1N/Aperl58delta - what is new for perl v5.8.0
1N/AThis document describes differences between the 5.6.0 release and
1N/AMany of the bug fixes in 5.8.0 were already seen in the 5.6.1
1N/Amaintenance release since the two releases were kept closely
1N/AChanges that were integrated into the 5.6.1 release are marked C<[561]>.
1N/AMany of these changes have been further developed since 5.6.1 was released,
1N/Athose are marked C<[561+]>.
1N/AYou can see the list of changes in the 5.6.1 release (both from the
1N/A5.005_03 release and the 5.6.0 release) by reading L<perl561delta>.
1N/A=head1 Highlights In 5.8.0
1N/ABetter Unicode support
1N/ANew IO Implementation
1N/ANew Thread Implementation
1N/ABetter Numeric Accuracy
1N/AMore Extensive Regression Testing
1N/A=head1 Incompatible Changes
1N/A=head2 Binary Incompatibility
1N/AB<Perl 5.8 is not binary compatible with earlier releases of Perl.>
1N/AB<You have to recompile your XS modules.>
1N/A(Pure Perl modules should continue to work.)
1N/AThe major reason for the discontinuity is the new IO architecture
1N/Acalled PerlIO. PerlIO is the default configuration because without
1N/Ait many new features of Perl 5.8 cannot be used. In other words:
1N/Ayou just have to recompile your modules containing XS code, sorry
1N/AIn future releases of Perl, non-PerlIO aware XS modules may become
1N/Acompletely unsupported. This shouldn't be too difficult for module
1N/Aauthors, however: PerlIO has been designed as a drop-in replacement
1N/A(at the source code level) for the stdio interface.
1N/ADepending on your platform, there are also other reasons why
1N/Awe decided to break binary compatibility, please read on.
1N/A=head2 64-bit platforms and malloc
1N/AIf your pointers are 64 bits wide, the Perl malloc is no longer being
1N/Aused because it does not work well with 8-byte pointers. Also,
1N/Ausually the system mallocs on such platforms are much better optimized
1N/Afor such large memory models than the Perl malloc. Some memory-hungry
1N/APerl applications like the PDL don't work well with Perl's malloc.
1N/AFinally, other applications than Perl (such as mod_perl) tend to prefer
1N/Athe system malloc. Such platforms include Alpha and 64-bit HPPA,
1N/AMIPS, PPC, and Sparc.
1N/A=head2 AIX Dynaloading
1N/AThe AIX dynaloading now uses in AIX releases 4.3 and newer the native
1N/Adlopen interface of AIX instead of the old emulated interface. This
1N/Achange will probably break backward compatibility with compiled
1N/Amodules. The change was made to make Perl more compliant with other
1N/Aapplications like mod_perl which are using the AIX native interface.
1N/A=head2 Attributes for C<my> variables now handled at run-time
1N/AThe C<my EXPR : ATTRS> syntax now applies variable attributes at
1N/Arun-time. (Subroutine and C<our> variables still get attributes applied
1N/Aat compile-time.) See L<attributes> for additional details. In particular,
1N/Ahowever, this allows variable attributes to be useful for C<tie> interfaces,
1N/Awhich was a deficiency of earlier releases. Note that the new semantics
1N/Adoesn't work with the Attribute::Handlers module (as of version 0.76).
1N/A=head2 Socket Extension Dynamic in VMS
1N/AThe Socket extension is now dynamically loaded instead of being
1N/Astatically built in. This may or may not be a problem with ancient
1N/ATCP/IP stacks of VMS: we do not know since we weren't able to test
1N/APerl in such configurations.
1N/A=head2 IEEE-format Floating Point Default on OpenVMS Alpha
1N/APerl now uses IEEE format (T_FLOAT) as the default internal floating
1N/Apoint format on OpenVMS Alpha, potentially breaking binary compatibility
1N/Awith external libraries or existing data. G_FLOAT is still available as
1N/Aa configuration option. The default on VAX (D_FLOAT) has not changed.
1N/A=head2 New Unicode Semantics (no more C<use utf8>, almost)
1N/APreviously in Perl 5.6 to use Unicode one would say "use utf8" and
1N/Athen the operations (like string concatenation) were Unicode-aware
1N/Ain that lexical scope.
1N/AThis was found to be an inconvenient interface, and in Perl 5.8 the
1N/AUnicode model has completely changed: now the "Unicodeness" is bound
1N/Ato the data itself, and for most of the time "use utf8" is not needed
1N/Aat all. The only remaining use of "use utf8" is when the Perl script
1N/Aitself has been written in the UTF-8 encoding of Unicode. (UTF-8 has
1N/Anot been made the default since there are many Perl scripts out there
1N/Athat are using various national eight-bit character sets, which would
1N/Abe illegal in UTF-8.)
1N/ASee L<perluniintro> for the explanation of the current model,
1N/Aand L<utf8> for the current use of the utf8 pragma.
1N/A=head2 New Unicode Properties
1N/AUnicode I<scripts> are now supported. Scripts are similar to (and superior
1N/Ato) Unicode I<blocks>. The difference between scripts and blocks is that
1N/Ascripts are the glyphs used by a language or a group of languages, while
1N/Athe blocks are more artificial groupings of (mostly) 256 characters based
1N/Aon the Unicode numbering.
1N/AIn general, scripts are more inclusive, but not universally so. For
1N/Aexample, while the script C<Latin> includes all the Latin characters and
1N/Atheir various diacritic-adorned versions, it does not include the various
1N/Apunctuation or digits (since they are not solely C<Latin>).
1N/AA number of other properties are now supported, including C<\p{L&}>,
1N/AC<\p{Any}> C<\p{Assigned}>, C<\p{Unassigned}>, C<\p{Blank}> [561] and
1N/AC<\p{SpacePerl}> [561] (along with their C<\P{...}> versions, of course).
1N/ASee L<perlunicode> for details, and more additions.
1N/AThe C<In> or C<Is> prefix to names used with the C<\p{...}> and C<\P{...}>
1N/Aare now almost always optional. The only exception is that a C<In> prefix
1N/Ais required to signify a Unicode block when a block name conflicts with a
1N/Ascript name. For example, C<\p{Tibetan}> refers to the script, while
1N/AC<\p{InTibetan}> refers to the block. When there is no name conflict, you
1N/Acan omit the C<In> from the block name (
e.g. C<\p{BraillePatterns}>), but
1N/Ato be safe, it's probably best to always use the C<In>).
1N/A=head2 REF(...) Instead Of SCALAR(...)
1N/AA reference to a reference now stringifies as "REF(0x81485ec)" instead
1N/Aof "SCALAR(0x81485ec)" in order to be more consistent with the return
1N/Afor better use: now they stand for long double (if supported by the
1N/Aplatform) and NV (Perl internal floating point type). (They used
1N/Ato be aliases for d/f, but you never knew that.)
1N/A=head2 glob() now returns filenames in alphabetical order
1N/AThe list of filenames from glob() (or <...>) is now by default sorted
1N/Aalphabetically to be csh-compliant (which is what happened before
1N/Ain most UNIX platforms). (bsd_glob() does still sort platform
1N/Anatively, ASCII or EBCDIC, unless GLOB_ALPHASORT is specified.) [561]
1N/AThe semantics of bless(REF, REF) were unclear and until someone proves
1N/Ait to make some sense, it is forbidden.
1N/AThe obsolete chat2 library that should never have been allowed
1N/Ato escape the laboratory has been decommissioned.
1N/AUsing chdir("") or chdir(undef) instead of explicit chdir() is
1N/Adoubtful. A failure (think chdir(some_function()) can lead into
1N/Aunintended chdir() to the home directory, therefore this behaviour
1N/AThe builtin dump() function has probably outlived most of its
1N/Ausefulness. The core-dumping functionality will remain in future
1N/Aavailable as an explicit call to C<CORE::dump()>, but in future
1N/Areleases the behaviour of an unqualified C<dump()> call may change.
1N/AThe very dusty examples in the eg/ directory have been removed.
1N/ASuggestions for new shiny examples welcome but the main issue is that
1N/Athe examples need to be documented, tested and (most importantly)
1N/AThe (bogus) escape sequences \8 and \9 now give an optional warning
1N/A("Unrecognized escape passed through"). There is no need to \-escape
1N/AThe *glob{FILEHANDLE} is deprecated, use *glob{IO} instead.
1N/AThe C<package;> syntax (C<package> without an argument) has been
1N/Adeprecated. Its semantics were never that clear and its
1N/Aimplementation even less so. If you have used that feature to
1N/Adisallow all but fully qualified variables, C<use strict;> instead.
1N/AThe unimplemented POSIX regex features [[.cc.]] and [[=c=]] are still
1N/Arecognised but now cause fatal errors. The previous behaviour of
1N/Aignoring them by default and warning if requested was unacceptable
1N/Asince it, in a way, falsely promised that the features could be used.
1N/AIn future releases, non-PerlIO aware XS modules may become completely
1N/Aunsupported. Since PerlIO is a drop-in replacement for stdio at the
1N/Asource code level, this shouldn't be that drastic a change.
1N/APrevious versions of perl and some readings of some sections of Camel
1N/AIII implied that the C<:raw> "discipline" was the inverse of C<:crlf>.
1N/ATurning off "clrfness" is no longer enough to make a stream truly
1N/Abinary. So the PerlIO C<:raw> layer (or "discipline", to use the Camel
1N/Abook's older terminology) is now formally defined as being equivalent
1N/Ato binmode(FH) - which is in turn defined as doing whatever is
1N/Anecessary to pass each byte as-is without any translation. In
1N/Aparticular binmode(FH) - and hence C<:raw> - will now turn off both
1N/ACRLF and UTF-8 translation and remove other layers (
e.g. :encoding())
1N/Awhich would modify byte stream.
1N/AThe current user-visible implementation of pseudo-hashes (the weird
1N/Ause of the first array element) is deprecated starting from Perl 5.8.0
1N/Aand will be removed in Perl 5.10.0, and the feature will be
1N/Aimplemented differently. Not only is the current interface rather
1N/Augly, but the current implementation slows down normal array and hash
1N/Ause quite noticeably. The C<fields> pragma interface will remain
1N/Aavailable. The I<restricted hashes> interface is expected to
1N/Abe the replacement interface (see L<Hash::Util>). If your existing
1N/Aprograms depends on the underlying implementation, consider using
1N/AL<Class::PseudoHash> from CPAN.
1N/AThe syntaxes C<< @a->[...] >> and C<< %h->{...} >> have now been deprecated.
1N/AAfter years of trying, suidperl is considered to be too complex to
1N/Aever be considered truly secure. The suidperl functionality is likely
1N/Ato be removed in a future release.
1N/AThe 5.005 threads model (module C<Thread>) is deprecated and expected
1N/Ato be removed in Perl 5.10. Multithreaded code should be migrated to
1N/Athe new ithreads model (see L<threads>, L<threads::shared> and
1N/AThe long deprecated uppercase aliases for the string comparison
1N/Aoperators (EQ, NE, LT, LE, GE, GT) have now been removed.
1N/AThe tr///C and tr///U features have been removed and will not return;
1N/Athe interface was a mistake. Sorry about that. For similar
1N/Afunctionality, see pack('U0', ...) and pack('C0', ...). [561]
1N/AEarlier Perls treated "sub foo (@bar)" as equivalent to "sub foo (@)".
1N/AThe prototypes are now checked better at compile-time for invalid
1N/Asyntax. An optional warning is generated ("Illegal character in
1N/Aprototype...") but this may be upgraded to a fatal error in a future
1N/AThe C<exec LIST> and C<system LIST> operations now produce warnings on
1N/Atainted data and in some future release they will produce fatal errors.
1N/AThe existing behaviour when localising tied arrays and hashes is wrong,
1N/Aand will be changed in a future release, so do not rely on the existing
1N/Abehaviour. See L<"Localising Tied Arrays and Hashes Is Broken">.
1N/A=head1 Core Enhancements
1N/A=head2 Unicode Overhaul
1N/AUnicode in general should be now much more usable than in Perl 5.6.0
1N/A(or even in 5.6.1). Unicode can be used in hash keys, Unicode in
1N/Aregular expressions should work now, Unicode in tr/// should work now,
1N/AUnicode in I/O should work now. See L<perluniintro> for introduction
1N/Aand L<perlunicode> for details.
1N/AThe Unicode Character Database coming with Perl has been upgraded
1N/A[561+] (5.6.1 has UCD 3.0.1.)
1N/AFor developers interested in enhancing Perl's Unicode capabilities:
1N/Aalmost all the UCD files are included with the Perl distribution in
1N/Aconsiderations, is the Unihan database.
1N/AThe properties \p{Blank} and \p{SpacePerl} have been added. "Blank" is like
1N/AC isblank(), that is, it contains only "horizontal whitespace" (the space
1N/Acharacter is, the newline isn't), and the "SpacePerl" is the Unicode
1N/Aequivalent of C<\s> (\p{Space} isn't, since that includes the vertical
1N/Atabulator character, whereas C<\s> doesn't.)
1N/ASee "New Unicode Properties" earlier in this document for additional
1N/Ainformation on changes with Unicode properties.
1N/A=head2 PerlIO is Now The Default
1N/AIO is now by default done via PerlIO rather than system's "stdio".
1N/APerlIO allows "layers" to be "pushed" onto a file handle to alter the
1N/Ahandle's behaviour. Layers can be specified at open time via 3-arg
1N/A open($fh,'>:crlf :utf8', $path) || ...
