++ed by:

146 PAUSE users
121 non-PAUSE users.

Steve Hay
and 1 contributors


perl5120delta - what is new for perl v5.12.0






This document describes differences between the 5.10.0 release and the 5.12.0 release.

Many of the bug fixes in 5.12.0 were already seen in the 5.10.1 maintenance release since the two releases were kept closely coordinated (while 5.12.0 was still called 5.11.something).

You can see the list of changes in the 5.10.1 release by reading perl5101delta.

Core Enhancements

qr overloading

It is now possible to overload the qr// operator, that is, conversion to regexp, like it was already possible to overload conversion to boolean, string or number of objects. It is invoked when an object appears on the right hand side of the =~ operator or when it is interpolated into a regexp. See overload.

Pluggable keywords

Extension modules can now cleanly hook into the Perl parser to define new kinds of keyword-headed expression and compound statement. The syntax following the keyword is defined entirely by the extension. This allow a completely non-Perl sublanguage to be parsed inline, with the correct ops cleanly generated. This feature is currently considered experimental.

See "PL_keyword_plugin" in perlapi for the mechanism. The Perl core source distribution also includes a new module XS::APItest::KeywordRPN, which implements reverse Polish notation arithmetic via pluggable keywords. This module is mainly used for test purposes, and is not normally installed, but also serves as an example of how to use the new mechanism.

APIs for more internals

The lowest layers of the lexer and parts of the pad system now have C APIs available to XS extensions. These are necessary to support proper use of pluggable keywords, but have other uses too. The new APIs are experimental, and only cover a small proportion of what would be necessary to take full advantage of the core's facilities in these areas. It is intended that the Perl 5.13 development cycle will see the addition of a full range of clean, supported interfaces.

Overridable function lookup

Where an extension module hooks the creation of rv2cv ops to modify the subroutine lookup process, this now works correctly for bareword subroutine calls. This means that prototypes on subroutines referenced this way will be processed correctly. (Previously bareword subroutine names were initially looked up, for parsing purposes, by an unhookable mechanism, so extensions could only properly influence subroutine names that appeared with an & sigil.)

Unicode version

Perl is shipped with the latest Unicode version, 5.2, dated October 2009. See http://www.unicode.org/versions/Unicode5.2.0 for details about this release of Unicode. See perlunicode for instructions on installing and using older versions of Unicode.

Unicode properties

A concerted effort has been made to update Perl to be in sync with the latest Unicode standard. Changes for this include:

Perl can now handle every Unicode character property. A new pod, perluniprops, lists all available non-Unihan character properties. By default the Unihan properties and certain others (deprecated and Unicode internal-only ones) are not exposed. See below for more details on these; there is also a section in the pod listing them, and explaining why they are not exposed.

Perl now fully supports the Unicode compound-style of using = and : in writing regular expressions: \p{property=value} and \p{property:value} (both of which mean the same thing).

Perl now fully supports the Unicode loose matching rules for text between the braces in \p{...} constructs. In addition, Perl allows underscores between digits of numbers.

All the Unicode-defined synonyms for properties and property values are now accepted.

qr/\X/, which matches a Unicode logical character, has been expanded to work better with various Asian languages. It now is defined as an extended grapheme cluster. (See http://www.unicode.org/reports/tr29/). Anything matched previously and that made sense will continue to be matched, but in addition:

  • \X will not break apart a CR LF sequence.

  • \X will now match a sequence which includes the ZWJ and ZWNJ characters.

  • \X will now always match at least one character, including an initial mark. Marks generally come after a base character, but it is possible in Unicode to have them in isolation, and \X will now handle that case, for example at the beginning of a line, or after a ZWSP. And this is the part where \X doesn't match the things that it used to that don't make sense. Formerly, for example, you could have the nonsensical case of an accented LF.

  • \X will now match a (Korean) Hangul syllable sequence, and the Thai and Lao exception cases.

Otherwise, this change should be transparent for the non-affected languages.

\p{...} matches using the Canonical_Combining_Class property were completely broken in previous Perls. This is now fixed.

In previous Perls, the Unicode Decomposition_Type=Compat property and a Perl extension had the same name, which led to neither matching all the correct values (with more than 100 mistakes in one, and several thousand in the other). The Perl extension has now been renamed to be Decomposition_Type=Noncanonical (short: dt=noncanon). It has the same meaning as was previously intended, namely the union of all the non-canonical Decomposition types, with Unicode Compat being just one of those.

\p{Uppercase} and \p{Lowercase} have been brought into line with the Unicode definitions. This means they each match a few more characters than previously.

\p{Cntrl} now matches the same characters as \p{Control}. This means it no longer will match Private Use (gc=co), Surrogates (gc=cs), nor Format (gc=cf) code points. The Format code points represent the biggest possible problem. All but 36 of them are either officially deprecated or strongly discouraged from being used. Of those 36, likely the most widely used are the soft hyphen (U+00AD), and BOM, ZWSP, ZWNJ, WJ, and similar characters, plus bidirectional controls.

\p{Alpha} now matches the same characters as \p{Alphabetic}. The Perl definition included a number of things that aren't really alpha (all marks), while omitting many that were. As a direct consequence, the definitions of \p{Alnum} and \p{Word} which depend on Alpha also change correspondingly.

\p{Word} also now doesn't match certain characters it wasn't supposed to, such as fractions.

\p{Print} no longer matches the line control characters: Tab, LF, CR, FF, VT, and NEL. This brings it in line with standards and the documentation.

\p{Decomposition_Type=Canonical} now includes the Hangul syllables.

\p{XDigit} now matches the same characters as \p{Hex_Digit}. This means that in addition to the characters it currently matches, [A-Fa-f0-9], it will also match the 22 fullwidth equivalents, for example U+FF10: FULLWIDTH DIGIT ZERO.

The Numeric type property has been extended to include the Unihan characters.

There is a new Perl extension, the 'Present_In', or simply 'In', property. This is an extension of the Unicode Age property, but \p{In=5.0} matches any code point whose usage has been determined as of Unicode version 5.0. The \p{Age=5.0} only matches code points added in precisely version 5.0.

A number of properties did not have the correct values for unassigned code points. This is now fixed. The affected properties are Bidi_Class, East_Asian_Width, Joining_Type, Decomposition_Type, Hangul_Syllable_Type, Numeric_Type, and Line_Break.

The Default_Ignorable_Code_Point, ID_Continue, and ID_Start properties have been updated to their current Unicode definitions.

Certain properties that are supposed to be Unicode internal-only were erroneously exposed by previous Perls. Use of these in regular expressions will now generate, if enabled, a deprecated warning message. The properties are: Other_Alphabetic, Other_Default_Ignorable_Code_Point, Other_Grapheme_Extend, Other_ID_Continue, Other_ID_Start, Other_Lowercase, Other_Math, and Other_Uppercase.

An installation can now fairly easily change which Unicode properties Perl understands. As mentioned above, certain properties are by default turned off. These include all the Unihan properties (which should be accessible via the CPAN module Unicode::Unihan) and any deprecated or Unicode internal-only property that Perl has never exposed.

