Security Issues

Note that suidperl is neither built nor installed by default in any recent version of perl. Use of suidperl is highly discouraged. If you think you need it, try alternatives such as sudo first. See http://www.courtesan.com/sudo/ .

Core bug fixes

"UNIVERSAL::isa()"

A bug in the caching mechanism used by "UNIVERSAL::isa()" that affected base.pm has been fixed. The bug has existed since the 5.005 releases, but wasn't tickled by base.pm in those releases.

Memory leaks

Various cases of memory leaks and attempts to access uninitialized memory have been cured. See "Known Problems" below for further issues.

Numeric conversions

Numeric conversions did not recognize changes in the string value properly in certain circumstances.

In other situations, large unsigned numbers (those above 2**31) could sometimes lose their unsignedness, causing bogus results in arithmetic operations.

Integer modulus on large unsigned integers sometimes returned incorrect values.

Perl 5.6.0 generated "not a number" warnings on certain conversions where previous versions didn't.

These problems have all been rectified.

Infinity is now recognized as a number.

qw(a\\b)

In Perl 5.6.0, qw(a\\b) produced a string with two backslashes instead of one, in a departure from the behavior in previous versions. The older behavior has been reinstated.

caller()

caller() could cause core dumps in certain situations. Carp was sometimes affected by this problem.

Bugs in regular expressions

Pattern matches on overloaded values are now handled correctly.

Perl 5.6.0 parsed m/\x{ab}/ incorrectly, leading to spurious warnings. This has been corrected.

The RE engine found in Perl 5.6.0 accidentally pessimised certain kinds of simple pattern matches. These are now handled better.

Regular expression debug output (whether through "use re 'debug'" or via "-Dr") now looks better.

Multi-line matches like ""a\nxb\n" =~ /(?!\A)x/m" were flawed. The bug has been fixed.

Use of $& could trigger a core dump under some situations. This is now avoided.

Match variables $1 et al., weren't being unset when a pattern match was backtracking, and the anomaly showed up inside "/...(?{ ... }).../" etc. These variables are now tracked correctly.

pos() did not return the correct value within s///ge in earlier versions. This is now handled correctly.

"slurp" mode

readline() on files opened in "slurp" mode could return an extra "" at the end in certain situations. This has been corrected.

Autovivification of symbolic references to special variables

Autovivification of symbolic references of special variables described in perlvar (as in "${$num}") was accidentally disabled. This works again now.

Lexical warnings

Lexical warnings now propagate correctly into "eval "..."".

"use warnings qw(FATAL all)" did not work as intended. This has been corrected.

Lexical warnings could leak into other scopes in some situations. This is now fixed.

warnings::enabled() now reports the state of $^W correctly if the caller isn't using lexical warnings.

Spurious warnings and errors

Perl 5.6.0 could emit spurious warnings about redefinition of dl_error() when statically building extensions into perl. This has been corrected.

"our" variables could result in bogus "Variable will not stay shared" warnings. This is now fixed.

"our" variables of the same name declared in two sibling blocks resulted in bogus warnings about "redeclaration" of the variables. The problem has been corrected.

glob()

Compatibility of the builtin glob() with old csh-based glob has been improved with the addition of GLOB_ALPHASORT option. See "File::Glob".

File::Glob::glob() has been renamed to File::Glob::bsd_glob() because the name clashes with the builtin glob(). The older name is still available for compatibility, but is deprecated.

Spurious syntax errors generated in certain situations, when glob() caused File::Glob to be loaded for the first time, have been fixed.

Tainting

Some cases of inconsistent taint propagation (such as within hash values) have been fixed.

The tainting behavior of sprintf() has been rationalized. It does not taint the result of floating point formats anymore, making the behavior consistent with that of string interpolation.

sort()

Arguments to sort() weren't being provided the right wantarray() context. The comparison block is now run in scalar context, and the arguments to be sorted are always provided list context.

sort() is also fully reentrant, in the sense that the sort function can itself call sort(). This did not work reliably in previous releases.

#line directives

#line directives now work correctly when they appear at the very beginning of "eval "..."".

Subroutine prototypes

The (\&) prototype now works properly.

map()

map() could get pathologically slow when the result list it generates is larger than the source list. The performance has been improved for common scenarios.

Debugger

Debugger exit code now reflects the script exit code.

Condition "0" in breakpoints is now treated correctly.

The "d" command now checks the line number.

$. is no longer corrupted by the debugger.

All debugger output now correctly goes to the socket if RemotePort is set.

PERL5OPT

PERL5OPT can be set to more than one switch group. Previously, it used to be limited to one group of options only.

chop()

chop(@list) in list context returned the characters chopped in reverse order. This has been reversed to be in the right order.

Unicode support

Unicode support has seen a large number of incremental improvements, but continues to be highly experimental. It is not expected to be fully supported in the 5.6.x maintenance releases.

substr(), join(), repeat(), reverse(), quotemeta() and string concatenation were all handling Unicode strings incorrectly in Perl 5.6.0. This has been corrected.

Support for "tr///CU" and "tr///UC" etc., have been removed since we realized the interface is broken. For similar functionality, see "pack" in perlfunc.

The Unicode Character Database has been updated to version 3.0.1 with additions made available to the public as of August 30, 2000.

The Unicode character classes \p{Blank} and \p{SpacePerl} have been added. "Blank" is like C isblank(), that is, it contains only "horizontal whitespace" (the space character is, the newline isn't), and the "SpacePerl" is the Unicode equivalent of "\s" (\p{Space} isn't, since that includes the vertical tabulator character, whereas "\s" doesn't.)

If you are experimenting with Unicode support in perl, the development versions of Perl may have more to offer. In particular, I/O layers are now available in the development track, but not in the maintenance track, primarily to do backward compatibility issues. Unicode support is also evolving rapidly on a daily basis in the development track--the maintenance track only reflects the most conservative of these changes.

64-bit support

Support for 64-bit platforms has been improved, but continues to be experimental. The level of support varies greatly among platforms.

Compiler

The B Compiler and its various backends have had many incremental improvements, but they continue to remain highly experimental. Use in production environments is discouraged.

The perlcc tool has been rewritten so that the user interface is much more like that of a C compiler.

The perlbc tools has been removed. Use "perlcc -B" instead.

Lvalue subroutines

There have been various bugfixes to support lvalue subroutines better. However, the feature still remains experimental.

IO::Socket

IO::Socket::INET failed to open the specified port if the service name was not known. It now correctly uses the supplied port number as is.

File::Find

File::Find now chdir()s correctly when chasing symbolic links.

xsubpp

xsubpp now tolerates embedded POD sections.

"no Module;"

"no Module;" does not produce an error even if Module does not have an unimport() method. This parallels the behavior of "use" vis-a-vis "import".

Tests

A large number of tests have been added.

Core features

The "-DT" command line switch outputs copious tokenizing information. See perlrun.

Arrays are now always interpolated in double-quotish strings. Previously, "foo@bar.com" used to be a fatal error at compile time, if an array @bar was not used or declared. This transitional behavior was intended to help migrate perl4 code, and is deemed to be no longer useful. See "Arrays now always interpolate into double-quoted strings".

keys(), each(), pop(), push(), shift(), splice() and unshift() can all be overridden now.

"my __PACKAGE__ $obj" now does the expected thing.

Configuration issues

"Configure" has been enhanced in various ways:

• Minimizes use of temporary files.
• By default, does not link perl with libraries not used by it, such as the various dbm libraries. SunOS 4.x hints preserve behavior on that platform.
• Support for pdp11-style memory models has been removed due to obsolescence.
• Building outside the source tree is supported on systems that have symbolic links. This is done by running

      sh /path/to/source/Configure -Dmksymlinks ...
      make all test install
      

  in a directory other than the perl source directory. See INSTALL.