1N/Aor on already opened handles via extended C<binmode>:
1N/A binmode($fh,':encoding(iso-8859-7)');
1N/AThe built-in layers are: unix (low level
read/write), stdio (as in
1N/Aprevious Perls), perlio (re-implementation of stdio buffering in a
1N/Aportable manner), crlf (does CRLF <=> "\n" translation as on Win32,
1N/Abut available on any platform). A mmap layer may be available if
1N/Aplatform supports it (mostly UNIXes).
1N/ALayers to be applied by default may be specified via the 'open' pragma.
1N/ASee L</"Installation and Configuration Improvements"> for the effects
1N/Aof PerlIO on your architecture name.
1N/AIf your platform supports fork(), you can use the list form of C<open>
1N/Afor pipes. For example:
1N/A open KID_PS, "-|", "ps", "aux" or die $!;
1N/Aforks the ps(1) command (without spawning a shell, as there are more
1N/Athan three arguments to open()), and reads its standard output via the
1N/AC<KID_PS> filehandle. See L<perlipc>.
1N/AFile handles can be marked as accepting Perl's internal encoding of Unicode
1N/A(UTF-8 or UTF-EBCDIC depending on platform) by a pseudo layer ":utf8" :
1N/ANote for EBCDIC users: the pseudo layer ":utf8" is erroneously named
1N/Afor you since it's not UTF-8 what you will be getting but instead
1N/AUTF-EBCDIC. See L<perlunicode>, L<utf8>, and
1N/AIn future releases this naming may change. See L<perluniintro>
1N/Afor more information about UTF-8.
1N/AIf your environment variables (LC_ALL, LC_CTYPE, LANG) look like you
1N/Awant to use UTF-8 (any of the the variables match C</utf-?8/i>), your
1N/ASTDIN, STDOUT, STDERR handles and the default open layer (see L<open>)
1N/Aare marked as UTF-8. (This feature, like other new features that
1N/Acombine Unicode and I/O, work only if you are using PerlIO, but that's
1N/ANote that after this Perl really does assume that everything is UTF-8:
1N/Afor example if some input handle is not, Perl will probably very soon
1N/Acomplain about the input data like this "Malformed UTF-8 ..." since
1N/Aany old eight-bit data is not legal UTF-8.
1N/ANote for code authors: if you want to enable your users to use UTF-8
1N/Aas their default encoding but in your code still have eight-bit I/O streams
1N/A(such as images or zip files), you need to explicitly open() or binmode()
1N/Acan just use C<binmode(FH)> (nice for pre-5.8.0 backward compatibility).
1N/AFile handles can translate character encodings
from/to Perl's internal
1N/AFile handles can be opened to "in memory" files held in Perl scalars via:
1N/A open($fh,'>', \$variable) || ...
1N/AAnonymous temporary files are available without need to
1N/A'use FileHandle' or other module via
1N/A open($fh,"+>", undef) || ...
1N/AThat is a literal undef, not an undefined value.
1N/AThe new interpreter threads ("ithreads" for short) implementation of
1N/Amultithreading, by Arthur Bergman, replaces the old "5.005 threads"
1N/Aimplementation. In the ithreads model any data sharing between
1N/Athreads must be explicit, as opposed to the model where data sharing
1N/Awas implicit. See L<threads> and L<threads::shared>, and
1N/AAs a part of the ithreads implementation Perl will also use
1N/Aany necessary and detectable reentrant libc interfaces.
1N/A=head2 Restricted Hashes
1N/AA restricted hash is restricted to a certain set of keys, no keys
1N/Aoutside the set can be added. Also individual keys can be restricted
1N/Aso that the key cannot be deleted and the value cannot be changed.
1N/ANo new syntax is involved: the Hash::Util module is the interface.
1N/APerl used to be fragile in that signals arriving at inopportune moments
1N/Acould corrupt Perl's internal state. Now Perl postpones handling of
1N/Asignals until it's safe (between opcodes).
1N/AThis change may have surprising side effects because signals no longer
1N/Ainterrupt Perl instantly. Perl will now first finish whatever it was
1N/Adoing, like finishing an internal operation (like sort()) or an
1N/Aexternal operation (like an I/O operation), and only then look at any
1N/Aarrived signals (and before starting the next operation). No more corrupt
1N/Ainternal state since the current operation is always finished first,
1N/Abut the signal may take more time to get heard. Note that breaking
1N/Aout from potentially blocking operations should still work, though.
1N/A=head2 Understanding of Numbers
1N/AIn general a lot of fixing has happened in the area of Perl's
1N/Aunderstanding of numbers, both integer and floating point. Since in
1N/Amany systems the standard number parsing functions like C<strtoul()>
1N/Aand C<atof()> seem to have bugs, Perl tries to work around their
1N/Adeficiencies. This results hopefully in more accurate numbers.
1N/APerl now tries internally to use integer values in numeric conversions
1N/Aand basic arithmetics (+ - * /) if the arguments are integers, and
1N/Atries also to keep the results stored internally as integers.
1N/AThis change leads to often slightly faster and always less lossy
1N/Aarithmetics. (Previously Perl always preferred floating point numbers
1N/A=head2 Arrays now always interpolate into double-quoted strings [561]
1N/AIn double-quoted strings, arrays now interpolate, no matter what. The
1N/Abehavior in earlier versions of perl 5 was that arrays would interpolate
1N/Ainto strings if the array had been mentioned before the string was
1N/Acompiled, and otherwise Perl would raise a fatal compile-time error.
1N/AIn versions 5.000 through 5.003, the error was
1N/A Literal @example now requires backslash
1N/AIn versions 5.004_01 through 5.6.0, the error was
1N/A In string, @example now must be written as \@example
1N/AThe idea here was to get people into the habit of writing
1N/Athey have always written C<"Give me back my \$5"> when they wanted a
1N/AStarting with 5.6.1, when Perl now sees an C<@> sign in a
1N/Adouble-quoted string, it I<always> attempts to interpolate an array,
1N/Aregardless of whether or not the array has been used or declared
1N/Aalready. The fatal error has been downgraded to an optional warning:
1N/A Possible unintended interpolation of @example in string
1N/AThis warns you that C<"fred@example.com"> is going to turn into
1N/Aabout the history here.
1N/A=head2 Miscellaneous Changes
1N/AAUTOLOAD is now lvaluable, meaning that you can add the :lvalue attribute
1N/Ato AUTOLOAD subroutines and you can assign to the AUTOLOAD return value.
1N/AThe $Config{byteorder} (and corresponding BYTEORDER in
config.h) was
1N/Apreviously wrong in platforms if sizeof(long) was 4, but sizeof(IV)
1N/Awas 8. The byteorder was only sizeof(long) bytes long (1234 or 4321),
1N/Abut now it is correctly sizeof(IV) bytes long, (12345678 or 87654321).
1N/A(This problem didn't affect Windows platforms.)
1N/AAlso, $Config{byteorder} is now computed dynamically--this is more
1N/Arobust with "fat binaries" where an executable image contains binaries
1N/Afor more than one binary platform, and when cross-compiling.
1N/AC<perl -d:Module=arg,arg,arg> now works (previously one couldn't pass
1N/Ain multiple arguments.)
1N/AC<do> followed by a bareword now ensures that this bareword isn't
1N/Aa keyword (to avoid a bug where C<do q(
foo.pl)> tried to call a
1N/Asubroutine called C<q>). This means that for example instead of
1N/AC<do format()> you must write C<do &format()>.
1N/AThe builtin dump() now gives an optional warning
1N/AC<dump() better written as CORE::dump()>,
1N/Ameaning that by default C<dump(...)> is resolved as the builtin
1N/Adump() which dumps core and aborts, not as (possibly) user-defined
1N/AC<sub dump>. To call the latter, qualify the call as C<&dump(...)>.
1N/A(The whole dump() feature is to considered deprecated, and possibly
1N/Achomp() and chop() are now overridable. Note, however, that their
1N/Aprototype (as given by C<prototype("CORE::chomp")> is undefined,
1N/Abecause it cannot be expressed and therefore one cannot really write
1N/Areplacements to override these builtins.
1N/AEND blocks are now run even if you
exit/die in a BEGIN block.
1N/AInternally, the execution of END blocks is now controlled by
1N/APL_exit_flags & PERL_EXIT_DESTRUCT_END. This enables the new
1N/Abehaviour for Perl embedders. This will default in 5.10. See
1N/AFormats now support zero-padded decimal fields.
1N/AAlthough "you shouldn't do that", it was possible to write code that
1N/Adepends on Perl's hashed key order (Data::Dumper does this). The new
1N/Aalgorithm "One-at-a-Time" produces a different hashed key order.
1N/AMore details are in L</"Performance Enhancements">.
1N/Alstat(FILEHANDLE) now gives a warning because the operation makes no sense.
1N/AIn future releases this may become a fatal error.
1N/ASpurious syntax errors generated in certain situations, when glob()
1N/Acaused File::Glob to be loaded for the first time, have been fixed. [561]
1N/ALvalue subroutines can now return C<undef> in list context. However,
1N/Athe lvalue subroutine feature still remains experimental. [561+]
1N/AA lost warning "Can't declare ... dereference in my" has been
1N/Arestored (Perl had it earlier but it became lost in later releases.)
1N/AA new special regular expression variable has been introduced:
1N/AC<$^N>, which contains the most-recently closed group (submatch).
1N/AC<no Module;> does not produce an error even if Module does not have an
1N/Aunimport() method. This parallels the behavior of C<use> vis-a-vis
1N/AThe numerical comparison operators return C<undef> if either operand
1N/Ais a NaN. Previously the behaviour was unspecified.
1N/AC<our> can now have an experimental optional attribute C<unique> that
1N/Aaffects how global variables are shared among multiple interpreters,
1N/AThe following builtin functions are now overridable: each(), keys(),
1N/Apop(), push(), shift(), splice(), unshift(). [561]
1N/AC<pack() / unpack()> can now group template letters with C<()> and then
1N/AC<pack() / unpack()> can now process the Perl internal numeric types:
1N/AIVs, UVs, NVs-- and also long doubles, if supported by the platform.
1N/AThe template letters are C<j>, C<J>, C<F>, and C<D>.
1N/AC<pack('U0a*', ...)> can now be used to force a string to UTF-8.
1N/Amy __PACKAGE__ $obj now works. [561]
1N/APOSIX::sleep() now returns the number of I<unslept> seconds
1N/A(as the POSIX standard says), as opposed to CORE::sleep() which
1N/Areturns the number of slept seconds.
1N/Aprintf() and sprintf() now support parameter reordering using the
1N/AC<%\d+\$> and C<*\d+\$> syntaxes. For example
1N/A printf "%2\$s %1\$s\n", "foo", "bar";
1N/Awill print "bar foo\n". This feature helps in writing
1N/Ainternationalised software, and in general when the order
1N/Aof the parameters can vary.
1N/AThe (\&) prototype now works properly. [561]
1N/Aprototype(\[$@%&]) is now available to implicitly create references
1N/A(useful for example if you want to emulate the tie() interface).
1N/AA new command-line option, C<-t> is available. It is the
1N/Alittle brother of C<-T>: instead of dying on taint violations,
1N/Alexical warnings are given. B<This is only meant as a temporary
1N/Adebugging aid while securing the code of old legacy applications.
1N/AThis is not a substitute for -T.>
1N/AIn other taint news, the C<exec LIST> and C<system LIST> have now been
1N/Aconsidered too risky (think C<exec @ARGV>: it can start any program
1N/Awith any arguments), and now the said forms cause a warning under
1N/Alexical warnings. You should carefully launder the arguments to
1N/Aguarantee their validity. In future releases of Perl the forms will
1N/Abecome fatal errors so consider starting laundering now.
1N/ATied hash interfaces are now required to have the EXISTS and DELETE
1N/Amethods (either own or inherited).
1N/AIf tr/// is just counting characters, it doesn't attempt to
1N/Auntie() will now call an UNTIE() hook if it exists. See L<perltie>
1N/AL<utime> now supports C<utime undef, undef, @files> to change the
1N/Afile timestamps to the current time.
1N/AThe rules for allowing underscores (underbars) in numeric constants
1N/Ahave been relaxed and simplified: now you can have an underscore
1N/Asimply B<between digits>.
1N/ARather than relying on C's argv[0] (which may not contain a full pathname)
1N/Awhere possible $^X is now set by asking the operating system.
1N/AA new variable, C<${^TAINT}>, indicates whether taint mode is enabled.
1N/AYou can now override the readline() builtin, and this overrides also
1N/Athe <FILEHANDLE> angle bracket operator.
1N/AThe command-line options -s and -F are now recognized on the shebang
1N/AUse of the C</c> match modifier without an accompanying C</g> modifier
1N/Aelicits a new warning: C<Use of /c modifier is meaningless without /g>.
1N/AUse of C</c> in substitutions, even with C</g>, elicits
1N/AC<Use of /c modifier is meaningless in s///>.
1N/AUse of C</g> with C<split> elicits C<Use of /g modifier is meaningless
1N/ASupport for the C<CLONE> special subroutine had been added.
1N/AWith ithreads, when a new thread is created, all Perl data is cloned,
1N/Ahowever non-Perl data cannot be cloned automatically. In C<CLONE> you
1N/Acan do whatever you need to do, like for example handle the cloning of
1N/Anon-Perl data, if necessary. C<CLONE> will be executed once for every
1N/Apackage that has it defined or inherited. It will be called in the
1N/Acontext of the new thread, so all modifications are made in the new area.
