+=head1 FUNCTIONS
+
+=head2 C<msg>
+
+ my $msg = msg($object, $method, $file, $line);
+
+Returns the default error message that C<indirect> generates when an indirect method call is reported.
+
+=cut
+
+sub msg {
+ my $obj = $_[0];
+
+ join ' ', "Indirect call of method \"$_[1]\" on",
+ ($obj =~ /^\s*\{/ ? "a block" : "object \"$obj\""),
+ "at $_[2] line $_[3].\n";
+};
+
+=head1 CONSTANTS
+
+=head2 C<I_THREADSAFE>
+
+True iff the module could have been built with thread-safety features enabled.
+
+=head2 C<I_FORKSAFE>
+
+True iff this module could have been built with fork-safety features enabled.
+This will always be true except on Windows where it's false for perl 5.10.0 and below .
+
+=head1 DIAGNOSTICS
+
+=head2 C<Indirect call of method "%s" on object "%s" at %s line %d.>
+
+The default warning/exception message thrown when an indirect method call on an object is found.
+
+=head2 C<Indirect call of method "%s" on a block at %s line %d.>
+
+The default warning/exception message thrown when an indirect method call on a block is found.
+
+=head1 ENVIRONMENT
+
+=head2 C<PERL_INDIRECT_PM_DISABLE>
+
+If this environment variable is set to true when the pragma is used for the first time, the XS code won't be loaded and, although the C<'indirect'> lexical hint will be set to true in the scope of use, the pragma itself won't do anything.
+In this case, the pragma will always be considered to be thread-safe, and as such L</I_THREADSAFE> will be true.
+This is useful for disabling C<indirect> in production environments.
+
+Note that clearing this variable after C<indirect> was loaded has no effect.
+If you want to re-enable the pragma later, you also need to reload it by deleting the C<'indirect.pm'> entry from C<%INC>.
+
+=head1 CAVEATS
+
+The implementation was tweaked to work around several limitations of vanilla C<perl> pragmas : it's thread safe, and does not suffer from a C<perl 5.8.x-5.10.0> bug that causes all pragmas to propagate into C<require>d scopes.
+
+Before C<perl> 5.12, C<meth $obj> (no semicolon) at the end of a file is not seen as an indirect method call, although it is as soon as there is another token before the end (as in C<meth $obj;> or C<meth $obj 1>).
+If you use C<perl> 5.12 or greater, those constructs are correctly reported.
+
+With 5.8 perls, the pragma does not propagate into C<eval STRING>.
+This is due to a shortcoming in the way perl handles the hints hash, which is addressed in perl 5.10.
+
+The search for indirect method calls happens before constant folding.
+Hence C<my $x = new Class if 0> will be caught.
+
+=head1 REFERENCES
+
+Numerous articles have been written about the quirks of the indirect object construct :
+
+=over 4
+
+=item *
+
+L<http://markmail.org/message/o7d5sxnydya7bwvv> : B<Far More Than Everything You've Ever Wanted to Know about the Indirect Object syntax>, Tom Christiansen, 1998-01-28.
+
+This historical post to the C<perl5-porters> mailing list raised awareness about the perils of this syntax.
+
+=item *
+
+L<http://www.shadowcat.co.uk/blog/matt-s-trout/indirect-but-still-fatal> : B<Indirect but still fatal>, Matt S. Trout, 2009-07-29.
+
+In this blog post, the author gives an example of an undesirable indirect method call on a block that causes a particularly bewildering error.
+
+=back
+