Update to 1.20:

1.20    Wed Mar 12 23:56:11 CDT 2008
-----------------------------------
[FIXES]
stuff_inputs() used to do nothing.  Now it works.
http://code.google.com/p/www-mechanize/issues/detail?id=9

Fixed punctuation in some error messages.

Fixed compatibility with WWW::Mechanize 1.36.

1.18    Thu Dec  6 10:12:14 CST 2007
------------------------------------
[ENHANCEMENTS]
Added default descriptions for most test assertions.

[FIXES]
HTML::Lint is now properly optional.

Added delays in all the tests that use HTTP::Server::Simple to give
it time to correctly fire up.

1.16    Mon Oct 29 15:34:21 CDT 2007
------------------------------------
[ENHANCEMENTS]
Added $mech->post_ok().  Thanks, Greg Sheard.

Added $mech->submit_form_ok().  Thanks, Mark Stosberg.

Added $mech->html_lint_ok()

[FIXES]
Fixed some bugs that were causing test failures.

1.14    Fri May 11 16:22:02 CDT 2007
------------------------------------
[FIXES]
Fixes test failures.  Thanks to Mark Blackman for RT #26602:

The module tests currently spawn a server (based on
HTTP::Server::Simple::CGI) which is killed when a __DIE__
signal is received, normally only when the invoking
perl dies. However, as perlvar makes clear, the __DIE__
signal is received when eval blocks die as well. The
new version (1.22) of WWW::Mechanize now calles
HTTP::Message::decoded_content which calls Encode.pm
which has a eval block that require's Encode::ConfigLocal
which is usually not present, so the eval dies and the
HTTP server is killed as soon as the $mech object tries
to do a 'get'. It's simple to use a system variable,
$^S to find out if the __DIE__ signal is being called
for an eval so we ignore that case with the patch
attached.

The stuff_inputs() function now actually works.

[DOCUMENTATION]
*  Made the synopsis show that T:W:M doesn't take the tests=>$x
like other Test::* modules.  It'd be nice if it did, though.

(wiz)