perlhack - How to hack on Perl
This document explains how Perl development works. It includes details about the Perl 5 Porters email list, the Perl repository, the Perl bug tracker, patch guidelines, and commentary on Perl development philosophy.
If you just want to submit a single small patch like a pod fix, a test for a bug, comment fixes, etc., it's easy! Here's how:
Check out the source repository
The perl source is in a git repository. You can clone the repository with the following command:
% git clone https://github.com/Perl/perl5.git perlEnsure you're following the latest advice
In case the advice in this guide has been updated recently, read the latest version directly from the perl source:
% perldoc pod/perlhack.podCreate a branch for your change
Create a branch based on blead to commit your change to, which will later be used to send it to the Perl issue tracker.
% git checkout -b mychangeMake your change
Hack, hack, hack. Keep in mind that Perl runs on many different platforms, with different operating systems that have different capabilities, different filesystem organizations, and even different character sets. perlhacktips gives advice on this.
Test your change
You can run all the tests with the following commands:
% ./Configure -des -Dusedevel
% make testKeep hacking until the tests pass.
Commit your change
Committing your work will save the change on your local system:
% git commit -a -m 'Commit message goes here'Make sure the commit message describes your change in a single sentence. For example, "Fixed spelling errors in perlhack.pod".
Send your change to the Perl issue tracker
The next step is to submit your patch to the Perl core ticket system.
Create a GitHub fork of the perl5 repository and add it as a remote, if you haven't already, as described in the GitHub documentation at https://help.github.com/en/articles/working-with-forks.
% git remote add fork [email protected]:MyUser/perl5.gitFor more information, see "Connecting to GitHub with SSH".
If you'd rather use an HTTPS URL for your git push see "Cloning with HTTPS URLs".
% git remote add fork https://github.com/MyUser/perl5.gitThen, push your new branch to your fork.
% git push -u fork mychangeFinally, create a Pull Request on GitHub from your branch to blead as described in the GitHub documentation at https://help.github.com/en/articles/creating-a-pull-request-from-a-fork.
Thank you
The porters appreciate the time you spent helping to make Perl better. Thank you!
Acknowledgement
All contributors are credited (by name and email address) in the AUTHORS file, which is part of the perl distribution, as well as the Git commit history.
If you don’t want to be included in the AUTHORS file, just let us know. Otherwise we will take your submission of a patch as permission to credit you in the AUTHORS file.
Next time
The next time you wish to make a patch, you need to start from the latest perl in a pristine state. Check you don't have any local changes or added files in your perl check-out which you wish to keep, then run these commands:
% git checkout blead
% git pull
% git reset --hard origin/blead
% git clean -dxfIf you want to report a bug in Perl, or browse existing Perl bugs and patches, use the GitHub issue tracker at https://github.com/perl/perl5/issues.
Please check the archive of the perl5-porters list (see below) and/or the bug tracking system before submitting a bug report. Often, you'll find that the bug has been reported already.
You can log in to the bug tracking system and comment on existing bug reports. If you have additional information regarding an existing bug, please add it. This will help the porters fix the bug.
The perl5-porters (p5p) mailing list is where the Perl standard distribution is maintained and developed. The people who maintain Perl are also referred to as the "Perl 5 Porters", "p5p" or just the "porters".
A searchable archive of the list is available at https://markmail.org/search/?q=perl5-porters. There is also an archive at https://archive.develooper.com/[email protected]/.
The perl5-changes mailing list receives a copy of each patch that gets submitted to the maintenance and development branches of the perl repository. See https://lists.perl.org/list/perl5-changes.html for subscription and archive information.
Many porters are also active on the irc://irc.perl.org/#p5p channel. Feel free to join the channel and ask questions about hacking on the Perl core.
All of Perl's source code is kept centrally in a Git repository at github.com. The repository contains many Perl revisions from Perl 1 onwards and all the revisions from Perforce, the previous version control system.
For much more detail on using git with the Perl repository, please see perlgit.
You will need a copy of Git for your computer. You can fetch a copy of the repository using the git protocol:
% git clone [email protected]:Perl/perl5.git perlThis clones the repository and makes a local copy in the perl directory.
If you cannot use the git protocol for firewall reasons, you can also clone via http:
% git clone https://github.com/Perl/perl5.git perlYou may access the repository over the web. This allows you to browse the tree, see recent commits, subscribe to repository notifications, search for particular commits and more. You may access it at https://github.com/Perl/perl5.
If you have a commit bit, please see perlgit for more details on using git.
If you're planning to do more extensive work than a single small fix, we encourage you to read the documentation below. This will help you focus your work and make your patches easier to incorporate into the Perl source.