• "Configure -S" can be run non-interactively.

Documentation

The following pod documents are brand new:

    perlclib    Internal replacements for standard C library functions
    perldebtut  Perl debugging tutorial
    perlebcdic  Considerations for running Perl on EBCDIC platforms
    perlnewmod  Perl modules: preparing a new module for distribution
    perlrequick Perl regular expressions quick start
    perlretut   Perl regular expressions tutorial
    perlutil    utilities packaged with the Perl distribution

The INSTALL file has been expanded to cover various issues, such as 64-bit support.

A longer list of contributors has been added to the source distribution. See the file "AUTHORS".

Numerous other changes have been made to the included documentation and FAQs.

Bundled modules

B::Concise

Walks Perl syntax tree, printing concise info about ops. See B::Concise.

File::Temp

Returns name and handle of a temporary file safely. See File::Temp.

Pod::LaTeX

Converts Pod data to formatted LaTeX. See Pod::LaTeX.

Pod::Text::Overstrike

Converts POD data to formatted overstrike text. See Pod::Text::Overstrike.

The following modules have been upgraded.

CGI

CGI v2.752 is now included.

CPAN

CPAN v1.59_54 is now included.

Class::Struct

Various bugfixes have been added.

DB_File

DB_File v1.75 supports newer Berkeley DB versions, among other improvements.

Devel::Peek

Devel::Peek has been enhanced to support dumping of memory statistics, when perl is built with the included malloc().

File::Find

File::Find now supports pre and post-processing of the files in order to sort() them, etc.

Getopt::Long

Getopt::Long v2.25 is included.

IO::Poll

Various bug fixes have been included.

IPC::Open3

IPC::Open3 allows use of numeric file descriptors.

Math::BigFloat

The fmod() function supports modulus operations. Various bug fixes have also been included.

Math::Complex

Math::Complex handles inf, NaN etc., better.

Net::Ping

ping() could fail on odd number of data bytes, and when the echo service isn't running. This has been corrected.

Opcode

A memory leak has been fixed.

Pod::Parser

Version 1.13 of the Pod::Parser suite is included.

Pod::Text

Pod::Text and related modules have been upgraded to the versions in podlators suite v2.08.

SDBM_File

On dosish platforms, some keys went missing because of lack of support for files with "holes". A workaround for the problem has been added.

Sys::Syslog

Various bug fixes have been included.

Tie::RefHash

Now supports Tie::RefHash::Nestable to automagically tie hashref values.

Tie::SubstrHash

Various bug fixes have been included.

Platform-specific improvements

NCR MP-RAS

NonStop-UX

Perl now builds under Amdahl UTS.

Perl has also been verified to build under Amiga OS.

Support for EPOC has been much improved. See README.epoc.

Building perl with -Duseithreads or -Duse5005threads now works under HP-UX 10.20 (previously it only worked under 10.30 or later). You will need a thread library package installed. See README.hpux.

Long doubles should now work under Linux.

Mac OS Classic is now supported in the mainstream source package. See README.macos.

Support for MPE/iX has been updated. See README.mpeix.

Support for OS/2 has been improved. See "os2/Changes" and README.os2.

Dynamic loading on z/OS (formerly OS/390) has been improved. See README.os390.

Support for VMS has seen many incremental improvements, including better support for operators like backticks and system(), and better %ENV handling. See "README.vms" and perlvms.

Support for Stratus VOS has been improved. See "vos/Changes" and README.vos.

Support for Windows has been improved.

• fork() emulation has been improved in various ways, but still continues to be experimental. See perlfork for known bugs and caveats.
• %SIG has been enabled under USE_ITHREADS, but its use is completely unsupported under all configurations.
• Borland C++ v5.5 is now a supported compiler that can build Perl. However, the generated binaries continue to be incompatible with those generated by the other supported compilers (GCC and Visual C++).
• Non-blocking waits for child processes (or pseudo-processes) are supported via "waitpid($pid, &POSIX::WNOHANG)".
• A memory leak in accept() has been fixed.
• wait(), waitpid() and backticks now return the correct exit status under Windows 9x.
• Trailing new %ENV entries weren't propagated to child processes. This is now fixed.
• Current directory entries in %ENV are now correctly propagated to child processes.
• Duping socket handles with open(F, ">&MYSOCK") now works under Windows 9x.
• The makefiles now provide a single switch to bulk-enable all the features enabled in ActiveState ActivePerl (a popular binary distribution).
• Win32::GetCwd() correctly returns C:\ instead of C: when at the drive root. Other bugs in chdir() and Cwd::cwd() have also been fixed.
• fork() correctly returns undef and sets EAGAIN when it runs out of pseudo-process handles.
• ExtUtils::MakeMaker now uses $ENV{LIB} to search for libraries.
• UNC path handling is better when perl is built to support fork().
• A handle leak in socket handling has been fixed.
• send() works from within a pseudo-process.

Unless specifically qualified otherwise, the remainder of this document covers changes between the 5.005 and 5.6.0 releases.

Interpreter cloning, threads, and concurrency

On the Windows platform, this feature is used to emulate fork() at the interpreter level. See perlfork for details about that.

This feature is still in evolution. It is eventually meant to be used to selectively clone a subroutine and data reachable from that subroutine in a separate interpreter and run the cloned subroutine in a separate thread. Since there is no shared data between the interpreters, little or no locking will be needed (unless parts of the symbol table are explicitly shared). This is obviously intended to be an easy-to-use replacement for the existing threads support.

Support for cloning interpreters and interpreter concurrency can be enabled using the -Dusethreads Configure option (see win32/Makefile for how to enable it on Windows.) The resulting perl executable will be functionally identical to one that was built with -Dmultiplicity, but the perl_clone() API call will only be available in the former.

-Dusethreads enables the cpp macro USE_ITHREADS by default, which in turn enables Perl source code changes that provide a clear separation between the op tree and the data it operates with. The former is immutable, and can therefore be shared between an interpreter and all of its clones, while the latter is considered local to each interpreter, and is therefore copied for each clone.

Note that building Perl with the -Dusemultiplicity Configure option is adequate if you wish to run multiple independent interpreters concurrently in different threads. -Dusethreads only provides the additional functionality of the perl_clone() API call and other support for running cloned interpreters concurrently.

    NOTE: This is an experimental feature.  Implementation details are
    subject to change.

Lexically scoped warning categories

Unicode and UTF-8 support

This feature is expected to evolve quickly to support some form of I/O disciplines that can be used to specify the kind of input and output data (bytes or characters). Until that happens, additional modules from CPAN will be needed to complete the toolkit for dealing with Unicode.

    NOTE: This should be considered an experimental feature.  Implementation
    details are subject to change.

Support for interpolating named characters

"our" declarations

Support for strings represented as a vector of ordinals

Strings written in this form are also useful to represent version "numbers". It is easy to compare such version "numbers" (which are really just plain strings) using any of the usual string comparison operators "eq", "ne", "lt", "gt", etc., or perform bitwise string operations on them using "|", "&", etc.

