EVOLUTION-MANAGER

Edit File: coding.pod

=head1 NAME

Writing mod_perl Handlers and Scripts

=head1 Description

This chapter covers the mod_perl coding specifics, different from
normal Perl coding. Most other perl coding issues are covered in the
perl manpages and rich literature.

=head1 Prerequisites

=head1 Where the Methods Live

mod_perl 2.0 has all its methods spread across many modules. In order
to use these methods the modules containing them have to be loaded
first. If you don't do that mod_perl will complain that it can't find
the methods in question. The module
C<L<ModPerl::MethodLookup|docs::2.0::api::ModPerl::MethodLookup>> can
be used to find out which modules need to be used.

=head1 Techniques

=head2 Method Handlers

In addition to function handlers method handlers can be used.  Method
handlers are useful when you want to write code that takes advantage
of inheritance. To make the handler act as a method under mod_perl 2,
use the C<method> attribute.

See the Perl I<attributes> manpage for details on the attributes
syntax (C<perldoc attributes>).

For example:

package Bird::Eagle;
  @ISA = qw(Bird);
  
  sub handler : method {
      my ($class_or_object, $r) = @_;
      ...;
  }
  
  sub new { bless {}, __PACKAGE__ }

and then register it as:

PerlResponseHandler Bird::Eagle

When mod_perl sees that the handler has a method attribute, it passes
two arguments to it: the calling object or a class, depending on how
it was called, and the request object, as shown above.

If C<Class-E<gt>method> syntax is used for a C<Perl*Handler>, e.g.:

PerlResponseHandler Bird::Eagle->handler;

the C<:method> attribute is not required.

In the preceding configuration example, the C<handler()> method will
be called as a class (static) method.

Also, you can use objects created at startup to call methods. For example:

<Perl>
      use Bird::Eagle;
      $Bird::Global::object = Bird::Eagle->new();
  </Perl>
  ...
  PerlResponseHandler $Bird::Global::object->handler

In this example, the C<handler()> method will be called as an instance
method on the global object $C<Bird::Global::object>.

=head2 Cleaning up

It's possible to arrange for cleanups to happen at the end of various
phases. One can't rely on C<END> blocks to do the job, since these
L<don't get executed|/C_END__Blocks> until the interpreter quits, with
an exception to the
L<Registry|docs::2.0::api::ModPerl::Registry/C_END__Blocks> handlers.

Module authors needing to run cleanups after each HTTP request, should
use
C<L<PerlCleanupHandler|docs::2.0::user::handlers::http/PerlCleanupHandler>>.

Module authors needing to run cleanups at other times can always
register a cleanup callback via
C<L<cleanup_register|docs::2.0::api::APR::Pool/C_cleanup_register_>>
on the pool object of choice. Here are some examples of its usage:

To run something at the server shutdown and restart use a cleanup
handler registered on
C<L<server_shutdown_cleanup_register()|docs::2.0::api::Apache2::ServerUtil/C_server_shutdown_cleanup_register_>>
in F<startup.pl>:

#PerlPostConfigRequire startup.pl
  use Apache2::ServerUtil ();
  use APR::Pool ();
  
  warn "parent pid is $$\n";
  Apache2::ServerUtil::server_shutdown_cleanup_register((\&cleanup);
  sub cleanup { warn "server cleanup in $$\n" }

This is usually useful when some server-wide cleanup should be
performed when the server is stopped or restarted.

To run a cleanup at the end of each connection phase, assign a cleanup
callback to the connection pool object:

use Apache2::Connection ();
  use APR::Pool ();
  
  my $pool = $c->pool;
  $pool->cleanup_register(\&my_cleanup);
  sub my_cleanup { ... }

You can also create your own pool object, register a cleanup callback
and it'll be called when the object is destroyed:

use APR::Pool ();
  
  {
      my @args = 1..3;
      my $pool = APR::Pool->new;
      $pool->cleanup_register(\&cleanup, \@args);
  }
  
  sub cleanup {
      my @args = @{ +shift };
      warn "cleanup was called with args: @args";
  }

In this example the cleanup callback gets called, when C<$pool> goes
out of scope and gets destroyed. This is very similar to OO C<DESTROY>
method.