If you have a small patch to submit, please submit it via the GitHub Pull Request workflow. You may also send patches to the p5p list.
Patches are reviewed and discussed on GitHub or the p5p list. Simple, uncontroversial patches will usually be applied without any discussion. When the patch is applied, the ticket will be updated and you will receive email.
In other cases, the patch will need more work or discussion. You are encouraged to participate in the discussion and advocate for your patch. Sometimes your patch may get lost in the shuffle. It's appropriate to send a reminder email to p5p if no action has been taken in a month. Please remember that the Perl 5 developers are all volunteers, and be polite.
Changes are always applied directly to the main development branch, called "blead". Some patches may be backported to a maintenance branch. If you think your patch is appropriate for the maintenance branch (see "MAINTENANCE BRANCHES" in perlpolicy), please explain why when you submit it.
If you are submitting a code patch there are several things that you can do to help the Perl 5 Porters accept your patch.
Using the GitHub Pull Request workflow, your patch will automatically be available in a suitable format. If you wish to submit a patch to the p5p list for review, make sure to create it appropriately.
If you used git to check out the Perl source, then using git format-patch will produce a patch in a style suitable for Perl. The format-patch command produces one patch file for each commit you made. If you prefer to send a single patch for all commits, you can use git diff.
% git checkout blead
% git pull
% git diff blead my-branch-nameThis produces a patch based on the difference between blead and your current branch. It's important to make sure that blead is up to date before producing the diff, that's why we call git pull first.
We strongly recommend that you use git if possible. It will make your life easier, and ours as well.
However, if you're not using git, you can still produce a suitable patch. You'll need a pristine copy of the Perl source to diff against. The porters prefer unified diffs. Using GNU diff, you can produce a diff like this:
% diff -Npurd perl.pristine perl.mineMake sure that you make realclean in your copy of Perl to remove any build artifacts, or you may get a confusing result.
As you craft each patch you intend to submit to the Perl core, it's important to write a good commit message. This is especially important if your submission will consist of a series of commits.
The first line of the commit message should be a short description without a period. It should be no longer than the subject line of an email, 50 characters being a good rule of thumb.
A lot of Git tools (Gitweb, GitHub, git log --pretty=oneline, ...) will only display the first line (cut off at 50 characters) when presenting commit summaries.
The commit message should include a description of the problem that the patch corrects or new functionality that the patch adds.
As a general rule of thumb, your commit message should help a programmer who knows the Perl core quickly understand what you were trying to do, how you were trying to do it, and why the change matters to Perl.
Why
Your commit message should describe why the change you are making is important. When someone looks at your change in six months or six years, your intent should be clear.
If you're deprecating a feature with the intent of later simplifying another bit of code, say so. If you're fixing a performance problem or adding a new feature to support some other bit of the core, mention that.
What
Your commit message should describe what part of the Perl core you're changing and what you expect your patch to do.
How
While it's not necessary for documentation changes, new tests or trivial patches, it's often worth explaining how your change works. Even if it's clear to you today, it may not be clear to a porter next month or next year.
A commit message isn't intended to take the place of comments in your code. Commit messages should describe the change you made, while code comments should describe the current state of the code.
If you've just implemented a new feature, complete with doc, tests and well-commented code, a brief commit message will often suffice. If, however, you've just changed a single character deep in the parser or lexer, you might need to write a small novel to ensure that future readers understand what you did and why you did it.
Be sure to adequately comment your code. While commenting every line is unnecessary, anything that takes advantage of side effects of operators, that creates changes that will be felt outside of the function being patched, or that others may find confusing should be documented. If you are going to err, it is better to err on the side of adding too many comments than too few.
The best comments explain why the code does what it does, not what it does.
In general, please follow the particular style of the code you are patching.
In particular, follow these general guidelines for patching Perl sources:
4-wide indents for code, 2-wide indents for nested CPP #defines, with 8-wide tabstops.
Use spaces for indentation, not tab characters.
The codebase is a mixture of tabs and spaces for indentation, and we are moving to spaces only. Converting lines you're patching from 8-wide tabs to spaces will help this migration.
Try not to exceed 79 columns
In general, we target 80 column lines. When sticking to 80 columns would lead to torturous code or rework, it's fine to go longer. Try to keep your excess past 80 to a minimum.
ANSI C prototypes
Uncuddled elses and "K&R" style for indenting control constructs
No C++ style (//) comments
Mark places that need to be revisited with XXX (and revisit often!)
