to use a function which directly looks up the scalar in question. The main optimizer is peep in op.c, and many ops have their own optimizing functions.

Running

Now we're finally ready to go: we have compiled Perl byte code, and all that's left to do is run it. The actual execution is done by the runops_standard function in run.c; more specifically, it's done by these three innocent looking lines:

while ((PL_op = CALL_FPTR(PL_op->op_ppaddr)(aTHX))) {
    PERL_ASYNC_CHECK();
}

You may be more comfortable with the Perl version of that:

PERL_ASYNC_CHECK() while $Perl::op = &{$Perl::op->{function}};

Well, maybe not. Anyway, each op contains a function pointer, which stipulates the function which will actually carry out the operation. This function will return the next op in the sequence - this allows for things like if which choose the next op dynamically at run time. The PERL_ASYNC_CHECK makes sure that things like signals interrupt execution if required.

The actual functions called are known as PP code, and they're spread between four files: pp_hot.c contains the "hot" code, which is most often used and highly optimized, pp_sys.c contains all the system-specific functions, pp_ctl.c contains the functions which implement control structures (if, while and the like) and pp.c contains everything else. These are, if you like, the C code for Perl's built-in functions and operators.

Note that each pp_ function is expected to return a pointer to the next op. Calls to perl subs (and eval blocks) are handled within the same runops loop, and do not consume extra space on the C stack. For example, pp_entersub and pp_entertry just push a CxSUB or CxEVAL block struct onto the context stack which contain the address of the op following the sub call or eval. They then return the first op of that sub or eval block, and so execution continues of that sub or block. Later, a pp_leavesub or pp_leavetry op pops the CxSUB or CxEVAL, retrieves the return op from it, and returns it.

Exception handing

Perl's exception handing (i.e. die etc.) is built on top of the low-level setjmp()/longjmp() C-library functions. These basically provide a way to capture the current PC and SP registers and later restore them; i.e. a longjmp() continues at the point in code where a previous setjmp() was done, with anything further up on the C stack being lost. This is why code should always save values using SAVE_FOO rather than in auto variables.

The perl core wraps setjmp() etc in the macros JMPENV_PUSH and JMPENV_JUMP. The basic rule of perl exceptions is that exit, and die (in the absence of eval) perform a JMPENV_JUMP(2), while die within eval does a JMPENV_JUMP(3).

At entry points to perl, such as perl_parse(), perl_run() and call_sv(cv, G_EVAL) each does a JMPENV_PUSH, then enter a runops loop or whatever, and handle possible exception returns. For a 2 return, final cleanup is performed, such as popping stacks and calling CHECK or END blocks. Amongst other things, this is how scope cleanup still occurs during an exit.

If a die can find a CxEVAL block on the context stack, then the stack is popped to that level and the return op in that block is assigned to PL_restartop; then a JMPENV_JUMP(3) is performed. This normally passes control back to the guard. In the case of perl_run and call_sv, a non-null PL_restartop triggers re-entry to the runops loop. The is the normal way that die or croak is handled within an eval.

Sometimes ops are executed within an inner runops loop, such as tie, sort or overload code. In this case, something like

sub FETCH { eval { die } }

would cause a longjmp right back to the guard in perl_run, popping both runops loops, which is clearly incorrect. One way to avoid this is for the tie code to do a JMPENV_PUSH before executing FETCH in the inner runops loop, but for efficiency reasons, perl in fact just sets a flag, using CATCH_SET(TRUE). The pp_require, pp_entereval and pp_entertry ops check this flag, and if true, they call docatch, which does a JMPENV_PUSH and starts a new runops level to execute the code, rather than doing it on the current loop.

As a further optimisation, on exit from the eval block in the FETCH, execution of the code following the block is still carried on in the inner loop. When an exception is raised, docatch compares the JMPENV level of the CxEVAL with PL_top_env and if they differ, just re-throws the exception. In this way any inner loops get popped.

Here's an example.

1: eval { tie @a, 'A' };
2: sub A::TIEARRAY {
3:     eval { die };
4:     die;
5: }

To run this code, perl_run is called, which does a JMPENV_PUSH then enters a runops loop. This loop executes the eval and tie ops on line 1, with the eval pushing a CxEVAL onto the context stack.

The pp_tie does a CATCH_SET(TRUE), then starts a second runops loop to execute the body of TIEARRAY. When it executes the entertry op on line 3, CATCH_GET is true, so pp_entertry calls docatch which does a JMPENV_PUSH and starts a third runops loop, which then executes the die op. At this point the C call stack looks like this:

Perl_pp_die
Perl_runops      # third loop
S_docatch_body
S_docatch
Perl_pp_entertry
Perl_runops      # second loop
S_call_body
Perl_call_sv
Perl_pp_tie
Perl_runops      # first loop
S_run_body
perl_run
main

and the context and data stacks, as shown by -Dstv, look like:

STACK 0: MAIN
  CX 0: BLOCK  =>
  CX 1: EVAL   => AV()  PV("A"\0)
  retop=leave
STACK 1: MAGIC
  CX 0: SUB    =>
  retop=(null)
  CX 1: EVAL   => *
retop=nextstate

The die pops the first CxEVAL off the context stack, sets PL_restartop from it, does a JMPENV_JUMP(3), and control returns to the top docatch. This then starts another third-level runops level, which executes the nextstate, pushmark and die ops on line 4. At the point that the second pp_die is called, the C call stack looks exactly like that above, even though we are no longer within an inner eval; this is because of the optimization mentioned earlier. However, the context stack now looks like this, ie with the top CxEVAL popped:

STACK 0: MAIN
  CX 0: BLOCK  =>
  CX 1: EVAL   => AV()  PV("A"\0)
  retop=leave
STACK 1: MAGIC
  CX 0: SUB    =>
  retop=(null)

The die on line 4 pops the context stack back down to the CxEVAL, leaving it as:

STACK 0: MAIN
  CX 0: BLOCK  =>

As usual, PL_restartop is extracted from the CxEVAL, and a JMPENV_JUMP(3) done, which pops the C stack back to the docatch:

S_docatch
Perl_pp_entertry
Perl_runops      # second loop
S_call_body
Perl_call_sv
Perl_pp_tie
Perl_runops      # first loop
S_run_body
perl_run
main

In this case, because the JMPENV level recorded in the CxEVAL differs from the current one, docatch just does a JMPENV_JUMP(3) and the C stack unwinds to:

perl_run
main

Because PL_restartop is non-null, run_body starts a new runops loop and execution continues.

Internal Variable Types

You should by now have had a look at perlguts, which tells you about Perl's internal variable types: SVs, HVs, AVs and the rest. If not, do that now.

These variables are used not only to represent Perl-space variables, but also any constants in the code, as well as some structures completely internal to Perl. The symbol table, for instance, is an ordinary Perl hash. Your code is represented by an SV as it's read into the parser; any program files you call are opened via ordinary Perl filehandles, and so on.

The core Devel::Peek module lets us examine SVs from a Perl program. Let's see, for instance, how Perl treats the constant "hello".

  % perl -MDevel::Peek -e 'Dump("hello")'
1 SV = PV(0xa041450) at 0xa04ecbc
2   REFCNT = 1
3   FLAGS = (POK,READONLY,pPOK)
4   PV = 0xa0484e0 "hello"\0
5   CUR = 5
6   LEN = 6

Reading Devel::Peek output takes a bit of practise, so let's go through it line by line.

Line 1 tells us we're looking at an SV which lives at 0xa04ecbc in memory. SVs themselves are very simple structures, but they contain a pointer to a more complex structure. In this case, it's a PV, a structure which holds a string value, at location 0xa041450. Line 2 is the reference count; there are no other references to this data, so it's 1.

Line 3 are the flags for this SV - it's OK to use it as a PV, it's a read-only SV (because it's a constant) and the data is a PV internally. Next we've got the contents of the string, starting at location 0xa0484e0.

Line 5 gives us the current length of the string - note that this does not include the null terminator. Line 6 is not the length of the string, but the length of the currently allocated buffer; as the string grows, Perl automatically extends the available storage via a routine called SvGROW.

You can get at any of these quantities from C very easily; just add Sv to the name of the field shown in the snippet, and you've got a macro which will return the value: SvCUR(sv) returns the current length of the string, SvREFCOUNT(sv) returns the reference count, SvPV(sv, len) returns the string itself with its length, and so on. More macros to manipulate these properties can be found in perlguts.

Let's take an example of manipulating a PV, from sv_catpvn, in sv.c

 1  void
 2  Perl_sv_catpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
 3  {
 4      STRLEN tlen;
 5      char *junk;

 6      junk = SvPV_force(sv, tlen);
 7      SvGROW(sv, tlen + len + 1);
 8      if (ptr == junk)
 9          ptr = SvPVX(sv);
10      Move(ptr,SvPVX(sv)+tlen,len,char);
11      SvCUR(sv) += len;
12      *SvEND(sv) = '\0';
13      (void)SvPOK_only_UTF8(sv);          /* validate pointer */
14      SvTAINT(sv);
15  }

This is a function which adds a string, ptr, of length len onto the end of the PV stored in sv. The first thing we do in line 6 is make sure that the SV has a valid PV, by calling the SvPV_force macro to force a PV. As a side effect, tlen gets set to the current value of the PV, and the PV itself is returned to junk.

In line 7, we make sure that the SV will have enough room to accommodate the old string, the new string and the null terminator. If LEN isn't big enough, SvGROW will reallocate space for us.

Now, if junk is the same as the string we're trying to add, we can grab the string directly from the SV; SvPVX is the address of the PV in the SV.

Line 10 does the actual catenation: the Move macro moves a chunk of memory around: we move the string ptr to the end of the PV - that's the start of the PV plus its current length. We're moving len bytes of type char. After doing so, we need to tell Perl we've extended the string, by altering CUR to reflect the new length. SvEND is a macro which gives us the end of the string, so that needs to be a "\0".

Line 13 manipulates the flags; since we've changed the PV, any IV or NV values will no longer be valid: if we have $a=10; $a.="6"; we don't want to use the old IV of 10. SvPOK_only_utf8 is a special UTF-8-aware version of SvPOK_only, a macro which turns off the IOK and NOK flags and turns on POK. The final SvTAINT is a macro which launders tainted data if taint mode is turned on.