=head1 Goodies Toolkit

=head2 Environment Variables

mod_perl sets the following environment variables:

=over

=item *

C<$ENV{MOD_PERL}> - is set to the mod_perl version the server is
running under. e.g.:

mod_perl/2.000002

If C<$ENV{MOD_PERL}> doesn't exist, most likely you are not running
under mod_perl.

die "I refuse to work without mod_perl!" unless exists $ENV{MOD_PERL};

However to check which version is used it's better to use the
following technique:

use mod_perl;
  use constant MP2 => ( exists $ENV{MOD_PERL_API_VERSION} and 
                        $ENV{MOD_PERL_API_VERSION} >= 2 );

# die "I want mod_perl 2.0!" unless MP2;

=back

mod_perl passes (exports) the following shell environment variables
(if they are set) :

=over

=item *

C<PATH> - Executables search path.

=item *

C<TZ> - Time Zone.

=back

Any of these environment variables can be accessed via C<%ENV>.

=head2 Threaded MPM or not?

If the code needs to behave differently depending on whether it's
running under one of the threaded MPMs, or not, the class method
C<Apache2::MPM-E<gt>is_threaded> can be used. For example:

use Apache2::MPM ();
  if (Apache2::MPM->is_threaded) {
      require APR::OS;
      my $tid = APR::OS::current_thread_id();
      print "current thread id: $tid (pid: $$)";
  }
  else {
      print "current process id: $$";
  }

This code prints the current thread id if running under a threaded
MPM, otherwise it prints the process id.

=head2 Writing MPM-specific Code

If you write a CPAN module it's a bad idea to write code that won't
run under all MPMs, and developers should strive to write a code that
works with all mpms. However it's perfectly fine to perform different
things under different mpms.

If you don't develop CPAN modules, it's perfectly fine to develop your
project to be run under a specific MPM.

use Apache2::MPM ();
  my $mpm = lc Apache2::MPM->show;
  if ($mpm eq 'prefork') {
      # prefork-specific code
  }
  elsif ($mpm eq 'worker') {
      # worker-specific code
  }
  elsif ($mpm eq 'winnt') {
      # winnt-specific code
  }
  else {
      # others...
  }

=head1 Code Developing Nuances

=head2 Auto-Reloading Modified Modules with Apache2::Reload

META: need to port Apache2::Reload notes from the guide here. but the
gist is:

PerlModule Apache2::Reload
  PerlInitHandler Apache2::Reload
  #PerlPreConnectionHandler Apache2::Reload
  PerlSetVar ReloadAll Off
  PerlSetVar ReloadModules "ModPerl::* Apache2::*"

Use:

PerlInitHandler Apache2::Reload

if you need to debug HTTP protocol handlers. Use:

PerlPreConnectionHandler Apache2::Reload

for any handlers.

Though notice that we have started to practice the following style in
our modules:

package Apache2::Whatever;
  
  use strict;
  use warnings FATAL => 'all';

C<FATAL =E<gt> 'all'> escalates all warnings into fatal errors. So
when C<Apache2::Whatever> is modified and reloaded by C<Apache2::Reload>
the request is aborted. Therefore if you follow this very healthy
style and want to use C<Apache2::Reload>, flex the strictness by
changing it to:

use warnings FATAL => 'all';
  no warnings 'redefine';

but you probably still want to get the I<redefine> warnings, but
downgrade them to be non-fatal. The following will do the trick:

use warnings FATAL => 'all';
  no warnings 'redefine';
  use warnings 'redefine';

Perl 5.8.0 allows to do all this in one line:

use warnings FATAL => 'all', NONFATAL => 'redefine';

but if your code may be used with older perl versions, you probably
don't want to use this new functionality.

Refer to the I<perllexwarn> manpage for more information.

=head1 Integration with Apache Issues

In the following sections we discuss the specifics of Apache behavior
relevant to mod_perl developers.

=head2 HTTP Response Headers

=head3 Generating HTTP Response Headers

The best approach for generating HTTP response headers is by using the
L<mod_perl API|docs::2.0::api::Apache2::RequestRec>. Some common
headers have dedicated methods, others are set by manipulating the
C<L<headers_out|docs::2.0::api::Apache2::RequestRec/C_headers_out_>>
table directly.

For example to set the I<Content-type> header you should call
C<L<$r-E<gt>content_type|docs::2.0::api::Apache2::RequestRec/C_content_type_>>:

use Apache2::RequestRec ();
  $r->content_type('text/html');

To C<L<set|docs::2.0::api::APR::Table/C_set_>> a custom header
I<My-Header> you should call:

use Apache2::RequestRec ();
  use APR::Table;
  $r->headers_out->set(My-Header => "SomeValue");

If you are inside a registry script L<you can still
access|docs::2.0::user::coding::coding/Getting_the_C__r__Object> the
C<L<Apache2::RequestRec|docs::2.0::api::Apache2::RequestRec>> object.

