my $ua = LWP::UserAgent->new();
my $response = $ua->get("http://www.barackobama.com/2008/08/07/obama_t
+alks_about_reviving_eco.php");
# print $response->decoded_content;                                   
+                                        
my $html = $response->decoded_content();
my $tree = HTML::TreeBuilder->new;
$tree->parse($html);
print $tree->as_HTML;
[download]

That gives no errors for me either. And gets and parses the contents fine. Both on LWP 5.805 with perl 5.10 and LWP 5.814 on Perl 5.8.8. Time to upgrade?

It still exists in the latest version. The docs say

(W) The first chunk parsed appears to contain undecoded UTF-8 and one
or more argspecs that decode entities are used for the callback
handlers.

The result of decoding will be a mix of encoded and decoded characters
for any entities that expand to characters with code above 127.  This
is not a good thing.

The solution is to use the Encode::encode_utf8() on the data before
feeding it to the $p->parse().  For $p->parse_file() pass a file that
has been opened in ":utf8" mode.

The parser can process raw undecoded UTF-8 sanely if the C<utf8_mode>
is enabled or if the "attr", "@attr" or "dtext" argspecs is avoided.
[download]

It could be that that server didn't specify the character encoding of the content.

Well I upgraded to 5.814 (from 5.805) and that does seem to have fixed the problem. Thanks.