AVs and HVs are more complicated, but SVs are by far the most common variable type being thrown around. Having seen something of how we manipulate these, let's go on and look at how the op tree is constructed.

Op Trees

First, what is the op tree, anyway? The op tree is the parsed representation of your program, as we saw in our section on parsing, and it's the sequence of operations that Perl goes through to execute your program, as we saw in "Running".

An op is a fundamental operation that Perl can perform: all the built-in functions and operators are ops, and there are a series of ops which deal with concepts the interpreter needs internally - entering and leaving a block, ending a statement, fetching a variable, and so on.

The op tree is connected in two ways: you can imagine that there are two "routes" through it, two orders in which you can traverse the tree. First, parse order reflects how the parser understood the code, and secondly, execution order tells perl what order to perform the operations in.

The easiest way to examine the op tree is to stop Perl after it has finished parsing, and get it to dump out the tree. This is exactly what the compiler backends B::Terse, B::Concise and B::Debug do.

Let's have a look at how Perl sees $a = $b + $c:

 % perl -MO=Terse -e '$a=$b+$c'
 1  LISTOP (0x8179888) leave
 2      OP (0x81798b0) enter
 3      COP (0x8179850) nextstate
 4      BINOP (0x8179828) sassign
 5          BINOP (0x8179800) add [1]
 6              UNOP (0x81796e0) null [15]
 7                  SVOP (0x80fafe0) gvsv  GV (0x80fa4cc) *b
 8              UNOP (0x81797e0) null [15]
 9                  SVOP (0x8179700) gvsv  GV (0x80efeb0) *c
10          UNOP (0x816b4f0) null [15]
11              SVOP (0x816dcf0) gvsv  GV (0x80fa460) *a

Let's start in the middle, at line 4. This is a BINOP, a binary operator, which is at location 0x8179828. The specific operator in question is sassign - scalar assignment - and you can find the code which implements it in the function pp_sassign in pp_hot.c. As a binary operator, it has two children: the add operator, providing the result of $b+$c, is uppermost on line 5, and the left hand side is on line 10.

Line 10 is the null op: this does exactly nothing. What is that doing there? If you see the null op, it's a sign that something has been optimized away after parsing. As we mentioned in "Optimization", the optimization stage sometimes converts two operations into one, for example when fetching a scalar variable. When this happens, instead of rewriting the op tree and cleaning up the dangling pointers, it's easier just to replace the redundant operation with the null op. Originally, the tree would have looked like this:

10          SVOP (0x816b4f0) rv2sv [15]
11              SVOP (0x816dcf0) gv  GV (0x80fa460) *a

That is, fetch the a entry from the main symbol table, and then look at the scalar component of it: gvsv (pp_gvsv into pp_hot.c) happens to do both these things.

The right hand side, starting at line 5 is similar to what we've just seen: we have the add op (pp_add also in pp_hot.c) add together two gvsvs.

Now, what's this about?

1  LISTOP (0x8179888) leave
2      OP (0x81798b0) enter
3      COP (0x8179850) nextstate

enter and leave are scoping ops, and their job is to perform any housekeeping every time you enter and leave a block: lexical variables are tidied up, unreferenced variables are destroyed, and so on. Every program will have those first three lines: leave is a list, and its children are all the statements in the block. Statements are delimited by nextstate, so a block is a collection of nextstate ops, with the ops to be performed for each statement being the children of nextstate. enter is a single op which functions as a marker.

That's how Perl parsed the program, from top to bottom:

 Program
    |
Statement
    |
    =
   / \
  /   \
 $a   +
     / \
   $b   $c

However, it's impossible to perform the operations in this order: you have to find the values of $b and $c before you add them together, for instance. So, the other thread that runs through the op tree is the execution order: each op has a field op_next which points to the next op to be run, so following these pointers tells us how perl executes the code. We can traverse the tree in this order using the exec option to B::Terse:

% perl -MO=Terse,exec -e '$a=$b+$c'
1  OP (0x8179928) enter
2  COP (0x81798c8) nextstate
3  SVOP (0x81796c8) gvsv  GV (0x80fa4d4) *b
4  SVOP (0x8179798) gvsv  GV (0x80efeb0) *c
5  BINOP (0x8179878) add [1]
6  SVOP (0x816dd38) gvsv  GV (0x80fa468) *a
7  BINOP (0x81798a0) sassign
8  LISTOP (0x8179900) leave

This probably makes more sense for a human: enter a block, start a statement. Get the values of $b and $c, and add them together. Find $a, and assign one to the other. Then leave.

The way Perl builds up these op trees in the parsing process can be unravelled by examining perly.y, the YACC grammar. Let's take the piece we need to construct the tree for $a = $b + $c

1 term    :   term ASSIGNOP term
2                { $$ = newASSIGNOP(OPf_STACKED, $1, $2, $3); }
3         |   term ADDOP term
4                { $$ = newBINOP($2, 0, scalar($1), scalar($3)); }

If you're not used to reading BNF grammars, this is how it works: You're fed certain things by the tokeniser, which generally end up in upper case. Here, ADDOP, is provided when the tokeniser sees + in your code. ASSIGNOP is provided when = is used for assigning. These are "terminal symbols", because you can't get any simpler than them.

The grammar, lines one and three of the snippet above, tells you how to build up more complex forms. These complex forms, "non-terminal symbols" are generally placed in lower case. term here is a non-terminal symbol, representing a single expression.

The grammar gives you the following rule: you can make the thing on the left of the colon if you see all the things on the right in sequence. This is called a "reduction", and the aim of parsing is to completely reduce the input. There are several different ways you can perform a reduction, separated by vertical bars: so, term followed by = followed by term makes a term, and term followed by + followed by term can also make a term.

So, if you see two terms with an = or +, between them, you can turn them into a single expression. When you do this, you execute the code in the block on the next line: if you see =, you'll do the code in line 2. If you see +, you'll do the code in line 4. It's this code which contributes to the op tree.

|   term ADDOP term
{ $$ = newBINOP($2, 0, scalar($1), scalar($3)); }

What this does is creates a new binary op, and feeds it a number of variables. The variables refer to the tokens: $1 is the first token in the input, $2 the second, and so on - think regular expression backreferences. $$ is the op returned from this reduction. So, we call newBINOP to create a new binary operator. The first parameter to newBINOP, a function in op.c, is the op type. It's an addition operator, so we want the type to be ADDOP. We could specify this directly, but it's right there as the second token in the input, so we use $2. The second parameter is the op's flags: 0 means "nothing special". Then the things to add: the left and right hand side of our expression, in scalar context.

Stacks

When perl executes something like addop, how does it pass on its results to the next op? The answer is, through the use of stacks. Perl has a number of stacks to store things it's currently working on, and we'll look at the three most important ones here.

Argument stack

Arguments are passed to PP code and returned from PP code using the argument stack, ST. The typical way to handle arguments is to pop them off the stack, deal with them how you wish, and then push the result back onto the stack. This is how, for instance, the cosine operator works:

NV value;
value = POPn;
value = Perl_cos(value);
XPUSHn(value);

We'll see a more tricky example of this when we consider Perl's macros below. POPn gives you the NV (floating point value) of the top SV on the stack: the $x in cos($x). Then we compute the cosine, and push the result back as an NV. The X in XPUSHn means that the stack should be extended if necessary - it can't be necessary here, because we know there's room for one more item on the stack, since we've just removed one! The XPUSH* macros at least guarantee safety.

Alternatively, you can fiddle with the stack directly: SP gives you the first element in your portion of the stack, and TOP* gives you the top SV/IV/NV/etc. on the stack. So, for instance, to do unary negation of an integer:

SETi(-TOPi);

Just set the integer value of the top stack entry to its negation.

Argument stack manipulation in the core is exactly the same as it is in XSUBs - see perlxstut, perlxs and perlguts for a longer description of the macros used in stack manipulation.

Mark stack

I say "your portion of the stack" above because PP code doesn't necessarily get the whole stack to itself: if your function calls another function, you'll only want to expose the arguments aimed for the called function, and not (necessarily) let it get at your own data. The way we do this is to have a "virtual" bottom-of-stack, exposed to each function. The mark stack keeps bookmarks to locations in the argument stack usable by each function. For instance, when dealing with a tied variable, (internally, something with "P" magic) Perl has to call methods for accesses to the tied variables. However, we need to separate the arguments exposed to the method to the argument exposed to the original function - the store or fetch or whatever it may be. Here's roughly how the tied push is implemented; see av_push in av.c:

1	PUSHMARK(SP);
2	EXTEND(SP,2);
3	PUSHs(SvTIED_obj((SV*)av, mg));
4	PUSHs(val);
5	PUTBACK;
6	ENTER;
7	call_method("PUSH", G_SCALAR|G_DISCARD);
8	LEAVE;

Let's examine the whole implementation, for practice:

1	PUSHMARK(SP);

Push the current state of the stack pointer onto the mark stack. This is so that when we've finished adding items to the argument stack, Perl knows how many things we've added recently.

2	EXTEND(SP,2);
3	PUSHs(SvTIED_obj((SV*)av, mg));
4	PUSHs(val);

We're going to add two more items onto the argument stack: when you have a tied array, the PUSH subroutine receives the object and the value to be pushed, and that's exactly what we have here - the tied object, retrieved with SvTIED_obj, and the value, the SV val.

5	PUTBACK;

Next we tell Perl to update the global stack pointer from our internal variable: dSP only gave us a local copy, not a reference to the global.

6	ENTER;
7	call_method("PUSH", G_SCALAR|G_DISCARD);
8	LEAVE;

ENTER and LEAVE localise a block of code - they make sure that all variables are tidied up, everything that has been localised gets its previous value returned, and so on. Think of them as the { and } of a Perl block.

To actually do the magic method call, we have to call a subroutine in Perl space: call_method takes care of that, and it's described in perlcall. We call the PUSH method in scalar context, and we're going to discard its return value. The call_method() function removes the top element of the mark stack, so there is nothing for the caller to clean up.

Save stack

C doesn't have a concept of local scope, so perl provides one. We've seen that ENTER and LEAVE are used as scoping braces; the save stack implements the C equivalent of, for example:

{
    local $foo = 42;
    ...
}

See "Localising Changes" in perlguts for how to use the save stack.