In conjunction with the new $^V magic variable (which contains the perl version as a string), such literals can be used as a readable way to check if you're running a particular version of Perl:

    # this will parse in older versions of Perl also
    if ($^V and $^V gt v5.6.0) {
        # new features supported
    }

"require" and "use" also have some special magic to support such literals. They will be interpreted as a version rather than as a module name:

    require v5.6.0;             # croak if $^V lt v5.6.0
    use v5.6.0;                 # same, but croaks at compile-time

Alternatively, the "v" may be omitted if there is more than one dot:

    require 5.6.0;
    use 5.6.0;

Also, "sprintf" and "printf" support the Perl-specific format flag %v to print ordinals of characters in arbitrary strings:

    printf "v%vd", $^V;         # prints current version, such as "v5.5.650"
    printf "%*vX", ":", $addr;  # formats IPv6 address
    printf "%*vb", " ", $bits;  # displays bitstring

See "Scalar value constructors" in perldata for additional information.

Improved Perl version numbering system

Maintenance versions of v5.6.0 will be released as v5.6.1, v5.6.2 etc. The next development series following v5.6.0 will be numbered v5.7.x, beginning with v5.7.0, and the next major production release following v5.6.0 will be v5.8.0.

The English module now sets $PERL_VERSION to $^V (a string value) rather than $] (a numeric value). (This is a potential incompatibility. Send us a report via perlbug if you are affected by this.)

The v1.2.3 syntax is also now legal in Perl. See "Support for strings represented as a vector of ordinals" for more on that.

To cope with the new versioning system's use of at least three significant digits for each version component, the method used for incrementing the subversion number has also changed slightly. We assume that versions older than v5.6.0 have been incrementing the subversion component in multiples of 10. Versions after v5.6.0 will increment them by 1. Thus, using the new notation, 5.005_03 is the "same" as v5.5.30, and the first maintenance version following v5.6.0 will be v5.6.1 (which should be read as being equivalent to a floating point value of 5.006_001 in the older format, stored in $]).

New syntax for declaring subroutine attributes

    sub mymethod : locked method;
    ...
    sub mymethod : locked method {
        ...
    }

    sub othermethod :locked :method;
    ...
    sub othermethod :locked :method {
        ...
    }

(Note how only the first ":" is mandatory, and whitespace surrounding the ":" is optional.)

AutoSplit.pm and SelfLoader.pm have been updated to keep the attributes with the stubs they provide. See attributes.

File and directory handles can be autovivified

    sub myopen {
        open my $fh, "@_"
             or die "Can't open '@_': $!";
        return $fh;
    }

    {
        my $f = myopen("</etc/motd");
        print <$f>;
        # $f implicitly closed here
    }

open() with more than two arguments

64-bit support

        (1) natively as longs or ints
        (2) via special compiler flags
        (3) using long long or int64_t

is able to use "quads" (64-bit integers) as follows:

• constants (decimal, hexadecimal, octal, binary) in the code
• arguments to oct() and hex()
• arguments to print(), printf() and sprintf() (flag prefixes ll, L, q)
• printed as such
• pack() and unpack() "q" and "Q" formats
• in basic arithmetics: + - * / % (NOTE: operating close to the limits of the integer values may produce surprising results)
• in bit arithmetics: & | ^ ~ << >> (NOTE: these used to be forced to be 32 bits wide but now operate on the full native width.)
• vec()

Note that unless you have the case (a) you will have to configure and compile Perl using the -Duse64bitint Configure flag.

    NOTE: The Configure flags -Duselonglong and -Duse64bits have been
    deprecated.  Use -Duse64bitint instead.

There are actually two modes of 64-bitness: the first one is achieved using Configure -Duse64bitint and the second one using Configure -Duse64bitall. The difference is that the first one is minimal and the second one maximal. The first works in more places than the second.

The "use64bitint" does only as much as is required to get 64-bit integers into Perl (this may mean, for example, using "long longs") while your memory may still be limited to 2 gigabytes (because your pointers could still be 32-bit). Note that the name "64bitint" does not imply that your C compiler will be using 64-bit "int"s (it might, but it doesn't have to): the "use64bitint" means that you will be able to have 64 bits wide scalar values.

The "use64bitall" goes all the way by attempting to switch also integers (if it can), longs (and pointers) to being 64-bit. This may create an even more binary incompatible Perl than -Duse64bitint: the resulting executable may not run at all in a 32-bit box, or you may have to reboot/reconfigure/rebuild your operating system to be 64-bit aware.

Natively 64-bit systems like Alpha and Cray need neither -Duse64bitint nor -Duse64bitall.

Last but not least: note that due to Perl's habit of always using floating point numbers, the quads are still not true integers. When quads overflow their limits (0...18_446_744_073_709_551_615 unsigned, -9_223_372_036_854_775_808...9_223_372_036_854_775_807 signed), they are silently promoted to floating point numbers, after which they will start losing precision (in their lower digits).

    NOTE: 64-bit support is still experimental on most platforms.
    Existing support only covers the LP64 data model.  In particular, the
    LLP64 data model is not yet supported.  64-bit libraries and system
    APIs on many platforms have not stabilized--your mileage may vary.

Large file support

    NOTE: The default action is to enable large file support, if
    available on the platform.

If the large file support is on, and you have a Fcntl constant O_LARGEFILE, the O_LARGEFILE is automatically added to the flags of sysopen().

Beware that unless your filesystem also supports "sparse files" seeking to umpteen petabytes may be inadvisable.

Note that in addition to requiring a proper file system to do large files you may also need to adjust your per-process (or your per-system, or per-process-group, or per-user-group) maximum filesize limits before running Perl scripts that try to handle large files, especially if you intend to write such files.

Finally, in addition to your process/process group maximum filesize limits, you may have quota limits on your filesystems that stop you (your user id or your user group id) from using large files.

Adjusting your process/user/group/file system/operating system limits is outside the scope of Perl core language. For process limits, you may try increasing the limits using your shell's limits/limit/ulimit command before running Perl. The BSD::Resource extension (not included with the standard Perl distribution) may also be of use, it offers the getrlimit/setrlimit interface that can be used to adjust process resource usage limits, including the maximum filesize limit.

Long doubles

"more bits"

Enhanced support for sort() subroutines

For unprototyped sort subroutines, the historical behavior of passing the elements to be compared as the global variables $a and $b remains unchanged.

"sort $coderef @foo" allowed

File globbing implemented internally

    NOTE: This is currently an experimental feature.  Interfaces and
    implementation are subject to change.

Support for CHECK blocks

POSIX character class syntax [: :] supported

Better pseudo-random number generator

These changes should result in better random numbers from rand().

Improved "qw//" operator

Thus:

    $foo = ($bar) = qw(a b c); print "$foo|$bar\n";

now correctly prints "3|a", instead of "2|a".

Better worst-case behavior of hashes

pack() format 'Z' supported

pack() format modifier '!' supported

pack() and unpack() support counted strings

Comments in pack() templates

Weak references

Another familiar problem is with circular references. When an object references itself, its reference count would never go down to zero, and it would not get destroyed until the program is about to exit.

Weak references solve this by allowing you to "weaken" any reference, that is, make it not count towards the reference count. When the last non-weak reference to an object is deleted, the object is destroyed and all the weak references to the object are automatically undef-ed.

To use this feature, you need the Devel::WeakRef package from CPAN, which contains additional documentation.

    NOTE: This is an experimental feature.  Details are subject to change.

Binary numbers supported

    $answer = 0b101010;
    printf "The answer is: %b\n", oct("0b101010");

Lvalue subroutines

    NOTE: This is an experimental feature.  Details are subject to change.