1N/A=head1 Modules and Pragmata
1N/A=head2 New Modules and Pragmata
1N/AC<Attribute::Handlers>, originally by Damian Conway and now maintained
1N/Aby Arthur Bergman, allows a class to define attribute handlers.
1N/A use Attribute::Handlers;
1N/A sub Wolf :ATTR(SCALAR) { print "howl!\n" }
1N/A # later, in some package using or inheriting from MyPack...
1N/A my MyPack $Fluffy : Wolf; # the attribute handler Wolf will be called
1N/ABoth variables and routines can have attribute handlers. Handlers can
1N/Abe specific to type (SCALAR, ARRAY, HASH, or CODE), or specific to the
1N/Aexact compilation phase (BEGIN, CHECK, INIT, or END).
1N/ASee L<Attribute::Handlers>.
1N/AC<B::Concise>, by Stephen McCamant, is a new compiler backend for
1N/Awalking the Perl syntax tree, printing concise info about ops.
1N/AThe output is highly customisable. See L<B::Concise>. [561+]
1N/AThe new bignum, bigint, and bigrat pragmas, by Tels, implement
1N/Atransparent bignum support (using the Math::BigInt, Math::BigFloat,
1N/Aand Math::BigRat backends).
1N/AC<Class::ISA>, by Sean Burke, is a module for reporting the search
1N/Apath for a class's ISA tree. See L<Class::ISA>.
1N/AC<Cwd> now has a split personality: if possible, an XS extension is
1N/Aused, (this will hopefully be faster, more secure, and more robust)
1N/Abut if not possible, the familiar Perl implementation is used.
1N/AC<Devel::PPPort>, originally by Kenneth Albanowski and now
1N/Amaintained by Paul Marquess, has been added. It is primarily used
1N/Aby C<h2xs> to enhance portability of XS modules between different
1N/Aversions of Perl. See L<Devel::PPPort>.
1N/AC<Digest>, frontend module for calculating digests (checksums), from
1N/AGisle Aas, has been added. See L<Digest>.
1N/AC<Digest::MD5> for calculating MD5 digests (checksums) as defined in
1N/ARFC 1321, from Gisle Aas, has been added. See L<Digest::MD5>.
1N/A use Digest::MD5 'md5_hex';
1N/A $digest = md5_hex("Thirsty Camel");
1N/A print $digest, "\n"; # 01d19d9d2045e005c3f1b80e8b164de1
1N/ANOTE: the C<MD5> backward compatibility module is deliberately not
1N/Aincluded since its further use is discouraged.
1N/ASee also L<PerlIO::via::QuotedPrint>.
1N/AC<Encode>, originally by Nick Ing-Simmons and now maintained by Dan
1N/AKogai, provides a mechanism to translate between different character
1N/Aencodings. Support for Unicode, ISO-8859-1, and ASCII are compiled in
1N/Ato the module. Several other encodings (like the rest of the
1N/AISO-8859, CP*/Win*, Mac, KOI8-R, three variants EBCDIC, Chinese,
1N/AJapanese, and Korean encodings) are included and can be loaded at
1N/Aruntime. (For space considerations, the largest Chinese encodings
1N/Ahave been separated into their own CPAN module, Encode::HanExtra,
1N/Awhich Encode will use if available). See L<Encode>.
1N/AAny encoding supported by Encode module is also available to the
1N/A":encoding()" layer if PerlIO is used.
1N/AC<Hash::Util> is the interface to the new I<restricted hashes>
1N/Afeature. (Implemented by Jeffrey Friedl, Nick Ing-Simmons, and
1N/AMichael Schwern.) See L<Hash::Util>.
1N/AC<I18N::Langinfo> can be used to query locale information.
1N/ASee L<I18N::Langinfo>.
1N/AC<I18N::LangTags>, by Sean Burke, has functions for dealing with
1N/ARFC3066-style language tags. See L<I18N::LangTags>.
1N/AC<ExtUtils::Constant>, by Nicholas Clark, is a new tool for extension
1N/Awriters for generating XS code to import C header constants.
1N/ASee L<ExtUtils::Constant>.
1N/AC<Filter::Simple>, by Damian Conway, is an easy-to-use frontend to
1N/AFilter::Util::Call. See L<Filter::Simple>.
1N/A use Filter::Simple sub {
1N/A while (my ($from, $to) = splice @_, 0, 2) {
1N/A print "red\n"; # this code is filtered, will print "green\n"
1N/A print "bored\n"; # this code is filtered, will print "bogreen\n"
1N/A print "red\n"; # this code is not filtered, will print "red\n"
1N/AC<File::Temp>, by Tim Jenness, allows one to create temporary files
1N/Aand directories in an easy, portable, and secure way. See L<File::Temp>.
1N/AC<Filter::Util::Call>, by Paul Marquess, provides you with the
1N/Aframework to write I<source filters> in Perl. For most uses, the
1N/Afrontend Filter::Simple is to be preferred. See L<Filter::Util::Call>.
1N/AC<if>, by Ilya Zakharevich, is a new pragma for conditional inclusion
1N/AL<libnet>, by Graham Barr, is a collection of perl5 modules related
1N/Ato network programming. See L<Net::FTP>, L<Net::NNTP>, L<Net::Ping>
1N/A(not part of libnet, but related), L<Net::POP3>, L<Net::SMTP>,
1N/APerl installation leaves libnet unconfigured; use F<libnetcfg>
1N/AC<List::Util>, by Graham Barr, is a selection of general-utility
1N/Alist subroutines, such as sum(), min(), first(), and shuffle().
1N/AC<Locale::Constants>, C<Locale::Country>, C<Locale::Currency>
1N/AC<Locale::Language>, and L<Locale::Script>, by Neil Bowers, have
1N/Abeen added. They provide the codes for various locale standards, such
1N/Aas "fr" for France, "usd" for US Dollar, and "ja" for Japanese.
1N/A use Locale::Country;
1N/A $country = code2country('jp'); # $country gets 'Japan'
1N/A $code = country2code('Norway'); # $code gets 'no'
1N/ASee L<Locale::Constants>, L<Locale::Country>, L<Locale::Currency>,
1N/Aand L<Locale::Language>.
1N/AC<Locale::Maketext>, by Sean Burke, is a localization framework. See
1N/AL<Locale::Maketext>, and L<Locale::Maketext::TPJ13>. The latter is an
1N/Aarticle about software localization, originally published in The Perl
1N/AJournal #13, and republished here with kind permission.
1N/AC<Math::BigRat> for big rational numbers, to accompany Math::BigInt and
1N/AMath::BigFloat, from Tels. See L<Math::BigRat>.
1N/AC<Memoize> can make your functions faster by trading space for time,
1N/Afrom Mark-Jason Dominus. See L<Memoize>.
1N/AC<MIME::Base64>, by Gisle Aas, allows you to encode data in base64,
1N/Aas defined in RFC 2045 - I<MIME (Multipurpose Internet Mail
1N/A $encoded = encode_base64('Aladdin:open sesame');
1N/A $decoded = decode_base64($encoded);
1N/A print $encoded, "\n"; # "QWxhZGRpbjpvcGVuIHNlc2FtZQ=="
1N/AC<MIME::QuotedPrint>, by Gisle Aas, allows you to encode data
1N/Ain quoted-printable encoding, as defined in RFC 2045 - I<MIME
1N/A(Multipurpose Internet Mail Extensions)>.
1N/A use MIME::QuotedPrint;
1N/A $encoded = encode_qp("\xDE\xAD\xBE\xEF");
1N/A $decoded = decode_qp($encoded);
1N/A print $encoded, "\n"; # "=DE=AD=BE=EF\n"
1N/A print $decoded, "\n"; # "\xDE\xAD\xBE\xEF\n"
1N/ASee also L<PerlIO::via::QuotedPrint>.
1N/AC<NEXT>, by Damian Conway, is a pseudo-class for method redispatch.
1N/AC<open> is a new pragma for setting the default I/O layers
1N/AC<PerlIO::scalar>, by Nick Ing-Simmons, provides the implementation
1N/Aof IO to "in memory" Perl scalars as discussed above. It also serves
1N/Aas an example of a loadable PerlIO layer. Other future possibilities
1N/Ainclude PerlIO::Array and PerlIO::Code. See L<PerlIO::scalar>.
1N/AC<PerlIO::via>, by Nick Ing-Simmons, acts as a PerlIO layer and wraps
1N/APerlIO layer functionality provided by a class (typically implemented
1N/AC<PerlIO::via::QuotedPrint>, by Elizabeth Mattijsen, is an example
1N/Aof a C<PerlIO::via> class:
1N/A use PerlIO::via::QuotedPrint;
1N/A open($fh,">:via(QuotedPrint)",$path);
1N/AThis will automatically convert everything output to C<$fh> to
1N/AQuoted-Printable. See L<PerlIO::via> and L<PerlIO::via::QuotedPrint>.
1N/AC<Pod::ParseLink>, by Russ Allbery, has been added,
1N/Ato parse LZ<><> links in pods as described in the new
1N/AC<Pod::Text::Overstrike>, by Joe Smith, has been added.
1N/AIt converts POD data to formatted overstrike text.
1N/ASee L<Pod::Text::Overstrike>. [561+]
1N/AC<Scalar::Util> is a selection of general-utility scalar subroutines,
1N/Asuch as blessed(), reftype(), and tainted(). See L<Scalar::Util>.
1N/AC<sort> is a new pragma for controlling the behaviour of sort().
1N/AC<Storable> gives persistence to Perl data structures by allowing the
1N/Astorage and retrieval of Perl data to and from files in a fast and
1N/Acompact binary format. Because in effect Storable does serialisation
1N/Aof Perl data structures, with it you can also clone deep, hierarchical
1N/Adatastructures. Storable was originally created by Raphael Manfredi,
1N/Abut it is now maintained by Abhijit Menon-Sen. Storable has been
1N/Aenhanced to understand the two new hash features, Unicode keys and
1N/Arestricted hashes. See L<Storable>.
1N/AC<Switch>, by Damian Conway, has been added. Just by saying
1N/Ayou have C<switch> and C<case> available in Perl.
1N/A case 1 { print "number 1" }
1N/A case "a" { print "string a" }
1N/A case [1..10,42] { print "number in list" }
1N/A case (@array) { print "number in list" }
1N/A case /\w+/ { print "pattern" }
1N/A case qr/\w+/ { print "pattern" }
1N/A case (%hash) { print "entry in hash" }
1N/A case (\%hash) { print "entry in hash" }
1N/A case (\&sub) { print "arg to subroutine" }
1N/A else { print "previous case not true" }
1N/AC<Test::More>, by Michael Schwern, is yet another framework for writing
1N/Atest scripts, more extensive than Test::Simple. See L<Test::More>.
1N/AC<Test::Simple>, by Michael Schwern, has basic utilities for writing
1N/Atests. See L<Test::Simple>.
1N/AC<Text::Balanced>, by Damian Conway, has been added, for extracting
1N/Adelimited text sequences from strings.
1N/A use Text::Balanced 'extract_delimited';
1N/A ($a, $b) = extract_delimited("'never say never', he never said", "'", '');
1N/A$a will be "'never say never'", $b will be ', he never said'.
1N/AIn addition to extract_delimited(), there are also extract_bracketed(),
1N/Aextract_quotelike(), extract_codeblock(), extract_variable(),
1N/Aextract_tagged(), extract_multiple(), gen_delimited_pat(), and
1N/Agen_extract_tagged(). With these, you can implement rather advanced
1N/Aparsing algorithms. See L<Text::Balanced>.
1N/AC<threads>, by Arthur Bergman, is an interface to interpreter threads.
1N/AInterpreter threads (ithreads) is the new thread model introduced in
1N/APerl 5.6 but only available as an internal interface for extension
1N/Awriters (and for Win32 Perl for C<fork()> emulation). See L<threads>,
1N/AL<threads::shared>, and L<perlthrtut>.
1N/AC<threads::shared>, by Arthur Bergman, allows data sharing for
1N/Ainterpreter threads. See L<threads::shared>.
1N/AC<Tie::File>, by Mark-Jason Dominus, associates a Perl array with the
1N/Alines of a file. See L<Tie::File>.
1N/AC<Tie::Memoize>, by Ilya Zakharevich, provides on-demand loaded hashes.
1N/AC<Tie::RefHash::Nestable>, by Edward Avis, allows storing hash
1N/Areferences (unlike the standard Tie::RefHash) The module is contained
1N/Awithin Tie::RefHash. See L<Tie::RefHash>.
1N/AC<Time::HiRes>, by Douglas E. Wegscheid, provides high resolution
1N/Atiming (ualarm, usleep, and gettimeofday). See L<Time::HiRes>.
1N/AC<Unicode::UCD> offers a querying interface to the Unicode Character
1N/ADatabase. See L<Unicode::UCD>.
1N/AC<Unicode::Collate>, by SADAHIRO Tomoyuki, implements the UCA
1N/A(Unicode Collation Algorithm) for sorting Unicode strings.
1N/ASee L<Unicode::Collate>.
1N/AC<Unicode::Normalize>, by SADAHIRO Tomoyuki, implements the various
1N/AUnicode normalization forms. See L<Unicode::Normalize>.