The generated files in the lib/unicore/To directory are now more clearly marked as being stable, directly usable by applications. New hash entries in them give the format of the normal entries, which allows for easier machine parsing. Perl can generate files in this directory for any property, though most are suppressed. An installation can choose to change which get written. Instructions are in perluniprops.

Regular Expressions

U+0FFFF is now a legal character in regular expressions.

A proper interface for pluggable Method Resolution Orders

As of Perl 5.11.0 there is a new interface for plugging and using method resolution orders other than the default (linear depth first search). The C3 method resolution order added in 5.10.0 has been re-implemented as a plugin, without changing its Perl-space interface. See perlmroapi for more information.

The overloading pragma

This pragma allows you to lexically disable or enable overloading for some or all operations. (Yuval Kogman)

\N regex escape

A new regex escape has been added, \N. It will match any character that is not a newline, independently from the presence or absence of the single line match modifier /s. It is not usable within a character class. (If \N is followed by an opening brace and by a letter, perl will still assume that a Unicode character name is coming, so compatibility is preserved.) (Rafael Garcia-Suarez).

This will break a custom charnames translator which allows numbers for character names, as \N{3} will now mean to match 3 non-newline characters, and not the character whose name is 3. (No standard name is a number, so only a custom translator would be affected.)

Implicit strictures

Using the use VERSION syntax with a version number greater or equal to 5.11.0 will also lexically enable strictures just like use strict would do (in addition to enabling features.) So, the following:

    use 5.11.0;

will now imply:

    use strict;
    use feature ':5.11';

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.

The ... operator

A new operator, ..., nicknamed the Yada Yada operator, has been added. It is intended to mark placeholder code that is not yet implemented. See "Yada Yada Operator" in perlop. (chromatic)

DTrace support

Some support for DTrace has been added. See "DTrace support" in INSTALL.

Support for configure_requires in CPAN module metadata

Both CPAN and CPANPLUS now support the configure_requires keyword in the META.yml metadata file included in most recent CPAN distributions. This allows distribution authors to specify configuration prerequisites that must be installed before running Makefile.PL or Build.PL.

See the documentation for ExtUtils::MakeMaker or Module::Build for more on how to specify configure_requires when creating a distribution for CPAN.

each is now more flexible

The each function can now operate on arrays.

Y2038 compliance

Perl's core time-related functions are now Y2038 compliant. (With 29 years to spare!)

$, flexibility

The variable $, may now be tied.

// in where clauses

// now behaves like || in when clauses

Enabling warnings from your shell environment

You can now set -W from the PERL5OPT environment variable

delete local

delete local now allows you to locally delete a hash entry.

New support for Abstract namespace sockets

Abstract namespace sockets are Linux-specific socket type that live in AF_UNIX family, slightly abusing it to be able to use arbitrary character arrays as addresses: They start with nul byte and are not terminated by nul byte, but with the length passed to the socket() system call.

New package NAME VERSION syntax

This new syntax allows a module author to set the $VERSION of a namespace when the namespace is declared with 'package'. It eliminates the need for our $VERSION = ... and similar constructs. E.g.

      package Foo::Bar 1.23;
      # $Foo::Bar::VERSION == 1.23

There are several advantages to this:

  • $VERSION is parsed in exactly the same way as use NAME VERSION

  • $VERSION is set at compile time

  • $VERSION is a version object that provides proper overloading of comparision operators so comparing $VERSION to decimal (1.23) or dotted-decimal (v1.2.3) version numbers works correctly.

  • Eliminates $VERSION = ... and eval $VERSION clutter

  • As it requires VERSION to be a numeric literal or v-string literal, it can be statically parsed by toolchain modules without eval the way MM->parse_version does for $VERSION = ...

  • It does not break old code with only package NAME, but code that uses package NAME VERSION will need to be restricted to perl 5.12.0 or newer This is analogous to the change to open from two-args to three-args. Users requiring the latest Perl will benefit, and perhaps after several years, it will become a standard practice.

However, package NAME VERSION requires a new, 'strict' version number format. See "Version number formats" for details.

Incompatible Changes

Version number formats

Acceptable version number formats have been formalized into "strict" and "lax" rules. package NAME VERSION takes a strict version number. UNIVERSAL::VERSION and the version object constructors take lax version numbers. Providing an invalid version will result in a fatal error. The version argument in use NAME VERSION is first parsed as a numeric literal or v-string and then passed to UNIVERSAL::VERSION (and must then pass the "lax" format test).

These formats are documented fully in the version module. To a first approximation, a "strict" version number is a positive decimal number (integer or decimal-fraction) without exponentiation or else a dotted-decimal v-string with a leading 'v' character and at least three components. A "lax" version number allows v-strings with fewer than three components or without a leading 'v'. Under "lax" rules, both decimal and dotted-decimal versions may have a trailing "alpha" component separated by an underscore character after a fractional or dotted-decimal component.

The version module adds version::is_strict and version::is_lax functions to check a scalar against these rules.

@INC reorganization

In @INC, ARCHLIB and PRIVLIB now occur after after the current version's site_perl and vendor_perl.

Switch statement changes

The handling of complex expressions by the given/when switch statement has been enhanced. These enhancements are also available in 5.10.1 and subsequent 5.10 releases. There are two new cases where when now interprets its argument as a boolean, instead of an expression to be used in a smart match:

flip-flop operators

The .. and ... flip-flop operators are now evaluated in boolean context, following their usual semantics; see "Range Operators" in perlop.

Note that, as in perl 5.10.0, when (1..10) will not work to test whether a given value is an integer between 1 and 10; you should use when ([1..10]) instead (note the array reference).

However, contrary to 5.10.0, evaluating the flip-flop operators in boolean context ensures it can now be useful in a when(), notably for implementing bistable conditions, like in:

    when (/^=begin/ .. /^=end/) {
      # do something

defined-or operator

A compound expression involving the defined-or operator, as in when (expr1 // expr2), will be treated as boolean if the first expression is boolean. (This just extends the existing rule that applies to the regular or operator, as in when (expr1 || expr2).)

Smart match changes

This section details more changes brought to the semantics to the smart match operator, that naturally also modify the behaviour of the switch statements where smart matching is implicitly used. These changes were also made for the 5.10.1 release, and will remain in subsequent 5.10 releases.

Changes to type-based dispatch

The smart match operator ~~ is no longer commutative. The behaviour of a smart match now depends primarily on the type of its right hand argument. Moreover, its semantics have been adjusted for greater consistency or usefulness in several cases. While the general backwards compatibility is maintained, several changes must be noted:

  • Code references with an empty prototype are no longer treated specially. They are passed an argument like the other code references (even if they choose to ignore it).

  • %hash ~~ sub {} and @array ~~ sub {} now test that the subroutine returns a true value for each key of the hash (or element of the array), instead of passing the whole hash or array as a reference to the subroutine.

  • Due to the commutativity breakage, code references are no longer treated specially when appearing on the left of the ~~ operator, but like any vulgar scalar.

  • undef ~~ %hash is always false (since undef can't be a key in a hash). No implicit conversion to "" is done (as was the case in perl 5.10.0).