Some arrows may be omitted in calls through references

Boolean assignment operators are legal lvalues

exists() is supported on subroutine names

exists() and delete() are supported on array elements

exists() can be used to check whether an array element has been initialized. This avoids autovivifying array elements that don't exist. If the array is tied, the EXISTS() method in the corresponding tied package will be invoked.

delete() may be used to remove an element from the array and return it. The array element at that position returns to its uninitialized state, so that testing for the same element with exists() will return false. If the element happens to be the one at the end, the size of the array also shrinks up to the highest element that tests true for exists(), or 0 if none such is found. If the array is tied, the DELETE() method in the corresponding tied package will be invoked.

See "exists" in perlfunc and "delete" in perlfunc for examples.

Pseudo-hashes work better

When applied to a pseudo-hash element, exists() now reports whether the specified value exists, not merely if the key is valid.

delete() now works on pseudo-hashes. When given a pseudo-hash element or slice it deletes the values corresponding to the keys (but not the keys themselves). See "Pseudo-hashes: Using an array as a hash" in perlref.

Pseudo-hash slices with constant keys are now optimized to array lookups at compile-time.

List assignments to pseudo-hash slices are now supported.

The "fields" pragma now provides ways to create pseudo-hashes, via fields::new() and fields::phash(). See fields.

    NOTE: The pseudo-hash data type continues to be experimental.
    Limiting oneself to the interface elements provided by the
    fields pragma will provide protection from any future changes.

Automatic flushing of output buffers

This is not supported on some platforms like Solaris where a suitably correct implementation of fflush(NULL) isn't available.

Better diagnostics on meaningless filehandle operations

Where possible, buffered data discarded from duped input filehandle

eof() has the same old magic as <>

binmode() can be used to set :crlf and :raw modes

"-T" filetest recognizes UTF-8 encoded files as "text"

system(), backticks and pipe open now reflect exec() failure

The child process now communicates with the parent about the error in launching the external command, which allows these constructs to return with their usual error value and set $!.

Improved diagnostics

Diagnostics emitted from code running in threads other than the main thread are now accompanied by the thread ID.

Embedded null characters in diagnostics now actually show up. They used to truncate the message in prior versions.

$foo::a and $foo::b are now exempt from "possible typo" warnings only if sort() is encountered in package "foo".

Unrecognized alphabetic escapes encountered when parsing quote constructs now generate a warning, since they may take on new semantics in later versions of Perl.

Many diagnostics now report the internal operation in which the warning was provoked, like so:

    Use of uninitialized value in concatenation (.) at (eval 1) line 1.
    Use of uninitialized value in print at (eval 1) line 1.

Diagnostics that occur within eval may also report the file and line number where the eval is located, in addition to the eval sequence number and the line number within the evaluated text itself. For example:

    Not enough arguments for scalar at (eval 4)[newlib/perl5db.pl:1411] line 2, at EOF

Diagnostics follow STDERR

More consistent close-on-exec behavior

syswrite() ease-of-use

Better syntax checks on parenthesized unary operators

    print defined(&foo,&bar,&baz);
    print uc("foo","bar","baz");
    undef($foo,&bar);

used to be accidentally allowed in earlier versions, and produced unpredictable behaviour. Some produced ancillary warnings when used in this way; others silently did the wrong thing.

The parenthesized forms of most unary operators that expect a single argument now ensure that they are not called with more than one argument, making the cases shown above syntax errors. The usual behaviour of:

    print defined &foo, &bar, &baz;
    print uc "foo", "bar", "baz";
    undef $foo, &bar;

remains unchanged. See perlop.

Bit operators support full native integer width

Improved security features

The "passwd" and "shell" fields returned by the getpwent(), getpwnam(), and getpwuid() are now tainted, because the user can affect their own encrypted password and login shell.

The variable modified by shmread(), and messages returned by msgrcv() (and its object-oriented interface IPC::SysV::Msg::rcv) are also tainted, because other untrusted processes can modify messages and shared memory segments for their own nefarious purposes.

More functional bareword prototype (*)

Arguments prototyped as "*" will now be visible within the subroutine as either a simple scalar or as a reference to a typeglob. See "Prototypes" in perlsub.

"require" and "do" may be overridden

$^X variables may now have names longer than one character

The old syntax has not changed. As before, `^X' may be either a literal control-X character or the two-character sequence `caret' plus `X'. When braces are omitted, the variable name stops after the control character. Thus "$^XYZ" continues to be synonymous with "$^X . "YZ"" as before.

As before, lexical variables may not have names beginning with control characters. As before, variables whose names begin with a control character are always forced to be in package `main'. All such variables are reserved for future extensions, except those that begin with "^_", which may be used by user programs and are guaranteed not to acquire special meaning in any future version of Perl.

New variable $^C reflects "-c" switch

New variable $^V contains Perl version as a string

See "Support for strings represented as a vector of ordinals" for an example.

Optional Y2K warnings

This behavior must be specifically enabled when running Configure. See INSTALL and README.Y2K.

Arrays now always interpolate into double-quoted strings

        Literal @example now requires backslash

In versions 5.004_01 through 5.6.0, the error was

        In string, @example now must be written as \@example

The idea here was to get people into the habit of writing "fred\@example.com" when they wanted a literal "@" sign, just as they have always written "Give me back my \$5" when they wanted a literal "$" sign.

Starting with 5.6.1, when Perl now sees an "@" sign in a double-quoted string, it always attempts to interpolate an array, regardless of whether or not the array has been used or declared already. The fatal error has been downgraded to an optional warning:

        Possible unintended interpolation of @example in string

This warns you that "fred@example.com" is going to turn into "fred.com" if you don't backslash the "@". See http://perl.plover.com/at-error.html for more details about the history here.

@- and @+ provide starting/ending offsets of regex submatches

Modules

attributes

While used internally by Perl as a pragma, this module also provides a way to fetch subroutine and variable attributes. See attributes.

B

The Perl Compiler suite has been extensively reworked for this release. More of the standard Perl test suite passes when run under the Compiler, but there is still a significant way to go to achieve production quality compiled executables.

    NOTE: The Compiler suite remains highly experimental.  The
    generated code may not be correct, even when it manages to execute
    without errors.
    

Benchmark

Overall, Benchmark results exhibit lower average error and better timing accuracy.

You can now run tests for n seconds instead of guessing the right number of tests to run: e.g., timethese(-5, ...) will run each code for at least 5 CPU seconds. Zero as the "number of repetitions" means "for at least 3 CPU seconds". The output format has also changed. For example:

   use Benchmark;$x=3;timethese(-5,{a=>sub{$x*$x},b=>sub{$x**2}})
    

will now output something like this:

   Benchmark: running a, b, each for at least 5 CPU seconds...
            a:  5 wallclock secs ( 5.77 usr +  0.00 sys =  5.77 CPU) @ 200551.91/s (n=1156516)
            b:  4 wallclock secs ( 5.00 usr +  0.02 sys =  5.02 CPU) @ 159605.18/s (n=800686)
    

New features: "each for at least N CPU seconds...", "wallclock secs", and the "@ operations/CPU second (n=operations)".

timethese() now returns a reference to a hash of Benchmark objects containing the test results, keyed on the names of the tests.

timethis() now returns the iterations field in the Benchmark result object instead of 0.

timethese(), timethis(), and the new cmpthese() (see below) can also take a format specifier of 'none' to suppress output.

A new function countit() is just like timeit() except that it takes a TIME instead of a COUNT.

A new function cmpthese() prints a chart comparing the results of each test returned from a timethese() call. For each possible pair of tests, the percentage speed difference (iters/sec or seconds/iter) is shown.

For other details, see Benchmark.

ByteLoader

The ByteLoader is a dedicated extension to generate and run Perl bytecode. See ByteLoader.

constant

References can now be used.

