PERL5400DELTA(1) Perl Programmers Reference Guide PERL5400DELTA(1)
NAME
perl5400delta - what is new for perl v5.40.0
DESCRIPTION
This document describes differences between the 5.38.0 release and the
5.40.0 release.
Core Enhancements
New "__CLASS__" Keyword
When using the new "class" feature, code inside a method, "ADJUST"
block or field initializer expression is now permitted to use the new
"__CLASS__" keyword. This yields a class name, similar to
"__PACKAGE__", but whereas that gives the compile-time package that the
code appears in, the "__CLASS__" keyword is aware of the actual run-
time class that the object instance is a member of. This makes it
useful for method dispatch on that class, especially during
constructors, where access to $self is not permitted.
For more information, see "__CLASS__" in perlfunc.
":reader" attribute for field variables
When using the "class" feature, field variables can now take a
":reader" attribute. This requests that an accessor method be
automatically created that simply returns the value of the field
variable from the given instance.
field $name :reader;
Is equivalent to
field $name;
method name () { return $name; }
An alternative name can also be provided:
field $name :reader(get_name);
For more detail, see ":reader" in perlclass.
Permit a space in "-M" command-line option
When processing command-line options, perl now allows a space between
the "-M" switch and the name of the module after it.
$ perl -M Data::Dumper=Dumper -E 'say Dumper [1,2,3]'
This matches the existing behaviour of the "-I" option.
Restrictions to "use VERSION" declarations
In Perl 5.36, a deprecation warning was added when downgrading a "use
VERSION" declaration from one above version 5.11, to below. This has
now been made a fatal error.
Additionally, it is now a fatal error to issue a subsequent "use
VERSION" declaration when another is in scope, when either version is
5.39 or above. This is to avoid complications surrounding imported
lexical functions from builtin. A deprecation warning has also been
added for any other subsequent "use VERSION" declaration below version
5.39, to warn that it will no longer be permitted in Perl version 5.44.
New "builtin::inf" and "builtin::nan" functions (experimental)
Two new functions, "inf" and "nan", have been added to the "builtin"
namespace. These act like constants that yield the floating-point
infinity and Not-a-Number value respectively.
New "^^" logical xor operator
Perl has always had three low-precedence logical operators "and", "or"
and "xor", as well as three high-precedence bitwise versions "&", "^"
and "|". Until this release, while the medium-precedence logical
operators of "&&" and "||" were also present, there was no exclusive-or
equivalent. This release of Perl adds the final "^^" operator,
completing the set.
$x ^^ $y and say "One of x or y is true, but not both";
"try"/"catch" feature is no longer experimental
Prior to this release, the "try"/"catch" feature for handling errors
was considered experimental. Introduced in Perl version 5.34.0, this is
now considered a stable language feature and its use no longer prints a
warning. It still must be enabled with the 'try' feature.
See "Try Catch Exception Handling" in perlsyn.
"for" iterating over multiple values at a time is no longer experimental
Prior to this release, iterating over multiple values at a time with
"for" was considered experimental. Introduced in Perl version 5.36.0,
this is now considered a stable language feature and its use no longer
prints a warning. See "Compound Statements" in perlsyn.
"builtin" module is no longer experimental
Prior to this release, the builtin module and all of its functions were
considered experimental. Introduced in Perl version 5.36.0, this module
is now considered stable its use no longer prints a warning. However,
several of its functions are still considered experimental.
The ":5.40" feature bundle adds "try"
The latest version feature bundle now contains the recently-stablized
feature "try". As this feature bundle is used by the "-E" commandline
switch, these are immediately available in "-E" scripts.
"use v5.40;" imports builtin functions
In addition to importing a feature bundle, "use v5.40;" (or later
versions) imports the corresponding builtin version bundle.
Security
CVE-2023-47038 - Write past buffer end via illegal user-defined Unicode
property
This vulnerability was reported directly to the Perl security team by
Nathan Mills "
[email protected]".
A crafted regular expression when compiled by perl 5.30.0 through
5.38.0 can cause a one-byte attacker controlled buffer overflow in a
heap allocated buffer.
CVE-2023-47039 - Perl for Windows binary hijacking vulnerability
This vulnerability was reported to the Intel Product Security Incident
Response Team (PSIRT) by GitHub user ycdxsb
<https://github.com/ycdxsb/WindowsPrivilegeEscalation>. PSIRT then
reported it to the Perl security team.
Perl for Windows relies on the system path environment variable to find
the shell ("cmd.exe"). When running an executable which uses Windows
Perl interpreter, Perl attempts to find and execute "cmd.exe" within
the operating system. However, due to path search order issues, Perl
initially looks for cmd.exe in the current working directory.
An attacker with limited privileges can exploit this behavior by
placing "cmd.exe" in locations with weak permissions, such as
"C:\ProgramData". By doing so, when an administrator attempts to use
this executable from these compromised locations, arbitrary code can be
executed.
Incompatible Changes
reset EXPR now calls set-magic on scalars
Previously "reset EXPR" did not call set magic when clearing scalar
variables. This meant that changes did not propagate to the underlying
internal state where needed, such as for $^W, and did not result in an
exception where the underlying magic would normally throw an exception,
such as for $1.
This means code that had no effect before may now actually have an
effect, including possibly throwing an exception.
"reset EXPR" already called set magic when modifying arrays and hashes.
This has no effect on plain "reset" used to reset one-match searches as
with "m?pattern?".