  • $scalar ~~ @array now always distributes the smart match across the elements of the array. It's true if one element in @array verifies $scalar ~~ $element. This is a generalization of the old behaviour that tested whether the array contained the scalar.

The full dispatch table for the smart match operator is given in "Smart matching in detail" in perlsyn.

Smart match and overloading

According to the rule of dispatch based on the rightmost argument type, when an object overloading ~~ appears on the right side of the operator, the overload routine will always be called (with a 3rd argument set to a true value, see overload.) However, when the object will appear on the left, the overload routine will be called only when the rightmost argument is a simple scalar. This way, distributivity of smart match across arrays is not broken, as well as the other behaviours with complex types (coderefs, hashes, regexes). Thus, writers of overloading routines for smart match mostly need to worry only with comparing against a scalar, and possibly with stringification overloading; the other common cases will be automatically handled consistently.

~~ will now refuse to work on objects that do not overload it (in order to avoid relying on the object's underlying structure). (However, if the object overloads the stringification or the numification operators, and if overload fallback is active, it will be used instead, as usual.)

Labels can't be keywords

Labels used as targets for the goto, last, next or redo statements cannot be keywords anymore. This restriction will prevent potential confusion between the goto LABEL and goto EXPR syntaxes: for example, a statement like goto print would jump to a label whose name would be the return value of print(), (usually 1), instead of a label named print. Moreover, the other control flow statements would just ignore any keyword passed to them as a label name. Since such labels cannot be defined anymore, this kind of error will be avoided.

Other incompatible changes

  • The definitions of a number of Unicode properties have changed to match those of the current Unicode standard. These are listed above under "Unicode properties". This could break code that is expecting the old definitions.

  • The boolkeys op moved to the group of hash ops. This breaks binary compatibility.

  • Filehandles are blessed directly into IO::Handle, as FileHandle is merely a wrapper around IO::Handle.

    The previous behaviour was to bless Filehandles into FileHandle (an empty proxy class) if it was loaded into memory and otherwise to bless them into IO::Handle.

  • The semantics of use feature :5.10* have changed slightly. See "Modules and Pragmata" for more information.

  • The version control system used for the development of the perl interpreter has been switched from Perforce to git. This is mainly an internal issue that only affects people actively working on the perl core; but it may have minor external visibility, for example in some of details of the output of perl -V. See perlrepository for more information.

  • The internal structure of the ext/ directory in the perl source has been reorganised. In general, a module Foo::Bar whose source was stored under ext/Foo/Bar/ is now located under ext/Foo-Bar/. Also, nearly all dual-life modules have been moved from lib/ to ext/. This is purely a source tarball change, and should make no difference to the compilation or installation of perl, unless you have a very customised build process that explicitly relies on this structure, or which hard-codes the nonxs_ext Configure parameter. Specifically, this change does not by default alter the location of any files in the final installation.

  • As part of the Test::Harness 2.x to 3.x upgrade, the experimental Test::Harness::Straps module has been removed. See "Updated Modules" for more details.

  • As part of the ExtUtils::MakeMaker upgrade, the ExtUtils::MakeMaker::bytes and ExtUtils::MakeMaker::vmsish modules have been removed from this distribution.

  • Module::CoreList no longer contains the %:patchlevel hash.

  • This one is actually a change introduced in 5.10.0, but it was missed from that release's perldelta, so it is mentioned here instead.

    A bugfix related to the handling of the /m modifier and qr resulted in a change of behaviour between 5.8.x and 5.10.0:

        # matches in 5.8.x, doesn't match in 5.10.0
        $re = qr/^bar/; "foo\nbar" =~ /$re/m;
  • length undef now returns undef.

  • Unsupported private C API functions are now declared "static" to prevent leakage to Perl's public API.

  • To support the bootstrapping process, miniperl no longer builds with UTF-8 support in the regexp engine.

    This allows a build to complete with PERL_UNICODE set and a UTF-8 locale. Without this there's a bootstrapping problem, as miniperl can't load the UTF-8 components of the regexp engine, because they're not yet built.

  • miniperl's @INC is now restricted to just -I..., the split of $ENV{PERL5LIB}, and "."

  • A space or a newline is now required after a "#line XXX" directive.

  • Tied filehandles now have an additional method EOF which provides the EOF type

  • To better match all other flow control statements, foreach may no longer be used as an attribute.


From time to time, Perl's developers find it necessary to deprecate features or modules we've previously shipped as part of the core distribution. We are well aware of the pain and frustration that a backwards-incompatible change to Perl can cause for developers building or maintaining software in Perl. You can be sure that when we deprecate a functionality or syntax, it isn't a choice we make lightly. Sometimes, we choose to deprecate functionality or syntax because it was found to be poorly designed or implemented. Sometimes, this is because they're holding back other features or causing performance problems. Sometimes, the reasons are more complex. Wherever possible, we try to keep deprecated functionality available to developers in its previous form for at least one major release. So long as a deprecated feature isn't actively disrupting our ability to maintain and extend Perl, we'll try to leave it in place as long as possible.

The following items are now deprecated.

Use of := to mean an empty attribute list is now deprecated.

An accident of Perl's parser meant that these constructions were all equivalent:

    my $pi := 4;
    my $pi : = 4;
    my $pi :  = 4;

with the : being treated as the start of an attribute list, which ends before the =. As whitespace is not significant here, all are parsed as an empty attribute list, hence all the above are equivalent to, and better written as

    my $pi = 4;

because no attribute processing is done for an empty list.

As is, this meant that := cannot be used as a new token, without silently changing the meaning of existing code. Hence that particular form is now deprecated, and will become a syntax error. If it is absolutely necessary to have empty attribute lists (for example, because of a code generator) then avoid the warning by adding a space before the =.


The method UNIVERSAL->import() is now deprecated. Attempting to pass import arguments to a use UNIVERSAL statement will result in a deprecation warning.

Use of "goto" to jump into a construct is deprecated

Using goto to jump from an outer scope into an inner scope is now deprecated. This rare use case was causing problems in the implementation of scopes.

Deprecated Modules

The following modules will be removed from the core distribution in a future release, and should be installed from CPAN instead. Distributions on CPAN which require these should add them to their prerequisites. The core versions of these modules warnings will issue a deprecation warning.

If you ship a packaged version of Perl, either alone or as part of a larger system, then you should carefully consider the reprecussions of core module deprecations. You may want to consider shipping your default build of Perl with packages for some or all deprecated modules which install into vendor or site perl library directories. This will inhibit the deprecation warnings.

Alternatively, you may want to consider patching lib/deprecate.pm to provide deprecation warnings specific to your packaging system or distribution of Perl.


Switch is buggy and should be avoided. See "Switch statements" in perlsyn for its replacement.


suidperl has been removed. It used to provide a mechanism to emulate setuid permission bits on systems that don't support it properly.

Assignment to $[

Remove attrs, which has been deprecated since 1999-10-02.

