My attempt to reproduce this has failed.

use strict;
use warnings;

use Test::More tests => 2;
use DBI;
use Data::Dumper;

my $dbh = DBI->connect ('dbi:SQLite:dbname=:memory:', '', '');

my $sth = $dbh->prepare ('SELECT 1 = 1;');
$sth->execute;
$sth->finish;
ok ! $sth->{Active}, 'Handle is inactive';
my $res = $sth->fetchall_arrayref ();

is $res, undef, 'Result is undef' or diag Dumper ($res);

diag "DBI $DBI::VERSION";
[download]

$ perl inactive.t
1..2
ok 1 - Handle is inactive
not ok 2 - Result is undef
#   Failed test 'Result is undef'
#   at /tmp/inactive2.t line 16.
#          got: 'ARRAY(0x2c41050)'
#     expected: undef
# $VAR1 = [];
# DBI 1.643
# Looks like you failed 1 test of 2.
[download]

On this apparently inactive handle it still returns an empty arrayref. Where am I going wrong?

I tried your test using MySQL and it gave the same result. But I checked the handle for errors and it produced "fetch() without execute()". So perhaps returning an empty arrayref is correct, as there was an error. Interesting that SQLite did not throw that error.

You can't do any fetch after calling finish(), I don't think

I'm not trying to perform a successful fetch - I'm trying to reproduce the documented behaviour. Hence calling finish first in order to ensure that the handle is inactive. And the fetch still doesn't return undef.

---

🦛

Thank for trying all that.

In the end, I decided it would just be easier if I checked for null in both places and move on.

checking for undef in the return, and no records when I check the array size

Another option is to enable RaiseError, and then you get an exception instead of undef, and cleaner code to wrap up errors.