1N/AC<XS::APItest>, by Tim Jenness, is a test extension that exercises XS
1N/AAPIs. Currently only C<printf()> is tested: how to output various
1N/Abasic data types from XS.
1N/AC<XS::Typemap>, by Tim Jenness, is a test extension that exercises
1N/AXS typemaps. Nothing gets installed, but the code is worth studying
1N/Afor extension writers.
1N/A=head2 Updated And Improved Modules and Pragmata
1N/AThe following independently supported modules have been updated to the
1N/Anewest versions from CPAN: CGI, CPAN, DB_File, File::Spec, File::Temp,
1N/AGetopt::Long, Math::BigFloat, Math::BigInt, the podlators bundle
1N/A(Pod::Man, Pod::Text), Pod::LaTeX [561+], Pod::Parser, Storable,
1N/ATerm::ANSIColor, Test, Text-Tabs+Wrap.
1N/Aattributes::reftype() now works on tied arguments.
1N/AAutoLoader can now be disabled with C<no AutoLoader;>.
1N/AB::Deparse has been significantly enhanced by Robin Houston. It can
1N/Anow deparse almost all of the standard test suite (so that the tests
1N/ACarp now has better interface documentation, and the @CARP_NOT
1N/Ainterface has been added to get optional control over where errors
1N/Aare reported independently of @ISA, by Ben Tilly.
1N/AClass::Struct can now define the classes in compile time.
1N/AThe return value of Cwd::fastcwd() is now tainted.
1N/AData::Dumper now has an option to sort hashes.
1N/AData::Dumper now has an option to dump code references
1N/ADB_File now supports newer Berkeley DB versions, among
1N/ADevel::Peek now has an interface for the Perl memory statistics
1N/A(this works only if you are using perl's malloc, and if you have
1N/Acompiled with debugging).
1N/AThe English module can now be used without the infamous performance
1N/A use English '-no_match_vars';
1N/A(Assuming, of course, that you don't need the troublesome variables
1N/AC<$`>, C<$&>, or C<$'>.) Also, introduced C<@LAST_MATCH_START> and
1N/AC<@LAST_MATCH_END> English aliases for C<@-> and C<@+>.
1N/AExtUtils::MakeMaker has been significantly cleaned up and fixed.
1N/AThe enhanced version has also been backported to earlier releases
1N/Aof Perl and submitted to CPAN so that the earlier releases can
1N/Afor sanity much more carefully than before. This may cause new
1N/Awarnings when modules are being installed. See L<ExtUtils::MakeMaker>
1N/AExtUtils::MakeMaker now uses File::Spec internally, which hopefully
1N/Aleads to better portability.
1N/AFcntl, Socket, and Sys::Syslog have been rewritten by Nicholas Clark
1N/Ato use the new-style constant dispatch section (see L<ExtUtils::Constant>).
1N/AThis means that they will be more robust and hopefully faster.
1N/AFile::Find now chdir()s correctly when chasing symbolic links. [561]
1N/AFile::Find now has pre- and post-processing callbacks. It also
1N/Acorrectly changes directories when chasing symbolic links. Callbacks
1N/A(naughtily) exiting with "next;" instead of "return;" now work.
1N/AFile::Find is now (again) reentrant. It also has been made
1N/AThe warnings issued by File::Find now belong to their own category.
1N/AFile::Glob::glob() has been renamed to File::Glob::bsd_glob()
1N/Abecause the name clashes with the builtin glob(). The older
1N/Aname is still available for compatibility, but is deprecated. [561]
1N/AFile::Glob now supports C<GLOB_LIMIT> constant to limit the size of
1N/Athe returned list of filenames.
1N/AIPC::Open3 now allows the use of numeric file descriptors.
1N/AIO::Socket now has an atmark() method, which returns true if the socket
1N/Ais positioned at the out-of-band mark. The method is also exportable
1N/Aas a sockatmark() function.
1N/AIO::Socket::INET failed to open the specified port if the service name
1N/Awas not known. It now correctly uses the supplied port number as is. [561]
1N/AIO::Socket::INET has support for the ReusePort option (if your
1N/Aplatform supports it). The Reuse option now has an alias, ReuseAddr.
1N/AFor clarity, you may want to prefer ReuseAddr.
1N/AIO::Socket::INET now supports a value of zero for C<LocalPort>
1N/A(usually meaning that the operating system will make one up.)
1N/A'use lib' now works identically to @INC. Removing directories
1N/Awith 'no lib' now works.
1N/AMath::BigFloat and Math::BigInt have undergone a full rewrite by Tels.
1N/AThey are now magnitudes faster, and they support various bignum
1N/Alibraries such as GMP and PARI as their backends.
1N/AMath::Complex handles inf, NaN etc., better.
1N/ANet::Ping has been considerably enhanced by Rob Brown: multihoming is
1N/Anow supported, Win32 functionality is better, there is now time
1N/Ameasuring functionality (optionally high-resolution using
1N/ATime::HiRes), and there is now "external" protocol which uses
1N/ANet::Ping::External module which runs your external ping utility and
1N/Aparses the output. A version of Net::Ping::External is available in
1N/ANote that some of the Net::Ping tests are disabled when running
1N/Aunder the Perl distribution since one cannot assume one or more
1N/Aof the following: enabled echo port at localhost, full Internet
1N/Aconnectivity, or sympathetic firewalls. You can set the environment
1N/Avariable PERL_TEST_Net_Ping to "1" (one) before running the Perl test
1N/Asuite to enable all the Net::Ping tests.
1N/APOSIX::sigaction() is now much more flexible and robust.
1N/AYou can now install coderef handlers, 'DEFAULT', and 'IGNORE'
1N/Ahandlers, installing new handlers was not atomic.
1N/AIn Safe, C<%INC> is now localised in a Safe compartment so that
1N/AIn SDBM_File on dosish platforms, some keys went missing because of
1N/Alack of support for files with "holes". A workaround for the problem
1N/AIn Search::Dict one can now have a pre-processing hook for the
1N/Alines being searched.
1N/AThe Shell module now has an OO interface.
1N/AIn Sys::Syslog there is now a failover mechanism that will go
1N/Athrough alternative connection mechanisms until the message
1N/Ais successfully logged.
1N/AThe Test module has been significantly enhanced.
1N/ATime::Local::timelocal() does not handle fractional seconds anymore.
1N/AThe rationale is that neither does localtime(), and timelocal() and
1N/Alocaltime() are supposed to be inverses of each other.
1N/AThe vars pragma now supports declaring fully qualified variables.
1N/A(Something that C<our()> does not and will not support.)
1N/AThe C<utf8::> name space (as in the pragma) provides various
1N/APerl-callable functions to provide low level access to Perl's
1N/Ainternal Unicode representation. At the moment only length()
1N/Ahas been implemented.
1N/A=head1 Utility Changes
1N/AC<enc2xs> is a tool for people adding their own encodings to the
1N/AC<h2ph> now supports C trigraphs.
1N/AC<h2xs> now produces a template README.
1N/AC<h2xs> now uses C<Devel::PPPort> for better portability between
1N/Adifferent versions of Perl.
1N/AC<h2xs> uses the new L<ExtUtils::Constant|ExtUtils::Constant> module
1N/Awhich will affect newly created extensions that define constants.
1N/ASince the new code is more correct (if you have two constants where the
1N/Afirst one is a prefix of the second one, the first constant B<never>
1N/Agot defined), less lossy (it uses integers for integer constant,
1N/Aas opposed to the old code that used floating point numbers even for
1N/Ainteger constants), and slightly faster, you might want to consider
1N/Aregenerating your extension code (the new scheme makes regenerating
1N/Aeasy). L<h2xs> now also supports C trigraphs.
1N/AC<libnetcfg> has been added to configure libnet.
1N/AC<perlbug> is now much more robust. It also sends the bug report to
1N/AC<perlcc> has been rewritten and its user interface (that is,
1N/Acommand line) is much more like that of the UNIX C compiler, cc.
1N/A(The perlbc tools has been removed. Use C<perlcc -B> instead.)
1N/AB<Note that perlcc is still considered very experimental and
1N/AC<perlivp> is a new Installation Verification Procedure utility
1N/Afor running any time after installing Perl.
1N/AC<piconv> is an implementation of the character conversion utility
1N/AC<iconv>, demonstrating the new Encode module.
1N/AC<pod2html> now allows specifying a cache directory.
1N/AC<pod2html> now produces XHTML 1.0.
1N/AC<pod2html> now understands POD written using different line endings
1N/A(PC-like CRLF versus UNIX-like LF versus MacClassic-like CR).
1N/AC<s2p> has been completely rewritten in Perl. (It is in fact a full
1N/Aimplementation of sed in Perl: you can use the sed functionality by
1N/Ausing the C<psed> utility.)
1N/AC<xsubpp> now understands POD documentation embedded in the *.xs
1N/AC<xsubpp> now supports the OUT keyword.
1N/A=head1 New Documentation
1N/Aperl56delta details the changes between the 5.005 release and the
1N/Aperlclib documents the internal replacements for standard C library
1N/Afunctions. (Interesting only for extension writers and Perl core
1N/Aperldebtut is a Perl debugging tutorial. [561+]
1N/Aperlebcdic contains considerations for running Perl on EBCDIC
1N/Aperlintro is a gentle introduction to Perl.
1N/Aperliol documents the internals of PerlIO with layers.
1N/Aperlmodstyle is a style guide for writing modules.
1N/Aperlnewmod tells about writing and submitting a new module. [561+]
1N/Aperlpacktut is a pack() tutorial.
1N/Aperlpod has been rewritten to be clearer and to record the best
1N/Apractices gathered over the years.
1N/Aperlpodspec is a more formal specification of the pod format,
1N/Amainly of interest for writers of pod applications, not to
1N/Apeople writing in pod.
1N/Aperlretut is a regular expression tutorial. [561+]
1N/Aperlrequick is a regular expressions quick-start guide.
1N/AYes, much quicker than perlretut. [561]
1N/Aperltodo has been updated.
1N/Aperltootc has been renamed as perltooc (to not to conflict
1N/Awith perltoot in filesystems restricted to "8.3" names).
1N/Aperluniintro is an introduction to using Unicode in Perl.
1N/A(perlunicode is more of a detailed reference and background
1N/Aperlutil explains the command line utilities packaged with the Perl
1N/AThe following platform-specific documents are available before
1N/Athe installation as
README.I<platform>, and after the installation
1N/A perlaix perlamiga perlapollo perlbeos perlbs2000
1N/A perlce perlcygwin perldgux perldos perlepoc perlfreebsd perlhpux
1N/A perlhurd perlirix perlmachten perlmacos perlmint perlmpeix
1N/A perlnetware perlos2 perlos390 perlplan9 perlqnx perlsolaris
1N/A perltru64 perluts perlvmesa perlvms perlvos perlwin32
1N/AThese documents usually detail one or more of the following subjects:
1N/Aconfiguring, building, testing, installing, and sometimes also using
1N/APerl on the said platform.
1N/AEastern Asian Perl users are now welcomed in their own languages:
1N/AChinese) and
README.tw (traditional Chinese), which are written in
1N/Anormal pod but encoded in EUC-JP, EUC-KR, EUC-CN and Big5. These
1N/Awill get installed as
1N/A perljp perlko perlcn perltw
1N/AThe documentation for the POSIX-BC platform is called "BS2000", to avoid
1N/Aconfusion with the Perl POSIX module.
1N/AThe documentation for the WinCE platform is called perlce (
README.ce 1N/Ain the source code kit), to avoid confusion with the perlwin32
1N/Adocumentation on 8.3-restricted filesystems.
1N/A=head1 Performance Enhancements
1N/Amap() could get pathologically slow when the result list it generates
1N/Ais larger than the source list. The performance has been improved for
1N/Acommon scenarios. [561]
1N/Asort() is also fully reentrant, in the sense that the sort function
1N/Acan itself call sort(). This did not work reliably in previous
1N/Asort() has been changed to use primarily mergesort internally as
1N/Aopposed to the earlier quicksort. For very small lists this may
1N/Aresult in slightly slower sorting times, but in general the speedup
1N/Ashould be at least 20%. Additional bonuses are that the worst case
1N/Abehaviour of sort() is now better (in computer science terms it now
1N/Aruns in time O(N log N), as opposed to quicksort's Theta(N**2)
1N/Aworst-case run time behaviour), and that sort() is now stable
1N/A(meaning that elements with identical keys will stay ordered as they
1N/Awere before the sort). See the C<sort> pragma for information.
1N/AThe story in more detail: suppose you want to serve yourself a little
1N/A @digits = ( 3,1,4,1,5,9 );
1N/AA numerical sort of the digits will yield (1,1,3,4,5,9), as expected.
1N/AWhich C<1> comes first is hard to know, since one C<1> looks pretty
1N/Amuch like any other. You can regard this as totally trivial,
1N/Aor somewhat profound. However, if you just want to sort the even
1N/Adigits ahead of the odd ones, then what will
1N/A sort { ($a % 2) <=> ($b % 2) } @digits;
1N/Ayield? The only even digit, C<4>, will come first. But how about
1N/Athe odd numbers, which all compare equal? With the quicksort algorithm
1N/Aused to implement Perl 5.6 and earlier, the order of ties is left up
1N/Ato the sort. So, as you add more and more digits of Pi, the order
1N/Ain which the sorted even and odd digits appear will change.