Howerever you can choose a slower method of generating headers by just
printing them out before printing any response. This will work only if
C<L<PerlOptions
+ParseHeaders|docs::2.0::user::config::config/C_ParseHeaders_>> is
in effect. For example:

print "Content-type: text/html\n";
   print "My-Header: SomeValue\n";
   print "\n";

This method is slower since Apache needs to parse the text to identify
certain headers it needs to know about. It also has several
limitations which we will now discuss.

When using this approach you must make sure that the C<STDOUT>
filehandle is not set to flush the data after each print (which is set
by the value of a special perl variable C<$|>). Here we assume that
STDOUT is the currently C<select()>ed filehandle and C<$|> affects it.

For example this code won't work:

local $| = 1;
   print "Content-type: text/html\n";
   print "My-Header: SomeValue\n";
   print "\n";

Having a true C<$|> causes the first print() call to flush its data
immediately, which is sent to the internal HTTP header parser, which
will fail since it won't see the terminating C<"\n\n">. One solution
is to make sure that STDOUT won't flush immediately, like so:

local $| = 0;
   print "Content-type: text/html\n";
   print "My-Header: SomeValue\n";
   print "\n";

Notice that we C<local()>ize that change, so it L<won't affect any
other
code|docs::general::perl_reference::perl_reference/The_Scope_of_the_Special_Perl_Variables>.

If you send headers line by line and their total length is bigger than
8k, you will have the header parser problem again, since mod_perl will
flush data when the 8k buffer gets full. In which case the solution is
not to print the headers one by one, but to buffer them all in a
variable and then print the whole set at once.

Notice that you don't have any of these problems with mod_cgi, because
it ignores any of the flush attempts by Perl. mod_cgi simply opens a
pipe to the external process and reads any output sent from that
process at once.

If you use C<$r> to set headers as explained at the beginning of this
section, you won't encounter any of these problems.

Finally, if you don't want Apache to send its own headers and you want 
to send your own set of headers (non-parsed headers handlers) use the
C<L<$r-E<gt>assbackwards|docs::2.0::api::Apache2::RequestRec/C_assbackwards_>>
method. Notice that registry handlers will do that for you if the
script's name start with the C<nph-> prefix.

=head3 Forcing HTTP Response Headers Out

Apache 2.0 doesn't provide a method to force HTTP response headers
sending (what used to be done by C<send_http_header()> in Apache
1.3). HTTP response headers are sent as soon as the first bits of the
response body are seen by the special core output filter that
generates these headers. When the response handler sends the first
chunks of body it may be cached by the mod_perl internal buffer or
even by some of the output filters. The response handler needs to
flush the output in order to tell all the components participating in
the sending of the response to pass the data out.

For example if the handler needs to perform a relatively long-running
operation (e.g. a slow db lookup) and the client may timeout if it
receives nothing right away, you may want to start the handler by
setting the I<Content-Type> header, following by an immediate flush:

sub handler {
      my $r = shift;
      $r->content_type('text/html');
      $r->rflush; # send the headers out
  
      $r->print(long_operation());
      return Apache2::Const::OK;
  }

If this doesn't work, check whether you have configured any
third-party output filters for the resource in question. L<Improperly
written
filter|docs::2.0::user::handlers::filters/Writing_Well_Behaving_Filters>
may ignore the command to flush the data.

=head2 Sending HTTP Response Body

In mod_perl 2.0 a response body can be sent only during the response
phase. Any attempts to do that in the earlier phases will fail with an
appropriate explanation logged into the I<error_log> file.

This happens due to the Apache 2.0 HTTP architecture specifics. One of
the issues is that the HTTP response filters are not setup before the
response phase.

=head2 Using Signal Handlers

3rd party Apache 2 modules should avoid using code relying on
signals. This is because typical signal use is not thread-safe and
modules which rely on signals may not work portably. Certain signals
may still work for non-threaded mpms. For example C<alarm()> can be
used under prefork MPM, but it won't work on any other MPM. Moreover
the Apache developers don'tq guarantee that the signals that currently
happen to work will continue to do so in the future Apache
releases. So use them at your own risk.