[GH #20763 <https://github.com/Perl/perl5/issues/20763>]
Calling the import method of an unknown package produces a warning
Historically, it has been possible to call the "import" or "unimport"
method of any class, including ones which have not been defined, with
an argument and not experience an error. For instance, this code will
not throw an error in Perl 5.38:
Class::That::Does::Not::Exist->import("foo");
However, as of Perl 5.39.1 this is deprecated and will issue a warning.
Note that calling these methods with no arguments continues to silently
succeed and do nothing. For instance,
Class::That::Does::Not::Exist->import();
will continue to not throw an error. This is because every class
implicitly inherits from the class UNIVERSAL which now defines an
"import" method. In older perls there was no such method defined, and
instead the method calls for "import" and "unimport" were special cased
to not throw errors if there was no such method defined.
This change has been added because it makes it easier to detect case
typos in "use" statements when running on case-insensitive file
systems. For instance, on Windows or other platforms with case-
insensitive file systems on older perls the following code
use STRICT 'refs';
would silently do nothing as the module is actually called strict.pm,
not STRICT.pm, so it would be loaded but its import method would never
be called. It will also detect cases where a user passes an argument
when using a package that does not provide its own import, for instance
most "pure" class definitions do not define an import method.
"return" no longer allows an indirect object
The "return" operator syntax now rejects indirect objects. In most
cases this would compile and even run, but wasn't documented and could
produce confusing results, for example:
# note that sum hasn't been defined
sub sum_positive {
return sum grep $_ > 0, @_;
# unexpectedly parsed as:
# return *sum, grep $_ > 0, @_;
# ... with the bareword acting like an extra (typeglob) argument
}
say for sum_positive(-1, 2, 3)
produced:
*main::sum
2
3
[GH #21716 <https://github.com/Perl/perl5/issues/21716>]
Class barewords no longer resolved as file handles in method calls under
"no feature "bareword_filehandles""
Under "no feature "bareword_filehandles"" bareword file handles
continued to be resolved in method calls:
open FH, "<", $somefile or die;
no feature 'bareword_filehandles';
FH->binmode;
This has been fixed, so the:
FH->binmode;
will attempt to resolve "FH" as a class, typically resulting in a
runtime error.
The standard file handles such as "STDOUT" continue to be resolved as a
handle:
no feature 'bareword_filehandles';
STDOUT->flush; # continues to work
Note that once perl resolves a bareword name as a class it will
continue to do so:
package SomeClass {
sub somemethod{}
}
open SomeClass, "<", "somefile" or die;
# SomeClass resolved as a handle
SomeClass->binmode;
{
no feature "bareword_filehandles";
SomeClass->somemethod;
}
# SomeClass resolved as a class
SomeClass->binmode;
[GH #19426 <https://github.com/Perl/perl5/issues/19426>]
Deprecations
o Using "goto" to jump from an outer scope into an inner scope is
deprecated and will be removed completely in Perl 5.42. [GH #21601
<https://github.com/Perl/perl5/issues/21601>]
Performance Enhancements
o The negation OPs have been modified to support the generic "TARGMY"
optimization. [GH #21442
<https://github.com/Perl/perl5/issues/21442>]
Modules and Pragmata
New Modules and Pragmata
o Term::Table 0.018 has been added to the Perl core.
This module is a dependency of Test2::Suite.
o Test2::Suite 0.000162 has been added to the Perl core.
This distribution contains a comprehensive set of test tools for
writing unit tests. It is the successor to Test::More and similar
modules. Its inclusion in the Perl core means that CPAN module
tests can be written using this suite of tools without extra
dependencies.
Updated Modules and Pragmata
o Archive::Tar has been upgraded from version 2.40 to 3.02_001.
o attributes has been upgraded from version 0.35 to 0.36.
o autodie has been upgraded from version 2.36 to 2.37.
o B has been upgraded from version 1.88 to 1.89.
o B::Deparse has been upgraded from version 1.74 to 1.76.
o Benchmark has been upgraded from version 1.24 to 1.25.
o bignum has been upgraded from version 0.66 to 0.67.
o builtin has been upgraded from version 0.008 to 0.014.
builtin now accepts a version bundle as an input argument,
requesting it to import all of the functions that are considered a
stable part of the module at the given Perl version. For example:
use builtin ':5.40';
Added the load_module() builtin function as per PPC 0006
<https://github.com/Perl/PPCs/blob/main/ppcs/ppc0006-load-
module.md>.
o bytes has been upgraded from version 1.08 to 1.09.
o Compress::Raw::Bzip2 has been upgraded from version 2.204_001 to
2.212.
o Compress::Raw::Zlib has been upgraded from version 2.204_001 to
2.212.
o CPAN::Meta::Requirements has been upgraded from version 2.140 to
2.143.
o Data::Dumper has been upgraded from version 2.188 to 2.189.
o DB_File has been upgraded from version 1.858 to 1.859.
o Devel::Peek has been upgraded from version 1.33 to 1.34.
o Devel::PPPort has been upgraded from version 3.71 to 3.72.
o diagnostics has been upgraded from version 1.39 to 1.40.
o DynaLoader has been upgraded from version 1.54 to 1.56.
o Encode has been upgraded from version 3.19 to 3.21.
o Errno has been upgraded from version 1.37 to 1.38.
The "osvers" and "archname" baked into the module to ensure Errno
is loaded by the perl that built it are now more comprehensively
escaped. [GH #21135 <https://github.com/Perl/perl5/issues/21135>]
o experimental has been upgraded from version 0.031 to 0.032.
o Exporter has been upgraded from version 5.77 to 5.78.
o ExtUtils::CBuilder has been upgraded from version 0.280238 to
0.280240.
o ExtUtils::Manifest has been upgraded from version 1.73 to 1.75.
o ExtUtils::Miniperl has been upgraded from version 1.13 to 1.14.
o Fcntl has been upgraded from version 1.15 to 1.18.