Use of the attribute :locked on subroutines.
Use of "locked" with the attributes pragma.
Use of "unique" with the attributes pragma.
Numerous Perl 4-era libraries:

termcap.pl, tainted.pl, stat.pl, shellwords.pl, pwd.pl, open3.pl, open2.pl, newgetopt.pl, look.pl, find.pl, finddepth.pl, importenv.pl, hostname.pl, getopts.pl, getopt.pl, getcwd.pl, flush.pl, fastcwd.pl, exceptions.pl, ctime.pl, complete.pl, cacheout.pl, bigrat.pl, bigint.pl, bigfloat.pl, assert.pl, abbrev.pl, dotsh.pl, and timelocal.pl are all now deprecated. Using them will incur a warning.

Modules and Pragmata

Dual-lifed modules moved

Dual-lifed modules maintained primarily in the Perl core now live in dist/. Dual-lifed modules maintained primarily on CPAN now live in cpan/

In previous releases of Perl, it was customary to enumerate all module changes in this section of the perldelta file. From 5.11.0 forward only notable updates (such as new or deprecated modules ) will be listed in this section. For a complete reference to the versions of modules shipped in a given release of perl, please see Module::CoreList.

New Modules and Pragmata

  • autodie

    This is a new lexically-scoped alternative for the Fatal module. The bundled version is 2.06_01. Note that in this release, using a string eval when autodie is in effect can cause the autodie behaviour to leak into the surrounding scope. See "BUGS" in autodie for more details.

  • Compress::Raw::Bzip2

    This has been added to the core (version 2.020).

  • parent

    This pragma establishes an ISA relationship with base classes at compile time. It provides the key feature of base without further unwanted behaviors.

  • Parse::CPAN::Meta

    This has been added to the core (version 1.39).

Pragmata Changes

  • overloading

    See "The overloading pragma" above.

  • attrs

    The attrs pragma has been removed. It had been marked as deprecated since 5.6.0.

  • charnames

    The Unicode NameAliases.txt database file has been added. This has the effect of adding some extra \N character names that formerly wouldn't have been recognised; for example, "\N{LATIN CAPITAL LETTER GHA}".

  • feature

    The meaning of the :5.10 and :5.10.X feature bundles has changed slightly. The last component, if any (i.e. X) is simply ignored. This is predicated on the assumption that new features will not, in general, be added to maintenance releases. So :5.10 and :5.10.X have identical effect. This is a change to the behaviour documented for 5.10.0.

  • mro

    Upgraded from version 1.00 to 1.01. Performance for single inheritance is 40% faster - see "Performance Enhancements" below.

    mro is now implemented as an XS extension. The documented interface has not changed. Code relying on the implementation detail that some mro:: methods happened to be available at all times gets to "keep both pieces".

  • diagnostics

    Supports %.0f formatting internally.

  • overload

    Allow overloading of 'qr'.

  • constant

    Upgraded from version 1.19 to 1.20.

  • diagnostics

    This pragma no longer suppresses Use of uninitialized value in range (or flip) warnings. [perl #71204]

  • feature

    Upgraded from 1.13 to 1.14. Added the unicode_strings feature:

        use feature "unicode_strings";

    This pragma turns on Unicode semantics for the case-changing operations (uc, lc, ucfirst, lcfirst) on strings that don't have the internal UTF-8 flag set, but that contain single-byte characters between 128 and 255.

  • threads

    Upgraded from version 1.74 to 1.75.

  • less

    Upgraded from version 0.02 to 0.03.

    This version introduces the stash_name method to allow subclasses of less to pick where in %^H to store their stash.

  • version

    Upgraded from version 0.77 to 0.81.

    This version adds support for "Version number formats" as described earlier in this document and in its own documentation.

  • warnings

    Upgraded from 1.07 to 1.09.

    Added new warnings::fatal_enabled() function. This version adds the illegalproto warning category. See also "New or Changed Diagnostics" for this change.

Updated Modules


Removed Modules and Pragmata

  • Devel::DProf::V

    Removed from the Perl core. Prior version was 'undef'.


New Documentation

  • perlhaiku

    This contains instructions on how to build perl for the Haiku platform.

  • perlmroapi

    This describes the new interface for pluggable Method Resolution Orders.

  • perlperf

    This document, by Richard Foley, provides an introduction to the use of performance and optimization techniques which can be used with particular reference to perl programs.

  • perlrepository

    This describes how to access the perl source using the git version control system.

  • perlpolicy extends the "Social contract about contributed modules" into the beginnings of a document on Perl porting policies.

Changes to Existing Documentation

The various large Changes* files (which listed every change made to perl over the last 18 years) have been removed, and replaced by a small file, also called Changes, which just explains how that same information may be extracted from the git version control system.

The file Porting/patching.pod has been deleted, as it mainly described interacting with the old Perforce-based repository, which is now obsolete. Information still relevant has been moved to perlrepository.

perlapi, perlintern, perlmodlib and perltoc are now all generated at build time, rather than being shipped as part of the release.

  • Documented -X overloading.

  • Documented that when() treats specially most of the filetest operators

  • Documented when as a syntax modifier

  • Eliminated "Old Perl threads tutorial", which described 5005 threads.

    pod/perlthrtut.pod is the same material reworked for ithreads.

  • Correct previous documentation: v-strings are not deprecated

    With version objects, we need them to use MODULE VERSION syntax. This patch removes the deprecation notice.

  • Added security contact information to perlsec

    A significant fraction of the core documentation has been updated to clarify the behavior of Perl's Unicode handling.

    Much of the remaining core documentation has been reviewed and edited for clarity, consistent use of language, and to fix the spelling of Tom Christiansen's name.

    The Pod specification (perlpodspec) has been updated to bring the specification in line with modern usage already supported by most Pod systems. A parameter string may now follow the format name in a "begin/end" region. Links to URIs with a text description are now allowed. The usage of L<"section"> has been marked as deprecated.

    if.pm has been documented in "use" in perlfunc as a means to get conditional loading of modules despite the implicit BEGIN block around use.

  • The documentation for $1 in perlvar.pod has been clarified.

Performance Enhancements

  • A new internal cache means that isa() will often be faster.

  • The implementation of C3 Method Resolution Order has been optimised - linearisation for classes with single inheritance is 40% faster. Performance for multiple inheritance is unchanged.

  • Under use locale, the locale-relevant information is now cached on read-only values, such as the list returned by keys %hash. This makes operations such as sort keys %hash in the scope of use locale much faster.

  • Empty DESTROY methods are no longer called.

  • Faster Perl_sv_utf8_upgrade()

  • Speed up keys on empty hash

  • if (%foo) has been optimized to be faster than if (keys %foo)

  • Reversing an array to itself (as in @a = reverse @a) in void context now happens in-place and is several orders of magnitude faster than it used to be. It will also preserve non-existent elements whenever possible, i.e. for non magical arrays or tied arrays with EXISTS and DELETE methods.

Installation and Configuration Improvements

ext/ reorganisation

The layout of directories in ext has been revised. Specifically, all extensions are now flat, and at the top level, with / in pathnames replaced by -, so that ext/Data/Dumper/ is now ext/Data-Dumper/, etc. The names of the extensions as specified to Configure, and as reported by %Config::Config under the keys dynamic_ext, known_extensions, nonxs_ext and static_ext have not changed, and still use /. Hence this change will not have any affect once perl is installed. Safe has been split out from being part of Opcode, and mro is now an extension in its own right.