It should be possible to rework the code using signals to use an
alternative solution, which works under threads. For example if you
were using C<alarm()> to trap potentially long running I/O, you can
modify the I/O logic for select/poll usage (or if you use APR I/O then
set timeouts on the apr pipes or sockets). For example, Apache 1.3 on
Unix made blocking I/O calls and relied on the parent process to send
the SIGALRM signal to break it out of the I/O after a timeout expired.
With Apache 2.0, APR support for timeouts on I/O operations is used so
that signals or other thread-unsafe mechanisms are not necessary.

CPU timeout handling is another example. It can be accomplished by
modifying the computation logic to explicitly check for the timeout at
intervals.

Talking about C<alarm()> under prefork mpm, POSIX signals seem to
work, but require Perl 5.8.x+. For example:

use POSIX qw(SIGALRM);
  my $mask      = POSIX::SigSet->new( SIGALRM );
  my $action    = POSIX::SigAction->new(sub { die "alarm" }, $mask);
  my $oldaction = POSIX::SigAction->new();
  POSIX::sigaction(SIGALRM, $action, $oldaction );
  eval {
      alarm 2;
      sleep 10 # some real code should be here
      alarm 0;
  };
  POSIX::sigaction(SIGALRM, $oldaction); # restore original
  warn "got alarm" if $@ and $@ =~ /alarm/;

For more details see:
http://search.cpan.org/dist/perl/ext/POSIX/POSIX.pod#POSIX::SigAction.

One could use the C<$SIG{ALRM}> technique, working for 5.6.x+, but it
works B<only> under DSO modperl build. Moreover starting from 5.8.0
Perl delays signal delivery, making signals safe. This change may
break previously working code.  For more information please see:
http://search.cpan.org/dist/perl/pod/perl58delta.pod#Safe_Signals and
http://search.cpan.org/dist/perl/pod/perlipc.pod#Deferred_Signals_%28Safe_Signals%29.

For example if you had the alarm code:

eval {
      local $SIG{ALRM} = sub { die "alarm" };
      alarm 3;
      sleep 10; # in reality some real code should be here
      alarm 0;
  };
  die "the operation was aborted" if $@ and $@ =~ /alarm/;

It may not work anymore. Starting from 5.8.1 it's possible to
circumvent the safeness of signals, by setting:

$ENV{PERL_SIGNALS} = "unsafe";

as soon as you start your program (e.g. in the case of mod_perl in
startup.pl). As of this writing, this workaround fails on MacOSX,
POSIX signals must be used instead.

For more information please refer to:
http://search.cpan.org/dist/perl/pod/perl581delta.pod#Unsafe_signals_again_available
and http://search.cpan.org/dist/perl/pod/perlrun.pod#PERL_SIGNALS.

Though if you use perl 5.8.x+ it's preferrable to use the POSIX
API technique explained earlier in this section.

=head1 Perl Specifics in the mod_perl Environment

In the following sections we discuss the specifics of Perl behavior
under mod_perl.

=head2 C<BEGIN> Blocks

Perl executes C<BEGIN> blocks as soon as possible, at the time of
compiling the code. The same is true under mod_perl. However, since
mod_perl normally only compiles scripts and modules once, either in
the parent server (at the server startup) or once per-child (on the
first request using a module), C<BEGIN> blocks in that code will only
be run once.  As the C<perlmod> manpage explains, once a C<BEGIN>
block has run, it is immediately undefined. In the mod_perl
environment, this means that C<BEGIN> blocks will not be run during
the response to an incoming request unless that request happens to be
the one that causes the compilation of the code, i.e. if it wasn't
loaded yet.

C<BEGIN> blocks in modules and files pulled in via C<require()> or
C<use()> will be executed:

=over 4

=item *

Only once, if pulled in by the parent process at the server startup.

=item *

Once per each child process or Perl interpreter if not pulled in by
the parent process.

=item *

An additional time, once per each child process or Perl interpreter if
the module is reloaded off disk again via
C<L<Apache2::Reload|docs::2.0::api::Apache2::Reload>>.

=item *

Unpredictable if you fiddle with C<%INC> yourself.

=back

The C<BEGIN> blocks behavior is different in
C<L<ModPerl::Registry|docs::2.0::api::ModPerl::Registry/C_BEGIN__Blocks>>
and
C<L<ModPerl::PerlRun|docs::2.0::api::ModPerl::PerlRun/C_BEGIN__Blocks>>
handlers, and their subclasses.