The old module documentation stub has been greatly expanded and
revised.
Adds support for the "O_TMPFILE" flag on Linux.
o feature has been upgraded from version 1.82 to 1.89.
It now documents the ":all" feature bundle, and suggests a reason
why you may not wish to use it.
o fields has been upgraded from version 2.24 to 2.25.
o File::Compare has been upgraded from version 1.1007 to 1.1008.
o File::Find has been upgraded from version 1.43 to 1.44.
o File::Glob has been upgraded from version 1.40 to 1.42.
o File::Spec has been upgraded from version 3.89 to 3.90.
o File::stat has been upgraded from version 1.13 to 1.14.
o FindBin has been upgraded from version 1.53 to 1.54.
o Getopt::Long has been upgraded from version 2.54 to 2.57.
o Getopt::Std has been upgraded from version 1.13 to 1.14.
Documentation and test improvements only; no change in
functionality.
o Hash::Util has been upgraded from version 0.30 to 0.32.
o Hash::Util::FieldHash has been upgraded from version 1.26 to 1.27.
o HTTP::Tiny has been upgraded from version 0.086 to 0.088.
o I18N::Langinfo has been upgraded from version 0.22 to 0.24.
It now handles the additional locale categories that Linux defines
beyond those in the POSIX Standard.
This fixes what is returned for the "ALT_DIGITS" item, which has
never before worked properly in Perl.
o IO has been upgraded from version 1.52 to 1.55.
Fixed "IO::Handle/blocking" on Windows, which has been non-
functional since IO 1.32. [GH #17455
<https://github.com/Perl/perl5/issues/17455>]
o IO-Compress has been upgraded from version 2.204 to 2.212.
o IO::Socket::IP has been upgraded from version 0.41_01 to 0.42.
o IO::Zlib has been upgraded from version 1.14 to 1.15.
o locale has been upgraded from version 1.10 to 1.12.
o Math::BigInt has been upgraded from version 1.999837 to 2.003002.
o Math::BigInt::FastCalc has been upgraded from version 0.5013 to
0.5018.
o Module::CoreList has been upgraded from version 5.20230520 to
5.20240609.
o Module::Metadata has been upgraded from version 1.000037 to
1.000038.
o mro has been upgraded from version 1.28 to 1.29.
o NDBM_File has been upgraded from version 1.16 to 1.17.
o Opcode has been upgraded from version 1.64 to 1.65.
o perl5db.pl has been upgraded from version 1.77 to 1.78.
Made parsing of the "l" command arguments saner. [GH #21350
<https://github.com/Perl/perl5/issues/21350>]
o perlfaq has been upgraded from version 5.20210520 to 5.20240218.
o PerlIO::encoding has been upgraded from version 0.30 to 0.31.
o PerlIO::scalar has been upgraded from version 0.31 to 0.32.
o PerlIO::via has been upgraded from version 0.18 to 0.19.
o Pod::Checker has been upgraded from version 1.75 to 1.77.
o Pod::Html has been upgraded from version 1.34 to 1.35.
o Pod::Simple has been upgraded from version 3.43 to 3.45.
o podlators has been upgraded from version 5.01 to 5.01_02.
o POSIX has been upgraded from version 2.13 to 2.20.
The "mktime" function now works correctly on 32-bit platforms even
if the platform's "time_t" type is larger than 32 bits. [GH #21551
<https://github.com/Perl/perl5/issues/21551>]
The "T_SIGNO" and "T_FD" typemap entries have been fixed so they
work with any variable name, rather than just the hardcoded "sig"
and "fd".
The mappings for "Mode_t", "pid_t", "Uid_t", "Gid_t" and "Time_t"
have been updated to be integer types; previously they were "NV"
floating-point.
Adjusted the signbit() on NaN test to handle the unusual bit
pattern returned for NaN by Oracle Developer Studio's compiler.
[GH #21533 <https://github.com/Perl/perl5/issues/21533>]
o re has been upgraded from version 0.44 to 0.47.
o Safe has been upgraded from version 2.44 to 2.46.
o SelfLoader has been upgraded from version 1.26 to 1.27.
o Socket has been upgraded from version 2.036 to 2.038.
o strict has been upgraded from version 1.12 to 1.13.
o Test::Harness has been upgraded from version 3.44 to 3.48.
o Test::Simple has been upgraded from version 1.302194 to 1.302199.
o Text::Tabs has been upgraded from version 2021.0814 to 2024.001.
o Text::Wrap has been upgraded from version 2021.0814 to 2024.001.
o threads has been upgraded from version 2.36 to 2.40.
An internal error has been made slightly more verbose ("Out of
memory in perl:threads:ithread_create").
o threads::shared has been upgraded from version 1.68 to 1.69.
o Tie::File has been upgraded from version 1.07 to 1.09.
Old compatibility code for perl 5.005 that was no longer functional
has been removed.
o Time::gmtime has been upgraded from version 1.04 to 1.05.
o Time::HiRes has been upgraded from version 1.9775 to 1.9777.
o Time::Local has been upgraded from version 1.30 to 1.35.
o Time::localtime has been upgraded from version 1.03 to 1.04.
o Time::tm has been upgraded from version 1.00 to 1.01.
o UNIVERSAL has been upgraded from version 1.15 to 1.17.
o User::grent has been upgraded from version 1.04 to 1.05.
o User::pwent has been upgraded from version 1.02 to 1.03.
o version has been upgraded from version 0.9929 to 0.9930.
o warnings has been upgraded from version 1.65 to 1.69.
o XS::APItest has been upgraded from version 1.32 to 1.36.
o XS::Typemap has been upgraded from version 0.19 to 0.20.