Nearly all dual-life modules have been moved from lib to ext, and will now appear as known nonxs_ext. This will made no difference to the structure of an installed perl, nor will the modules installed differ, unless you run Configure with options to specify an exact list of extensions to build. In this case, you will rapidly become aware that you need to add to your list, because various modules needed to complete the build, such as ExtUtils::ParseXS, have now become extensions, and without them the build will fail well before it attempts to run the regression tests.

Configuration improvements

If vendorlib and vendorarch are the same, then they are only added to @INC once.

$Config{usedevel} and the C-level PERL_USE_DEVEL are now defined if perl is built with -Dusedevel.

Configure will enable use of -fstack-protector, to provide protection against stack-smashing attacks, if the compiler supports it.

Configure will now determine the correct prototypes for re-entrant functions and for gconvert if you are using a C++ compiler rather than a C compiler.

On Unix, if you build from a tree containing a git repository, the configuration process will note the commit hash you have checked out, for display in the output of perl -v and perl -V. Unpushed local commits are automatically added to the list of local patches displayed by perl -V.

USE_ATTRIBUTES_FOR_PERLIO is now reported in the compile-time options listed by the -V switch.

Compilation improvements

As part of the flattening of ext, all extensions on all platforms are built by make_ext.pl. This replaces the Unix-specific ext/util/make_ext, VMS-specific make_ext.com and Win32-specific win32/buildext.pl.

Changed Internals

  • Perl_pmflag has been removed from the public API. Calling it now generates a deprecation warning, and it will be removed in a future release. Although listed as part of the API, it was never documented, and only ever used in toke.c, and prior to 5.10, regcomp.c. In core, it has been replaced by a static function.

  • Perl_magic_setmglob now knows about globs, fixing RT #71254.

  • TODO: SVt_RV is gone. RVs are now stored in IVs

  • TODO: REGEXPs are first class

  • TODO: OOK is reworked, such that an OOKed scalar is PV not PVIV

  • The J.R.R. Tolkien quotes at the head of C source file have been checked and proper citations added, thanks to a patch from Tom Christiansen.

  • Perl_vcroak() now accepts a null first argument. In addition, a full audit was made of the "not NULL" compiler annotations, and those for several other internal functions were corrected.

  • New macros dSAVEDERRNO, dSAVE_ERRNO, SAVE_ERRNO, RESTORE_ERRNO have been added to formalise the temporary saving of the errno variable.

  • The function Perl_sv_insert_flags has been added to augment Perl_sv_insert.

  • The function Perl_newSV_type(type) has been added, equivalent to Perl_newSV() followed by Perl_sv_upgrade(type).

  • The function Perl_newSVpvn_flags() has been added, equivalent to Perl_newSVpvn() and then performing the action relevant to the flag.

    Two flag bits are currently supported.

    • SVf_UTF8

      This will call SvUTF8_on() for you. (Note that this does not convert an sequence of ISO 8859-1 characters to UTF-8). A wrapper, newSVpvn_utf8() is available for this.

    • SVs_TEMP

      Call Perl_sv_2mortal() on the new SV.

    There is also a wrapper that takes constant strings, newSVpvs_flags().

  • The function Perl_croak_xs_usage has been added as a wrapper to Perl_croak.

  • The functions PerlIO_find_layer and PerlIO_list_alloc are now exported.

  • PL_na has been exterminated from the core code, replaced by local STRLEN temporaries, or *_nolen() calls. Either approach is faster than PL_na, which is a pointer dereference into the interpreter structure under ithreads, and a global variable otherwise.

  • Perl_mg_free() used to leave freed memory accessible via SvMAGIC() on the scalar. It now updates the linked list to remove each piece of magic as it is freed.

  • Under ithreads, the regex in PL_reg_curpm is now reference counted. This eliminates a lot of hackish workarounds to cope with it not being reference counted.

  • Perl_mg_magical() would sometimes incorrectly turn on SvRMAGICAL(). This has been fixed.

  • The public IV and NV flags are now not set if the string value has trailing "garbage". This behaviour is consistent with not setting the public IV or NV flags if the value is out of range for the type.

  • SV allocation tracing has been added to the diagnostics enabled by -Dm. The tracing can alternatively output via the PERL_MEM_LOG mechanism, if that was enabled when the perl binary was compiled.

  • Smartmatch resolution tracing has been added as a new diagnostic. Use -DM to enable it.

  • A new debugging flag -DB now dumps subroutine definitions, leaving -Dx for its original purpose of dumping syntax trees.

  • Uses of Nullav, Nullcv, Nullhv, Nullop, Nullsv etc have been replaced by NULL in the core code, and non-dual-life modules, as NULL is clearer to those unfamiliar with the core code.

  • A macro MUTABLE_PTR(p) has been added, which on (non-pedantic) gcc will not cast away const, returning a void *. Macros MUTABLE_SV(av), MUTABLE_SV(cv) etc build on this, casting to AV * etc without casting away const. This allows proper compile-time auditing of const correctness in the core, and helped picked up some errors (now fixed).

  • Macros mPUSHs() and mXPUSHs() have been added, for pushing SVs on the stack and mortalizing them.

  • Use of the private structure mro_meta has changed slightly. Nothing outside the core should be accessing this directly anyway.

  • A new tool, Porting/expand-macro.pl has been added, that allows you to view how a C preprocessor macro would be expanded when compiled. This is handy when trying to decode the macro hell that is the perl guts.


New Tests

Many modules updated from CPAN incorporate new tests. Several tests that have the potential to hang forever if they fail now incorporate a "watchdog" functionality that will kill them after a timeout, which helps ensure that make test and make test_harness run to completion automatically. (Jerry Hedden).

Some core-specific tests have been added:

  • Significant cleanups to core tests to ensure that language and interpreter features are not used before they're tested.

  • make test_porting now runs a number of important pre-commit checks which might be of use to anyone working on the Perl core.

  • t/porting/podcheck.t automatically checks the well-formedness of POD found in all .pl, .pm and .pod files in the MANIFEST, other than in dual-lifed modules which are primarily maintained outside the Perl core.

  • t/porting/manifest.t now tests that all files listed in MANIFEST are present.

  • t/op/while_readdir.t

    Test that a bare readdir in while loop sets $_.

  • t/comp/retainedlines.t

    Check that the debugger can retain source lines from eval.

  • t/io/perlio_fail.t

    Check that bad layers fail.

  • t/io/perlio_leaks.t

    Check that PerlIO layers are not leaking.

  • t/io/perlio_open.t

    Check that certain special forms of open work.

  • t/io/perlio.t

    General PerlIO tests.

  • t/io/pvbm.t

    Check that there is no unexpected interaction between the internal types PVBM and PVGV.

  • t/mro/package_aliases.t

    Check that mro works properly in the presence of aliased packages.

  • t/op/dbm.t

    Tests for dbmopen and dbmclose.

  • t/op/index_thr.t

    Tests for the interaction of index and threads.

  • t/op/pat_thr.t

    Tests for the interaction of esoteric patterns and threads.