1N/Aand, for sufficiently large slices of Pi, the quicksort algorithm
1N/Ain Perl 5.8 won't return the same results even if reinvoked with the
1N/Asame input. The justification for this rests with quicksort's
1N/Aworst case behavior. If you run
1N/A sort { $a <=> $b } ( 1 .. $N , 1 .. $N );
1N/A(something you might approximate if you wanted to merge two sorted
1N/Aarrays using sort), doubling $N doesn't just double the quicksort time,
1N/Ait I<quadruples> it. Quicksort has a worst case run time that can
1N/Agrow like N**2, so-called I<quadratic> behaviour, and it can happen
1N/Aon patterns that may well arise in normal use. You won't notice this
1N/Afor small arrays, but you I<will> notice it with larger arrays,
1N/Aand you may not live long enough for the sort to complete on arrays
1N/Aof a million elements. So the 5.8 quicksort scrambles large arrays
1N/Abefore sorting them, as a statistical defence against quadratic behaviour.
1N/ABut that means if you sort the same large array twice, ties may be
1N/Abroken in different ways.
1N/ABecause of the unpredictability of tie-breaking order, and the quadratic
1N/Aworst-case behaviour, quicksort was I<almost> replaced completely with
1N/Aa stable mergesort. I<Stable> means that ties are broken to preserve
1N/Athe original order of appearance in the input array. So
1N/A sort { ($a % 2) <=> ($b % 2) } (3,1,4,1,5,9);
1N/Awill yield (4,3,1,1,5,9), guaranteed. The even and odd numbers
1N/Aappear in the output in the same order they appeared in the input.
1N/AMergesort has worst case O(N log N) behaviour, the best value
1N/Aattainable. And, ironically, this mergesort does particularly
1N/Awell where quicksort goes quadratic: mergesort sorts (1..$N, 1..$N)
1N/Ain O(N) time. But quicksort was rescued at the last moment because
1N/Ait is faster than mergesort on certain inputs and platforms.
1N/AFor example, if you really I<don't> care about the order of even
1N/Aand odd digits, quicksort will run in O(N) time; it's very good
1N/Aat sorting many repetitions of a small number of distinct elements.
1N/AThe quicksort divide and conquer strategy works well on platforms
1N/Awith relatively small, very fast, caches. Eventually, the problem gets
1N/Awhittled down to one that fits in the cache, from which point it
1N/Abenefits from the increased memory speed.
1N/AQuicksort was rescued by implementing a sort pragma to control aspects
1N/Aof the sort. The B<stable> subpragma forces stable behaviour,
1N/Aregardless of algorithm. The B<_quicksort> and B<_mergesort>
1N/Asubpragmas are heavy-handed ways to select the underlying implementation.
1N/AThe leading C<_> is a reminder that these subpragmas may not survive
1N/Abeyond 5.8. More appropriate mechanisms for selecting the implementation
1N/Aexist, but they wouldn't have arrived in time to save quicksort.
1N/AHashes now use Bob Jenkins "One-at-a-Time" hashing key algorithm
1N/Areasonably fast while producing a much better spread of values than
1N/Athe old hashing algorithm (originally by Chris Torek, later tweaked by
1N/AIlya Zakharevich). Hash values output from the algorithm on a hash of
1N/Aall 3-char printable ASCII keys comes much closer to passing the
1N/ADIEHARD random number generation tests. According to perlbench, this
1N/Achange has not affected the overall speed of Perl.
1N/Aunshift() should now be noticeably faster.
1N/A=head1 Installation and Configuration Improvements
1N/A=head2 Generic Improvements
1N/AINSTALL now explains how you can configure Perl to use 64-bit
1N/Aintegers even on non-64-bit platforms.
1N/A(see INSTALL) and you use Configure -Dprefix=
/foo/bar and in the old
1N/APolicy $prefix eq $siteprefix and $prefix eq $vendorprefix, all of
1N/Athem will now be changed to the new prefix,
/foo/bar. (Previously
1N/Aonly $prefix changed.) If you do not like this new behaviour,
1N/Aspecify prefix, siteprefix, and vendorprefix explicitly.
1N/AA new optional location for Perl libraries, otherlibdirs, is available.
1N/AIt can be used for example for vendor add-ons without disturbing Perl's
1N/Aown library directories.
1N/AIn many platforms, the vendor-supplied 'cc' is too stripped-down to
1N/Abuild Perl (basically, 'cc' doesn't do ANSI C). If this seems
1N/Ato be the case and 'cc' does not seem to be the GNU C compiler
1N/A'gcc', an automatic attempt is made to find and use 'gcc' instead.
1N/Agcc needs to closely track the operating system release to avoid
1N/Abuild problems. If Configure finds that gcc was built for a different
1N/Aoperating system release than is running, it now gives a clearly visible
1N/Awarning that there may be trouble ahead.
1N/ASince Perl 5.8 is not binary-compatible with previous releases
1N/Aof Perl, Configure no longer suggests including the 5.005
1N/AConfigure C<-S> can now run non-interactively. [561]
1N/AConfigure support for pdp11-style memory models has been removed due
1N/Ato obsolescence. [561]
1N/Ainstallperl now outputs everything to STDERR.
1N/ABecause PerlIO is now the default on most platforms, "-perlio" doesn't
1N/Aget appended to the $Config{archname} (also known as $^O) anymore.
1N/AInstead, if you explicitly choose not to use perlio (Configure command
1N/Aline option -Uuseperlio), you will get "-stdio" appended.
1N/AAnother change related to the architecture name is that "-64all"
1N/A(-Duse64bitall, or "maximally 64-bit") is appended only if your
1N/Apointers are 64 bits wide. (To be exact, the use64bitall is ignored.)
1N/AIn AFS installations, one can configure the root of the AFS to be
1N/Asomewhere else than the default F</afs> by using the Configure
1N/AAPPLLIB_EXP, a lesser-known configuration-time definition, has been
1N/Adocumented. It can be used to prepend site-specific directories
1N/Ato Perl's default search path (@INC); see INSTALL for information.
1N/AThe version of Berkeley DB used when the Perl (and, presumably, the
1N/ADB_File extension) was built is now available as
1N/AC<@Config{qw(db_version_major db_version_minor db_version_patch)}>
1N/Afrom Perl and as C<DB_VERSION_MAJOR_CFG DB_VERSION_MINOR_CFG
1N/ADB_VERSION_PATCH_CFG> from C.
1N/ABuilding Berkeley DB3 for compatibility modes for DB, NDBM, and ODBM
1N/Ahas been documented in INSTALL.
1N/AIf you have CPAN access (either network or a local copy such as a
1N/ACD-ROM) you can during specify extra modules to Configure to build and
1N/Ainstall with Perl using the -Dextras=... option. See INSTALL for
1N/Aavailable. This file is supposed to be used by hints file writers
1N/Afor site-wide changes).
1N/AIf your file system supports symbolic links, you can build Perl outside
1N/Aof the source directory by
1N/Aunaffected. After Configure has finished, you can just say
1N/AFor Perl developers, several new make targets for profiling
1N/Aand debugging have been added; see L<perlhack>.
1N/AUse of the F<gprof> tool to profile Perl has been documented in
1N/Agenerating a gprofiled Perl executable.
1N/AIf you have GCC 3, there is a make target called "
perl.gcov" for
1N/Acreating a gcoved Perl executable for coverage analysis. See
1N/Ahave been added; see L<perlhack> for more information about pixie and
1N/AGuidelines of how to construct minimal Perl installations have
1N/Abeen added to INSTALL.
1N/AThe Thread extension is now not built at all under ithreads
1N/A(C<Configure -Duseithreads>) because it wouldn't work anyway (the
1N/AThread extension requires being Configured with C<-Duse5005threads>).
1N/AB<Note that the 5.005 threads are unsupported and deprecated: if you
1N/Ahave code written for the old threads you should migrate it to the
1N/AThe Gconvert macro ($Config{d_Gconvert}) used by perl for stringifying
1N/Afloating-point numbers is now more picky about using sprintf %.*g
1N/Arules for the conversion. Some platforms that used to use gcvt may
1N/Anow resort to the slower sprintf.
1N/AThe obsolete method of making a special (
e.g., debugging) flavor
1N/Ahas been removed. Use -DDEBUGGING instead.
1N/A=head2 New Or Improved Platforms
1N/AFor the list of platforms known to support Perl,
1N/Asee L<perlport/"Supported Platforms">.
1N/AAIX dynamic loading should be now better supported.
1N/AAIX should now work better with gcc, threads, and 64-bitness. Also the
1N/Along doubles support in AIX should be better now. See L<perlaix>.
1N/ABeOS has been reclaimed.
1N/AThe
DG/UX platform now supports 5.005-style threads.
1N/AThe
DYNIX/ptx platform (also known as dynixptx) is supported at or
1N/AEBCDIC platforms (
z/OS (also known as OS/390), POSIX-BC, and
VM/ESA)
1N/Ahave been regained. Many test suite tests still fail and the
1N/Aco-existence of Unicode and EBCDIC isn't quite settled, but the
1N/Asituation is much better than with Perl 5.6. See L<perlos390>,
1N/AL<perlbs2000> (for POSIX-BC), and L<perlvmesa> for more information.
1N/ABuilding perl with -Duseithreads or -Duse5005threads now works under
1N/AHP-UX 10.20 (previously it only worked under 10.30 or later). You will
1N/AMac OS Classic is now supported in the mainstream source package
1N/A(MacPerl has of course been available since perl 5.004 but now the
1N/Asource code bases of standard Perl and MacPerl have been synchronised)
1N/AMac OS X (or Darwin) should now be able to build Perl even on HFS+
1N/Afilesystems. (The case-insensitivity used to confuse the Perl build
1N/ANCR MP-RAS is now supported. [561]
1N/AAll the NetBSD specific patches (except for the installation
1N/Aspecific ones) have been merged back to the main distribution.
1N/ANetWare from Novell is now supported. See L<perlnetware>.
1N/ANonStop-UX is now supported. [561]
1N/ANEC SUPER-UX is now supported.
1N/AAll the OpenBSD specific patches (except for the installation
1N/Aspecific ones) have been merged back to the main distribution.
1N/APerl has been tested with the GNU pth userlevel thread package
1N/Aof Perl now work, but not without adding some yield()s to the tests,
1N/Aso while pth (and other userlevel thread implementations) can be
1N/Aconsidered to be "working" with Perl ithreads, keep in mind the
1N/Apossible non-preemptability of the underlying thread implementation.
1N/AStratus VOS is now supported using Perl's native build method
1N/A(Configure). This is the recommended method to build Perl on
1N/AVOS. The older methods, which build miniperl, are still
1N/Aavailable. See L<perlvos>. [561+]
1N/AThe Amdahl UTS UNIX mainframe platform is now supported. [561]
1N/AWinCE is now supported. See L<perlce>.
1N/Az/OS (formerly known as OS/390, formerly known as MVS OE) now has
1N/Asupport for dynamic loading. This is not selected by default,
1N/Ahowever, you must specify -Dusedl in the arguments of Configure. [561]
1N/A=head1 Selected Bug Fixes
1N/ANumerous memory leaks and uninitialized memory accesses have been
1N/Ahunted down. Most importantly, anonymous subs used to leak quite
1N/AThe autouse pragma didn't work for Multi::Part::Function::Names.
1N/Acaller() could cause core dumps in certain situations. Carp was
1N/Asometimes affected by this problem. In particular, caller() now
1N/Areturns a subroutine name of C<(unknown)> for subroutines that have
1N/Abeen removed from the symbol table.
1N/Achop(@list) in list context returned the characters chopped in
1N/Areverse order. This has been reversed to be in the right order. [561]
1N/AConfigure no longer includes the DBM libraries (dbm, gdbm, db, ndbm)
1N/Awhen building the Perl binary. The only exception to this is SunOS
4.x,
1N/Awhich needs them. [561]
1N/AThe behaviour of non-decimal but numeric string constants such as
1N/A"0x23" was platform-dependent: in some platforms that was seen as 35,
1N/Ain some as 0, in some as a floating point number (don't ask). This
1N/Awas caused by Perl's using the operating system libraries in a situation
1N/Awhere the result of the string to number conversion is undefined: now
1N/APerl consistently handles such strings as zero in numeric contexts.
1N/ASeveral debugger fixes: exit code now reflects the script exit code,
1N/Acondition C<"0"> now treated correctly, the C<d> command now checks
1N/Aline number, C<$.> no longer gets corrupted, and all debugger output
1N/Anow goes correctly to the socket if RemotePort is set. [561]
1N/Aconsistent commands interface, via (CommandSet=580).
perl5db.t was
1N/Aalso added to test the changes, and as a placeholder for further tests.
1N/AThe debugger has a new C<dumpDepth> option to control the maximum
1N/Adepth to which nested structures are dumped. The C<x> command has
1N/Abeen extended so that C<x N EXPR> dumps out the value of I<EXPR> to a
1N/Adepth of at most I<N> levels.
1N/AThe debugger can now show lexical variables if you have the CPAN
1N/Amodule PadWalker installed.
1N/AThe order of DESTROYs has been made more predictable.
1N/APerl 5.6.0 could emit spurious warnings about redefinition of
1N/Adl_error() when statically building extensions into perl.
1N/AThis has been corrected. [561]
1N/AL<dprofpp> -R didn't work.
1N/AC<*foo{FORMAT}> now works.
1N/AInfinity is now recognized as a number.