The new version also allows a leading underscore in constant names, but disallows a double leading underscore (as in "__LINE__"). Some other names are disallowed or warned against, including BEGIN, END, etc. Some names which were forced into main:: used to fail silently in some cases; now they're fatal (outside of main::) and an optional warning (inside of main::). The ability to detect whether a constant had been set with a given name has been added.

See constant.

charnames

This pragma implements the "\N" string escape. See charnames.

Data::Dumper

A "Maxdepth" setting can be specified to avoid venturing too deeply into deep data structures. See Data::Dumper.

The XSUB implementation of Dump() is now automatically called if the "Useqq" setting is not in use.

Dumping "qr//" objects works correctly.

DB

"DB" is an experimental module that exposes a clean abstraction to Perl's debugging API.

DB_File

DB_File can now be built with Berkeley DB versions 1, 2 or 3. See "ext/DB_File/Changes".

Devel::DProf

Devel::DProf, a Perl source code profiler has been added. See Devel::DProf and dprofpp.

Devel::Peek

The Devel::Peek module provides access to the internal representation of Perl variables and data. It is a data debugging tool for the XS programmer.

Dumpvalue

The Dumpvalue module provides screen dumps of Perl data.

DynaLoader

DynaLoader now supports a dl_unload_file() function on platforms that support unloading shared objects using dlclose().

Perl can also optionally arrange to unload all extension shared objects loaded by Perl. To enable this, build Perl with the Configure option "-Accflags=-DDL_UNLOAD_ALL_AT_EXIT". (This maybe useful if you are using Apache with mod_perl.)

English

$PERL_VERSION now stands for $^V (a string value) rather than for $] (a numeric value).

Env

Env now supports accessing environment variables like PATH as array variables.

Fcntl

More Fcntl constants added: F_SETLK64, F_SETLKW64, O_LARGEFILE for large file (more than 4GB) access (NOTE: the O_LARGEFILE is automatically added to sysopen() flags if large file support has been configured, as is the default), Free/Net/OpenBSD locking behaviour flags F_FLOCK, F_POSIX, Linux F_SHLCK, and O_ACCMODE: the combined mask of O_RDONLY, O_WRONLY, and O_RDWR. The seek()/sysseek() constants SEEK_SET, SEEK_CUR, and SEEK_END are available via the ":seek" tag. The chmod()/stat() S_IF* constants and S_IS* functions are available via the ":mode" tag.

File::Compare

A compare_text() function has been added, which allows custom comparison functions. See File::Compare.

File::Find

File::Find now works correctly when the wanted() function is either autoloaded or is a symbolic reference.

A bug that caused File::Find to lose track of the working directory when pruning top-level directories has been fixed.

File::Find now also supports several other options to control its behavior. It can follow symbolic links if the "follow" option is specified. Enabling the "no_chdir" option will make File::Find skip changing the current directory when walking directories. The "untaint" flag can be useful when running with taint checks enabled.

See File::Find.

File::Glob

This extension implements BSD-style file globbing. By default, it will also be used for the internal implementation of the glob() operator. See File::Glob.

File::Spec

New methods have been added to the File::Spec module: devnull() returns the name of the null device (/dev/null on Unix) and tmpdir() the name of the temp directory (normally /tmp on Unix). There are now also methods to convert between absolute and relative filenames: abs2rel() and rel2abs(). For compatibility with operating systems that specify volume names in file paths, the splitpath(), splitdir(), and catdir() methods have been added.

File::Spec::Functions

The new File::Spec::Functions modules provides a function interface to the File::Spec module. Allows shorthand

    $fullname = catfile($dir1, $dir2, $file);
    

instead of

    $fullname = File::Spec->catfile($dir1, $dir2, $file);
    

Getopt::Long

Getopt::Long licensing has changed to allow the Perl Artistic License as well as the GPL. It used to be GPL only, which got in the way of non-GPL applications that wanted to use Getopt::Long.

Getopt::Long encourages the use of Pod::Usage to produce help messages. For example:

    use Getopt::Long;
    use Pod::Usage;
    my $man = 0;
    my $help = 0;
    GetOptions('help|?' => \$help, man => \$man) or pod2usage(2);
    pod2usage(1) if $help;
    pod2usage(-exitstatus => 0, -verbose => 2) if $man;

    __END__

    =head1 NAME

    sample - Using Getopt::Long and Pod::Usage

    =head1 SYNOPSIS

    sample [options] [file ...]

     Options:
       -help            brief help message
       -man             full documentation

    =head1 OPTIONS

    =over 8

    =item B<-help>

    Print a brief help message and exits.

    =item B<-man>

    Prints the manual page and exits.

    =back

    =head1 DESCRIPTION

    B<This program> will read the given input file(s) and do something
    useful with the contents thereof.

    =cut
    

See Pod::Usage for details.

A bug that prevented the non-option call-back <> from being specified as the first argument has been fixed.

To specify the characters < and > as option starters, use ><. Note, however, that changing option starters is strongly deprecated.

IO

write() and syswrite() will now accept a single-argument form of the call, for consistency with Perl's syswrite().

You can now create a TCP-based IO::Socket::INET without forcing a connect attempt. This allows you to configure its options (like making it non-blocking) and then call connect() manually.

A bug that prevented the IO::Socket::protocol() accessor from ever returning the correct value has been corrected.

IO::Socket::connect now uses non-blocking IO instead of alarm() to do connect timeouts.

IO::Socket::accept now uses select() instead of alarm() for doing timeouts.

IO::Socket::INET->new now sets $! correctly on failure. $@ is still set for backwards compatibility.

JPL

Java Perl Lingo is now distributed with Perl. See jpl/README for more information.

lib

"use lib" now weeds out any trailing duplicate entries. "no lib" removes all named entries.

Math::BigInt

The bitwise operations "<<", ">>", "&", "|", and "~" are now supported on bigints.

Math::Complex

The accessor methods Re, Im, arg, abs, rho, and theta can now also act as mutators (accessor $z->Re(), mutator $z->Re(3)).

The class method "display_format" and the corresponding object method "display_format", in addition to accepting just one argument, now can also accept a parameter hash. Recognized keys of a parameter hash are "style", which corresponds to the old one parameter case, and two new parameters: "format", which is a printf()-style format string (defaults usually to "%.15g", you can revert to the default by setting the format string to "undef") used for both parts of a complex number, and "polar_pretty_print" (defaults to true), which controls whether an attempt is made to try to recognize small multiples and rationals of pi (2pi, pi/2) at the argument (angle) of a polar complex number.

The potentially disruptive change is that in list context both methods now return the parameter hash, instead of only the value of the "style" parameter.

Math::Trig

A little bit of radial trigonometry (cylindrical and spherical), radial coordinate conversions, and the great circle distance were added.

Pod::Parser, Pod::InputObjects

Pod::Parser is a base class for parsing and selecting sections of pod documentation from an input stream. This module takes care of identifying pod paragraphs and commands in the input and hands off the parsed paragraphs and commands to user-defined methods which are free to interpret or translate them as they see fit.

Pod::InputObjects defines some input objects needed by Pod::Parser, and for advanced users of Pod::Parser that need more about a command besides its name and text.

As of release 5.6.0 of Perl, Pod::Parser is now the officially sanctioned "base parser code" recommended for use by all pod2xxx translators. Pod::Text (pod2text) and Pod::Man (pod2man) have already been converted to use Pod::Parser and efforts to convert Pod::HTML (pod2html) are already underway. For any questions or comments about pod parsing and translating issues and utilities, please use the pod-people@perl.org mailing list.

For further information, please see Pod::Parser and Pod::InputObjects.