  • t/op/qr_gc.t

    Test that qr doesn't leak.

  • t/op/reg_email_thr.t

    Tests for the interaction of regex recursion and threads.

  • t/op/regexp_qr_embed_thr.t

    Tests for the interaction of patterns with embedded qr// and threads.

  • t/op/regexp_unicode_prop.t

    Tests for Unicode properties in regular expressions.

  • t/op/regexp_unicode_prop_thr.t

    Tests for the interaction of Unicode properties and threads.

  • t/op/reg_nc_tie.t

    Test the tied methods of Tie::Hash::NamedCapture.

  • t/op/reg_posixcc.t

    Check that POSIX character classes behave consistently.

  • t/op/re.t

    Check that exportable re functions in universal.c work.

  • t/op/setpgrpstack.t

    Check that setpgrp works.

  • t/op/substr_thr.t

    Tests for the interaction of substr and threads.

  • t/op/upgrade.t

    Check that upgrading and assigning scalars works.

  • t/uni/lex_utf8.t

    Check that Unicode in the lexer works.

  • t/uni/tie.t

    Check that Unicode and tie work.

  • t/comp/final_line_num.t

    See if line numbers are correct at EOF

  • t/comp/form_scope.t

    See if format scoping works

  • t/comp/line_debug.t

    See if @{"_<$file"} works

  • t/op/filetest_t.t

    See if -t file test works

  • t/op/qr.t

    See if qr works

  • t/op/utf8cache.t

    Tests malfunctions of utf8 cache

  • t/re/uniprops.t

    Test unicode \p{} regex constructs

Testing improvements

  • It's now possible to override PERL5OPT and friends in t/TEST

New or Changed Diagnostics

Several new diagnostics, see perldiag for details.

  • Bad plugin affecting keyword '%s'

  • gmtime(%.0f) too large

  • Lexing code attempted to stuff non-Latin-1 character into Latin-1 input

  • Lexing code internal error (%s)

  • localtime(%.0f) too large

  • Overloaded dereference did not return a reference

  • Overloaded qr did not return a REGEXP

  • Perl_pmflag() is deprecated, and will be removed from the XS API

  • New warning category illegalproto

    The two warnings :

      Illegal character in prototype for %s : %s
      Prototype after '%c' for %s : %s

    have been moved from the syntax top-level warnings category into a new first-level category, illegalproto. These two warnings are currently the only ones emitted during parsing of an invalid/illegal prototype, so one can now do

      no warnings 'illegalproto';

    to suppress only those, but not other syntax-related warnings. Warnings where prototypes are changed, ignored, or not met are still in the prototype category as before. (Matt S. Trout)

  • lvalue attribute ignored after the subroutine has been defined

    This new warning is issued when one attempts to mark a subroutine as lvalue after it has been defined.

  • warn if ++ or -- are unable to change the value because it's beyond the limit of representation

    This uses a new warnings category: "imprecision".

  • lc, uc, lcfirst, and ucfirst warn when passed undef.

  • Show constant in "Useless use of a constant in void context"

  • Make the new warning report undef constants as undef

  • Add a new warning, "Prototype after '%s'"

  • Tweak the "Illegal character in prototype" warning so it's more precise when reporting illegal characters after _

  • Correct the unintended interpolation of $\ in regex

  • Make overflow warnings in gmtime and localtime only occur when warnings are enabled

  • Improve mro merging error messages.

    They are now very similar to those produced by Algorithm::C3.

  • Amelioration of the error message "Unrecognized character %s in column %d"

    Changes the error message to "Unrecognized character %s; marked by <-- HERE after %s<-- HERE near column %d". This should make it a little simpler to spot and correct the suspicious character.

  • Explicitely point to $. when it causes an uninitialized warning for ranges in scalar context

  • split now warns when called in void context

  • printf-style functions called with too few arguments will now issue the warning "Missing argument in %s" [perl #71000]

  • panic: sv_chop %s

    This new fatal error occurs when the C routine Perl_sv_chop() was passed a position that is not within the scalar's string buffer. This could be caused by buggy XS code, and at this point recovery is not possible.

  • Deep recursion on subroutine "%s"

    It is now possible to change the depth threshold for this warning from the default of 100, by recompiling the perl binary, setting the C pre-processor macro PERL_SUB_DEPTH_WARN to the desired value.

  • Perl now properly returns a syntax error instead of segfaulting if each, keys, or values is used without an argument.

  • tell() now fails properly if called without an argument and when no previous file was read.

    tell() now returns -1, and sets errno to EBADF, thus restoring the 5.8.x behaviour.

  • overload no longer implicitly unsets fallback on repeated 'use overload' lines.

  • POSIX::strftime() can now handle Unicode characters in the format string.

  • The Windows select() implementation now supports all empty fd_sets more correctly.

  • The "syntax" category was removed from 5 warnings that should only be in "deprecated".

  • Three fatal pack/unpack error messages have been normalized to "panic: %s"

  • "Unicode character is illegal" has been rephrased to be more accurate

    It now reads Unicode non-character is illegal in interchange and the perldiag documentation has been expanded a bit.

  • Perl now defaults to issuing a warning if a deprecated language feature is used.

    To disable this feature in a given lexical scope, you should use no warnings 'deprecated'; For information about which language features are deprecated and explanations of various deprecation warnings, please see perldiag.pod

The following diagnostics have been removed:

  • Runaway format

  • Can't locate package %s for the parents of %s

    This warning has been removed. In general, it only got produced in conjunction with other warnings, and removing it allowed an ISA lookup optimisation to be added.

  • v-string in use/require is non-portable

Utility Changes

  • h2ph

    Now looks in include-fixed too, which is a recent addition to gcc's search path.

  • h2xs

    No longer incorrectly treats enum values like macros (Daniel Burr).

    Now handles C++ style constants (//) properly in enums. (A patch from Rainer Weikusat was used; Daniel Burr also proposed a similar fix).

  • perl5db.pl

    LVALUE subroutines now work under the debugger.

    The debugger now correctly handles proxy constant subroutines, and subroutine stubs.

  • perlbug

    perlbug now uses %Module::CoreList::bug_tracker to print out upstream bug tracker URLs.

    Where the user names a module that their bug report is about, and we know the URL for its upstream bug tracker, provide a message to the user explaining that the core copies the CPAN version directly, and provide the URL for reporting the bug directly to upstream.

  • perlthanks

    Perl 5.11.0 added a new utility perlthanks, which is a variant of perlbug, but for sending non-bug-reports to the authors and maintainers of Perl. Getting nothing but bug reports can become a bit demoralising: we'll see if this changes things.

  • perlbug

    No longer reports "Message sent" when it hasn't actually sent the message

  • a2p

    Fixed bugs with the match() operator in list context, remove mention of $[.

Selected Bug Fixes

  • Ensure that pp_qr returns a new regexp SV each time. Resolves RT #69852.

    Instead of returning a(nother) reference to the (pre-compiled) regexp in the optree, use reg_temp_copy() to create a copy of it, and return a reference to that. This resolves issues about Regexp::DESTROY not being called in a timely fashion (the original bug tracked by RT #69852), as well as bugs related to blessing regexps, and of assigning to regexps, as described in correspondence added to the ticket.

