[% setvar title MT-Safe Autovariables in perl 5.005 Threading %]
<div id="archive-notice">
    <h3>This file is part of the Perl 6 Archive</h3>
    <table bgcolor="red"><tr><td><font color="white"> Note: these documents may be out of date. Do <b>not</b> use as reference!</font></tr></td></table>
    <p>To see what is currently happening visit <a href="http://www.perl6.org/">http://www.perl6.org/</a></p>
</div>
<div class='pod'>
<a name='TITLE'></a><h1>TITLE</h1>
<p>MT-Safe Autovariables in perl 5.005 Threading</p>
<a name='VERSION'></a><h1>VERSION</h1>
<pre>  Maintainer: Michael Maraist &lt;<a href='mailto:maraist@udel.edu'>maraist@udel.edu</a>&gt;
  Date: 25 Sep 2000
  Mailing List: <a href='mailto:perl6-internals@perl.org'>perl6-internals@perl.org</a>
  Number: 293
  Version: 1
  Status: Developing</pre>
<a name='ABSTRACT'></a><h1>ABSTRACT</h1>
<p>There are currently two threading models in Perl.  The first was 5.005s
true-threads.  This obviously caused problems for numerous modules (and
programs alike).  It was later replaced by the &quot;virtual&quot; process, who's main
benifit was to provide efficient forks through the use of spawning a
perl-thread and seperating environment spaces..</p>
<p>Perl5.005's Threading may go away, or it may continue in an experimental stage
for a while.  At the very least, perl's threading provides an easy way to test
CIS OS theory.  This RFC attempts to identify at least one existing
implementation problem with 5.005 threads, that of auto-variables. To my
understanding, there is only one copy of each auto-variable, and thus, multiple
threads would have race conditions for them.</p>
<a name='DESCRIPTION'></a><h1>DESCRIPTION</h1>
<p>Make auto-variables in a threading environment &quot;thread-specific&quot;.  The current
implementation of:</p>
<pre> for my $i ( 0 .. 100000 ) { do something }</pre>
<p>causes core dumps when there are race conditions.  It is my guess that the
problem is because this produces a lazy iteration which unfortunately makes use
of an auto-variable other than $i (above).  It is these sort of auto-vars
(especially in light of lazy operators), that can cause race-conditions.</p>
<p>Normally you would provide locks to protect such code, but it is not obvious to
the developer that a simple for-loop using a lexical variable would even have a
race condition.  Beyond that, it might be possible that the generation of such
auto-vars might not be thread-safe.</p>
<a name='IMPLEMENTATION'></a><h1>IMPLEMENTATION</h1>
<p>For all auto-generated non-lexical meta-data, either apply a lock around usage,
or make heavy use of thread-specific keys/variables.</p>
<a name='REFERENCES'></a><h1>REFERENCES</h1>
<p>RFC 185: Thread Programming Model</p>
<p>perldoc Thread: 5.005 threading</p>
<p>RFC 178: Lightweight Threads</p>
<p>RFC 1: Implementation of Threads in Perl</p>
<p>man perlop: foreach lazy iteration of 1..10</p>
<p>RFC 24: Data types: Semi-finite (lazy) lists</p>
<p>RFC 123: Builtin: lazy</p>
<p>RFC 128: Subroutines: Extend subroutine contexts to include name parameters and lazy arguments</p>
<p>RFC 285: Lazy Input</p>
</div>