Documentation
Changes to Existing Documentation
We have attempted to update the documentation to reflect the changes
listed in this document. If you find any we have missed, open an issue
at <https://github.com/Perl/perl5/issues>.
Additionally, the following selected changes have been made:
perlapi
o Corrected the documentation for "Perl_form", "form_nocontext", and
"vform", which claimed that any later call to one of them will
destroy the previous returns from any. This hasn't been true since
5.6.0, except it does remain true if these are called during global
destruction. With that caveat, the return of each of these is a
fresh string in a temporary that will automatically be freed by a
call to ""FREETMPS"" in perlapi or at at places such as statement
boundaries.
o Several internal functions now have documentation - the various
"newSUB" functions, newANONLIST(), newANONHASH(), newSVREF() and
similar.
perlclass
o Added a list of known bugs in the experimental "class" feature.
perlfunc
o The documentation for "local", "my", "our", and "state", has been
updated to include examples and descriptions of their effects
within a statement.
perlguts
o A new section has been added which describes the experimental
reference-counted argument stack build option ("PERL_RC_STACK").
perlclib
o Extensive guidance has been added for interfacing with the standard
C library, including many more functions to avoid, and how to cope
with locales and threads.
perlhacktips
o Document we can't use compound literals or array designators due to
C++ compatibility. [GH #21073
<https://github.com/Perl/perl5/issues/21073>]
o Document new functions sv_mark_arenas() and sv_sweep_arenas()
(which only exist on "DEBUGGING" builds)
o Added brief documentation for some tools useful when developing
perl itself on Windows or Cygwin.
perllol
o Removed indirect object syntax in "Dumpvalue" example
perlre
o Removed statement suggesting "/p" is a no-op.
perlref
o Documented ref assignment in list context (as part of the
"refaliasing" feature)
perlop
o The section on the empty pattern "//" has been amended to mention
that the current dynamic scope is used to find the last successful
match.
perlport
o The "-S" file test has been meaningful on Win32 since 5.37.6
o The "-l" file test is now meaningful on Win32
o Some strange behaviour with "." at the end of names under Windows
has been documented
perlvar
o Added documentation for an alternative to "${^CAPTURE}"
Diagnostics
The following additions or changes have been made to diagnostic output,
including warnings and fatal error messages. For the complete list of
diagnostic messages, see perldiag.
New Diagnostics
New Errors
o Cannot use __CLASS__ outside of a method or field initializer
expression
(F) A "__CLASS__" expression yields the class name of the object
instance executing the current method, and therefore it can only be
placed inside an actual method (or method-like expression, such as
a field initializer expression).
o get_layers: unknown argument '%s'
(F) You called PerlIO::get_layers() with an unknown argument. Legal
arguments are provided in key/value pairs, with the keys being one
of "input", "output" or "detail", followed by a boolean.
o UNIVERSAL does not export anything
(F) You asked UNIVERSAL to export something, but UNIVERSAL is the
base class for all classes and contains no exportable symbols.
o Builtin version bundle "%s" is not supported by Perl
(F) You attempted to "use builtin :ver" for a version number that
is either older than 5.39 (when the ability was added), or newer
than the current perl version.
o Invalid version bundle "%s"
(F) A version number that is used to specify an import bundle
during a "use builtin ..." statement must be formatted as
":MAJOR.MINOR" with an optional third component, which is ignored.
Each component must be a number of 1 to 3 digits. No other
characters are permitted. The value that was specified does not
conform to these rules.
o Missing comma after first argument to return
(F) While certain operators allow you to specify a filehandle or an
"indirect object" before the argument list, "return" isn't one of
them.
o Out of memory during vec in lvalue context
(F) An attempt was made to extend a string beyond the largest
possible memory allocation by assigning to vec() called with a
large second argument.
(This case used to throw a generic "Out of memory!" error.)
o Cannot create an object of incomplete class "%s"
(F) An attempt was made to create an object of a class where the
start of the class definition has been seen, but the class has not
been completed.
This can happen for a failed eval, or if you attempt to create an
object at compile time before the class is complete:
eval "class Foo {"; Foo->new; # error
class Bar { BEGIN { Bar->new } }; # error
Previously perl would assert or crash. [GH #22159
<https://github.com/Perl/perl5/issues/22159>]
New Warnings
o Forked open '%s' not meaningful in <>
(S inplace) You had "|-" or "-|" in @ARGV and tried to use "<>" to
read from it.
Previously this would fork and produce a confusing error message.
[GH #21176 <https://github.com/Perl/perl5/issues/21176>]
o Attempt to call undefined %s method with arguments ("%s"%s) via
package "%s" (Perhaps you forgot to load the package?)
(D deprecated::missing_import_called_with_args) You called the
import() or unimport() method of a class that has no import method
defined in its inheritance graph, and passed an argument to the
method. This is very often the sign of a misspelled package name
in a use or require statement that has silently succeeded due to a
case insensitive file system.
Another common reason this may happen is when mistakenly attempting
to import or unimport a symbol from a class definition or package
which does not use "Exporter" or otherwise define its own "import"
or "unimport" method.