Opening brace lines up with "if" when conditional spans multiple lines; should be at end-of-line otherwise
In function definitions, name starts in column 0 (return value-type is on previous line)
Single space after keywords that are followed by parens, no space between function name and following paren
Avoid assignments in conditionals, but if they're unavoidable, use extra paren, e.g. "if (a && (b = c)) ..."
"return foo;" rather than "return(foo);"
"if (!foo) ..." rather than "if (foo == FALSE) ..." etc.
Do not declare variables using "register". It may be counterproductive with modern compilers, and is deprecated in C++, under which the Perl source is regularly compiled.
In-line functions that are in headers that are accessible to XS code need to be able to compile without warnings with commonly used extra compilation flags, such as gcc's -Wswitch-default which warns whenever a switch statement does not have a "default" case. The use of these extra flags is to catch potential problems in legal C code, and is often used by Perl aggregators, such as Linux distributors.
If your patch changes code (rather than just changing documentation), you should also include one or more test cases which illustrate the bug you're fixing or validate the new functionality you're adding. In general, you should update an existing test file rather than create a new one.
Your test suite additions should generally follow these guidelines (courtesy of Gurusamy Sarathy <[email protected]>):
Know what you're testing. Read the docs, and the source.
Tend to fail, not succeed.
Interpret results strictly.
Use unrelated features (this will flush out bizarre interactions).
Use non-standard idioms (otherwise you are not testing TIMTOWTDI).
Avoid using hardcoded test numbers whenever possible (the EXPECTED/GOT found in t/op/tie.t is much more maintainable, and gives better failure reports).
Give meaningful error messages when a test fails.
Avoid using qx// and system() unless you are testing for them. If you do use them, make sure that you cover _all_ perl platforms.
Unlink any temporary files you create.
Promote unforeseen warnings to errors with $SIG{__WARN__}.
Be sure to use the libraries and modules shipped with the version being tested, not those that were already installed.
Add comments to the code explaining what you are testing for.
Make updating the '1..42' string unnecessary. Or make sure that you update it.
Test _all_ behaviors of a given operator, library, or function.
Test all optional arguments.
Test return values in various contexts (boolean, scalar, list, lvalue).
Use both global and lexical variables.
Don't forget the exceptional, pathological cases.
This works just like patching anything else, with one extra consideration.
Modules in the cpan/ directory of the source tree are maintained outside of the Perl core. When the author updates the module, the updates are simply copied into the core. See that module's documentation or its listing on https://metacpan.org/ for more information on reporting bugs and submitting patches.
In most cases, patches to modules in cpan/ should be sent upstream and should not be applied to the Perl core individually. If a patch to a file in cpan/ absolutely cannot wait for the fix to be made upstream, released to CPAN and copied to blead, you must add (or update) a CUSTOMIZED entry in the Porting/Maintainers.pl file to flag that a local modification has been made. See Porting/Maintainers.pl for more details.
In contrast, modules in the dist/ directory are maintained in the core.
For changes significant enough to warrant a pod/perldelta.pod entry, the porters will greatly appreciate it if you submit a delta entry along with your actual change. Significant changes include, but are not limited to:
Adding, deprecating, or removing core features
Adding, deprecating, removing, or upgrading core or dual-life modules
Adding new core tests
Fixing security issues and user-visible bugs in the core
Changes that might break existing code, either on the perl or C level
Significant performance improvements
Adding, removing, or significantly changing documentation in the pod/ directory
Important platform-specific changes
Please make sure you add the perldelta entry to the right section within pod/perldelta.pod. More information on how to write good perldelta entries is available in the Style section of Porting/how_to_write_a_perldelta.pod.
New features and extensions to the language can be contentious. There is no specific set of criteria which determine what features get added, but here are some questions to consider when developing a patch:
Our goals include, but are not limited to:
Keep it fast, simple, and useful.
Keep features/concepts as orthogonal as possible.
No arbitrary limits (platforms, data sizes, cultures).
Keep it open and exciting to use/patch/advocate Perl everywhere.
Either assimilate new technologies, or build bridges to them.
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.
It's a cardinal sin to break existing Perl programs. New warnings can be 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.
The Perl 5 core includes mechanisms to help porters make backwards incompatible changes more compatible such as the feature and deprecate modules. Please use them when appropriate.
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.
Whenever possible, new features should be prototyped in a CPAN module before they will be considered for the core.
Is this something that only the submitter wants added to the language, or is it 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.
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. Similarly, a series of small patches is greatly preferred over a single large patch.
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.
Good patches (tight code, complete, correct) stand more chance of going in. Sloppy or incorrect patches might be placed on the back burner until fixes can be made, or they might be discarded altogether without further notice.