=head2 C<CHECK> and C<INIT> Blocks

C<CHECK> and C<INIT> blocks run when the source code compilation is
complete, but before the program starts. C<CHECK> can mean
"checkpoint" or "double-check" or even just "stop". C<INIT> stands for
"initialization". The difference is subtle; C<CHECK> blocks are run
just after the compilation ends, C<INIT> just before the runtime
begins. (Hence the C<-c> command-line perl option runs C<CHECK> blocks
but not C<INIT> blocks.)

Perl only calls these blocks during I<perl_parse()>, which mod_perl
calls once at startup time. Under threaded mpm, these blocks will be
called once per C<L<parent perl interpreter
startup|docs::2.0::user::config::config/C_Parent_>>. Therefore
C<CHECK> and C<INIT> blocks don't work after the server is started,
for the same reason these code samples don't work:

% perl -e 'eval qq(CHECK { print "ok\n" })'
  % perl -e 'eval qq(INIT  { print "ok\n" })'

=head2 C<END> Blocks

As the C<perlmod> manpage explains, an C<END> block is executed as
late as possible, that is, when the interpreter exits. So for example
mod_cgi will run its C<END> blocks on each invocation, since on every
invocation it starts a new interpreter and then kills it when the
request processing is done.

In the mod_perl environment, the interpreter does not exit after
serving a single request (unless it is configured to do so) and hence
it will run its C<END> blocks only when it exits, which usually
happens during the server shutdown, but may also happen earlier than
that (e.g. a process exits because it has served a
C<MaxRequestsPerChild> number of requests).

mod_perl does L<make a special
case|docs::2.0::api::ModPerl::Registry/C_END__Blocks>
for scripts running under
C<L<ModPerl::Registry|docs::2.0::api::ModPerl::Registry>> and friends.

The L<Cleaning up|/Cleaning_up> section explains how to deal with
cleanups for non-Registry handlers.

C<L<ModPerl::Global|docs::2.0::api::ModPerl::Global>> API:
C<L<special_list_register|docs::2.0::api::ModPerl::Global/C_special_list_register_>>,
C<L<special_list_call|docs::2.0::api::ModPerl::Global/C_special_list_call_>>
and
C<L<special_list_clear|docs::2.0::api::ModPerl::Global/C_special_list_clear_>>,
internally used by registry handlers, can be used to run C<END> blocks
at arbitrary times.

=head2 Request-localized Globals

mod_perl 2.0 provides two types of C<SetHandler> handlers:
C<L<modperl|docs::2.0::user::config::config/C_modperl_>> and
C<L<perl-script|docs::2.0::user::config::config/C_perl_script_>>.
Remember that the C<SetHandler> directive is only relevant for the
response phase handlers, it neither needed nor affects non-response
phases.

Under the handler:

SetHandler perl-script

several special global Perl variables are saved before the handler is
called and restored afterwards. This includes: C<%ENV>, C<@INC>,
C<$/>, C<STDOUT>'s C<$|> and C<END> blocks array (C<PL_endav>).

Under:

SetHandler modperl

nothing is restored, so you should be especially careful to remember
localize all special Perl variables so the local changes won't affect
other handlers.

=head2 C<exit>

In the normal Perl code exit() is used to stop the program flow and
exit the Perl interpreter. However under mod_perl we only want the
stop the program flow without killing the Perl interpreter.

You should take no action if your code includes exit() calls and it's
OK to continue using them. mod_perl worries to override the exit()
function with L<its own version|docs::2.0::api::ModPerl::Util/C_exit_>
which stops the program flow, and performs all the necessary cleanups,
but doesn't kill the server. This is done by overriding:

*CORE::GLOBAL::exit = \&ModPerl::Util::exit;

so if you mess up with C<*CORE::GLOBAL::exit> yourself you better know
what you are doing.

You can still call C<CORE::exit> to kill the interpreter, again if you
know what you are doing.

One caveat is when C<exit> is called inside C<eval> -- L<the
ModPerl::Util::exit
documentation|docs::2.0::api::ModPerl::Util/C_exit_> explains how to
deal with this situation.