Millions of Macros

One thing you'll notice about the Perl source is that it's full of macros. Some have called the pervasive use of macros the hardest thing to understand, others find it adds to clarity. Let's take an example, the code which implements the addition operator:

1  PP(pp_add)
2  {
3      dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
4      {
5        dPOPTOPnnrl_ul;
6        SETn( left + right );
7        RETURN;
8      }
9  }

Every line here (apart from the braces, of course) contains a macro. The first line sets up the function declaration as Perl expects for PP code; line 3 sets up variable declarations for the argument stack and the target, the return value of the operation. Finally, it tries to see if the addition operation is overloaded; if so, the appropriate subroutine is called.

Line 5 is another variable declaration - all variable declarations start with d - which pops from the top of the argument stack two NVs (hence nn) and puts them into the variables right and left, hence the rl. These are the two operands to the addition operator. Next, we call SETn to set the NV of the return value to the result of adding the two values. This done, we return - the RETURN macro makes sure that our return value is properly handled, and we pass the next operator to run back to the main run loop.

Most of these macros are explained in perlapi, and some of the more important ones are explained in perlxs as well. Pay special attention to "Background and PERL_IMPLICIT_CONTEXT" in perlguts for information on the [pad]THX_? macros.

The .i Targets

You can expand the macros in a foo.c file by saying

make foo.i

which will expand the macros using cpp. Don't be scared by the results.

SOURCE CODE STATIC ANALYSIS

Various tools exist for analysing C source code statically, as opposed to dynamically, that is, without executing the code. It is possible to detect resource leaks, undefined behaviour, type mismatches, portability problems, code paths that would cause illegal memory accesses, and other similar problems by just parsing the C code and looking at the resulting graph, what does it tell about the execution and data flows. As a matter of fact, this is exactly how C compilers know to give warnings about dubious code.

lint, splint

The good old C code quality inspector, lint, is available in several platforms, but please be aware that there are several different implementations of it by different vendors, which means that the flags are not identical across different platforms.

There is a lint variant called splint (Secure Programming Lint) available from http://www.splint.org/ that should compile on any Unix-like platform.

There are lint and <splint> targets in Makefile, but you may have to diddle with the flags (see above).

Coverity

Coverity (http://www.coverity.com/) is a product similar to lint and as a testbed for their product they periodically check several open source projects, and they give out accounts to open source developers to the defect databases.

cpd (cut-and-paste detector)

The cpd tool detects cut-and-paste coding. If one instance of the cut-and-pasted code changes, all the other spots should probably be changed, too. Therefore such code should probably be turned into a subroutine or a macro.

cpd (http://pmd.sourceforge.net/cpd.html) is part of the pmd project (http://pmd.sourceforge.net/). pmd was originally written for static analysis of Java code, but later the cpd part of it was extended to parse also C and C++.

Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the pmd-X.Y.jar from it, and then run that on source code thusly:

java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD --minimum-tokens 100 --files /some/where/src --language c > cpd.txt

You may run into memory limits, in which case you should use the -Xmx option:

java -Xmx512M ...

gcc warnings

Though much can be written about the inconsistency and coverage problems of gcc warnings (like -Wall not meaning "all the warnings", or some common portability problems not being covered by -Wall, or -ansi and -pedantic both being a poorly defined collection of warnings, and so forth), gcc is still a useful tool in keeping our coding nose clean.

The -Wall is by default on.

The -ansi (and its sidekick, -pedantic) would be nice to be on always, but unfortunately they are not safe on all platforms, they can for example cause fatal conflicts with the system headers (Solaris being a prime example). If Configure -Dgccansipedantic is used, the cflags frontend selects -ansi -pedantic for the platforms where they are known to be safe.

Starting from Perl 5.9.4 the following extra flags are added:

• -Wendif-labels

• -Wextra

• -Wdeclaration-after-statement

The following flags would be nice to have but they would first need their own Augean stablemaster:

• -Wpointer-arith

• -Wshadow

• -Wstrict-prototypes

The -Wtraditional is another example of the annoying tendency of gcc to bundle a lot of warnings under one switch -- it would be impossible to deploy in practice because it would complain a lot -- but it does contain some warnings that would be beneficial to have available on their own, such as the warning about string constants inside macros containing the macro arguments: this behaved differently pre-ANSI than it does in ANSI, and some C compilers are still in transition, AIX being an example.

Warnings of other C compilers

Other C compilers (yes, there are other C compilers than gcc) often have their "strict ANSI" or "strict ANSI with some portability extensions" modes on, like for example the Sun Workshop has its -Xa mode on (though implicitly), or the DEC (these days, HP...) has its -std1 mode on.

DEBUGGING

You can compile a special debugging version of Perl, which allows you to use the -D option of Perl to tell more about what Perl is doing. But sometimes there is no alternative than to dive in with a debugger, either to see the stack trace of a core dump (very useful in a bug report), or trying to figure out what went wrong before the core dump happened, or how did we end up having wrong or unexpected results.

Poking at Perl

To really poke around with Perl, you'll probably want to build Perl for debugging, like this:

./Configure -d -D optimize=-g
make

-g is a flag to the C compiler to have it produce debugging information which will allow us to step through a running program, and to see in which C function we are at (without the debugging information we might see only the numerical addresses of the functions, which is not very helpful).

Configure will also turn on the DEBUGGING compilation symbol which enables all the internal debugging code in Perl. There are a whole bunch of things you can debug with this: perlrun lists them all, and the best way to find out about them is to play about with them. The most useful options are probably

l  Context (loop) stack processing
t  Trace execution
o  Method and overloading resolution
c  String/numeric conversions

Some of the functionality of the debugging code can be achieved using XS modules.

-Dr => use re 'debug'
-Dx => use O 'Debug'

Using a source-level debugger

If the debugging output of -D doesn't help you, it's time to step through perl's execution with a source-level debugger.

• We'll use gdb for our examples here; the principles will apply to any debugger (many vendors call their debugger dbx), but check the manual of the one you're using.

To fire up the debugger, type

gdb ./perl

Or if you have a core dump:

gdb ./perl core

You'll want to do that in your Perl source tree so the debugger can read the source code. You should see the copyright message, followed by the prompt.

(gdb)

help will get you into the documentation, but here are the most useful commands:

run [args]

Run the program with the given arguments.

break function_name

break source.c:xxx

Tells the debugger that we'll want to pause execution when we reach either the named function (but see "Internal Functions" in perlguts!) or the given line in the named source file.

step

Steps through the program a line at a time.

next

Steps through the program a line at a time, without descending into functions.

continue

Run until the next breakpoint.

finish

Run until the end of the current function, then stop again.

'enter'

Just pressing Enter will do the most recent operation again - it's a blessing when stepping through miles of source code.

print

Execute the given C code and print its results. WARNING: Perl makes heavy use of macros, and gdb does not necessarily support macros (see later "gdb macro support"). You'll have to substitute them yourself, or to invoke cpp on the source code files (see "The .i Targets") So, for instance, you can't say

print SvPV_nolen(sv)

but you have to say

print Perl_sv_2pv_nolen(sv)

You may find it helpful to have a "macro dictionary", which you can produce by saying cpp -dM perl.c | sort. Even then, cpp won't recursively apply those macros for you.

gdb macro support

Recent versions of gdb have fairly good macro support, but in order to use it you'll need to compile perl with macro definitions included in the debugging information. Using gcc version 3.1, this means configuring with -Doptimize=-g3. Other compilers might use a different switch (if they support debugging macros at all).

Dumping Perl Data Structures

One way to get around this macro hell is to use the dumping functions in dump.c; these work a little like an internal Devel::Peek, but they also cover OPs and other structures that you can't get at from Perl. Let's take an example. We'll use the $a = $b + $c we used before, but give it a bit of context: $b = "6XXXX"; $c = 2.3;. Where's a good place to stop and poke around?

What about pp_add, the function we examined earlier to implement the + operator:

(gdb) break Perl_pp_add
Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.

Notice we use Perl_pp_add and not pp_add - see "Internal Functions" in perlguts. With the breakpoint in place, we can run our program:

(gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'

Lots of junk will go past as gdb reads in the relevant source files and libraries, and then:

Breakpoint 1, Perl_pp_add () at pp_hot.c:309
309         dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
(gdb) step
311           dPOPTOPnnrl_ul;
(gdb)

We looked at this bit of code before, and we said that dPOPTOPnnrl_ul arranges for two NVs to be placed into left and right - let's slightly expand it:

#define dPOPTOPnnrl_ul  NV right = POPn; \
                        SV *leftsv = TOPs; \
                        NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0

POPn takes the SV from the top of the stack and obtains its NV either directly (if SvNOK is set) or by calling the sv_2nv function. TOPs takes the next SV from the top of the stack - yes, POPn uses TOPs - but doesn't remove it. We then use SvNV to get the NV from leftsv in the same way as before - yes, POPn uses SvNV.

Since we don't have an NV for $b, we'll have to use sv_2nv to convert it. If we step again, we'll find ourselves there:

Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
1669        if (!sv)
(gdb)

We can now use Perl_sv_dump to investigate the SV:

SV = PV(0xa057cc0) at 0xa0675d0
REFCNT = 1
FLAGS = (POK,pPOK)
PV = 0xa06a510 "6XXXX"\0
CUR = 5
LEN = 6
$1 = void

We know we're going to get 6 from this, so let's finish the subroutine:

(gdb) finish
Run till exit from #0  Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
0x462669 in Perl_pp_add () at pp_hot.c:311
311           dPOPTOPnnrl_ul;

We can also dump out this op: the current op is always stored in PL_op, and we can dump it with Perl_op_dump. This'll give us similar output to B::Debug.

{
13  TYPE = add  ===> 14
    TARG = 1
    FLAGS = (SCALAR,KIDS)
    {
        TYPE = null  ===> (12)
          (was rv2sv)
        FLAGS = (SCALAR,KIDS)
        {
11          TYPE = gvsv  ===> 12
            FLAGS = (SCALAR)
            GV = main::b
        }
    }

# finish this later #

Patching

All right, we've now had a look at how to navigate the Perl sources and some things you'll need to know when fiddling with them. Let's now get on and create a simple patch. Here's something Larry suggested: if a U is the first active format during a pack, (for example, pack "U3C8", @stuff) then the resulting string should be treated as UTF-8 encoded.