The worst patches make use of system-specific features. It's highly unlikely that non-portable additions to the Perl language will be accepted.
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?
Patches without documentation are probably ill-thought out or incomplete. No features can be added or changed without documentation, so submitting a patch for the appropriate pod docs as well as the source code is important.
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.
Work for the committers, work for Perl programmers, work for module authors, ... Perl is supposed to be easy.
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.
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.
The t/test.pl library provides some of the features of Test::More, but avoids loading most modules and uses as few core features as possible.
If you write your own test, use the Test Anything Protocol.
t/base, t/comp and t/opbasic
Since we don't know if require works, or even subroutines, use ad hoc tests for these three. Step carefully to avoid using the feature being tested. Tests in t/opbasic, for instance, have been placed there rather than in t/op because they test functionality which t/test.pl presumes has already been demonstrated to work.
All other subdirectories of t/
Now that basic require() and subroutines are tested, you can use the t/test.pl library.
You can also use certain libraries like Config conditionally, but be sure to skip the test gracefully if it's not there.
Test files not found under t/
This category includes .t files underneath directories such as dist, ext and lib. Since the core of Perl has now been tested, Test::More can and now should be used. You can also use the full suite of core modules in the tests. (As noted in "Patching a core module" above, changes to .t files found under cpan/ should be submitted to the upstream maintainers of those modules.)
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, avoiding things like fork() and system() unless absolutely necessary, and not assuming that a given character has a particular ordinal value (code point) or that its UTF-8 representation is composed of particular bytes.
There are several functions available to specify characters and code points portably in tests. The always-preloaded functions utf8::unicode_to_native() and its inverse utf8::native_to_unicode() take code points and translate appropriately. The file t/charset_tools.pl has several functions that can be useful. It has versions of the previous two functions that take strings as inputs -- not single numeric code points: uni_to_native() and native_to_uni(). If you must look at the individual bytes comprising a UTF-8 encoded string, byte_utf8a_to_utf8n() takes as input a string of those bytes encoded for an ASCII platform, and returns the equivalent string in the native platform. For example, byte_utf8a_to_utf8n("\xC2\xA0") returns the byte sequence on the current platform that form the UTF-8 for U+00A0, since "\xC2\xA0" are the UTF-8 bytes on an ASCII platform for that code point. This function returns "\xC2\xA0" on an ASCII platform, and "\x80\x41" on an EBCDIC 1047 one.
But easiest is, if the character is specifiable as a literal, like "A" or "%", to use that; if not so specificable, you can use \N{} , if the side effects aren't troublesome. Simply specify all your characters in hex, using \N{U+ZZ} instead of \xZZ. \N{} is the Unicode name, and so it always gives you the Unicode character. \N{U+41} is the character whose Unicode code point is 0x41, hence is 'A' on all platforms. The side effects are:
These select Unicode rules. That means that in double-quotish strings, the string is always converted to UTF-8 to force a Unicode interpretation (you can utf8::downgrade() afterwards to convert back to non-UTF8, if possible). In regular expression patterns, the conversion isn't done, but if the character set modifier would otherwise be /d, it is changed to /u.
If you use the form \N{character name}, the charnames module gets automatically loaded. This may not be suitable for the test level you are doing.
If you are testing locales (see perllocale), there are helper functions in t/loc_tools.pl to enable you to see what locales there are on the current platform.
make test targetsThere 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.
test_porting
This runs some basic sanity tests on the source tree and helps catch basic errors before you submit a patch.
minitest
Run miniperl on t/base, t/comp, t/cmd, t/run, t/io, t/op, t/uni and t/mro tests.
miniperl is a minimalistic perl built to bootstrap building extensions, utilties, documentation etc. It doesn't support dynamic loading and depending on the point in the build process will only have access to a limited set of core modules. miniperl is not intended for day to day use.
test.valgrind check.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_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 the Unix build process you may use the TEST_ARGS and TEST_FILES parameters to pass arguments through to the underlying harness call. This means that for instance you could do
make test_harness TEST_ARGS="-v -re pat"which would make, and then run the test harness in verbose mode over files which contain "pat". Or you could do
make test_harness TEST_ARGS="-torture" TEST_FILES="op/*.t"and run torture tests on files matching the glob "op/*.t".
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"Note that for compatibility with the unix build process TEST_ARGS may also be used instead of the traditional TEST_SWITCHES argument.
test-notty test_notty
Sets PERL_SKIP_TTY_TEST to true before running normal test.
The core distribution can now run its regression tests in parallel on Unix-like and Windows platforms. On Unix, 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 parallelAn environment variable is used, rather than parallel make itself, because