The undef shouldn't allow this.

I normally only use ->fetchall_hashref

Update

Did you show us how you checked on this error after the fetch?

Or do you auto-raise them with RaiseError of DBI ?

Update

In general it's better to supply a SSCCE to reproduce the issue instead of all this less relevant CGI Code

Depending on the number of rows, fetching them all can by quite memory consuming. My code (although for DBD::Pg) usually looks like this:

use DBI;
use Carp;
use English;
...
my $dbh = DBI->connect($self->{dburl}, $self->{dbuser}, $self->{dbpass
+word},
                               {
                                   AutoCommit => 0, # This is a proper
+ database, not a key-value store
                                   RaiseError => 0, # Child protection
+ mode off. We are doing our own error checking, thank you very much
                                   AutoInactiveDestroy => 1, # Just do
+ the right thing when using fork()
                               }) or croak("$EVAL_ERROR");
...
# sub bla($dbh) {
my $selsth = $dbh->prepare_cached("SELECT foo, bar, baz FROM bli WHERE
+ bla = ? ORDER BY blub")
        or croak($dbh->errstr); # If the statement is wrong, no point 
+of internal error handling

if(!$selsth->execute("foobar")) {
    $reph->debuglog($dbh->errstr);
    $dbh->rollback;
    return 0; # Tell caller we failed
}

# Work on the data row-by-row
while((my $line = $selsth->fetchrow_array)) {
    ...
    # Do something with the data
    ...
}

$selsth->finish; # Make sure we don't leave database pointers and stuf
+f around

$dbh->commit; # Finish the current transaction, unlock all locked tabl
+es/rows and stuff.
              # commit() may or may not happen in THIS function, depen
+ding on if the caller needs to
              # chain multiple functions in a single transaction


return 1; # tell caller we succeeded.
...
[download]

PerlMonks XP is useless? Not anymore: XPD - Do more with your PerlMonks XP
Also check out my sisters artwork and my weekly webcomics