1N/AUNIVERSAL::isa no longer caches methods incorrectly. (This broke
1N/Athe Tk extension with 5.6.0.) [561]
1N/ALexicals I: lexicals outside an eval "" weren't resolved
1N/Acorrectly inside a subroutine definition inside the eval "" if they
1N/Awere not already referenced in the top level of the eval""ed code.
1N/ALexicals II: lexicals leaked at file scope into subroutines that
1N/Awere declared before the lexicals.
1N/ALexical warnings now propagating correctly between scopes
1N/Aand into C<eval "...">.
1N/AC<use warnings qw(FATAL all)> did not work as intended. This has been
1N/Awarnings::enabled() now reports the state of $^W correctly if the caller
1N/Aisn't using lexical warnings. [561]
1N/ALine renumbering with eval and C<#line> now works. [561]
1N/AFixed numerous memory leaks, especially in eval "".
1N/ALocalised tied variables no longer leak memory
1N/A tie my %tied_hash => 'Tie::StdHash';
1N/A # Used to leak memory every time local() was called;
1N/A # in a loop, this added up.
1N/A local($tied_hash{Foo}) = 1;
1N/ALocalised hash elements (and %ENV) are correctly unlocalised to not
1N/Aexist, if they didn't before they were localised.
1N/A tie my %tied_hash => 'Tie::StdHash';
1N/A # Nothing has set the FOO element so far
1N/A { local $tied_hash{FOO} = 'Bar' }
1N/A # This used to print, but not now.
1N/A print "exists!\n" if exists $tied_hash{FOO};
1N/AAs a side effect of this fix, tied hash interfaces B<must> define
1N/Athe EXISTS and DELETE methods.
1N/Amkdir() now ignores trailing slashes in the directory name,
1N/Aas mandated by POSIX.
1N/ASome versions of glibc have a broken modfl(). This affects builds
1N/Awith C<-Duselongdouble>. This version of Perl detects this brokenness
1N/Aand has a workaround for it. The glibc release 2.2.2 is known to have
1N/Afixed the modfl() bug.
1N/AModulus of unsigned numbers now works (4063328477 % 65535 used to
1N/Areturn 27406, instead of 27047). [561]
1N/ASome "not a number" warnings introduced in 5.6.0 eliminated to be
1N/Amore compatible with 5.005. Infinity is now recognised as a number. [561]
1N/ANumeric conversions did not recognize changes in the string value
1N/Aproperly in certain circumstances. [561]
1N/AAttributes (such as :shared) didn't work with our().
1N/Aour() variables will not cause bogus "Variable will not stay shared"
1N/A"our" variables of the same name declared in two sibling blocks
1N/Aresulted in bogus warnings about "redeclaration" of the variables.
1N/AThe problem has been corrected. [561]
1N/Apack "Z" now correctly terminates the string with "\0".
1N/AFix password routines which in some shadow password platforms
1N/A(
e.g. HP-UX) caused getpwent() to return every other entry.
1N/AThe PERL5OPT environment variable (for passing command line arguments
1N/Ato Perl) didn't work for more than a single group of options. [561]
1N/APERL5OPT with embedded spaces didn't work.
1N/Aprintf() no longer resets the numeric locale to "C".
1N/AC<qw(a\\b)> now parses correctly as C<'a\\b'>: that is, as three
1N/Acharacters, not four. [561]
1N/Apos() did not return the correct value within s///ge in earlier
1N/Aversions. This is now handled correctly. [561]
1N/Awithout the q L ll prefixes (assuming you are on a quad-capable platform).
1N/ARegular expressions on references and overloaded scalars now work. [561+]
1N/ARight-hand side magic (GMAGIC) could in many cases such as string
1N/Aconcatenation be invoked too many times.
1N/Ascalar() now forces scalar context even when used in void context.
1N/ASOCKS support is now much more robust.
1N/Asort() arguments are now compiled in the right wantarray context
1N/A(they were accidentally using the context of the sort() itself).
1N/AThe comparison block is now run in scalar context, and the arguments
1N/Ato be sorted are always provided list context. [561]
1N/AChanged the POSIX character class C<[[:space:]]> to include the (very
1N/Ararely used) vertical tab character. Added a new POSIX-ish character
1N/Aclass C<[[:blank:]]> which stands for horizontal whitespace
1N/A(currently, the space and the tab).
1N/AThe tainting behaviour of sprintf() has been rationalized. It does
1N/Anot taint the result of floating point formats anymore, making the
1N/Abehaviour consistent with that of string interpolation. [561]
1N/ASome cases of inconsistent taint propagation (such as within hash
1N/Avalues) have been fixed.
1N/AThe RE engine found in Perl 5.6.0 accidentally pessimised certain kinds
1N/Aof simple pattern matches. These are now handled better. [561]
1N/ARegular expression debug output (whether through C<use re 'debug'>
1N/Aor via C<-Dr>) now looks better. [561]
1N/AMulti-line matches like C<"a\nxb\n" =~ /(?!\A)x/m> were flawed. The
1N/Abug has been fixed. [561]
1N/AUse of $& could trigger a core dump under some situations. This
1N/Ais now avoided. [561]
1N/AThe regular expression captured submatches ($1, $2, ...) are now
1N/Amore consistently unset if the match fails, instead of leaving false
1N/Adata lying around in them. [561]
1N/Areadline() on files opened in "slurp" mode could return an extra
1N/A"" (blank line) at the end in certain situations. This has been
1N/AAutovivification of symbolic references of special variables described
1N/Ain L<perlvar> (as in C<${$num}>) was accidentally disabled. This works
1N/ASys::Syslog ignored the C<LOG_AUTH> constant.
1N/A$AUTOLOAD, sort(), lock(), and spawning subprocesses
1N/Ain multiple threads simultaneously are now thread-safe.
1N/ATie::Array's SPLICE method was broken.
1N/AAllow a read-only string on the left-hand side of a non-modifying tr///.
1N/AIf C<STDERR> is tied, warnings caused by C<warn> and C<die> now
1N/Acorrectly pass to it.
1N/ASeveral Unicode fixes.
1N/ABOMs (byte order marks) at the beginning of Perl files
1N/A(scripts, modules) should now be transparently skipped.
1N/AUTF-16 and UCS-2 encoded Perl files should now be read correctly.
1N/AThe character tables have been updated to Unicode 3.2.0.
1N/AComparing with utf8 data does not magically upgrade non-utf8 data
1N/Ainto utf8. (This was a problem for example if you were mixing data
1N/Afrom I/O and Unicode data: your output might have got magically encoded
1N/AGenerating illegal Unicode code points such as U+FFFE, or the UTF-16
1N/Asurrogates, now also generates an optional warning.
1N/AC<IsAlnum>, C<IsAlpha>, and C<IsWord> now match titlecase.
1N/AConcatenation with the C<.> operator or via variable interpolation,
1N/AC<eq>, C<substr>, C<reverse>, C<quotemeta>, the C<x> operator,
1N/Asubstitution with C<s///>, single-quoted UTF-8, should now work.
1N/AThe C<tr///> operator now works. Note that the C<tr///CU>
1N/Afunctionality has been removed (but see pack('U0', ...)).
1N/AC<eval "v200"> now works.
1N/APerl 5.6.0 parsed m/\x{ab}/ incorrectly, leading to spurious warnings.
1N/AThis has been corrected. [561]
1N/AZero entries were missing from the Unicode classes such as C<IsDigit>.
1N/ALarge unsigned numbers (those above 2**31) could sometimes lose their
1N/Aunsignedness, causing bogus results in arithmetic operations. [561]
1N/AThe Perl parser has been stress tested using both random input and
1N/AMarkov chain input and the few found crashes and lockups have been
1N/A=head2 Platform Specific Changes and Fixes
1N/ASetting C<$0> now works (as much as possible; see L<perlvar> for details).
1N/ANumerous updates; currently synchronised with Cygwin 1.3.10.
1N/APreviously
DYNIX/ptx had problems in its Configure probe for non-blocking I/O.
1N/APerl now works on post-3.0 FreeBSDs.
1N/Anow uses HP-UX malloc instead of Perl malloc.
1N/ANumerous compilation flag and hint enhancements; accidental mixing
1N/Aof 32-bit and 64-bit libraries (a doomed attempt) made much harder.
1N/ALong doubles should now work (see INSTALL). [561]
1N/ALinux previously had problems related to sockaddrlen when using
1N/Aaccept(), recvfrom() (in Perl: recv()), getpeername(), and
1N/ACompilation of the standard Perl distribution in Mac OS Classic should
1N/Anow work if you have the Metrowerks development environment and the
1N/Amissing Mac-specific toolkit bits. Contact the macperl mailing list
1N/Aand Configure with -Duseithreads.
1N/ANow works with usethreads (see INSTALL). [561]
1N/A64-bitness using the Sun Workshop compiler now works.
1N/AThe native build method requires at least VOS Release 14.5.0
1N/Aand GNU C++/GNU Tools 2.0.1 or later. The Perl pack function
1N/Anow maps overflowed values to +infinity and underflowed values
1N/ATru64 (aka Digital UNIX, aka DEC OSF/1)
1N/AThe operating system version letter now recorded in $Config{osvers}.
1N/AAllow compiling with gcc (previously explicitly forbidden). Compiling
1N/Awith gcc still not recommended because buggy code results, even with
1N/AFixed various alignment problems that lead into core dumps either
1N/Aduring build or later; no longer dies on math errors at runtime;
1N/Anow using full quad integers (64 bits), previously was using
1N/Aonly 46 bit integers for speed.
1N/ASee L</"Socket Extension Dynamic in VMS"> and L</"IEEE-format Floating Point
1N/ADefault on OpenVMS Alpha"> for important changes not otherwise listed here.
1N/Achdir() now works better despite a CRT bug; now works with MULTIPLICITY
1N/A(see INSTALL); now works with Perl's malloc.
1N/AThe tainting of C<%ENV> elements via C<keys> or C<values> was previously
1N/Aunimplemented. It now works as documented.
1N/AThe C<waitpid> emulation has been improved. The worst bug (now fixed)
1N/Awas that a pid of -1 would cause a wildcard search of all processes on
1N/APOSIX-style signals are now emulated much better on VMS versions prior
1N/AThe C<system> function and backticks operator have improved
1N/Afunctionality and better error handling. [561]
1N/AFile access tests now use current process privileges rather than the
1N/Auser's default privileges, which could sometimes result in a mismatch
1N/Abetween reported access and actual access. This improvement is only
1N/Aavailable on VMS v6.0 and later.
1N/AThere is a new C<kill> implementation based on C<sys$sigprc> that allows
1N/Aolder VMS systems (pre-7.0) to use C<kill> to send signals rather than
1N/Asimply force exit. This implementation also allows later systems to
1N/Acall C<kill> from within a signal handler.
1N/AIterative logical name translations are now limited to 10 iterations in
1N/Aimitation of SHOW LOGICAL and other OpenVMS facilities.
1N/ASignal handling now works better than it used to. It is now implemented
1N/Ausing a Windows message loop, and is therefore less prone to random
1N/Afork() emulation is now more robust, but still continues to have a few
1N/Aesoteric bugs and caveats. See L<perlfork> for details. [561+]
1N/AA failed (pseudo)fork now returns undef and sets errno to EAGAIN. [561]
1N/AThe following modules now work on Windows:
1N/A ExtUtils::Embed [561]
1N/AIO::File::new_tmpfile() is no longer limited to 32767 invocations
1N/ABetter chdir() return value for a non-existent directory.
1N/ACompiling perl using the 64-bit Platform SDK tools is now supported.
1N/AThe Win32::SetChildShowWindow() builtin can be used to control the
1N/Avisibility of windows created by child processes. See L<Win32> for
1N/ANon-blocking waits for child processes (or pseudo-processes) are
1N/Asupported via C<waitpid($pid, &POSIX::WNOHANG)>.
1N/AThe behavior of system() with multiple arguments has been rationalized.
1N/AEach unquoted argument will be automatically quoted to protect whitespace,
1N/Aand any existing whitespace in the arguments will be preserved. This
1N/Aimproves the portability of system(@args) by avoiding the need for
1N/AWindows C<cmd> shell specific quoting in perl programs.
1N/ANote that this means that some scripts that may have relied on earlier
1N/Abuggy behavior may no longer work correctly. For example,
1N/AC<system("nmake /nologo", @args)> will now attempt to run the file
1N/AC<nmake /nologo> and will fail when such a file isn't found.
1N/AOn the other hand, perl will now execute code such as
1N/AThe perl header files no longer suppress common warnings from the
1N/AMicrosoft Visual C++ compiler. This means that additional warnings may
1N/Anow show up when compiling XS code.
1N/ABorland C++ v5.5 is now a supported compiler that can build Perl.
1N/AHowever, the generated binaries continue to be incompatible with those
1N/Agenerated by the other supported compilers (GCC and Visual C++). [561]
1N/ADuping socket handles with open(F, ">&MYSOCK") now works under Windows 9x.
1N/ACurrent directory entries in %ENV are now correctly propagated to child
1N/ANew %ENV entries now propagate to subprocesses. [561]
1N/AWin32::GetCwd() correctly returns C:\ instead of C: when at the drive root.