Changes to Existing Diagnostics
o Name "%s::%s" used only once: possible typo
This warning now honors being marked as fatal. [GH #13814
<https://github.com/Perl/perl5/issues/13814>]
o Out of memory in perl:%s
There used to be several places in the perl core that would print a
generic "Out of memory!" message and abort when memory allocation
failed, giving no indication which program it was that ran out of
memory. These have been modified to include the word "perl" and
the general area of the allocation failure, e.g. "Out of memory in
perl:util:safesysrealloc". [GH #21672
<https://github.com/Perl/perl5/issues/21672>]
o Possible precedence issue with control flow operator (%s)
This warning now mentions the name of the control flow operator
that triggered the diagnostic (e.g. "return", "exit", "die", etc).
It also covers more cases: Previously, the warning was only
triggered if a low-precedence logical operator (like "and", "or",
"xor") was involved. Now it is also shown for misleading code like
this:
exit $x ? 0 : 1; # actually parses as: exit($x) ? 0 : 1;
exit $x == 0; # actually parses as: exit($x) == 0;
o Use of uninitialized value%s
This warning is now slightly more accurate in cases involving
"length", "pop", "shift", or "splice":
my $x;
length($x) == 0
# Before:
# Use of uninitialized value $x in numeric eq (==) at ...
# Now:
# Use of uninitialized value length($x) in numeric eq (==) at ...
That is, the warning no longer implies that $x was used directly as
an operand of "==", which it wasn't.
Similarly:
my @xs;
shift @xs == 0
# Before:
# Use of uninitialized value within @xs in numeric eq (==) at ...
# Now:
# Use of uninitialized value shift(@xs) in numeric eq (==) at ...
This is more accurate because there never was an "undef" within @xs
as the warning implied. (The warning for "pop" works analogously.)
Finally:
my @xs = (1, 2, 3);
splice(@xs, 0, 0) == 0
# Before:
# Use of uninitialized value within @xs in numeric eq (==) at ...
# Now:
# Use of uninitialized value in numeric eq (==) at ...
That is, in cases where "splice" returns "undef", it no longer
unconditionally blames its first argument. This was misleading
because "splice" can return "undef" even if none of its arguments
contain "undef".
[GH #21930 <https://github.com/Perl/perl5/issues/21930>]
o Old package separator "'" deprecated
Prevent this warning appearing spuriously when checking the
heuristic for the You need to quote "%s" warning.
[GH #22145 <https://github.com/Perl/perl5/issues/22145>]
Configuration and Compilation
o "microperl", long broken and of unclear present purpose, has been
removed as promised in Perl 5.18.
o Fix here-doc used for code to probe "LC_ALL" syntax for disparate
locales introduced in 5.39.2. [GH #21451
<https://github.com/Perl/perl5/issues/21451>]
o You can now separately enable high water mark checks for non-
DEBUGGING or disable them for DEBUGGING builds with
"-Accflags=-DPERL_USE_HWM" or "-Accflags=-DPERL_NO_HWM"
respectively. The default remains the same. [GH #16607
<https://github.com/Perl/perl5/issues/16607>]
Testing
Tests were added and changed to reflect the other additions and changes
in this release. Furthermore, these significant changes were made:
o Update nm output parsing for Darwin in t/porting/libperl.t to
handle changes in the output of nm on Darwin. [GH #21117
<https://github.com/Perl/perl5/issues/21117>]
o t/op/magic.t would fail when "ps" was the BusyBox implementation,
since that doesn't support the "-p" flag and otherwise ignores a
process id on the command-line. This caused TEST failures on
BusyBox systems such as Alpine Linux. [GH #17542
<https://github.com/Perl/perl5/issues/17542>]
o porting/globvar.t now uses the more portable "nm -P ..." to fetch
the names defined in an object file. The parsing of the names
found in the object is now separated from processing them to handle
the duplication between local and global definitions on AIX. [GH
#21637 <https://github.com/Perl/perl5/issues/21637>]
o A test was added to lib/locale_threads.t that extensively stress
tests locale handling. It turns out that the libc implementations
on various platforms have bugs in this regard, including Linux,
Windows, *BSD derivatives including Darwin, and others.
Experimental versions of this test have been used in the past few
years to find bugs in the Perl implementation and in those
platforms, as well as to develop workarounds in the Perl
implementation, where feasible, for the platform bugs. Multiple
bug report tickets have been filed against platforms, and some have
been fixed. The test checks that platforms that purport to support
thread-safe locale handling actually do so (and that perl works
properly on those that do; The read-only variable
"${^SAFE_LOCALES}" is set to 1 if perl thinks the platform can
handle this, whatever the platform's documentation says).
Also tested for is if the various locale categories can indeed be
set independently to disparate locales. (An example of where you
might want to do this is if you are a Western Canadian living and
working in Holland. You likely will want to have the "LC_MONETARY"
locale be set to where you are living, but have the other parts of
your locale retain your native English values. Later, as you get a
bit more comfortable with Dutch, and in order to communicate better
with your colleagues, you might want to change "LC_TIME" and
"LC_NUMERIC" to Dutch, while leaving "LC_CTYPE" and "LC_COLLATE"
set to English indefinitely.)
o The test t/porting/libperl.t will no longer run in maint releases.
This test is sensitive to changes in the output of nm on various
platforms, and tarballs aren't updated as we update this test in
blead. [GH #21677 <https://github.com/Perl/perl5/issues/21677>]
Platform Support
New Platforms
Serenity OS
Out of the box support for Serenity OS was added.