If you are working with a git clone of the Perl repository, you will want to create a branch for your changes. This will make creating a proper patch much simpler. See the perlrepository for details on how to do this.

How do we prepare to fix this up? First we locate the code in question - the pack happens at runtime, so it's going to be in one of the pp files. Sure enough, pp_pack is in pp.c. Since we're going to be altering this file, let's copy it to pp.c~.

[Well, it was in pp.c when this tutorial was written. It has now been split off with pp_unpack to its own file, pp_pack.c]

Now let's look over pp_pack: we take a pattern into pat, and then loop over the pattern, taking each format character in turn into datum_type. Then for each possible format character, we swallow up the other arguments in the pattern (a field width, an asterisk, and so on) and convert the next chunk input into the specified format, adding it onto the output SV cat.

How do we know if the U is the first format in the pat? Well, if we have a pointer to the start of pat then, if we see a U we can test whether we're still at the start of the string. So, here's where pat is set up:

STRLEN fromlen;
register char *pat = SvPVx(*++MARK, fromlen);
register char *patend = pat + fromlen;
register I32 len;
I32 datumtype;
SV *fromstr;

We'll have another string pointer in there:

   STRLEN fromlen;
   register char *pat = SvPVx(*++MARK, fromlen);
   register char *patend = pat + fromlen;
+  char *patcopy;
   register I32 len;
   I32 datumtype;
   SV *fromstr;

And just before we start the loop, we'll set patcopy to be the start of pat:

   items = SP - MARK;
   MARK++;
   sv_setpvn(cat, "", 0);
+  patcopy = pat;
   while (pat < patend) {

Now if we see a U which was at the start of the string, we turn on the UTF8 flag for the output SV, cat:

+  if (datumtype == 'U' && pat==patcopy+1)
+      SvUTF8_on(cat);
   if (datumtype == '#') {
       while (pat < patend && *pat != '\n')
           pat++;

Remember that it has to be patcopy+1 because the first character of the string is the U which has been swallowed into datumtype!

Oops, we forgot one thing: what if there are spaces at the start of the pattern? pack(" U*", @stuff) will have U as the first active character, even though it's not the first thing in the pattern. In this case, we have to advance patcopy along with pat when we see spaces:

if (isSPACE(datumtype))
    continue;

needs to become

if (isSPACE(datumtype)) {
    patcopy++;
    continue;
}

OK. That's the C part done. Now we must do two additional things before this patch is ready to go: we've changed the behaviour of Perl, and so we must document that change. We must also provide some more regression tests to make sure our patch works and doesn't create a bug somewhere else along the line.

The regression tests for each operator live in t/op/, and so we make a copy of t/op/pack.t to t/op/pack.t~. Now we can add our tests to the end. First, we'll test that the U does indeed create Unicode strings.

t/op/pack.t has a sensible ok() function, but if it didn't we could use the one from t/test.pl.

require './test.pl';
plan( tests => 159 );

so instead of this:

print 'not ' unless "1.20.300.4000" eq sprintf "%vd", pack("U*",1,20,300,4000);
print "ok $test\n"; $test++;

we can write the more sensible (see Test::More for a full explanation of is() and other testing functions).

is( "1.20.300.4000", sprintf "%vd", pack("U*",1,20,300,4000),
                                      "U* produces Unicode" );

Now we'll test that we got that space-at-the-beginning business right:

is( "1.20.300.4000", sprintf "%vd", pack("  U*",1,20,300,4000),
                                      "  with spaces at the beginning" );

And finally we'll test that we don't make Unicode strings if U is not the first active format:

isnt( v1.20.300.4000, sprintf "%vd", pack("C0U*",1,20,300,4000),
                                      "U* not first isn't Unicode" );

Mustn't forget to change the number of tests which appears at the top, or else the automated tester will get confused. This will either look like this:

print "1..156\n";

or this:

plan( tests => 156 );

We now compile up Perl, and run it through the test suite. Our new tests pass, hooray!

Finally, the documentation. The job is never done until the paperwork is over, so let's describe the change we've just made. The relevant place is pod/perlfunc.pod; again, we make a copy, and then we'll insert this text in the description of pack:

=item *

If the pattern begins with a C<U>, the resulting string will be treated
as UTF-8-encoded Unicode. You can force UTF-8 encoding on in a string
with an initial C<U0>, and the bytes that follow will be interpreted as
Unicode characters. If you don't want this to happen, you can begin your
pattern with C<C0> (or anything else) to force Perl not to UTF-8 encode your
string, and then follow this with a C<U*> somewhere in your pattern.

Patching a core module

This works just like patching anything else, with an extra consideration. Many core modules also live on CPAN. If this is so, patch the CPAN version instead of the core and send the patch off to the module maintainer (with a copy to p5p). This will help the module maintainer keep the CPAN version in sync with the core version without constantly scanning p5p.

The list of maintainers of core modules is usefully documented in Porting/Maintainers.pl.

Adding a new function to the core

If, as part of a patch to fix a bug, or just because you have an especially good idea, you decide to add a new function to the core, discuss your ideas on p5p well before you start work. It may be that someone else has already attempted to do what you are considering and can give lots of good advice or even provide you with bits of code that they already started (but never finished).

You have to follow all of the advice given above for patching. It is extremely important to test any addition thoroughly and add new tests to explore all boundary conditions that your new function is expected to handle. If your new function is used only by one module (e.g. toke), then it should probably be named S_your_function (for static); on the other hand, if you expect it to accessible from other functions in Perl, you should name it Perl_your_function. See "Internal Functions" in perlguts for more details.

The location of any new code is also an important consideration. Don't just create a new top level .c file and put your code there; you would have to make changes to Configure (so the Makefile is created properly), as well as possibly lots of include files. This is strictly pumpking business.

It is better to add your function to one of the existing top level source code files, but your choice is complicated by the nature of the Perl distribution. Only the files that are marked as compiled static are located in the perl executable. Everything else is located in the shared library (or DLL if you are running under WIN32). So, for example, if a function was only used by functions located in toke.c, then your code can go in toke.c. If, however, you want to call the function from universal.c, then you should put your code in another location, for example util.c.

In addition to writing your c-code, you will need to create an appropriate entry in embed.pl describing your function, then run 'make regen_headers' to create the entries in the numerous header files that perl needs to compile correctly. See "Internal Functions" in perlguts for information on the various options that you can set in embed.pl. You will forget to do this a few (or many) times and you will get warnings during the compilation phase. Make sure that you mention this when you post your patch to P5P; the pumpking needs to know this.

When you write your new code, please be conscious of existing code conventions used in the perl source files. See perlstyle for details. Although most of the guidelines discussed seem to focus on Perl code, rather than c, they all apply (except when they don't ;). Also see perlrepository for lots of details about both formatting and submitting patches of your changes.

Lastly, TEST TEST TEST TEST TEST any code before posting to p5p. Test on as many platforms as you can find. Test as many perl Configure options as you can (e.g. MULTIPLICITY). If you have profiling or memory tools, see "EXTERNAL TOOLS FOR DEBUGGING PERL" below for how to use them to further test your code. Remember that most of the people on P5P are doing this on their own time and don't have the time to debug your code.

Writing a test

Every module and built-in function has an associated test file (or should...). If you add or change functionality, you have to write a test. If you fix a bug, you have to write a test so that bug never comes back. If you alter the docs, it would be nice to test what the new documentation says.

In short, if you submit a patch you probably also have to patch the tests.

For modules, the test file is right next to the module itself. lib/strict.t tests lib/strict.pm. This is a recent innovation, so there are some snags (and it would be wonderful for you to brush them out), but it basically works that way. Everything else lives in t/.

If you add a new test directory under t/, it is imperative that you add that directory to t/HARNESS and t/TEST.

t/base/

Testing of the absolute basic functionality of Perl. Things like if, basic file reads and writes, simple regexes, etc. These are run first in the test suite and if any of them fail, something is really broken.

t/cmd/

These test the basic control structures, if/else, while, subroutines, etc.

t/comp/

Tests basic issues of how Perl parses and compiles itself.

t/io/

Tests for built-in IO functions, including command line arguments.

t/lib/

The old home for the module tests, you shouldn't put anything new in here. There are still some bits and pieces hanging around in here that need to be moved. Perhaps you could move them? Thanks!

t/mro/

Tests for perl's method resolution order implementations (see mro).

t/op/

Tests for perl's built in functions that don't fit into any of the other directories.

t/re/

Tests for regex related functions or behaviour. (These used to live in t/op).

t/run/

Testing features of how perl actually runs, including exit codes and handling of PERL* environment variables.

t/uni/

Tests for the core support of Unicode.

t/win32/

Windows-specific tests.

t/x2p

A test suite for the s2p converter.

The core uses the same testing style as the rest of Perl, a simple "ok/not ok" run through Test::Harness, but there are a few special considerations.

There are three ways to write a test in the core. Test::More, t/test.pl and ad hoc print $test ? "ok 42\n" : "not ok 42\n". The decision of which to use depends on what part of the test suite you're working on. This is a measure to prevent a high-level failure (such as Config.pm breaking) from causing basic functionality tests to fail.

t/base t/comp

Since we don't know if require works, or even subroutines, use ad hoc tests for these two. Step carefully to avoid using the feature being tested.

t/cmd t/run t/io t/op

Now that basic require() and subroutines are tested, you can use the t/test.pl library which emulates the important features of Test::More while using a minimum of core features.

You can also conditionally use certain libraries like Config, but be sure to skip the test gracefully if it's not there.

t/lib ext lib

Now that the core of Perl is tested, Test::More can be used. You can also use the full suite of core modules in the tests.

When you say "make test" Perl uses the t/TEST program to run the test suite (except under Win32 where it uses t/harness instead.) All tests are run from the t/ directory, not the directory which contains the test. This causes some problems with the tests in lib/, so here's some opportunity for some patching.

You must be triply conscious of cross-platform concerns. This usually boils down to using File::Spec and avoiding things like fork() and system() unless absolutely necessary.

Special Make Test Targets

There are various special make targets that can be used to test Perl slightly differently than the standard "test" target. Not all them are expected to give a 100% success rate. Many of them have several aliases, and many of them are not available on certain operating systems.

coretest

Run perl on all core tests (t/* and lib/[a-z]* pragma tests).

(Not available on Win32)

test.deparse

Run all the tests through B::Deparse. Not all tests will succeed.

