How can I avoid encoding errors using Net::OpenID::Consumer with Yahoo OpenIDs?

I've written a Dancer web app that utilizes Net::OpenID::Consumer to consume OpenIDs for authentication. It works well with Google and MyOpenID, but not Yahoo. When a user tries to authenticate using their Yahoo account, HTML::Parser warns:

Parsing of undecoded UTF-8 will give garbage when decoding entities

and this warning kills my app (rightfully so).

I don't see any existing bugs with Net::OpenID::Consumer (or Common) that relate to this.
The HTTP headers and the HTML meta tags both specify UTF-8 for the 'claimed id' URI.
Why would the response not be decoded for HTML::Parser? Am I missing something obvious?

Here's the relevant code:

get '/openid_landing' => sub {
    my $params = params();
    my $csr = Net::OpenID::Consumer->new(
        ua => LWP::UserAgent->new(),
        consumer_secret => $secret,
        params => $params,
    );  
    my $id = $params->{'openid.claimed_id'};

    if (my $setup_url = $csr->user_setup_url) {
        redirect $setup_url;

    } elsif ($csr->user_cancel) {
        redirect uri_for('/');

    } elsif (my $vident = $csr->verified_identity) {
       # verified identity, log in or register user
       ...

    } else {
        die "Error validating identity: " . $csr->err;
    } 
};

Solution

The bug is in Net/OpenID/URIFetch.pm on lines 122-128 of version 1.14 (latest) It's using the raw content instead of the decoded content of the response object. Just remove the manual gzip decoding and use the decoded_content method in the response.

I haven't filed a bug report yet, feel free. :)

Here's a diff you can apply to fix it:

122c122
<         my $content = $res->decoded_content;
---
>         my $content = $res->content;
125a126,129
>         if ($res->content_encoding && $res->content_encoding eq 'gzip') {
>             $content = Compress::Zlib::memGunzip($content);
>         }
>