Platform-Specific Notes
Windows
Eliminated several header build warnings under MSVC with "/W4" to
reduce noise for embedders. [GH #21031
<https://github.com/Perl/perl5/issues/21031>]
Work around a bug in most 32-bit Mingw builds, where the generated
code, including the code in the gcc support library, assumes
16-byte stack alignment, which 32-bit Windows does not preserve.
[GH #21313 <https://github.com/Perl/perl5/issues/21313>]
Enable "copysign", "signbit", "acosh", "asinh", "atanh", "exp2",
"tgamma" in the bundled configuration used for MSVC. [GH #21610
<https://github.com/Perl/perl5/issues/21610>]
The build process no longer supports Visual Studio 2013. This was
failing to build at a very basic level and there have been no
reports of such failures. [GH #21624
<https://github.com/Perl/perl5/issues/21624>]
Linux
The hints file has been updated to handle the Intel oneAPI
DPC++/C++ compiler.
MacOS/Darwin
Don't set "MACOSX_DEPLOYMENT_TARGET" when building on OS X 10.5.
[GH #21367 <https://github.com/Perl/perl5/issues/21367>]
VMS Fixed the configure "installation prefix" prompt to accept a string
rather than yes/no.
Fixed compilation by defining proper value for
"perl_lc_all_category_positions_init".
Increased buffer size when reading config_H.SH to fix compilation
under clang.
Oracle Developer Studio (Solaris, Oracle Linux)
Due to an apparent code generation bug, the default optimization
level for the Oracle Developer Studio (formerly Sun Workshop)
compiler is now "-xO1". [GH #21535
<https://github.com/Perl/perl5/issues/21535>]
Internal Changes
o "PERL_RC_STACK" build option added.
This new build option is highly experimental and is not enabled by
default. Perl can be built with it by using the Configure option
"-Accflags='-DPERL_RC_STACK'".
It makes the argument stack bump the reference count of SVs pushed
onto it. It is mostly functional, but currently slow and
incomplete.
It is intended in the long term that this build option will become
the default option, and then finally the only option; but this will
be many releases away.
In particular, there is currently no support within XS code for
using these new features. So under this build option, all XS
functions are called via a backwards-compatibility wrapper which
slows down such calls.
In future releases, better support for XS code is intended to be
added. It is expected that straightforward XS code will eventually
be able to make use of a reference-counted stack without
modification, with any heavy lifting being handled by the XS
compiler ("xsubpp") and the macros which it outputs. But code which
implements PP() functions will eventually have to be modified to
use a new PP API: rpp_foo() rather than PUSHs() etc. But this new
API is not yet stable, nor has it yet been back-ported via
"Devel::PPPort".
See perlguts for more details.
o A new API function has been added that simplifies C (or XS) code
that creates "LISTOP" optree fragments. newLISTOPn() is a variadic
function that takes a "NULL"-terminated list of child op pointers,
and constructs a new checked "LISTOP" to contain them all. This is
simpler than creating a new plain "OP_LIST", adding each child
individually, and finally calling op_convert_list() in most code
fragments.
o The eval_sv() API now accepts the "G_USEHINTS" flag, which uses the
hints such as strict and features from "PL_curcop" instead of the
default, which is to use default hints, e.g. no "use vX.XX;", no
strict, default features.
Beware if you use this flag in XS code: your evaluated code will
need to support whatever strictness or features are in effect at
the point your XS function is called.
[GH #21415 <https://github.com/Perl/perl5/issues/21415>]
o "PERL_VERSION_LE" has been fixed to properly check for "less than
or equal" rather than "less than".
o "dAX", "dITEMS" and hence "dXSARGS" now declare "AX" and "items" as
"Stack_off_t" rather than "SSize_t". This reverts back to
compatibility with pre-64-bit stack support for default builds of
perl where "Stack_off_t" is "I32". [GH #21782
<https://github.com/Perl/perl5/issues/21782>]
o A new function is now available to "XS" code, "sv_langinfo" in
perlapi. This provides the same information as the existing
"Perl_langinfo8" in perlapi, but returns an SV instead of a
"char *", so that programmers don't have to concern themselves with
the UTF-8ness of the result. This new function is now the
preferred interface for "XS" code to the nl_langinfo(3) "libc"
function. From Perl space, this information continues to be
provided by the I18N::Langinfo module.
o glibc has an undocumented equivalent function to querylocale(),
which our experience indicates is reliable. When this is function
is used, it removes the need for perl to keep its own records,
hence is more efficient and guaranteed to be accurate. Use of this
function can be disabled by defining the "NO_NL_LOCALE_NAME" build
option
Selected Bug Fixes
o The delimiter "SYRIAC COLON SKEWED LEFT/RIGHT" pair has been
removed from the ones recognized by the "extra_paired_delimiters"
feature. (See "Quote and Quote-like Operators" in perlop.) This
is because those characters are normally written right-to-left, and
this could be visually confusing [GH #22228
<https://github.com/Perl/perl5/issues/22228>]. The change was
actually to forbid any right-to-left delimiters, but this pair is
the only current instance that meets this criterion. By policy,
this change means that the "extra_paired_delimiters" feature cannot
be considered to have been stable long enough for its experimental
status to be removed.
o "use 5.36;" or later didn't enable the post parse reporting of Name
"%s::%s" used only once: possible typo warnings when enabling
warnings. [GH #21271 <https://github.com/Perl/perl5/issues/21271>]
o Fix a crash or assertion when cleaning up a closure that refers to
an outside "our" sub. [GH #21067
<https://github.com/Perl/perl5/issues/21067>]
o Fixed a number of issues where "I32" was used as a string offset or
size rather than "SSize_t" or "STRLEN"/"size_t" [GH #21012
<https://github.com/Perl/perl5/issues/21012>]
o "~$str" when $str was more than 2GB in size would do nothing or
produce an incomplete result.