(Not available on Win32)

test.taintwarn

Run all tests with the -t command-line switch. Not all tests are expected to succeed (until they're specifically fixed, of course).

(Not available on Win32)

minitest

Run miniperl on t/base, t/comp, t/cmd, t/run, t/io, t/op, t/uni and t/mro tests.

test.valgrind check.valgrind utest.valgrind ucheck.valgrind

(Only in Linux) Run all the tests using the memory leak + naughty memory access tool "valgrind". The log files will be named testname.valgrind.

test.third check.third utest.third ucheck.third

(Only in Tru64) Run all the tests using the memory leak + naughty memory access tool "Third Degree". The log files will be named perl.3log.testname.

test.torture torturetest

Run all the usual tests and some extra tests. As of Perl 5.8.0 the only extra tests are Abigail's JAPHs, t/japh/abigail.t.

You can also run the torture test with t/harness by giving -torture argument to t/harness.

utest ucheck test.utf8 check.utf8

Run all the tests with -Mutf8. Not all tests will succeed.

(Not available on Win32)

minitest.utf16 test.utf16

Runs the tests with UTF-16 encoded scripts, encoded with different versions of this encoding.

make utest.utf16 runs the test suite with a combination of -utf8 and -utf16 arguments to t/TEST.

(Not available on Win32)

test_harness

Run the test suite with the t/harness controlling program, instead of t/TEST. t/harness is more sophisticated, and uses the Test::Harness module, thus using this test target supposes that perl mostly works. The main advantage for our purposes is that it prints a detailed summary of failed tests at the end. Also, unlike t/TEST, it doesn't redirect stderr to stdout.

Note that under Win32 t/harness is always used instead of t/TEST, so there is no special "test_harness" target.

Under Win32's "test" target you may use the TEST_SWITCHES and TEST_FILES environment variables to control the behaviour of t/harness. This means you can say

nmake test TEST_FILES="op/*.t"
nmake test TEST_SWITCHES="-torture" TEST_FILES="op/*.t"

Parallel tests

The core distribution can now run its regression tests in parallel on Unix-like platforms. Instead of running make test, set TEST_JOBS in your environment to the number of tests to run in parallel, and run make test_harness. On a Bourne-like shell, this can be done as

TEST_JOBS=3 make test_harness  # Run 3 tests in parallel

An environment variable is used, rather than parallel make itself, because TAP::Harness needs to be able to schedule individual non-conflicting test scripts itself, and there is no standard interface to make utilities to interact with their job schedulers.

Note that currently some test scripts may fail when run in parallel (most notably ext/IO/t/io_dir.t). If necessary run just the failing scripts again sequentially and see if the failures go away. =item test-notty test_notty

Sets PERL_SKIP_TTY_TEST to true before running normal test.

Running tests by hand

You can run part of the test suite by hand by using one the following commands from the t/ directory :

./perl -I../lib TEST list-of-.t-files

or

./perl -I../lib harness list-of-.t-files

(if you don't specify test scripts, the whole test suite will be run.)

Using t/harness for testing

If you use harness for testing you have several command line options available to you. The arguments are as follows, and are in the order that they must appear if used together.

harness -v -torture -re=pattern LIST OF FILES TO TEST
harness -v -torture -re LIST OF PATTERNS TO MATCH

If LIST OF FILES TO TEST is omitted the file list is obtained from the manifest. The file list may include shell wildcards which will be expanded out.

-v

Run the tests under verbose mode so you can see what tests were run, and debug output.

-torture

Run the torture tests as well as the normal set.

-re=PATTERN

Filter the file list so that all the test files run match PATTERN. Note that this form is distinct from the -re LIST OF PATTERNS form below in that it allows the file list to be provided as well.

-re LIST OF PATTERNS

Filter the file list so that all the test files run match /(LIST|OF|PATTERNS)/. Note that with this form the patterns are joined by '|' and you cannot supply a list of files, instead the test files are obtained from the MANIFEST.

You can run an individual test by a command similar to

./perl -I../lib patho/to/foo.t

except that the harnesses set up some environment variables that may affect the execution of the test :

PERL_CORE=1

indicates that we're running this test part of the perl core test suite. This is useful for modules that have a dual life on CPAN.

PERL_DESTRUCT_LEVEL=2

is set to 2 if it isn't set already (see "PERL_DESTRUCT_LEVEL")

PERL

(used only by t/TEST) if set, overrides the path to the perl executable that should be used to run the tests (the default being ./perl).

PERL_SKIP_TTY_TEST

if set, tells to skip the tests that need a terminal. It's actually set automatically by the Makefile, but can also be forced artificially by running 'make test_notty'.

Other environment variables that may influence tests

PERL_TEST_Net_Ping

Setting this variable runs all the Net::Ping modules tests, otherwise some tests that interact with the outside world are skipped. See perl58delta.

PERL_TEST_NOVREXX

Setting this variable skips the vrexx.t tests for OS2::REXX.

PERL_TEST_NUMCONVERTS

This sets a variable in op/numconvert.t.

See also the documentation for the Test and Test::Harness modules, for more environment variables that affect testing.

Common problems when patching Perl source code

Perl source plays by ANSI C89 rules: no C99 (or C++) extensions. In some cases we have to take pre-ANSI requirements into consideration. You don't care about some particular platform having broken Perl? I hear there is still a strong demand for J2EE programmers.

Perl environment problems

• Not compiling with threading

  Compiling with threading (-Duseithreads) completely rewrites the function prototypes of Perl. You better try your changes with that. Related to this is the difference between "Perl_-less" and "Perl_-ly" APIs, for example:

  Perl_sv_setiv(aTHX_ ...);
  sv_setiv(...);

  The first one explicitly passes in the context, which is needed for e.g. threaded builds. The second one does that implicitly; do not get them mixed. If you are not passing in a aTHX_, you will need to do a dTHX (or a dVAR) as the first thing in the function.

  See "How multiple interpreters and concurrency are supported" in perlguts for further discussion about context.

• Not compiling with -DDEBUGGING

  The DEBUGGING define exposes more code to the compiler, therefore more ways for things to go wrong. You should try it.

• Introducing (non-read-only) globals

  Do not introduce any modifiable globals, truly global or file static. They are bad form and complicate multithreading and other forms of concurrency. The right way is to introduce them as new interpreter variables, see intrpvar.h (at the very end for binary compatibility).

  Introducing read-only (const) globals is okay, as long as you verify with e.g. nm libperl.a|egrep -v ' [TURtr] ' (if your nm has BSD-style output) that the data you added really is read-only. (If it is, it shouldn't show up in the output of that command.)

  If you want to have static strings, make them constant:

  static const char etc[] = "...";

  If you want to have arrays of constant strings, note carefully the right combination of consts:

      static const char * const yippee[] =
  	{"hi", "ho", "silver"};

  There is a way to completely hide any modifiable globals (they are all moved to heap), the compilation setting -DPERL_GLOBAL_STRUCT_PRIVATE. It is not normally used, but can be used for testing, read more about it in "Background and PERL_IMPLICIT_CONTEXT" in perlguts.

• Not exporting your new function

  Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any function that is part of the public API (the shared Perl library) to be explicitly marked as exported. See the discussion about embed.pl in perlguts.

• Exporting your new function

  The new shiny result of either genuine new functionality or your arduous refactoring is now ready and correctly exported. So what could possibly go wrong?

  Maybe simply that your function did not need to be exported in the first place. Perl has a long and not so glorious history of exporting functions that it should not have.

  If the function is used only inside one source code file, make it static. See the discussion about embed.pl in perlguts.

  If the function is used across several files, but intended only for Perl's internal use (and this should be the common case), do not export it to the public API. See the discussion about embed.pl in perlguts.

Portability problems

The following are common causes of compilation and/or execution failures, not common to Perl as such. The C FAQ is good bedtime reading. Please test your changes with as many C compilers and platforms as possible -- we will, anyway, and it's nice to save oneself from public embarrassment.

If using gcc, you can add the -std=c89 option which will hopefully catch most of these unportabilities. (However it might also catch incompatibilities in your system's header files.)

Use the Configure -Dgccansipedantic flag to enable the gcc -ansi -pedantic flags which enforce stricter ANSI rules.

If using the gcc -Wall note that not all the possible warnings (like -Wunitialized) are given unless you also compile with -O.

Note that if using gcc, starting from Perl 5.9.5 the Perl core source code files (the ones at the top level of the source code distribution, but not e.g. the extensions under ext/) are automatically compiled with as many as possible of the -std=c89, -ansi, -pedantic, and a selection of -W flags (see cflags.SH).

Also study perlport carefully to avoid any bad assumptions about the operating system, filesystems, and so forth.

You may once in a while try a "make microperl" to see whether we can still compile Perl with just the bare minimum of interfaces. (See README.micro.)

Do not assume an operating system indicates a certain compiler.

• Casting pointers to integers or casting integers to pointers

  void castaway(U8* p)
  {
    IV i = p;

  or

  void castaway(U8* p)
  {
    IV i = (IV)p;

  Both are bad, and broken, and unportable. Use the PTR2IV() macro that does it right. (Likewise, there are PTR2UV(), PTR2NV(), INT2PTR(), and NUM2PTR().)

• Casting between data function pointers and data pointers

  Technically speaking casting between function pointers and data pointers is unportable and undefined, but practically speaking it seems to work, but you should use the FPTR2DPTR() and DPTR2FPTR() macros. Sometimes you can also play games with unions.

• Assuming sizeof(int) == sizeof(long)

  There are platforms where longs are 64 bits, and platforms where ints are 64 bits, and while we are out to shock you, even platforms where shorts are 64 bits. This is all legal according to the C standard. (In other words, "long long" is not a portable way to specify 64 bits, and "long long" is not even guaranteed to be any wider than "long".)

  Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth. Avoid things like I32 because they are not guaranteed to be exactly 32 bits, they are at least 32 bits, nor are they guaranteed to be int or long. If you really explicitly need 64-bit variables, use I64 and U64, but only if guarded by HAS_QUAD.

• Assuming one can dereference any type of pointer for any type of data

  char *p = ...;
  long pony = *p;    /* BAD */

  Many platforms, quite rightly so, will give you a core dump instead of a pony if the p happens not be correctly aligned.