1N/AOther bugs in chdir() and Cwd::cwd() have also been fixed. [561]
1N/AThe makefiles now default to the features enabled in ActiveState ActivePerl
1N/A(a popular Win32 binary distribution). [561]
1N/AHTML files will now be installed in c:\perl\html instead of
1N/AREG_EXPAND_SZ keys are now allowed in registry settings used by perl. [561]
1N/ACan now send() from all threads, not just the first one. [561]
1N/AExtUtils::MakeMaker now uses $ENV{LIB} to search for libraries. [561]
1N/ALess stack reserved per thread so that more threads can run
1N/Aconcurrently. (Still 16M per thread.) [561]
1N/AC<< File::Spec->tmpdir() >> now prefers C:/temp over /tmp
1N/A(works better when perl is running as service).
1N/ABetter UNC path handling under ithreads. [561]
1N/Await(), waitpid(), and backticks now return the correct exit status
1N/Aunder Windows 9x. [561]
1N/AA socket handle leak in accept() has been fixed. [561]
1N/A=head1 New or Changed Diagnostics
1N/APlease see L<perldiag> for more details.
1N/AAmbiguous range in the transliteration operator (like a-z-9) now
1N/Achdir("") and chdir(undef) now give a deprecation warning because they
1N/Acause a possible unintentional chdir to the home directory.
1N/ASay chdir() if you really mean that.
1N/ATwo new debugging options have been added: if you have compiled your
1N/APerl with debugging, you can use the -DT [561] and -DR options to trace
1N/Atokenising and to add reference counts to displaying variables,
1N/AThe lexical warnings category "deprecated" is no longer a sub-category
1N/Aof the "syntax" category. It is now a top-level category in its own
1N/AUnadorned dump() will now give a warning suggesting to
1N/Ause explicit CORE::dump() if that's what really is meant.
1N/AThe "Unrecognized escape" warning has been extended to include C<\8>,
1N/AC<\9>, and C<\_>. There is no need to escape any of the C<\w> characters.
1N/AAll regular expression compilation error messages are now hopefully
1N/Aeasier to understand both because the error message now comes before
1N/Athe failed regex and because the point of failure is now clearly
1N/Amarked by a C<E<lt>-- HERE> marker.
1N/AVarious I/O (and socket) functions like binmode(), close(), and so
1N/Aforth now more consistently warn if they are used illogically either
1N/Aon a yet unopened or on an already closed filehandle (or socket).
1N/AUsing lstat() on a filehandle now gives a warning. (It's a non-sensical
1N/AThe C<-M> and C<-m> options now warn if you didn't supply the module name.
1N/AIf you in C<use> specify a required minimum version, modules matching
1N/Athe name and but not defining a $VERSION will cause a fatal failure.
1N/AUsing negative offset for vec() in lvalue context is now a warnable offense.
1N/AOdd number of arguments to overload::constant now elicits a warning.
1N/AOdd number of elements in anonymous hash now elicits a warning.
1N/AThe various "opened only for", "on closed", "never opened" warnings
1N/Adrop the C<main::> prefix for filehandles in the C<main> package,
1N/Afor example C<STDIN> instead of C<main::STDIN>.
1N/ASubroutine prototypes are now checked more carefully, you may
1N/Aget warnings for example if you have used non-prototype characters.
1N/AIf an attempt to use a (non-blessed) reference as an array index
1N/Ais made, a warning is given.
1N/AC<push @a;> and C<unshift @a;> (with no values to push or unshift)
1N/Anow give a warning. This may be a problem for generated and evaled
1N/Ausing the C<"C"> format you will get an optional warning. Similarly
1N/Afor the C<"c"> format and a number less than -128 or more than 127.
1N/Apack C<P> format now demands an explicit size.
1N/Aunpack C<w> now warns of unterminated compressed integers.
1N/AWarnings relating to the use of PerlIO have been added.
1N/ACertain regex modifiers such as C<(?o)> make sense only if applied to
1N/Athe entire regex. You will get an optional warning if you try to do
1N/AVariable length lookbehind has not yet been implemented, trying to
1N/Ause it will tell that.
1N/AUsing arrays or hashes as references (
e.g. C<< %foo->{bar} >>
1N/Ahas been deprecated for a while. Now you will get an optional warning.
1N/AWarnings relating to the use of the new restricted hashes feature
1N/ASelf-ties of arrays and hashes are not supported and fatal errors
1N/Awill happen even at an attempt to do so.
1N/AUsing C<sort> in scalar context now issues an optional warning.
1N/AThis didn't do anything useful, as the sort was not performed.
1N/AUsing the /g modifier in split() is meaningless and will cause a warning.
1N/AUsing splice() past the end of an array now causes a warning.
1N/AMalformed Unicode encodings (UTF-8 and UTF-16) cause a lot of warnings,
1N/Aad doestrying to use UTF-16 surrogates (which are unimplemented).
1N/ATrying to use Unicode characters on an I/O stream without marking the
1N/Astream's encoding (using open() or binmode()) will cause "Wide character"
1N/AUse of v-strings in
use/require causes a (backward) portability warning.
1N/AWarnings relating to the use interpreter threads and their shared data
1N/A=head1 Changed Internals
1N/APerlIO is now the default.
1N/AYou can now build a really minimal perl called microperl.
1N/ABuilding microperl does not require even running Configure;
1N/Amany assumptions, some of which may be too bold; the resulting
1N/Aexecutable may crash or otherwise misbehave in wondrous ways.
1N/AFor careful hackers only.
1N/AAdded rsignal(), whichsig(), do_join(), op_clear, op_null,
1N/Aptr_table_clear(), ptr_table_free(), sv_setref_uv(), and several UTF-8
1N/Ainterfaces to the publicised API. For the full list of the available
1N/AMade possible to propagate customised exceptions via croak()ing.
1N/ANow xsubs can have attributes just like subs. (Well, at least the
1N/Abuilt-in attributes.)
1N/AdTHR and djSP have been obsoleted; the former removed (because it's
1N/Aa no-op) and the latter replaced with dSP.
1N/APERL_OBJECT has been completely removed.
1N/AThe MAGIC constants (
e.g. C<'P'>) have been macrofied
1N/A(
e.g. C<PERL_MAGIC_TIED>) for better source code readability
1N/AThe regex compiler now maintains a structure that identifies nodes in
1N/Athe compiled bytecode with the corresponding syntactic features of the
1N/Aoriginal regex expression. The information is attached to the new
1N/AC<offsets> member of the C<struct regexp>. See L<perldebguts> for more
1N/Acomplete information.
1N/AThe C code has been made much more C<gcc -Wall> clean. Some warning
1N/Amessages still remain in some platforms, so if you are compiling with
1N/Agcc you may see some warnings about dubious practices. The warnings
1N/ADocumentation on how to use the Perl source repository has been added
1N/AThere are now several profiling make targets.
1N/A=head1 Security Vulnerability Closed [561]
1N/A(This change was already made in 5.7.0 but bears repeating here.)
1N/A(5.7.0 came out before 5.6.1: the development branch 5.7 released
1N/Aearlier than the maintenance branch 5.6)
1N/AA potential security vulnerability in the optional suidperl component
1N/Aof Perl was identified in August 2000. suidperl is neither built nor
1N/Ainstalled by default. As of November 2001 the only known vulnerable
1N/Aplatform is Linux, most likely all Linux distributions. CERT and
1N/Avarious vendors and distributors have been alerted about the vulnerability.
1N/Afor more information.
1N/AThe problem was caused by Perl trying to report a suspected security
1N/Aexploit attempt using an external program,
/bin/mail. On Linux
1N/Aplatforms the
/bin/mail program had an undocumented feature which
1N/Awhen combined with suidperl gave access to a root shell, resulting in
1N/Aa serious compromise instead of reporting the exploit attempt. If you
1N/Adon't have
/bin/mail, or if you have 'safe setuid scripts', or if
1N/Asuidperl is not installed, you are safe.
1N/AThe exploit attempt reporting feature has been completely removed from
1N/APerl 5.8.0 (and the maintenance release 5.6.1, and it was removed also
1N/Afrom all the Perl 5.7 releases), so that particular vulnerability
1N/Aisn't there anymore. However, further security vulnerabilities are,
1N/Aunfortunately, always possible. The suidperl functionality is most
1N/Aprobably going to be removed in Perl 5.10. In any case, suidperl
1N/Ashould only be used by security experts who know exactly what they are
1N/Adoing and why they are using suidperl instead of some other solution
1N/ASeveral new tests have been added, especially for the F<lib> and
1N/AF<ext> subsections. There are now about 69 000 individual tests
1N/A(spread over about 700 test scripts), in the regression suite (5.6.1
1N/Ahas about 11 700 tests, in 258 test scripts) The exact numbers depend
1N/Aon the platform and Perl configuration used. Many of the new tests
1N/Aare of course introduced by the new modules, but still in general Perl
1N/Ais now more thoroughly tested.
1N/ABecause of the large number of tests, running the regression suite
1N/Awill take considerably longer time than it used to: expect the suite
1N/Ato take up to 4-5 times longer to run than in perl 5.6. On a really
1N/Afast machine you can hope to finish the suite in about 6-8 minutes
1N/AThe tests are now reported in a different order than in earlier Perls.
1N/A(This happens because the test scripts from under
t/lib have been moved
1N/A=head1 Known Problems
1N/A=head2 The Compiler Suite Is Still Very Experimental
1N/AThe compiler suite is slowly getting better but it continues to be
1N/Ahighly experimental. Use in production environments is discouraged.
1N/A=head2 Localising Tied Arrays and Hashes Is Broken
1N/Adoesn't work as one would expect: the old value is restored
1N/Aincorrectly. This will be changed in a future release, but we don't
1N/Aknow yet what the new semantics will exactly be. In any case, the
1N/Achange will break existing code that relies on the current
1N/A(ill-defined) semantics, so just avoid doing this in general.
1N/A=head2 Building Extensions Can Fail Because Of Largefiles
1N/ASome extensions like mod_perl are known to have issues with
1N/A`largefiles', a change brought by Perl 5.6.0 in which file offsets
1N/Adefault to 64 bits wide, where supported. Modules may fail to compile
1N/Aat all, or they may compile and work incorrectly. Currently, there
1N/Ais no good solution for the problem, but Configure now provides
1N/Aappropriate non-largefile ccflags, ldflags, libswanted, and libs
1N/Ain the %Config hash (
e.g., $Config{ccflags_nolargefiles}) so the
1N/Aextensions that are having problems can try configuring themselves
1N/Awithout the largefileness. This is admittedly not a clean solution,
1N/Aand the solution may not even work at all. One potential failure is
1N/Awhether one can (or, if one can, whether it's a good idea to) link
1N/Atogether at all binaries with different ideas about file offsets;
1N/Aall this is platform-dependent.
1N/A=head2 Modifying $_ Inside for(..)
1N/Aworks without complaint. It shouldn't. (You should be able to
1N/Amodify only lvalue elements inside the loops.) You can see the
1N/Acorrect behaviour by replacing the 1..5 with 1, 2, 3, 4, 5.
1N/A=head2 mod_perl 1.26 Doesn't Build With Threaded Perl
1N/AUse mod_perl 1.27 or higher.
1N/ADon't panic. Read the 'make test' section of INSTALL instead.
1N/AUse libwww-perl 5.65 or later.
1N/A=head2 PDL failing some tests
1N/AUse PDL 2.3.4 or later.
1N/AYou may get errors like 'Undefined symbol "Perl_get_sv"' or "can't
1N/Aresolve symbol 'Perl_get_sv'", or the symbol may be "Perl_sv_2pv".
1N/AThis probably means that you are trying to use an older shared Perl
1N/Alibrary (or extensions linked with such) with Perl 5.8.0 executable.
1N/APerl used to have such a subroutine, but that is no more the case.
1N/ACheck your shared library path, and any shared Perl libraries in those
1N/ASometimes this problem may also indicate a partial Perl 5.8.0
1N/Ainstallation, see L</"Mac OS X dyld undefined symbols"> for an
1N/Aexample and how to deal with it.
1N/A=head2 Self-tying Problems
1N/ASelf-tying of arrays and hashes is broken in rather deep and
1N/Ahard-to-fix ways. As a stop-gap measure to avoid people from getting
1N/Afrustrated at the mysterious results (core dumps, most often), it is
1N/Aforbidden for now (you will get a fatal error even from an attempt).
1N/AA change to self-tying of globs has caused them to be recursively
1N/Areferenced (see: L<perlobj/"Two-Phased Garbage Collection">). You
1N/Awill now need an explicit untie to destroy a self-tied glob. This
1N/Abehaviour may be fixed at a later date.
1N/ASelf-tying of scalars and IO thingies works.
1N/AIf this test fails, it indicates that your libc (C library) is not
1N/Athreadsafe. This particular test stress tests the localtime() call to
1N/Afind out whether it is threadsafe. See L<perlthrtut> for more information.
1N/A=head2 Failure of Thread (5.005-style) tests
1N/AB<Note that support for 5.005-style threading is deprecated,
1N/Aexperimental and practically unsupported. In 5.10, it is expected
1N/Ato be removed. You should migrate your code to ithreads.>
1N/AThe following tests are known to fail due to fundamental problems in
1N/Athe 5.005 threading implementation. These are not new failures--Perl
1N/A5.005_0x has the same bugs, but didn't have these tests.
1N/AThese failures are unlikely to get fixed as 5.005-style threads
1N/Aare considered fundamentally broken. (Basically what happens is that
1N/Acompeting threads can corrupt shared global state, one good example
1N/Abeing regular expression engine's state.)