=head1 C<ModPerl::Registry> Handlers Family

=head2 A Look Behind the Scenes

If you have a CGI script F<test.pl>:

#!/usr/bin/perl
  print "Content-type: text/plain\n\n";
  print "Hello";

a typical registry family handler turns it into something like:

package foo_bar_baz;
  sub handler {
      local $0 = "/full/path/to/test.pl";
  #line 1 test.pl
      #!/usr/bin/perl
      print "Content-type: text/plain\n\n";
      print "Hello";
  }

Turning it into an almost full-fledged mod_perl handler. The only
difference is that it handles the return status for you. (META: more
details on return status needed.)

It then executes it as:

foo_bar_baz::handler($r);

passing the C<L<$r|docs::2.0::api::Apache2::RequestRec>> object as the
only argument to the C<handler()> function.

Depending on the used registry handler the package is made of the file
path, the uri or anything else. Check the handler's documentation to
learn which method is used.

=head2 Getting the C<$r> Object

As explained in L<A Look Behind the Scenes|/A_Look_Behind_the_Scenes>
the C<$r> object is always passed to the registry script's special
function C<handler> as the first and the only argument, so you can get
this object by accessing C<@_>, since:

my $r = shift;
  print "Content-type: text/plain\n\n";
  print "Hello";

is turned into:

sub handler {
      my $r = shift;
      print "Content-type: text/plain\n\n";
      print "Hello";
  }

behind the scenes. Now you can use C<$r> to call various mod_perl
methods, e.g. rewriting the script as:

my $r = shift;
  $r->content_type('text/plain');
  $r->print();

If you are deep inside some code and can't get to the entry point to
reach for C<$r>, you can use
C<L<Apache2-E<gt>request|docs::2.0::api::Apache2::RequestUtil/C_request_>>.

=head1 Threads Coding Issues Under mod_perl

The following sections discuss threading issues when running mod_perl
under a threaded MPM.

=head2 Thread-environment Issues

The "only" thing you have to worry about your code is that it's
thread-safe and that you don't use functions that affect all threads
in the same process.

Perl 5.8.0 itself is thread-safe. That means that operations like
C<push()>, C<map()>, C<chomp()>, C<=>, C</>, C<+=>, etc. are
thread-safe. Operations that involve system calls, may or may not be
thread-safe. It all depends on whether the underlying C libraries used
by the perl functions are thread-safe.

For example the function C<localtime()> is not thread-safe when the
implementation of C<asctime(3)> is not thread-safe. Other usually
problematic functions include C<readdir()>, C<srand()>, etc.

Another important issue that shouldn't be missed is what some people
refer to as I<thread-locality>. Certain functions executed in a single
thread affect the whole process and therefore all other threads
running inside that process. For example if you C<chdir()> in one
thread, all other thread now see the current working directory of that
thread that C<chdir()>'ed to that directory. Other functions with
similar effects include C<umask()>, C<chroot()>, etc. Currently there
is no cure for this problem. You have to find these functions in your
code and replace them with alternative solutions which don't incur
this problem.

For more information refer to the I<perlthrtut>
(I<http://perldoc.perl.org/perlthrtut.html>) manpage.

=head2 Deploying Threads

This is actually quite unrelated to mod_perl 2.0. You don't have to
know much about Perl threads, other than L<Thread-environment
Issues|/Thread_environment_Issues>, to have your code properly work 
under threaded MPM mod_perl.

If you want to spawn your own threads, first of all study how the new
ithreads Perl model works, by reading the I<perlthrtut>, I<threads>
(I<http://search.cpan.org/search?query=threads>) and
I<threads::shared>
(I<http://search.cpan.org/search?query=threads%3A%3Ashared>) manpages.

Artur Bergman wrote an article which explains how to port pure Perl
modules to work properly with Perl ithreads. Issues with C<chdir()>
and other functions that rely on shared process' datastructures are
discussed.  I<http://www.perl.com/lpt/a/2002/06/11/threads.html>.

=head2 Shared Variables

Global variables are only global to the interpreter in which they are
created. Other interpreters from other threads can't access that
variable. Though it's possible to make existing variables shared
between several threads running in the same process by using the
function C<threads::shared::share()>. New variables can be shared by
using the I<shared> attribute when creating them. This feature is
documented in the I<threads::shared>
(I<http://search.cpan.org/search?query=threads%3A%3Ashared>) manpage.

=head1 Maintainers

Maintainer is the person(s) you should contact with updates,
corrections and patches.

=over

=item *

=back

=head1 Authors

=over

=item *

=back

Only the major authors are listed above. For contributors see the
Changes file.

=cut