• Lvalue casts

  (int)*p = ...;    /* BAD */

  Simply not portable. Get your lvalue to be of the right type, or maybe use temporary variables, or dirty tricks with unions.

• Assume anything about structs (especially the ones you don't control, like the ones coming from the system headers)

  • That a certain field exists in a struct

  • That no other fields exist besides the ones you know of

  • That a field is of certain signedness, sizeof, or type

  • That the fields are in a certain order

    • While C guarantees the ordering specified in the struct definition, between different platforms the definitions might differ

  • That the sizeof(struct) or the alignments are the same everywhere

    • There might be padding bytes between the fields to align the fields - the bytes can be anything

    • Structs are required to be aligned to the maximum alignment required by the fields - which for native types is for usually equivalent to sizeof() of the field

• Assuming the character set is ASCIIish

  Perl can compile and run under EBCDIC platforms. See perlebcdic. This is transparent for the most part, but because the character sets differ, you shouldn't use numeric (decimal, octal, nor hex) constants to refer to characters. You can safely say 'A', but not 0x41. You can safely say '\n', but not \012. If a character doesn't have a trivial input form, you can create a #define for it in both utfebcdic.h and utf8.h, so that it resolves to different values depending on the character set being used. (There are three different EBCDIC character sets defined in utfebcdic.h, so it might be best to insert the #define three times in that file.)

  Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper case alphabetic characters. That is not true in EBCDIC. Nor for 'a' to 'z'. But '0' - '9' is an unbroken range in both systems. Don't assume anything about other ranges.

  Many of the comments in the existing code ignore the possibility of EBCDIC, and may be wrong therefore, even if the code works. This is actually a tribute to the successful transparent insertion of being able to handle EBCDIC without having to change pre-existing code.

  UTF-8 and UTF-EBCDIC are two different encodings used to represent Unicode code points as sequences of bytes. Macros with the same names (but different definitions) in utf8.h and utfebcdic.h are used to allow the calling code to think that there is only one such encoding. This is almost always referred to as utf8, but it means the EBCDIC version as well. Again, comments in the code may well be wrong even if the code itself is right. For example, the concept of invariant characters differs between ASCII and EBCDIC. On ASCII platforms, only characters that do not have the high-order bit set (i.e. whose ordinals are strict ASCII, 0 - 127) are invariant, and the documentation and comments in the code may assume that, often referring to something like, say, hibit. The situation differs and is not so simple on EBCDIC machines, but as long as the code itself uses the NATIVE_IS_INVARIANT() macro appropriately, it works, even if the comments are wrong.

• Assuming the character set is just ASCII

  ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra characters have different meanings depending on the locale. Absent a locale, currently these extra characters are generally considered to be unassigned, and this has presented some problems. This is scheduled to be changed in 5.12 so that these characters will be considered to be Latin-1 (ISO-8859-1).

• Mixing #define and #ifdef

  #define BURGLE(x) ... \
  #ifdef BURGLE_OLD_STYLE        /* BAD */
  ... do it the old way ... \
  #else
  ... do it the new way ... \
  #endif

  You cannot portably "stack" cpp directives. For example in the above you need two separate BURGLE() #defines, one for each #ifdef branch.

• Adding non-comment stuff after #endif or #else

  #ifdef SNOSH
  ...
  #else !SNOSH    /* BAD */
  ...
  #endif SNOSH    /* BAD */

  The #endif and #else cannot portably have anything non-comment after them. If you want to document what is going (which is a good idea especially if the branches are long), use (C) comments:

  #ifdef SNOSH
  ...
  #else /* !SNOSH */
  ...
  #endif /* SNOSH */

  The gcc option -Wendif-labels warns about the bad variant (by default on starting from Perl 5.9.4).

• Having a comma after the last element of an enum list

  enum color {
    CERULEAN,
    CHARTREUSE,
    CINNABAR,     /* BAD */
  };

  is not portable. Leave out the last comma.

  Also note that whether enums are implicitly morphable to ints varies between compilers, you might need to (int).

• Using //-comments

  // This function bamfoodles the zorklator.    /* BAD */

  That is C99 or C++. Perl is C89. Using the //-comments is silently allowed by many C compilers but cranking up the ANSI C89 strictness (which we like to do) causes the compilation to fail.

• Mixing declarations and code

  void zorklator()
  {
    int n = 3;
    set_zorkmids(n);    /* BAD */
    int q = 4;

  That is C99 or C++. Some C compilers allow that, but you shouldn't.

  The gcc option -Wdeclaration-after-statements scans for such problems (by default on starting from Perl 5.9.4).

• Introducing variables inside for()

  for(int i = ...; ...; ...) {    /* BAD */

  That is C99 or C++. While it would indeed be awfully nice to have that also in C89, to limit the scope of the loop variable, alas, we cannot.

• Mixing signed char pointers with unsigned char pointers

  int foo(char *s) { ... }
  ...
  unsigned char *t = ...; /* Or U8* t = ... */
  foo(t);   /* BAD */

  While this is legal practice, it is certainly dubious, and downright fatal in at least one platform: for example VMS cc considers this a fatal error. One cause for people often making this mistake is that a "naked char" and therefore dereferencing a "naked char pointer" have an undefined signedness: it depends on the compiler and the flags of the compiler and the underlying platform whether the result is signed or unsigned. For this very same reason using a 'char' as an array index is bad.

• Macros that have string constants and their arguments as substrings of the string constants

  #define FOO(n) printf("number = %d\n", n)    /* BAD */
  FOO(10);

  Pre-ANSI semantics for that was equivalent to

  printf("10umber = %d\10");

  which is probably not what you were expecting. Unfortunately at least one reasonably common and modern C compiler does "real backward compatibility" here, in AIX that is what still happens even though the rest of the AIX compiler is very happily C89.

• Using printf formats for non-basic C types

  IV i = ...;
  printf("i = %d\n", i);    /* BAD */

  While this might by accident work in some platform (where IV happens to be an int), in general it cannot. IV might be something larger. Even worse the situation is with more specific types (defined by Perl's configuration step in config.h):

  Uid_t who = ...;
  printf("who = %d\n", who);    /* BAD */

  The problem here is that Uid_t might be not only not int-wide but it might also be unsigned, in which case large uids would be printed as negative values.

  There is no simple solution to this because of printf()'s limited intelligence, but for many types the right format is available as with either 'f' or '_f' suffix, for example:

  IVdf /* IV in decimal */
  UVxf /* UV is hexadecimal */
  
  printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
  
  Uid_t_f /* Uid_t in decimal */
  
  printf("who = %"Uid_t_f"\n", who);

  Or you can try casting to a "wide enough" type:

  printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);

  Also remember that the %p format really does require a void pointer:

  U8* p = ...;
  printf("p = %p\n", (void*)p);

  The gcc option -Wformat scans for such problems.

• Blindly using variadic macros

  gcc has had them for a while with its own syntax, and C99 brought them with a standardized syntax. Don't use the former, and use the latter only if the HAS_C99_VARIADIC_MACROS is defined.

• Blindly passing va_list

  Not all platforms support passing va_list to further varargs (stdarg) functions. The right thing to do is to copy the va_list using the Perl_va_copy() if the NEED_VA_COPY is defined.

• Using gcc statement expressions

  val = ({...;...;...});    /* BAD */

  While a nice extension, it's not portable. The Perl code does admittedly use them if available to gain some extra speed (essentially as a funky form of inlining), but you shouldn't.

• Binding together several statements in a macro

  Use the macros STMT_START and STMT_END.

  STMT_START {
     ...
  } STMT_END

• Testing for operating systems or versions when should be testing for features

  #ifdef __FOONIX__    /* BAD */
  foo = quux();
  #endif

  Unless you know with 100% certainty that quux() is only ever available for the "Foonix" operating system and that is available and correctly working for all past, present, and future versions of "Foonix", the above is very wrong. This is more correct (though still not perfect, because the below is a compile-time check):

  #ifdef HAS_QUUX
  foo = quux();
  #endif

  How does the HAS_QUUX become defined where it needs to be? Well, if Foonix happens to be UNIXy enough to be able to run the Configure script, and Configure has been taught about detecting and testing quux(), the HAS_QUUX will be correctly defined. In other platforms, the corresponding configuration step will hopefully do the same.

  In a pinch, if you cannot wait for Configure to be educated, or if you have a good hunch of where quux() might be available, you can temporarily try the following:

  #if (defined(__FOONIX__) || defined(__BARNIX__))
  # define HAS_QUUX
  #endif
  
  ...
  
  #ifdef HAS_QUUX
  foo = quux();
  #endif

  But in any case, try to keep the features and operating systems separate.

Problematic System Interfaces

• malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable allocate at least one byte. (In general you should rarely need to work at this low level, but instead use the various malloc wrappers.)

• snprintf() - the return type is unportable. Use my_snprintf() instead.

Security problems

Last but not least, here are various tips for safer coding.

• Do not use gets()

  Or we will publicly ridicule you. Seriously.

• Do not use strcpy() or strcat() or strncpy() or strncat()

  Use my_strlcpy() and my_strlcat() instead: they either use the native implementation, or Perl's own implementation (borrowed from the public domain implementation of INN).

• Do not use sprintf() or vsprintf()

  If you really want just plain byte strings, use my_snprintf() and my_vsnprintf() instead, which will try to use snprintf() and vsnprintf() if those safer APIs are available. If you want something fancier than a plain byte string, use SVs and Perl_sv_catpvf().

EXTERNAL TOOLS FOR DEBUGGING PERL

Sometimes it helps to use external tools while debugging and testing Perl. This section tries to guide you through using some common testing and debugging tools with Perl. This is meant as a guide to interfacing these tools with Perl, not as any kind of guide to the use of the tools themselves.

NOTE 1: Running under memory debuggers such as Purify, valgrind, or Third Degree greatly slows down the execution: seconds become minutes, minutes become hours. For example as of Perl 5.8.1, the ext/Encode/t/Unicode.t takes extraordinarily long to complete under e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more than six hours, even on a snappy computer-- the said test must be doing something that is quite unfriendly for memory debuggers. If you don't feel like waiting, that you can simply kill away the perl process.