Pod::Checker, podchecker

This utility checks pod files for correct syntax, according to perlpod. Obvious errors are flagged as such, while warnings are printed for mistakes that can be handled gracefully. The checklist is not complete yet. See Pod::Checker.

Pod::ParseUtils, Pod::Find

These modules provide a set of gizmos that are useful mainly for pod translators. Pod::Find traverses directory structures and returns found pod files, along with their canonical names (like "File::Spec::Unix"). Pod::ParseUtils contains Pod::List (useful for storing pod list information), Pod::Hyperlink (for parsing the contents of "L<>" sequences) and Pod::Cache (for caching information about pod files, e.g., link nodes).

Pod::Select, podselect

Pod::Select is a subclass of Pod::Parser which provides a function named "podselect()" to filter out user-specified sections of raw pod documentation from an input stream. podselect is a script that provides access to Pod::Select from other scripts to be used as a filter. See Pod::Select.

Pod::Usage, pod2usage

Pod::Usage provides the function "pod2usage()" to print usage messages for a Perl script based on its embedded pod documentation. The pod2usage() function is generally useful to all script authors since it lets them write and maintain a single source (the pods) for documentation, thus removing the need to create and maintain redundant usage message text consisting of information already in the pods.

There is also a pod2usage script which can be used from other kinds of scripts to print usage messages from pods (even for non-Perl scripts with pods embedded in comments).

For details and examples, please see Pod::Usage.

Pod::Text and Pod::Man

Pod::Text has been rewritten to use Pod::Parser. While pod2text() is still available for backwards compatibility, the module now has a new preferred interface. See Pod::Text for the details. The new Pod::Text module is easily subclassed for tweaks to the output, and two such subclasses (Pod::Text::Termcap for man-page-style bold and underlining using termcap information, and Pod::Text::Color for markup with ANSI color sequences) are now standard.

pod2man has been turned into a module, Pod::Man, which also uses Pod::Parser. In the process, several outstanding bugs related to quotes in section headers, quoting of code escapes, and nested lists have been fixed. pod2man is now a wrapper script around this module.

SDBM_File

An EXISTS method has been added to this module (and sdbm_exists() has been added to the underlying sdbm library), so one can now call exists on an SDBM_File tied hash and get the correct result, rather than a runtime error.

A bug that may have caused data loss when more than one disk block happens to be read from the database in a single FETCH() has been fixed.

Sys::Syslog

Sys::Syslog now uses XSUBs to access facilities from syslog.h so it no longer requires syslog.ph to exist.

Sys::Hostname

Sys::Hostname now uses XSUBs to call the C library's gethostname() or uname() if they exist.

Term::ANSIColor

Term::ANSIColor is a very simple module to provide easy and readable access to the ANSI color and highlighting escape sequences, supported by most ANSI terminal emulators. It is now included standard.

Time::Local

The timelocal() and timegm() functions used to silently return bogus results when the date fell outside the machine's integer range. They now consistently croak() if the date falls in an unsupported range.

Win32

The error return value in list context has been changed for all functions that return a list of values. Previously these functions returned a list with a single element "undef" if an error occurred. Now these functions return the empty list in these situations. This applies to the following functions:

    Win32::FsType
    Win32::GetOSVersion
    

The remaining functions are unchanged and continue to return "undef" on error even in list context.

The Win32::SetLastError(ERROR) function has been added as a complement to the Win32::GetLastError() function.

The new Win32::GetFullPathName(FILENAME) returns the full absolute pathname for FILENAME in scalar context. In list context it returns a two-element list containing the fully qualified directory name and the filename. See Win32.

XSLoader

The XSLoader extension is a simpler alternative to DynaLoader. See XSLoader.

DBM Filters

A new feature called "DBM Filters" has been added to all the DBM modules--DB_File, GDBM_File, NDBM_File, ODBM_File, and SDBM_File. DBM Filters add four new methods to each DBM module:

    filter_store_key
    filter_store_value
    filter_fetch_key
    filter_fetch_value
    

These can be used to filter key-value pairs before the pairs are written to the database or just after they are read from the database. See perldbmfilter for further information.

Pragmata

Lexical warnings pragma, "use warnings;", to control optional warnings. See perllexwarn.

"use filetest" to control the behaviour of filetests ("-r" "-w" ...). Currently only one subpragma implemented, "use filetest 'access';", that uses access(2) or equivalent to check permissions instead of using stat(2) as usual. This matters in filesystems where there are ACLs (access control lists): the stat(2) might lie, but access(2) knows better.

The "open" pragma can be used to specify default disciplines for handle constructors (e.g. open()) and for qx//. The two pseudo-disciplines ":raw" and ":crlf" are currently supported on DOS-derivative platforms (i.e. where binmode is not a no-op). See also "binmode() can be used to set :crlf and :raw modes".

dprofpp

find2perl

h2xs

perlcc

Support for non-Unix platforms has been improved.

perldoc

The Perl Debugger

Simple sort() using { $a <=> $b } and the like are optimized

Optimized assignments to lexical variables

Faster subroutine calls

delete(), each(), values() and hash iteration are faster

-Dusethreads means something different

As of v5.6.0, interpreter-threads support is still lacking a way to create new threads from Perl (i.e., "use Thread;" will not work with interpreter threads). "use Thread;" continues to be available when you specify the -Duse5005threads option to Configure, bugs and all.

    NOTE: Support for threads continues to be an experimental feature.
    Interfaces and implementation are subject to sudden and drastic changes.

New Configure flags

    usemultiplicity
    usethreads useithreads      (new interpreter threads: no Perl API yet)
    usethreads use5005threads   (threads as they were in 5.005)

    use64bitint                 (equal to now deprecated 'use64bits')
    use64bitall

    uselongdouble
    usemorebits
    uselargefiles
    usesocks                    (only SOCKS v5 supported)

Threadedness and 64-bitness now more daring

Long Doubles

-Dusemorebits

-Duselargefiles

See "Large file support" for more information.

installusrbinperl

SOCKS support

    http://www.socks.nec.com/

"-A" flag

Enhanced Installation Directories

If you previously used "Configure -Dsitelib" or "-Dsitearch" to set special values for library directories, you might wish to consider using the new "-Dsiteprefix" setting instead. Also, if you wish to re-use a config.sh file from an earlier version of perl, you should be sure to check that Configure makes sensible choices for the new directories. See INSTALL for complete details.

gcc automatically tried if 'cc' does not seem to be working

Supported platforms

• The Mach CThreads (NEXTSTEP, OPENSTEP) are now supported by the Thread extension.
• GNU/Hurd is now supported.
• Rhapsody/Darwin is now supported.
• EPOC is now supported (on Psion 5).
• The cygwin port (formerly cygwin32) has been greatly improved.

DOS

• Perl now works with djgpp 2.02 (and 2.03 alpha).
• Environment variable names are not converted to uppercase any more.
• Incorrect exit codes from backticks have been fixed.
• This port continues to use its own builtin globbing (not File::Glob).

OS390 (OpenEdition MVS)

It is unclear whether future versions will renew support for this platform, but the possibility exists.

VMS

Expand %ENV-handling code to allow runtime mapping to logical names, CLI symbols, and CRTL environ array.

Extension of subprocess invocation code to accept filespecs as command "verbs".

Add to Perl command line processing the ability to use default file types and to recognize Unix-style "2>&1".

Expansion of File::Spec::VMS routines, and integration into ExtUtils::MM_VMS.

Extension of ExtUtils::MM_VMS to handle complex extensions more flexibly.

