This patch adds a ->kill() object method to the 'threads' package to
support sending signals to threads.

A few comments on your patch :

Jerry D. Hedden wrote:
diff -urN
2006-04-28 08:11:04.000000000 -0400
2006-05-10 14:05:42.000000000 -0400
@@ -16,10 +50,21 @@
'NRECURS' =1);
} else {
# CPAN
- push(@conditional_params, 'DEFINE' ='-DHAS_PPPRT_H');
+
+ # Verify that a 'C' compiler is available
+ require File::Spec;
+ require Config;
+ import Config;

This require/import seems to be unnecessary, since Config is loaded in
have_cc().

diff -urN
2006-05-09 04:04:58.000000000 -0400
2006-05-10 14:23:42.000000000 -0400
@@ -146,6 +146,8 @@
$stack_size = threads->get_stack_size();
$old_size = threads->set_stack_size(32*4096);

+ $thr->kill('SIGUSR1');
+
=head1 DESCRIPTIN

Perl 5.6 introduced something called interpreter threads. Interpreter threads
@@ -405,6 +407,88 @@

=back

+=head1 THREAD SIGNALLING
+
+If Perl has been compiled to use safe signals (i.e., was not built with
+C<PERLLD_SIGNALS- see C<perl -V>), then signals may be sent and acted upon
+by individual threads.

I think you shouldn't mention PERLLD_SIGNALS at all in threads.pm's
doc, it doesn't exist anymore in bleadperl or maintperl. However maybe
you can mention than using unsafe signals (see perlrun) breaks code that
uses thread signalling.

+=over 4
+
+=item $thr->kill('SIG');

Some suggestions that may or may not make sense :

Potential cause of error : SIGF isn't a valid %SIG key (one writes
%SIG{USR1} for example), and it often gets me, so maybe you should allow
F as well as SIGF here ?

Also, the error message could be "No such signal: SIG%s" like in
perldiag. (documented later in your pod)

Should we croak or warn when an unknown signal name is used ? Assigning
to $SIG{F} only warns. Maybe this should do likewise, for consistency.
Not all signals exist everywhere.