NOTE 2: To minimize the number of memory leak false alarms (see "PERL_DESTRUCT_LEVEL" for more information), you have to have environment variable PERL_DESTRUCT_LEVEL set to 2. The TEST and harness scripts do that automatically. But if you are running some of the tests manually-- for csh-like shells:

setenv PERL_DESTRUCT_LEVEL 2

and for Bourne-type shells:

PERL_DESTRUCT_LEVEL=2
export PERL_DESTRUCT_LEVEL

or in UNIXy environments you can also use the env command:

env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...

NOTE 3: There are known memory leaks when there are compile-time errors within eval or require, seeing S_doeval in the call stack is a good sign of these. Fixing these leaks is non-trivial, unfortunately, but they must be fixed eventually.

NOTE 4: DynaLoader will not clean up after itself completely unless Perl is built with the Configure option -Accflags=-DDL_UNLOAD_ALL_AT_EXIT.

Rational Software's Purify

Purify is a commercial tool that is helpful in identifying memory overruns, wild pointers, memory leaks and other such badness. Perl must be compiled in a specific way for optimal testing with Purify. Purify is available under Windows NT, Solaris, HP-UX, SGI, and Siemens Unix.

Purify on Unix

On Unix, Purify creates a new Perl binary. To get the most benefit out of Purify, you should create the perl to Purify using:

sh Configure -Accflags=-DPURIFY -Doptimize='-g' \
 -Uusemymalloc -Dusemultiplicity

where these arguments mean:

-Accflags=-DPURIFY

Disables Perl's arena memory allocation functions, as well as forcing use of memory allocation functions derived from the system malloc.

-Doptimize='-g'

Adds debugging information so that you see the exact source statements where the problem occurs. Without this flag, all you will see is the source filename of where the error occurred.

-Uusemymalloc

Disable Perl's malloc so that Purify can more closely monitor allocations and leaks. Using Perl's malloc will make Purify report most leaks in the "potential" leaks category.

-Dusemultiplicity

Enabling the multiplicity option allows perl to clean up thoroughly when the interpreter shuts down, which reduces the number of bogus leak reports from Purify.

Once you've compiled a perl suitable for Purify'ing, then you can just:

make pureperl

which creates a binary named 'pureperl' that has been Purify'ed. This binary is used in place of the standard 'perl' binary when you want to debug Perl memory problems.

As an example, to show any memory leaks produced during the standard Perl testset you would create and run the Purify'ed perl as:

make pureperl
cd t
../pureperl -I../lib harness

which would run Perl on test.pl and report any memory problems.

Purify outputs messages in "Viewer" windows by default. If you don't have a windowing environment or if you simply want the Purify output to unobtrusively go to a log file instead of to the interactive window, use these following options to output to the log file "perl.log":

setenv PURIFYOPTIONS "-chain-length=25 -windows=no \
 -log-file=perl.log -append-logfile=yes"

If you plan to use the "Viewer" windows, then you only need this option:

setenv PURIFYOPTIONS "-chain-length=25"

In Bourne-type shells:

PURIFYOPTIONS="..."
export PURIFYOPTIONS

or if you have the "env" utility:

env PURIFYOPTIONS="..." ../pureperl ...

Purify on NT

Purify on Windows NT instruments the Perl binary 'perl.exe' on the fly. There are several options in the makefile you should change to get the most use out of Purify:

DEFINES

You should add -DPURIFY to the DEFINES line so the DEFINES line looks something like:

DEFINES = -DWIN32 -D_CONSOLE -DNO_STRICT $(CRYPT_FLAG) -DPURIFY=1

to disable Perl's arena memory allocation functions, as well as to force use of memory allocation functions derived from the system malloc.

USE_MULTI = define

Enabling the multiplicity option allows perl to clean up thoroughly when the interpreter shuts down, which reduces the number of bogus leak reports from Purify.

#PERL_MALLOC = define

Disable Perl's malloc so that Purify can more closely monitor allocations and leaks. Using Perl's malloc will make Purify report most leaks in the "potential" leaks category.

CFG = Debug

Adds debugging information so that you see the exact source statements where the problem occurs. Without this flag, all you will see is the source filename of where the error occurred.

As an example, to show any memory leaks produced during the standard Perl testset you would create and run Purify as:

cd win32
make
cd ../t
purify ../perl -I../lib harness

which would instrument Perl in memory, run Perl on test.pl, then finally report any memory problems.

valgrind

The excellent valgrind tool can be used to find out both memory leaks and illegal memory accesses. As of version 3.3.0, Valgrind only supports Linux on x86, x86-64 and PowerPC. The special "test.valgrind" target can be used to run the tests under valgrind. Found errors and memory leaks are logged in files named testfile.valgrind.

Valgrind also provides a cachegrind tool, invoked on perl as:

VG_OPTS=--tool=cachegrind make test.valgrind

As system libraries (most notably glibc) are also triggering errors, valgrind allows to suppress such errors using suppression files. The default suppression file that comes with valgrind already catches a lot of them. Some additional suppressions are defined in t/perl.supp.

To get valgrind and for more information see

http://developer.kde.org/~sewardj/

Compaq's/Digital's/HP's Third Degree

Third Degree is a tool for memory leak detection and memory access checks. It is one of the many tools in the ATOM toolkit. The toolkit is only available on Tru64 (formerly known as Digital UNIX formerly known as DEC OSF/1).

When building Perl, you must first run Configure with -Doptimize=-g and -Uusemymalloc flags, after that you can use the make targets "perl.third" and "test.third". (What is required is that Perl must be compiled using the -g flag, you may need to re-Configure.)

The short story is that with "atom" you can instrument the Perl executable to create a new executable called perl.third. When the instrumented executable is run, it creates a log of dubious memory traffic in file called perl.3log. See the manual pages of atom and third for more information. The most extensive Third Degree documentation is available in the Compaq "Tru64 UNIX Programmer's Guide", chapter "Debugging Programs with Third Degree".

The "test.third" leaves a lot of files named foo_bar.3log in the t/ subdirectory. There is a problem with these files: Third Degree is so effective that it finds problems also in the system libraries. Therefore you should used the Porting/thirdclean script to cleanup the *.3log files.

There are also leaks that for given certain definition of a leak, aren't. See "PERL_DESTRUCT_LEVEL" for more information.

PERL_DESTRUCT_LEVEL

If you want to run any of the tests yourself manually using e.g. valgrind, or the pureperl or perl.third executables, please note that by default perl does not explicitly cleanup all the memory it has allocated (such as global memory arenas) but instead lets the exit() of the whole program "take care" of such allocations, also known as "global destruction of objects".

There is a way to tell perl to do complete cleanup: set the environment variable PERL_DESTRUCT_LEVEL to a non-zero value. The t/TEST wrapper does set this to 2, and this is what you need to do too, if you don't want to see the "global leaks": For example, for "third-degreed" Perl:

env PERL_DESTRUCT_LEVEL=2 ./perl.third -Ilib t/foo/bar.t

(Note: the mod_perl apache module uses also this environment variable for its own purposes and extended its semantics. Refer to the mod_perl documentation for more information. Also, spawned threads do the equivalent of setting this variable to the value 1.)

If, at the end of a run you get the message N scalars leaked, you can recompile with -DDEBUG_LEAKING_SCALARS, which will cause the addresses of all those leaked SVs to be dumped along with details as to where each SV was originally allocated. This information is also displayed by Devel::Peek. Note that the extra details recorded with each SV increases memory usage, so it shouldn't be used in production environments. It also converts new_SV() from a macro into a real function, so you can use your favourite debugger to discover where those pesky SVs were allocated.

If you see that you're leaking memory at runtime, but neither valgrind nor -DDEBUG_LEAKING_SCALARS will find anything, you're probably leaking SVs that are still reachable and will be properly cleaned up during destruction of the interpreter. In such cases, using the -Dm switch can point you to the source of the leak. If the executable was built with -DDEBUG_LEAKING_SCALARS, -Dm will output SV allocations in addition to memory allocations. Each SV allocation has a distinct serial number that will be written on creation and destruction of the SV. So if you're executing the leaking code in a loop, you need to look for SVs that are created, but never destroyed between each cycle. If such an SV is found, set a conditional breakpoint within new_SV() and make it break only when PL_sv_serial is equal to the serial number of the leaking SV. Then you will catch the interpreter in exactly the state where the leaking SV is allocated, which is sufficient in many cases to find the source of the leak.

As -Dm is using the PerlIO layer for output, it will by itself allocate quite a bunch of SVs, which are hidden to avoid recursion. You can bypass the PerlIO layer if you use the SV logging provided by -DPERL_MEM_LOG instead.

PERL_MEM_LOG

If compiled with -DPERL_MEM_LOG, both memory and SV allocations go through logging functions, which is handy for breakpoint setting.

Unless -DPERL_MEM_LOG_NOIMPL is also compiled, the logging functions read $ENV{PERL_MEM_LOG} to determine whether to log the event, and if so how:

$ENV{PERL_MEM_LOG} =~ /m/		Log all memory ops
$ENV{PERL_MEM_LOG} =~ /s/		Log all SV ops
$ENV{PERL_MEM_LOG} =~ /t/		include timestamp in Log
$ENV{PERL_MEM_LOG} =~ /^(\d+)/	write to FD given (default is 2)

Memory logging is somewhat similar to -Dm but is independent of -DDEBUGGING, and at a higher level; all uses of Newx(), Renew(), and Safefree() are logged with the caller's source code file and line number (and C function name, if supported by the C compiler). In contrast, -Dm is directly at the point of malloc(). SV logging is similar.

Since the logging doesn't use PerlIO, all SV allocations are logged and no extra SV allocations are introduced by enabling the logging. If compiled with -DDEBUG_LEAKING_SCALARS, the serial number for each SV allocation is also logged.

Profiling

Depending on your platform there are various of profiling Perl.

There are two commonly used techniques of profiling executables: statistical time-sampling and basic-block counting.

The first method takes periodically samples of the CPU program counter, and since the program counter can be correlated with the code generated for functions, we get a statistical view of in which functions the program is spending its time. The caveats are that very small/fast functions have lower probability of showing up in the profile, and that periodically interrupting the program (this is usually done rather frequently, in the scale of milliseconds) imposes an additional overhead that may skew the results. The first problem can be alleviated by running the code for longer (in general this is a good idea for profiling), the second problem is usually kept in guard by the profiling tools themselves.