1N/A=head2 Timing problems
1N/AThe following tests may fail intermittently because of timing
1N/Aproblems, for example if the system is heavily loaded.
1N/AIn case of failure please try running them manually, for example
1N/AFor normal arrays C<$foo = \$bar[1]> will assign C<undef> to
1N/AC<$bar[1]> (assuming that it didn't exist before), but for
1N/Abecause there is currently no way to catch the reference creation.
1N/Asubroutine names. While some limited functionality towards this does
1N/Aexist as of Perl 5.8.0, that is more accidental than designed; use of
1N/AUnicode for the said purposes is unsupported.
1N/AOne reason of this unfinishedness is its (currently) inherent
1N/Aunportability: since both package names and subroutine names may
1N/Aneed to be mapped to file and directory names, the Unicode capability
1N/Aof the filesystem becomes important-- and there unfortunately aren't
1N/A=head1 Platform Specific Problems
1N/AIf using the AIX native make command, instead of just "make" issue
1N/A"make all". In some setups the former has been known to spuriously
1N/Aalso try to run "make install". Alternatively, you may want to use
1N/AIn AIX 4.2, Perl extensions that use C++ functions that use statics
1N/Amay have problems in that the statics are not getting initialized.
1N/AIn newer AIX releases, this has been solved by linking Perl with
1N/Athe libC_r library, but unfortunately in AIX 4.2 the said library
1N/Ahas an obscure bug where the various functions related to time
1N/A(such as time() and gettimeofday()) return broken values, and
1N/Atherefore in AIX 4.2 Perl is not linked against libC_r.
1N/Avac 5.0.0.0 May Produce Buggy Code For Perl
1N/AThe AIX C compiler vac version 5.0.0.0 may produce buggy code,
1N/Aresulting in a few random tests failing when run as part of "make
1N/Atest", but when the failing tests are run by hand, they succeed.
1N/AWe suggest upgrading to at least vac version 5.0.1.0, that has been
1N/Aknown to compile Perl correctly. "lslpp -L|grep
vac.C" will tell
1N/AIf building threaded Perl, you may get compilation warning from
pp_sys.c:
1N/A "
pp_sys.c", line 4651.39: 1506-280 (W) Function argument assignment between types "unsigned char*" and "const void*" is not allowed.
1N/AThis is harmless; it is caused by the getnetbyaddr() and getnetbyaddr_r()
1N/Ahaving slightly different types for their first argument.
1N/A=head2 Alpha systems with old gccs fail several tests
1N/Agccs prior to 2.95.3 are definitely not good enough, and gcc 3.1 may
1N/Aas did Linux 2.4.18 with gcc 2.95.4.) (In Tru64, it is preferable to
1N/Ause the bundled C compiler.)
1N/APerl 5.8.0 doesn't build in AmigaOS. It broke at some point during
1N/Athe ithreads work and we could not find Amiga experts to unbreak the
1N/Aproblems. Perl 5.6.1 still works for AmigaOS (as does the the 5.7.2
1N/Adevelopment release).
1N/AThe following tests fail on 5.8.0 Perl in BeOS Personal 5.03:
1N/A=head2 Cygwin "unable to remap"
1N/AFor example when building the Tk extension for Cygwin,
1N/Ayou may get an error message saying "unable to remap".
1N/AThis is known problem with Cygwin, and a workaround is
1N/A=head2 Cygwin ndbm tests fail on FAT
1N/AOne can build but not install (or test the build of) the NDBM_File
1N/Aon FAT filesystems. Installation (or build) on NTFS works fine.
1N/AIf one attempts the test on a FAT install (or build) the following
1N/Afailures are expected:
1N/ANDBM_File fails and ODBM_File just coredumps.
1N/AIf you intend to run only on FAT (or if using AnyDBM_File on FAT),
1N/Arun Configure with the -Ui_ndbm and -Ui_dbm options to prevent
1N/ANDBM_File and ODBM_File being built.
1N/A=head2 DJGPP Failures
1N/AThe above failures are known as of 5.8.0 with native builds with long
1N/Afilenames, but there are a few more if running under dosemu because of
1N/Alimitations (and maybe bugs) of dosemu:
1N/Afailures that work fine with long filenames. So you really might
1N/Aprefer native builds and long filenames.
1N/A=head2 FreeBSD built with ithreads coredumps reading large directories
1N/AThis is a known bug in FreeBSD 4.5's readdir_r(), it has been fixed in
1N/A=head2 FreeBSD Failing locale Test 117 For ISO 8859-15 Locales
1N/AThe ISO 8859-15 locales may fail the locale test 117 in FreeBSD.
1N/AThis is caused by the characters \xFF (y with diaeresis) and \xBE
1N/A(Y with diaeresis) not behaving correctly when being matched
1N/Acase-insensitively. Apparently this problem has been fixed in
1N/Athe latest FreeBSD releases.
1N/AIRIX with MIPSpro 7.3.1.2m or 7.3.1.3m compiler may fail the List::Util
1N/Aa compiler error since if compiled with gcc no core dump ensues, and
1N/Ano failures have been seen on the said test on any other platform.
1N/ASimilarly, building the Digest::MD5 extension has been
1N/Aknown to fail with "*** Termination code 139 (bu21)".
1N/AThe cure is to drop optimization level (Configure -Doptimize=-O2).
1N/AIf perl is configured with -Duse64bitall, the successful result of the
1N/Asubtest 10 of
lib/posix may arrive before the successful result of the
1N/Asubtest 9, which confuses the test harness so much that it thinks the
1N/A=head2 Linux with glibc 2.2.5 fails
t/op/int subtest #6 with -Duse64bitint
1N/AThis is a known bug in the glibc 2.2.5 with long long integers.
1N/APlease remember to set your environment variable LC_ALL to "C"
1N/A(setenv LC_ALL C) before running "make test" to avoid a lot of
1N/Awarnings about the broken locales of Mac OS X.
1N/AThe following tests are known to fail in Mac OS X 10.1.5 because of
1N/Abuggy (old) implementations of Berkeley DB included in Mac OS X:
1N/A Failed Test Stat Wstat Total Fail Failed List of Failed
1N/A -------------------------------------------------------------------------
1N/AIf you are building on a UFS partition, you will also probably see
1N/Asupporting inode change time.
1N/Anow because the failure is Apple's fault, not Perl's (blocked signals
1N/Athis is not Perl's fault-- the libc of Mac OS X is not threadsafe
1N/A(in this particular test, the localtime() call is found to be
1N/A=head2 Mac OS X dyld undefined symbols
1N/AIf after installing Perl 5.8.0 you are getting warnings about missing
1N/A dyld: perl Undefined symbols
1N/Ayou probably have an old pre-Perl-5.8.0 installation (or parts of one)
1N/AIt seems that for some reason "make install" doesn't always completely
1N/Ashared library out of the way like this:
1N/Aand then reissue "make install". Note that the above of course is
1N/AIf that doesn't help, you may have to try removing all the .bundle
1N/A=head2 OS/2 Test Failures
1N/AThe following tests are known to fail on OS/2 (for clarity
1N/Aonly the failures are shown, not the full error messages):
1N/AThe
op/sprintf tests 91, 129, and 130 are known to fail on some platforms.
1N/ATest 91 is known to fail on QNX6 (nto), because C<sprintf '%e',0>
1N/Aincorrectly produces C<0.000000e+0> instead of C<0.000000e+00>.
1N/AFor tests 129 and 130, the failing platforms do not comply with
1N/Athe ANSI C Standard: lines 19ff on page 134 of ANSI X3.159 1989, to
1N/Abe exact. (They produce something other than "1" and "-1" when
1N/Aformatting 0.6 and -0.6 using the printf format "%.0f"; most often,
1N/Athey produce "0" and "-0".)
1N/AThe socketpair tests are known to be unhappy in SCO 3.2v5.0.4:
1N/AIn case you are still using Solaris 2.5 (aka SunOS 5.5), you may
1N/AThe suggested cure is to upgrade your Solaris.
1N/A=head2 Solaris x86 Fails Tests With -Duse64bitint
1N/AThe following tests are known to fail in Solaris x86 with Perl
1N/Aconfigured to use 64 bit integers:
1N/A=head2 SUPER-UX (NEC SX)
1N/AThe following tests are known to fail on SUPER-UX:
1N/Ais serious but as of yet unsolved. It points at some problems with the
1N/Asignedness handling of the C compiler, as do the 64bitint, arith, and pow
1N/Afailures. Most of the rest point at problems with SysV IPC.
1N/A=head2 Term::ReadKey not working on Win32
1N/AUse Term::ReadKey 2.20 or later.
1N/ADuring Configure, the test
1N/A Guessing which symbols your C compiler and preprocessor define...
1N/Awill probably fail with error messages like
1N/A The identifier "bad" is undefined.
1N/A bad switch yylook 79bad switch yylook 79bad switch yylook 79bad switch yylook 79#ifdef A29K
1N/A A semicolon is expected at this point.
1N/AThis is caused by a bug in the awk utility of
UNICOS/mk. You can ignore
1N/Athe error, but it does cause a slight problem: you cannot fully
1N/Abenefit from the h2ph utility (see L<h2ph>) that can be used to
1N/Aconvert C headers to Perl libraries, mainly used to be able to access
1N/Afrom Perl the constants defined using C preprocessor, cpp. Because of
1N/Athe above error, parts of the converted headers will be invisible.
1N/ALuckily, these days the need for h2ph is rare.
1N/AIf building Perl with interpreter threads (ithreads), the
1N/Agetgrent(), getgrnam(), and getgrgid() functions cannot return the
1N/Alist of the group members due to a bug in the multithreaded support of
1N/AUNICOS/mk. What this means is that in list context the functions will
1N/Areturn only three values, not four.
1N/AWhen Perl is built using the native build process on VOS Release
1N/A14.5.0 and GNU C++/GNU Tools 2.0.1, all attempted tests either
1N/Apass or result in TODO (ignored) failures.
1N/AThere should be no reported test failures with a default configuration,
1N/Athough there are a number of tests marked TODO that point to areas
1N/Aneeding further debugging
and/or porting work.
1N/AIn multi-CPU boxes, there are some problems with the I/O buffering:
1N/Asome output may appear twice.
1N/A=head2 XML::Parser not working
1N/AUse XML::Parser 2.31 or later.
1N/Az/OS has rather many test failures but the situation is actually much
1N/Abetter than it was in 5.6.0; it's just that so many new modules and
1N/Atests have been added.
1N/A Failed Test Stat Wstat Total Fail Failed List of Failed
1N/A ---------------------------------------------------------------------------
1N/Athose in io_unix and sprintf are problems in the USS (UDP sockets and
1N/Aprintf formats). The pat, tr, and fold failures are genuine Perl
1N/Aproblems caused by EBCDIC (and in the pat and fold cases, combining
1N/Athat with Unicode). The Constant and Embed are probably problems in
1N/Athe tests (since they test Perl's ability to build extensions, and
1N/Athat seems to be working reasonably well.)
1N/A=head2 Unicode Support on EBCDIC Still Spotty
1N/AThough mostly working, Unicode support still has problem spots on
1N/AEBCDIC platforms. One such known spot are the C<\p{}> and C<\P{}>
1N/Aregular expression constructs for code points less than 256: the
1N/AC<pP> are testing for Unicode code points, not knowing about EBCDIC.
1N/A=head2 Seen In Perl 5.7 But Gone Now
1N/AC<Time::Piece> (previously known as C<Time::Object>) was removed
1N/Abecause it was felt that it didn't have enough value in it to be a
1N/Acore module. It is still a useful module, though, and is available
1N/APerl 5.8 unfortunately does not build anymore on AmigaOS; this broke
1N/Aaccidentally at some point. Since there are not that many Amiga
1N/Adevelopers available, we could not get this fixed and tested in time
1N/Afor 5.8.0. Perl 5.6.1 still works for AmigaOS (as does the the 5.7.2
1N/Adevelopment release).
1N/AThe C<PerlIO::Scalar> and C<PerlIO::Via> (capitalised) were renamed as
1N/AC<PerlIO::scalar> and C<PerlIO::via> (all lowercase) just before 5.8.0.
1N/AThe main rationale was to have all core PerlIO layers to have all
1N/Alowercase names. The "plugins" are named as usual, for example
1N/AC<PerlIO::via::QuotedPrint>.
1N/AThe C<threads::shared::queue> and C<threads::shared::semaphore> were
1N/Arenamed as C<Thread::Queue> and C<Thread::Semaphore> just before 5.8.0.
1N/AThe main rationale was to have thread modules to obey normal naming,
1N/AC<Thread::> (the C<threads> and C<threads::shared> themselves are
1N/Amore pragma-like, they affect compile-time, so they stay lowercase).
1N/A=head1 Reporting Bugs
1N/AIf you find what you think is a bug, you might check the articles
1N/AIf you believe you have an unreported bug, please run the B<perlbug>
1N/Aprogram included with your release. Be sure to trim your bug down
1N/Ato a tiny but sufficient test case. Your bug report, along with the
1N/Aoutput of C<perl -V>, will be sent off to perlbug@perl.org to be
1N/Aanalysed by the Perl porting team.
1N/AThe F<Changes> file for exhaustive details on what changed.
1N/AThe F<INSTALL> file for how to build Perl.
1N/AThe F<README> file for general stuff.
1N/AThe F<Artistic> and F<Copying> files for copyright information.
1N/AWritten by Jarkko Hietaniemi <F<jhi@iki.fi>>.