Sorry in advance for such a long post here...

I'm using select() to select an item from a 'select' list (duh! :-) - but I do not know ahead of time what the names of said items are.

So, I wrote something like this:

my $ans = 1; # or whatever.
while ($ans)
{
    %pickme = ( n => 1 ); #yes, this should be outside the while. So s
+ue me ;-)
    $ans = $mech->select( 'server', \%pickme );
    if ($ans)
    {
        $inputobj = $mech->current_form()->find_input( 'remove', 'subm
+it');
        die 'argh! Cannot find delete button on the iSCSI Access page!
+' unless $inputobj;
        $ans = $mech->click_button( input => $inputobj );
    }
}
print "Done!  Now, accept the empty list\n";
# ok, now accept a (hopefully) empty list of allowed accessors
$inputobj = $mech->current_form()->find_input( 'accept', 'submit');
die 'argh! Cannot find Accept button on the iSCSI Access page!' unless
+ $inputobj;
$ans = $mech->click_button( input => $inputobj );
[download]

However, it turns out that select will ALWAYS think that it can select entry 1, even if the list is empty. In other words, even when the html looks like this:

<td>
<input type="hidden" name="serverList" value="">

<select size="10" name="server" style="width: 330px">

</select>
</td>
[download]

The above code will get '1' back from select and go running on forever. Not a good thing...

I discovered that if entry 2 didn't exist, and you tried to select THAT entry you'd get an appropriate error, so I ended up with this as a workaround:

while ($ans)
{
    %pickme = ( n => 2 );
    eval { $ans = $mech->select( 'server', \%pickme ); };
    if ($ans eq 1)
    {
        $inputobj = $mech->current_form()->find_input( 'remove', 'subm
+it');
        die 'argh! Cannot find delete button on the iSCSI Access page!
+' unless $inputobj;
        $ans = $mech->click_button( input => $inputobj );
    }
    else
    {
        $ans = 0;
    }
}
%pickme = ( n => 1 );
$ans = $mech->select( 'server', \%pickme );
$inputobj = $mech->current_form()->find_input( 'remove', 'submit');
die 'argh! Cannot find delete button on the iSCSI Access page!' unless
+ $inputobj;
$ans = $mech->click_button( input => $inputobj );

print "Done!  Now, accept the empty list\n";
# ok, now accept a (hopefully) empty list of allowed accessors
$inputobj = $mech->current_form()->find_input( 'accept', 'submit');
die 'argh! Cannot find Accept button on the iSCSI Access page!' unless
+ $inputobj;
$ans = $mech->click_button( input => $inputobj );
[download]

It seems to me that the first form should have worked (see the Mechanize page under select() for details of why - won't bore everyone here with that), so I'm looking at the Mechanize and Form code to see why it allows a 'select' of 1 when there are no items.

I've looked at WWW::Mechanize, and the issue seems to be in Form instead. Here is the section that seems to be involved here:

        for (@{$self->{menu}}) {
            if ((defined($val) && defined($_->{value}) && $val eq $_->
+{value}) ||
                (!defined($val) && !defined($_->{value}))
               )
            {
                $cur = $i;
                $disabled = $_->{disabled};
                last unless $disabled;
            }
            $i++;
        }
[download]

When there is nothing in the select list, the above test still passes! I note, when stepping through the code, that if I print $val when I'm looking at the 'big if' above, I get a blank line, BUT defined($val) is true!

So, am I just confused, is the above test wrong, is my code wrong, or ???

(Conclusion: the HTML is wrong - its illegal to have a select with no option. You should have a single empty option, and detect that its an empty option (I'll have to RTFM for how to do that, but that's just Fine :-)

Thanks!

Final comment - thanks to Hermit (psini for his/her help - and sorry for not thinking to RTFM on how HTTP select fields work, once the comment was made (too focussed on Mechanize::select and figuring out how to get around having no (obvious) way to tell when there's nothing there....

Anyway, I consider this thread finished (and now I know what bug to file with the authors). I've modified the subject, at the suggestion of ambrus, to make it better. Thanks again!