Barewords at start of Unix-syntax paths may be treated as text rather than only as logical names.

Optional secure translation of several logical names used internally by Perl.

Miscellaneous bugfixing and porting of new core code to VMS.

Thanks are gladly extended to the many people who have contributed VMS patches, testing, and ideas.

Win32

When given a pathname that consists only of a drivename, such as "A:", opendir() and stat() now use the current working directory for the drive rather than the drive root.

The builtin XSUB functions in the Win32:: namespace are documented. See Win32.

$^X now contains the full path name of the running executable.

A Win32::GetLongPathName() function is provided to complement Win32::GetFullPathName() and Win32::GetShortPathName(). See Win32.

POSIX::uname() is supported.

system(1,...) now returns true process IDs rather than process handles. kill() accepts any real process id, rather than strictly return values from system(1,...).

For better compatibility with Unix, "kill(0, $pid)" can now be used to test whether a process exists.

The "Shell" module is supported.

Better support for building Perl under command.com in Windows 95 has been added.

Scripts are read in binary mode by default to allow ByteLoader (and the filter mechanism in general) to work properly. For compatibility, the DATA filehandle will be set to text mode if a carriage return is detected at the end of the line containing the __END__ or __DATA__ token; if not, the DATA filehandle will be left open in binary mode. Earlier versions always opened the DATA filehandle in text mode.

The glob() operator is implemented via the "File::Glob" extension, which supports glob syntax of the C shell. This increases the flexibility of the glob() operator, but there may be compatibility issues for programs that relied on the older globbing syntax. If you want to preserve compatibility with the older syntax, you might want to run perl with "-MFile::DosGlob". For details and compatibility information, see File::Glob.

<HANDLE> on empty files

This means that the following will append "foo" to an empty file (it used to do nothing):

    perl -0777 -pi -e 's/^/foo/' empty_file

The behaviour of:

    perl -pi -e 's/^/foo/' empty_file

is unchanged (it continues to leave the file empty).

"eval '...'" improvements

Lexical lookups for variables appearing in "eval '...'" within functions that were themselves called within an "eval '...'" were searching the wrong place for lexicals. The lexical search now correctly ends at the subroutine's block boundary.

The use of "return" within "eval {...}" caused $@ not to be reset correctly when no exception occurred within the eval. This has been fixed.

Parsing of here documents used to be flawed when they appeared as the replacement expression in "eval 's/.../.../e'". This has been fixed.

All compilation errors are true errors

The mechanism for reporting such errors has been reimplemented to queue compile-time errors and report them at the end of the compilation as true errors rather than as warnings. This fixes cases where error messages leaked through in the form of warnings when code was compiled at run time using "eval STRING", and also allows such errors to be reliably trapped using "eval "..."".

Implicitly closed filehandles are safer

Behavior of list slices is more consistent

The new behavior is to produce an empty list if (and only if) the original list was empty. Consider the following example:

    @a = (1,undef,undef,2)[2,1,2];

The old behavior would have resulted in @a having no elements. The new behavior ensures it has three undefined elements.

Note in particular that the behavior of slices of the following cases remains unchanged:

    @a = ()[1,2];
    @a = (getpwent)[7,0];
    @a = (anything_returning_empty_list())[2,1,2];
    @a = @b[2,1,2];
    @a = @c{'a','b','c'};

See perldata.

"(\$)" prototype and $foo{a}

"goto &sub" and AUTOLOAD

"-bareword" allowed under "use integer"

Failures in DESTROY()

Locale bugs fixed

Numbers formatted according to the local numeric locale (such as using a decimal comma instead of a decimal dot) caused "isn't numeric" warnings, even while the operations accessing those numbers produced correct results. These warnings have been discontinued.

Memory leaks

Operations that aren't filehandle constructors used to leak memory when used on invalid filehandles. This has been fixed.

Constructs that modified @_ could fail to deallocate values in @_ and thus leak memory. This has been corrected.

Spurious subroutine stubs after failed subroutine calls

Taint failures under "-U"

END blocks and the "-c" switch

See "Support for CHECK blocks" for how to run things when the compile phase ends.

Potential to leak DATA filehandles

This caveat is now better explained in the documentation. See perldata.

Perl Source Incompatibilities

Since all new warnings must be explicitly requested via the "-w" switch or the "warnings" pragma, it is ultimately the programmer's responsibility to ensure that warnings are enabled judiciously.

CHECK is a new keyword

All subroutine definitions named CHECK are now special. See "/"Support for CHECK blocks"" for more information.

Treatment of list slices of undef has changed

There is a potential incompatibility in the behavior of list slices that are comprised entirely of undefined values. See "Behavior of list slices is more consistent".

Format of $English::PERL_VERSION is different

The English module now sets $PERL_VERSION to $^V (a string value) rather than $] (a numeric value). This is a potential incompatibility. Send us a report via perlbug if you are affected by this.

See "Improved Perl version numbering system" for the reasons for this change.

Literals of the form 1.2.3 parse differently

Previously, numeric literals with more than one dot in them were interpreted as a floating point number concatenated with one or more numbers. Such "numbers" are now parsed as strings composed of the specified ordinals.

For example, "print 97.98.99" used to output 97.9899 in earlier versions, but now prints "abc".

See "Support for strings represented as a vector of ordinals".

Possibly changed pseudo-random number generator

Perl programs that depend on reproducing a specific set of pseudo-random numbers may now produce different output due to improvements made to the rand() builtin. You can use "sh Configure -Drandfunc=rand" to obtain the old behavior.

See "Better pseudo-random number generator".

Hashing function for hash keys has changed

Even though Perl hashes are not order preserving, the apparently random order encountered when iterating on the contents of a hash is actually determined by the hashing algorithm used. Improvements in the algorithm may yield a random order that is different from that of previous versions, especially when iterating on hashes.

See "Better worst-case behavior of hashes" for additional information.

"undef" fails on read only values

Using the "undef" operator on a readonly value (such as $1) has the same effect as assigning "undef" to the readonly value--it throws an exception.

Close-on-exec bit may be set on pipe and socket handles

Pipe and socket handles are also now subject to the close-on-exec behavior determined by the special variable $^F.

See "More consistent close-on-exec behavior".

Writing "$$1" to mean "${$}1" is unsupported

Perl 5.004 deprecated the interpretation of $$1 and similar within interpolated strings to mean "$$ . "1"", but still allowed it.

In Perl 5.6.0 and later, "$$1" always means "${$1}".

delete(), each(), values() and "\(%h)"

operate on aliases to values, not copies

delete(), each(), values() and hashes (e.g. "\(%h)") in a list context return the actual values in the hash, instead of copies (as they used to in earlier versions). Typical idioms for using these constructs copy the returned values, but this can make a significant difference when creating references to the returned values. Keys in the hash are still returned as copies when iterating on a hash.

See also "delete(), each(), values() and hash iteration are faster".

vec(EXPR,OFFSET,BITS) enforces powers-of-two BITS

vec() generates a run-time error if the BITS argument is not a valid power-of-two integer.

Text of some diagnostic output has changed

Most references to internal Perl operations in diagnostics have been changed to be more descriptive. This may be an issue for programs that may incorrectly rely on the exact text of diagnostics for proper functioning.

"%@" has been removed

The undocumented special variable "%@" that used to accumulate "background" errors (such as those that happen in DESTROY()) has been removed, because it could potentially result in memory leaks.

Parenthesized not() behaves like a list operator

The "not" operator now falls under the "if it looks like a function, it behaves like a function" rule.