The second method divides up the generated code into basic blocks. Basic blocks are sections of code that are entered only in the beginning and exited only at the end. For example, a conditional jump starts a basic block. Basic block profiling usually works by instrumenting the code by adding enter basic block #nnnn book-keeping code to the generated code. During the execution of the code the basic block counters are then updated appropriately. The caveat is that the added extra code can skew the results: again, the profiling tools usually try to factor their own effects out of the results.

Gprof Profiling

gprof is a profiling tool available in many UNIX platforms, it uses statistical time-sampling.

You can build a profiled version of perl called "perl.gprof" by invoking the make target "perl.gprof" (What is required is that Perl must be compiled using the -pg flag, you may need to re-Configure). Running the profiled version of Perl will create an output file called gmon.out is created which contains the profiling data collected during the execution.

The gprof tool can then display the collected data in various ways. Usually gprof understands the following options:

-a

Suppress statically defined functions from the profile.

-b

Suppress the verbose descriptions in the profile.

-e routine

Exclude the given routine and its descendants from the profile.

-f routine

Display only the given routine and its descendants in the profile.

-s

Generate a summary file called gmon.sum which then may be given to subsequent gprof runs to accumulate data over several runs.

-z

Display routines that have zero usage.

For more detailed explanation of the available commands and output formats, see your own local documentation of gprof.

quick hint:

$ sh Configure -des -Dusedevel -Doptimize='-pg' && make perl.gprof
$ ./perl.gprof someprog # creates gmon.out in current directory
$ gprof ./perl.gprof > out
$ view out

GCC gcov Profiling

Starting from GCC 3.0 basic block profiling is officially available for the GNU CC.

You can build a profiled version of perl called perl.gcov by invoking the make target "perl.gcov" (what is required that Perl must be compiled using gcc with the flags -fprofile-arcs -ftest-coverage, you may need to re-Configure).

Running the profiled version of Perl will cause profile output to be generated. For each source file an accompanying ".da" file will be created.

To display the results you use the "gcov" utility (which should be installed if you have gcc 3.0 or newer installed). gcov is run on source code files, like this

gcov sv.c

which will cause sv.c.gcov to be created. The .gcov files contain the source code annotated with relative frequencies of execution indicated by "#" markers.

Useful options of gcov include -b which will summarise the basic block, branch, and function call coverage, and -c which instead of relative frequencies will use the actual counts. For more information on the use of gcov and basic block profiling with gcc, see the latest GNU CC manual, as of GCC 3.0 see

http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc.html

and its section titled "8. gcov: a Test Coverage Program"

http://gcc.gnu.org/onlinedocs/gcc-3.0/gcc_8.html#SEC132

quick hint:

$ sh Configure -des  -Doptimize='-g' -Accflags='-fprofile-arcs -ftest-coverage' \
    -Aldflags='-fprofile-arcs -ftest-coverage' && make perl.gcov
$ rm -f regexec.c.gcov regexec.gcda
$ ./perl.gcov
$ gcov regexec.c
$ view regexec.c.gcov

Pixie Profiling

Pixie is a profiling tool available on IRIX and Tru64 (aka Digital UNIX aka DEC OSF/1) platforms. Pixie does its profiling using basic-block counting.

You can build a profiled version of perl called perl.pixie by invoking the make target "perl.pixie" (what is required is that Perl must be compiled using the -g flag, you may need to re-Configure).

In Tru64 a file called perl.Addrs will also be silently created, this file contains the addresses of the basic blocks. Running the profiled version of Perl will create a new file called "perl.Counts" which contains the counts for the basic block for that particular program execution.

To display the results you use the prof utility. The exact incantation depends on your operating system, "prof perl.Counts" in IRIX, and "prof -pixie -all -L. perl" in Tru64.

In IRIX the following prof options are available:

-h

Reports the most heavily used lines in descending order of use. Useful for finding the hotspot lines.

-l

Groups lines by procedure, with procedures sorted in descending order of use. Within a procedure, lines are listed in source order. Useful for finding the hotspots of procedures.

In Tru64 the following options are available:

-p[rocedures]

Procedures sorted in descending order by the number of cycles executed in each procedure. Useful for finding the hotspot procedures. (This is the default option.)

-h[eavy]

Lines sorted in descending order by the number of cycles executed in each line. Useful for finding the hotspot lines.

-i[nvocations]

The called procedures are sorted in descending order by number of calls made to the procedures. Useful for finding the most used procedures.

-l[ines]

Grouped by procedure, sorted by cycles executed per procedure. Useful for finding the hotspots of procedures.

-testcoverage

The compiler emitted code for these lines, but the code was unexecuted.

-z[ero]

Unexecuted procedures.

For further information, see your system's manual pages for pixie and prof.

Miscellaneous tricks

• Those debugging perl with the DDD frontend over gdb may find the following useful:

  You can extend the data conversion shortcuts menu, so for example you can display an SV's IV value with one click, without doing any typing. To do that simply edit ~/.ddd/init file and add after:

  ! Display shortcuts.
  Ddd*gdbDisplayShortcuts: \
  /t ()   // Convert to Bin\n\
  /d ()   // Convert to Dec\n\
  /x ()   // Convert to Hex\n\
  /o ()   // Convert to Oct(\n\

  the following two lines:

  ((XPV*) (())->sv_any )->xpv_pv  // 2pvx\n\
  ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx

  so now you can do ivx and pvx lookups or you can plug there the sv_peek "conversion":

  Perl_sv_peek(my_perl, (SV*)()) // sv_peek

  (The my_perl is for threaded builds.) Just remember that every line, but the last one, should end with \n\

  Alternatively edit the init file interactively via: 3rd mouse button -> New Display -> Edit Menu

  Note: you can define up to 20 conversion shortcuts in the gdb section.

• If you see in a debugger a memory area mysteriously full of 0xABABABAB or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros, see perlclib.

• Under ithreads the optree is read only. If you want to enforce this, to check for write accesses from buggy code, compile with -DPL_OP_SLAB_ALLOC to enable the OP slab allocator and -DPERL_DEBUG_READONLY_OPS to enable code that allocates op memory via mmap, and sets it read-only at run time. Any write access to an op results in a SIGBUS and abort.

  This code is intended for development only, and may not be portable even to all Unix variants. Also, it is an 80% solution, in that it isn't able to make all ops read only. Specifically it

  1. Only sets read-only on all slabs of ops at CHECK time, hence ops allocated later via require or eval will be re-write

  2. Turns an entire slab of ops read-write if the refcount of any op in the slab needs to be decreased.

  3. Turns an entire slab of ops read-write if any op from the slab is freed.

  It's not possible to turn the slabs to read-only after an action requiring read-write access, as either can happen during op tree building time, so there may still be legitimate write access.

  However, as an 80% solution it is still effective, as currently it catches a write access during the generation of Config.pm, which means that we can't yet build perl with this enabled.

CONCLUSION

We've had a brief look around the Perl source, how to maintain quality of the source code, an overview of the stages perl goes through when it's running your code, how to use debuggers to poke at the Perl guts, and finally how to analyse the execution of Perl. We took a very simple problem and demonstrated how to solve it fully - with documentation, regression tests, and finally a patch for submission to p5p. Finally, we talked about how to use external tools to debug and test Perl.

I'd now suggest you read over those references again, and then, as soon as possible, get your hands dirty. The best way to learn is by doing, so:

• Subscribe to perl5-porters, follow the patches and try and understand them; don't be afraid to ask if there's a portion you're not clear on - who knows, you may unearth a bug in the patch...

• Keep up to date with the bleeding edge Perl distributions and get familiar with the changes. Try and get an idea of what areas people are working on and the changes they're making.

• Do read the README associated with your operating system, e.g. README.aix on the IBM AIX OS. Don't hesitate to supply patches to that README if you find anything missing or changed over a new OS release.

• Find an area of Perl that seems interesting to you, and see if you can work out how it works. Scan through the source, and step over it in the debugger. Play, poke, investigate, fiddle! You'll probably get to understand not just your chosen area but a much wider range of perl's activity as well, and probably sooner than you'd think.

The Road goes ever on and on, down from the door where it began.

If you can do these things, you've started on the long road to Perl porting. Thanks for wanting to help make Perl better - and happy hacking!

Metaphoric Quotations

If you recognized the quote about the Road above, you're in luck.

Most software projects begin each file with a literal description of each file's purpose. Perl instead begins each with a literary allusion to that file's purpose.

Like chapters in many books, all top-level Perl source files (along with a few others here and there) begin with an epigramic inscription that alludes, indirectly and metaphorically, to the material you're about to read.

Quotations are taken from writings of J.R.R Tolkien pertaining to his Legendarium, almost always from The Lord of the Rings. Chapters and page numbers are given using the following editions:

• The Hobbit, by J.R.R. Tolkien. The hardcover, 70th-anniversary edition of 2007 was used, published in the UK by Harper Collins Publishers and in the US by the Houghton Mifflin Company.

• The Lord of the Rings, by J.R.R. Tolkien. The hardcover, 50th-anniversary edition of 2004 was used, published in the UK by Harper Collins Publishers and in the US by the Houghton Mifflin Company.

• The Lays of Beleriand, by J.R.R. Tolkien and published posthumously by his son and literary executor, C.J.R. Tolkien, being the 3rd of the 12 volumes in Christopher's mammoth History of Middle Earth. Page numbers derive from the hardcover edition, first published in 1983 by George Allen & Unwin; no page numbers changed for the special 3-volume omnibus edition of 2002 or the various trade-paper editions, all again now by Harper Collins or Houghton Mifflin.

Other JRRT books fair game for quotes would thus include The Adventures of Tom Bombadil, The Silmarillion, Unfinished Tales, and The Tale of the Children of Hurin, all but the first posthumously assembled by CJRT. But The Lord of the Rings itself is perfectly fine and probably best to quote from, provided you can find a suitable quote there.

So if you were to supply a new, complete, top-level source file to add to Perl, you should conform to this peculiar practice by yourself selecting an appropriate quotation from Tolkien, retaining the original spelling and punctuation and using the same format the rest of the quotes are in. Indirect and oblique is just fine; remember, it's a metaphor, so being meta is, after all, what it's for.

AUTHOR

This document was written by Nathan Torkington, and is maintained by the perl5-porters mailing list.

SEE ALSO

perlrepository

cpanm TestInit

perl -MCPAN -e shell
install TestInit