    It transpires that we also need to undo the SvPVX() sharing when ithreads cloning a Regexp SV, because mother_re is set to NULL, instead of a cloned copy of the mother_re. This change might fix bugs with regexps and threads in certain other situations, but as yet neither tests nor bug reports have indicated any problems, so it might not actually be an edge case that it's possible to reach.

  • Several compilation errors and segfaults when perl was built with -Dmad were fixed.

  • Fixes for lexer API changes in 5.11.2 which broke NYTProf's savesrc option.

  • -t should only return TRUE for file handles connected to a TTY

    The Microsoft C version of isatty() returns TRUE for all character mode devices, including the /dev/null-style "nul" device and printers like "lpt1".

  • Fixed a regression caused by commit fafafbaf which caused a panic during parameter passing [perl #70171]

  • On systems which in-place edits without backup files, -i'*' now works as the documentation says it does [perl #70802]

  • Saving and restoring magic flags no longer loses readonly flag.

  • The malformed syntax grep EXPR LIST (note the missing comma) no longer causes abrupt and total failure.

  • Regular expressions compiled with qr{} literals properly set $' when matching again.

  • Using named subroutines with sort should no longer lead to bus errors [perl #71076]

  • Numerous bugfixes catch small issues caused by the recently-added Lexer API.

  • Smart match against @_ sometimes gave false negatives. [perl #71078]

  • $@ may now be assigned a read-only value (without error or busting the stack).

  • sort called recursively from within an active comparison subroutine no longer causes a bus error if run multiple times. [perl #71076]

  • Tie::Hash::NamedCapture::* will not abort if passed bad input (RT #71828)

  • @_ and $_ no longer leak under threads (RT #34342 and #41138, also #70602, #70974)

  • -I on shebang line now adds directories in front of @INC as documented, and as does -I when specified on the command-line.

  • kill is now fatal when called on non-numeric process identifiers. Previously, an undef process identifier would be interpreted as a request to kill process 0, which would terminate the current process group on POSIX systems. Since process identifiers are always integers, killing a non-numeric process is now fatal.

  • 5.10.0 inadvertently disabled an optimisation, which caused a measurable performance drop in list assignment, such as is often used to assign function parameters from @_. The optimisation has been re-instated, and the performance regression fixed. (This fix is also present in 5.10.1)

  • Fixed memory leak on while (1) { map 1, 1 } [RT #53038].

  • Some potential coredumps in PerlIO fixed [RT #57322,54828].

  • The debugger now works with lvalue subroutines.

  • The debugger's m command was broken on modules that defined constants [RT #61222].

  • crypt and string complement could return tainted values for untainted arguments [RT #59998].

  • The -i.suffix command-line switch now recreates the file using restricted permissions, before changing its mode to match the original file. This eliminates a potential race condition [RT #60904].

  • On some Unix systems, the value in $? would not have the top bit set ($? & 128) even if the child core dumped.

  • Under some circumstances, $^R could incorrectly become undefined [RT #57042].

  • In the XS API, various hash functions, when passed a pre-computed hash where the key is UTF-8, might result in an incorrect lookup.

  • XS code including XSUB.h before perl.h gave a compile-time error [RT #57176].

  • $object->isa('Foo') would report false if the package Foo didn't exist, even if the object's @ISA contained Foo.

  • Various bugs in the new-to 5.10.0 mro code, triggered by manipulating @ISA, have been found and fixed.

  • Bitwise operations on references could crash the interpreter, e.g. $x=\$y; $x |= "foo" [RT #54956].

  • Patterns including alternation might be sensitive to the internal UTF-8 representation, e.g.

        my $byte = chr(192);
        my $utf8 = chr(192); utf8::upgrade($utf8);
        $utf8 =~ /$byte|X}/i;       # failed in 5.10.0
  • Within UTF8-encoded Perl source files (i.e. where use utf8 is in effect), double-quoted literal strings could be corrupted where a \xNN, \0NNN or \N{} is followed by a literal character with ordinal value greater than 255 [RT #59908].

  • B::Deparse failed to correctly deparse various constructs: readpipe STRING [RT #62428], CORE::require(STRING) [RT #62488], sub foo(_) [RT #62484].

  • Using setpgrp with no arguments could corrupt the perl stack.

  • The block form of eval is now specifically trappable by Safe and ops. Previously it was erroneously treated like string eval.

  • In 5.10.0, the two characters [~ were sometimes parsed as the smart match operator (~~) [RT #63854].

  • In 5.10.0, the * quantifier in patterns was sometimes treated as {0,32767} [RT #60034, #60464]. For example, this match would fail:

        ("ab" x 32768) =~ /^(ab)*$/
  • shmget was limited to a 32 bit segment size on a 64 bit OS [RT #63924].

  • Using next or last to exit a given block no longer produces a spurious warning like the following:

        Exiting given via last at foo.pl line 123
  • On Windows, '.\foo' and '..\foo' were treated differently than './foo' and '../foo' by do and require [RT #63492].

  • Assigning a format to a glob could corrupt the format; e.g.:

         *bar=*foo{FORMAT}; # foo format now bad
  • Attempting to coerce a typeglob to a string or number could cause an assertion failure. The correct error message is now generated, Can't coerce GLOB to $type.

  • Under use filetest 'access', -x was using the wrong access mode. This has been fixed [RT #49003].

  • length on a tied scalar that returned a Unicode value would not be correct the first time. This has been fixed.

  • Using an array tie inside in array tie could SEGV. This has been fixed. [RT #51636]

  • A race condition inside PerlIOStdio_close() has been identified and fixed. This used to cause various threading issues, including SEGVs.

  • In unpack, the use of () groups in scalar context was internally placing a list on the interpreter's stack, which manifested in various ways, including SEGVs. This is now fixed [RT #50256].

  • Magic was called twice in substr, \&$x, tie $x, $m and chop. These have all been fixed.

  • A 5.10.0 optimisation to clear the temporary stack within the implicit loop of s///ge has been reverted, as it turned out to be the cause of obscure bugs in seemingly unrelated parts of the interpreter [commit ef0d4e17921ee3de].

  • The line numbers for warnings inside elsif are now correct.

  • The .. operator now works correctly with ranges whose ends are at or close to the values of the smallest and largest integers.

  • binmode STDIN, ':raw' could lead to segmentation faults on some platforms. This has been fixed [RT #54828].

  • An off-by-one error meant that index $str, ... was effectively being executed as index "$str\0", .... This has been fixed [RT #53746].

  • Various leaks associated with named captures in regexes have been fixed [RT #57024].

  • A weak reference to a hash would leak. This was affecting DBI [RT #56908].

  • Using (?|) in a regex could cause a segfault [RT #59734].

  • Use of a UTF-8 tr// within a closure could cause a segfault [RT #61520].

  • Calling Perl_sv_chop() or otherwise upgrading an SV could result in an unaligned 64-bit access on the SPARC architecture [RT #60574].

  • In the 5.10.0 release, inc_version_list would incorrectly list 5.10.* after 5.8.*; this affected the @INC search order [RT #67628].