As a result, the parenthesized form can be used with "grep" and "map". The following construct used to be a syntax error before, but it works as expected now:

    grep not($_), @things;
    

On the other hand, using "not" with a literal list slice may not work. The following previously allowed construct:

    print not (1,2,3)[0];
    

needs to be written with additional parentheses now:

    print not((1,2,3)[0]);
    

The behavior remains unaffected when "not" is not followed by parentheses.

Semantics of bareword prototype "(*)" have changed

The semantics of the bareword prototype "*" have changed. Perl 5.005 always coerced simple scalar arguments to a typeglob, which wasn't useful in situations where the subroutine must distinguish between a simple scalar and a typeglob. The new behavior is to not coerce bareword arguments to a typeglob. The value will always be visible as either a simple scalar or as a reference to a typeglob.

See "More functional bareword prototype (*)".

Semantics of bit operators may have changed on 64-bit platforms

If your platform is either natively 64-bit or if Perl has been configured to used 64-bit integers, i.e., $Config{ivsize} is 8, there may be a potential incompatibility in the behavior of bitwise numeric operators (& | ^ ~ << >>). These operators used to strictly operate on the lower 32 bits of integers in previous versions, but now operate over the entire native integral width. In particular, note that unary "~" will produce different results on platforms that have different $Config{ivsize}. For portability, be sure to mask off the excess bits in the result of unary "~", e.g., "~$x & 0xffffffff".

See "Bit operators support full native integer width".

More builtins taint their results

As described in "Improved security features", there may be more sources of taint in a Perl program.

To avoid these new tainting behaviors, you can build Perl with the Configure option "-Accflags=-DINCOMPLETE_TAINTS". Beware that the ensuing perl binary may be insecure.

C Source Incompatibilities

"PERL_POLLUTE"

Release 5.005 grandfathered old global symbol names by providing preprocessor macros for extension source compatibility. As of release 5.6.0, these preprocessor definitions are not available by default. You need to explicitly compile perl with "-DPERL_POLLUTE" to get these definitions. For extensions still using the old symbols, this option can be specified via MakeMaker:

    perl Makefile.PL POLLUTE=1
    

"PERL_IMPLICIT_CONTEXT"

This new build option provides a set of macros for all API functions such that an implicit interpreter/thread context argument is passed to every API function. As a result of this, something like "sv_setsv(foo,bar)" amounts to a macro invocation that actually translates to something like "Perl_sv_setsv(my_perl,foo,bar)". While this is generally expected to not have any significant source compatibility issues, the difference between a macro and a real function call will need to be considered.

This means that there is a source compatibility issue as a result of this if your extensions attempt to use pointers to any of the Perl API functions.

Note that the above issue is not relevant to the default build of Perl, whose interfaces continue to match those of prior versions (but subject to the other options described here).

See "Background and PERL_IMPLICIT_CONTEXT" in perlguts for detailed information on the ramifications of building Perl with this option.

    NOTE: PERL_IMPLICIT_CONTEXT is automatically enabled whenever Perl is built
    with one of -Dusethreads, -Dusemultiplicity, or both.  It is not
    intended to be enabled by users at this time.
    

"PERL_POLLUTE_MALLOC"

Enabling Perl's malloc in release 5.005 and earlier caused the namespace of the system's malloc family of functions to be usurped by the Perl versions, since by default they used the same names. Besides causing problems on platforms that do not allow these functions to be cleanly replaced, this also meant that the system versions could not be called in programs that used Perl's malloc. Previous versions of Perl have allowed this behaviour to be suppressed with the HIDEMYMALLOC and EMBEDMYMALLOC preprocessor definitions.

As of release 5.6.0, Perl's malloc family of functions have default names distinct from the system versions. You need to explicitly compile perl with "-DPERL_POLLUTE_MALLOC" to get the older behaviour. HIDEMYMALLOC and EMBEDMYMALLOC have no effect, since the behaviour they enabled is now the default.

Note that these functions do not constitute Perl's memory allocation API. See "Memory Allocation" in perlguts for further information about that.

Compatible C Source API Changes

"PATCHLEVEL" is now "PERL_VERSION"

The cpp macros "PERL_REVISION", "PERL_VERSION", and "PERL_SUBVERSION" are now available by default from perl.h, and reflect the base revision, patchlevel, and subversion respectively. "PERL_REVISION" had no prior equivalent, while "PERL_VERSION" and "PERL_SUBVERSION" were previously available as "PATCHLEVEL" and "SUBVERSION".

The new names cause less pollution of the cpp namespace and reflect what the numbers have come to stand for in common practice. For compatibility, the old names are still supported when patchlevel.h is explicitly included (as required before), so there is no source incompatibility from the change.

Binary Incompatibilities

The usethreads or usemultiplicity builds are not binary compatible with the corresponding builds in 5.005.

On platforms that require an explicit list of exports (AIX, OS/2 and Windows, among others), purely internal symbols such as parser functions and the run time opcodes are not exported by default. Perl 5.005 used to export all functions irrespective of whether they were considered part of the public API or not.

For the full list of public API functions, see perlapi.

Localizing a tied hash element may leak memory

    use Tie::Hash;
    tie my %tie_hash => 'Tie::StdHash';

    ...

    local($tie_hash{Foo}) = 1; # leaks

Known test failures

• 64-bit builds

  Subtest #15 of lib/b.t may fail under 64-bit builds on platforms such as HP-UX PA64 and Linux IA64. The issue is still being investigated.

  The lib/io_multihomed test may hang in HP-UX if Perl has been configured to be 64-bit. Because other 64-bit platforms do not hang in this test, HP-UX is suspect. All other tests pass in 64-bit HP-UX. The test attempts to create and connect to "multihomed" sockets (sockets which have multiple IP addresses).

  Note that 64-bit support is still experimental.

• Failure of Thread tests

  The subtests 19 and 20 of lib/thr5005.t test are known to fail due to fundamental problems in the 5.005 threading implementation. These are not new failures--Perl 5.005_0x has the same bugs, but didn't have these tests. (Note that support for 5.005-style threading remains experimental.)

• NEXTSTEP 3.3 POSIX test failure

  In NEXTSTEP 3.3p2 the implementation of the strftime(3) in the operating system libraries is buggy: the %j format numbers the days of a month starting from zero, which, while being logical to programmers, will cause the subtests 19 to 27 of the lib/posix test may fail.

• Tru64 (aka Digital UNIX, aka DEC OSF/1) lib/sdbm test failure with gcc

  If compiled with gcc 2.95 the lib/sdbm test will fail (dump core). The cure is to use the vendor cc, it comes with the operating system and produces good code.

EBCDIC platforms not fully supported

The 5.6.1 release improves support for EBCDIC platforms, but they are not fully supported yet.

UNICOS/mk CC failures during Configure run

        Guessing which symbols your C compiler and preprocessor define...
        CC-20 cc: ERROR File = try.c, Line = 3
        ...
          bad switch yylook 79bad switch yylook 79bad switch yylook 79bad switch yylook 79#ifdef A29K
        ...
        4 errors detected in the compilation of "try.c".

The culprit is the broken awk of UNICOS/mk. The effect is fortunately rather mild: Perl itself is not adversely affected by the error, only the h2ph utility coming with Perl, and that is rather rarely needed these days.

Arrow operator and arrays

    @x->[2]
    scalar(@x)->[2]

These expressions will get run-time errors in some future release of Perl.

Experimental features

Threads

Unicode

64-bit support

Lvalue subroutines

Weak references

The pseudo-hash data type

The Compiler suite

Internal implementation of file globbing

The DB module

The regular expression code constructs:

"(?{ code })" and "(??{ code })"