o String repeat, "$str x $count", didn't handle $str over 2GB in
size, throwing an error. Now such strings are repeated.
o Complex substitution after the 2GB point in a string could access
incorrect or invalid offsets in the string.
o sv_utf8_decode() would truncate the SVs pos() value. This wasn't
visible via utf8::decode().
o When compiling a constant folded hash key, the length was truncated
when creating the shared SV. Since hash keys over 2GB are not
supported, throw a compilation error instead.
o msgrcv() incorrectly called get magic on the buffer SV and failed
to call set magic on completion. [GH #21012
<https://github.com/Perl/perl5/issues/21012>]
o msgrcv() used the size parameter to resize the buffer before
validating it. [GH #21012
<https://github.com/Perl/perl5/issues/21012>]
o Inheriting from a class that was hierarchically an ancestor of the
new class, eg. " class A::B :isa(A) { ... } ", would not attempt to
load the parent class. [GH #21332
<https://github.com/Perl/perl5/issues/21332>]
o Declared references can now be used with "state" variables. [GH
#21351 <https://github.com/Perl/perl5/issues/21351>]
o Trailing elements in an "unshift"ed and resized array will now
always be initialized. [GH #21265
<https://github.com/Perl/perl5/issues/21265>]
o Make "use 5.036" respect the -X flag
perl's -X flag disables all warnings globally, but use 5.036 didn't
respect that until now. [GH #21431
<https://github.com/Perl/perl5/issues/21431>]
o Fixed an OP leak when an error was produced for initializer for a
class field. [GH #20812
<https://github.com/Perl/perl5/issues/20812>]
o Fixed a leak of the return value when smartmatching against a code
reference.
o Fixed a slowdown in repeated substitution replacements using
special variables, such as "s/....x$1/g". It actually makes all
string concatenations involving such "magic" variables less slow,
but the slowdown was more noticeable on repeated substitutions due
to extra memory usage that was only freed after the last iteration.
The slowdown started in perl 5.28.0 - which generally sped up
string concatenation but slowed down when using special variables.
[GH #21360 <https://github.com/Perl/perl5/issues/21360>]
o Lexical names from the enclosing scope in a lexical sub or closure
weren't visible to code executed by calling "eval EXPR;" from the
"DB" package. This was introduced in 5.18 in an attempt to prevent
subs from retaining a reference to their outer scope, but this
broke the special behaviour of "eval EXPR;" in package DB.
This incidentally fixed a TODO test for "B::Deparse". [GH #19370
<https://github.com/Perl/perl5/pull/19370>]
o Optionally support an argument stack over 2**32 entries on 64-bit
platforms. This requires 32GB of memory just for the argument
stack pointers itself, so you will require a significantly more
memory to take advantage of this.
To enable this add "-Accflags=-DPERL_STACK_OFFSET_SSIZET" or
equivalent to the "Configure" command-line.
[GH #20917 <https://github.com/Perl/perl5/issues/20917>] [GH #21523
<https://github.com/Perl/perl5/issues/21523>]
o Fixed various problems with join() where modifications to the
separator could be handled inconsistently, or could access released
memory. Changes to the separator from magic or overloading for
values in the "LIST" no longer have an effect on the resulting
joined string. [GH #21458
<https://github.com/Perl/perl5/issues/21458>]
o Don't clear the integer flag "IOK" from lines in the
"@{"_<$sourcefile"}" array when a "dbstate" op is removed for that
line. This was broken when fixing [GH #19198
<https://github.com/Perl/perl5/issues/19198>]. [GH #21564
<https://github.com/Perl/perl5/issues/21564>]
o Many bug fixes have been made for using locales under threads and
in embedded perls. And workarounds for libc bugs have been added.
As a result thread-safe locale handling is now the default under
OpenBSD, and MingW when compiled with UCRT.
However, testing has shown that Darwin's implementation of thread-
safe locale handling has bugs. So now Perl doesn't attempt to use
the thread-safe operations when compiled on Darwin.
As before, you can check to see if your program is running with
thread-safe locales by checking if the value of "${^SAFE_LOCALES}"
is 1.
o Various bugs have been fixed when perl is configured with
"-Accflags=-DNO_LOCALE_NUMERIC" or any other locale category (or
categories).
o Not all locale categories need be set to the same locale. Perl now
works around bugs in the libc implementations of locale handling on
some platforms that previously could result in mojibake.
o "LC_ALL" is represented in one of two ways when not all locale
categories are set to the same locale. On some platforms, such as
Linux and Windows, the representation is of the form of a series of
'category=locale-name' pairs. On other platforms, such as *BSD,
the representation is positional like "name1 / name2 / ... ".
name1 is always for a particular category as defined by the
platform, as are the other names. The sequence that separates the
names (the " / " above) also varies by platform. Previously, perl
had problems with platforms that used the positional notation.
This is now fixed.
o A bug has been fixed in the regexp engine with an optimisation that
applies to the "+" quantifier where it was followed by a "(*SKIP)"
pattern.