  • In 5.10.0, pack "a*", $tainted_value returned a non-tainted value [RT #52552].

  • In 5.10.0, printf and sprintf could produce the fatal error panic: utf8_mg_pos_cache_update when printing UTF-8 strings [RT #62666].

  • In the 5.10.0 release, a dynamically created AUTOLOAD method might be missed (method cache issue) [RT #60220,60232].

  • In the 5.10.0 release, a combination of use feature and //ee could cause a memory leak [RT #63110].

  • -C on the shebang (#!) line is once more permitted if it is also specified on the command line. -C on the shebang line used to be a silent no-op if it was not also on the command line, so perl 5.10.0 disallowed it, which broke some scripts. Now perl checks whether it is also on the command line and only dies if it is not [RT #67880].

  • In 5.10.0, certain types of re-entrant regular expression could crash, or cause the following assertion failure [RT #60508]:

        Assertion rx->sublen >= (s - rx->subbeg) + i failed
  • Previously missing files from Unicode 5.1 Character Database are now included.

  • TMPDIR is now honored when opening an anonymous temporary file

Platform Specific Changes

New Platforms


Patches from the Haiku maintainers have been merged in. Perl should now build on Haiku.


Perl should now build on MirOS BSD.

Discontinued Platforms


Support for Apollo DomainOS was removed in Perl 5.11.0


Support for Tenon Intersystems MachTen Unix layer for MacOS Classic was removed in Perl 5.11.0


Support for Atari MiNT was removed in Perl 5.11.0.

Updated Platforms

Darwin (Mac OS X)
  • Skip testing the be_BY.CP1131 locale on Darwin 10 (Mac OS X 10.6), as it's still buggy.

  • Correct infelicities in the regexp used to identify buggy locales on Darwin 8 and 9 (Mac OS X 10.4 and 10.5, respectively).

DragonFly BSD
  • Fix thread library selection [perl #69686]

  • Initial support for mingw64 is now available

  • Various bits of Perl's build infrastructure are no longer converted to win32 line endings at release time. If this hurts you, please report the problem with the perlbug program included with perl.

  • Always add a manifest resource to perl.exe to specify the trustInfo settings for Windows Vista and later. Without this setting Windows will treat perl.exe as a legacy application and apply various heuristics like redirecting access to protected file system areas (like the "Program Files" folder) to the users "VirtualStore" instead of generating a proper "permission denied" error.

    For VC8 and VC9 this manifest setting is automatically generated by the compiler/linker (together with the binding information for their respective runtime libraries); for all other compilers we need to embed the manifest resource explicitly in the external resource file.

    This change also requests the Microsoft Common-Controls version 6.0 (themed controls introduced in Windows XP) via the dependency list in the assembly manifest. For VC8 and VC9 this is specified using the /manifestdependency linker commandline option instead.

  • Improved message window handling means that alarm and kill messages will no longer be dropped under race conditions.

  • Enable IPv6 support on cygwin 1.7 and newer

  • Make -UDEBUGGING the default on VMS for 5.12.0.

    Like it has been everywhere else for ages and ages. Also make command-line selection of -UDEBUGGING and -DDEBUGGING work in configure.com; before the only way to turn it off was by saying no in answer to the interactive question.

  • The default pipe buffer size on VMS has been updated to 8192 on 64-bit systems.

  • Reads from the in-memory temporary files of PerlIO::scalar used to fail if $/ was set to a numeric reference (to indicate record-style reads). This is now fixed.

  • VMS now supports getgrgid.

  • Many improvements and cleanups have been made to the VMS file name handling and conversion code.

  • Enabling the PERL_VMS_POSIX_EXIT logical name now encodes a POSIX exit status in a VMS condition value for better interaction with GNV's bash shell and other utilities that depend on POSIX exit values. See "$?" in perlvms for details.

  • File::Copy now detects Unix compatibility mode on VMS.


Removed libbsd for AIX 5L and 6.1. Only flock() was used from libbsd.

Removed libgdbm for AIX 5L and 6.1. The libgdbm is delivered as an optional package with the AIX Toolbox. Unfortunately the 64 bit version is broken.

Hints changes mean that AIX 4.2 should work again.


On Cygwin we now strip the last number from the DLL. This has been the behaviour in the cygwin.com build for years. The hints files have been updated.


The hints files now identify the correct threading libraries on FreeBSD 7 and later.


We now work around a bizarre preprocessor bug in the Irix 6.5 compiler: cc -E - unfortunately goes into K&R mode, but cc -E file.c doesn't.


Hints now supports versions 5.*.

Stratus VOS

Various changes from Stratus have been merged in.


There is now support for Symbian S60 3.2 SDK and S60 5.0 SDK.

Known Problems

This is a list of some significant unfixed bugs, which are regressions from either 5.10.0 or 5.8.x.

  • List::Util::first misbehaves in the presence of a lexical $_ (typically introduced by my $_ or implicitly by given). The variable which gets set for each iteration is the package variable $_, not the lexical $_ [RT #67694].

    A similar issue may occur in other modules that provide functions which take a block as their first argument, like

        foo { ... $_ ...} list
  • Some regexes may run much more slowly when run in a child thread compared with the thread the pattern was compiled into [RT #55600].

  • Untriaged test crashes on Windows 2000

    Several porters have reported mysterious crashes when Perl's entire test suite is run after a build on certain Windows 2000 systems. When run by hand, the individual tests reportedly work fine.

  • Known test failures on VMS

    Perl 5.11.1 fails a small set of core and CPAN tests as of this release. With luck, that'll be sorted out for 5.11.2

  • Known test failures on VMS

    Perl 5.11.2 fails a small set of core and CPAN tests as of this release. With luck, that'll be sorted out for 5.11.3.


Perl 5.12.0 represents approximately two years of development since Perl 5.10.0 and contains __ lines of changes across ___ files from __ authors and committers:


Many of the changes included in this version originated in the CPAN modules included in Perl's core. We're grateful to the entire CPAN community for helping Perl to flourish.

Reporting Bugs

If you find what you think is a bug, you might check the articles recently posted to the comp.lang.perl.misc newsgroup and the perl bug database at http://rt.perl.org/perlbug/. There may also be information at http://www.perl.org/, the Perl Home Page.

If you believe you have an unreported bug, please run the perlbug program included with your release. Be sure to trim your bug down to a tiny but sufficient test case. Your bug report, along with the output of perl -V, will be sent off to perlbug@perl.org to be analyzed by the Perl porting team.

If the bug you are reporting has security implications, which make it inappropriate to send to a publicly archived mailing list, then please send it to perl5-security-report@perl.org. This points to a closed subscription unarchived mailing list, which includes all the core committers, who be able to help assess the impact of issues, figure out a resolution, and help co-ordinate the release of patches to mitigate or fix the problem across all platforms on which Perl is supported. Please only use this address for security issues in the Perl core, not for modules independently distributed on CPAN.


The Changes file for an explanation of how to view exhaustive details on what changed.

The INSTALL file for how to build Perl.

The README file for general stuff.

The Artistic and Copying files for copyright information.