[GH #21534 <https://github.com/Perl/perl5/issues/21534>]
o The tmps (mortal) stack now grows exponentially. Previously it
grew linearly, so if it was growing incrementally, such as through
many calls to sv_2mortal(), on a system where realloc() is O(size),
the performance would be O(n*n). With exponential grows this
changes to amortized O(n). [GH #21654
<https://github.com/Perl/perl5/issues/21654>]
o Lexical subs now have a new stub in the pad for each recursive call
into the containing function. This fixes two problems:
o If the lexical sub called the containing function, a "Can't
undef active subroutine" error would be thrown. For example:
use v5.36.0;
sub outer($oc) {
my sub inner ($c) {
outer($c-1) if $c; # Can't undef active subroutine
}
inner($oc);
}
outer(2);
[GH #18606 <https://github.com/Perl/perl5/issues/18606>]
o If the lexical sub was called from a recursive call into the
containing function, this would overwrite the bindings to the
closed over variables in the lexical sub, so calls into the
lexical sub from the outer recursive call would have access to
the variables from the inner recursive call:
use v5.36.0;
sub outer ($x) {
my sub inner ($label) {
say "$label $x";
}
inner("first");
outer("inner") if $x eq "outer";
# this call to inner() sees the wrong $x
inner("second");
}
outer("outer");
[GH #21987 <https://github.com/Perl/perl5/issues/21987>]
o prepare_export_lexical() was separately saving "PL_comppad" and
"PL_curpad", this could result in "PL_curpad" being restored to a
no longer valid value, resulting in a panic when importing lexicals
in some cases. [GH #21981
<https://github.com/Perl/perl5/issues/21981>]
o A string eval() operation in the scope of a "use VERSION"
declaration would sometimes emit spurious "Downgrading a use
VERSION declaration" warnings due to an inconsistency in the way
the version number was stored. This is now fixed. [GH #22121
<https://github.com/Perl/perl5/issues/22121>]
Known Problems
o perlivp is missing streamzip on Windows
The "streamzip" utility does not get installed on Windows but
should get installed.
Errata From Previous Releases
o perl5300delta has been updated to include the removal of the
"arybase" module that happened at the same time as the removal of
$[.
Acknowledgements
Perl 5.40.0 represents approximately 11 months of development since
Perl 5.38.0 and contains approximately 160,000 lines of changes across
1,500 files from 75 authors.
Excluding auto-generated files, documentation and release tools, there
were approximately 110,000 lines of changes to 1,200 .pm, .t, .c and .h
files.
Perl continues to flourish into its fourth decade thanks to a vibrant
community of users and developers. The following people are known to
have contributed the improvements that became Perl 5.40.0:
Abe Timmerman, Alexander Kanavin, Amory Meltzer, Aristotle Pagaltzis,
Arne Johannessen, Beckett Normington, Bernard Quatermass, Bernd, Bruno
Meneguele, Chad Granum, Chris 'BinGOs' Williams, Christoph Lamprecht,
Craig A. Berry, Dagfinn Ilmari Mannsker, Dan Book, Dan Church, Daniel
Bhmer, Dan Jacobson, Dan Kogai, David Golden, David Mitchell, E.
Choroba, Elvin Aslanov, Erik Huelsmann, Eugen Konkov, Gianni
Ceccarelli, Graham Knop, Greg Kennedy, guoguangwu, Hauke D, H.Merijn
Brand, Hugo van der Sanden, iabyn, Jake Hamby, Jakub Wilk, James E
Keenan, James Raspass, Joe McMahon, Johan Vromans, John Karr, Karen
Etheridge, Karl Williamson, Leon Timmermans, Lukas Mai, Marco Fontani,
Marek Rouchal, Martijn Lievaart, Mathias Kende, Matthew Horsfall, Max
Maischein, Nicolas Mendoza, Nicolas R, OpossumPetya, Paul Evans, Paul
Marquess, Peter John Acklam, Philippe Bruhat (BooK), Raul E Rangel,
Renee Baecker, Ricardo Signes, Richard Leach, Scott Baker, Sevan
Janiyan, Sisyphus, Steve Hay, TAKAI Kousuke, Todd Rinaldo, Tomasz
Konojacki, Tom Hughes, Tony Cook, William Lyu, x-yuri, Yves Orton,
Zakariyya Mughal, .
The list above is almost certainly incomplete as it is automatically
generated from version control history. In particular, it does not
include the names of the (very much appreciated) contributors who
reported issues to the Perl bug tracker.
Many of the changes included in this version originated in the CPAN
modules included in Perl's core. We're grateful to the entire CPAN
community for helping Perl to flourish.
For a more complete list of all of Perl's historical contributors,
please see the AUTHORS file in the Perl source distribution.
Reporting Bugs
If you find what you think is a bug, you might check the perl bug
database at <https://github.com/Perl/perl5/issues>. There may also be
information at <https://www.perl.org/>, the Perl Home Page.
If you believe you have an unreported bug, please open an issue at
<https://github.com/Perl/perl5/issues>. Be sure to trim your bug down
to a tiny but sufficient test case.
If the bug you are reporting has security implications which make it
inappropriate to send to a public issue tracker, then see "SECURITY
VULNERABILITY CONTACT INFORMATION" in perlsec for details of how to
report the issue.
Give Thanks
If you wish to thank the Perl 5 Porters for the work we had done in
Perl 5, you can do so by running the "perlthanks" program:
perlthanks
This will send an email to the Perl 5 Porters list with your show of
thanks.
SEE ALSO
The Changes file for an explanation of how to view exhaustive details
on what changed.
The INSTALL file for how to build Perl.
The README file for general stuff.
The Artistic and Copying files for copyright information.
perl v5.40.1 2025-01-28 PERL5400DELTA(1)