On 2023-01-02 Mo 09:45, Andrew Dunstan wrote:
> On 2023-01-01 Su 18:31, Andrew Dunstan wrote:
>> On 2023-01-01 Su 14:02, Thomas Munro wrote:
>>> On Mon, Jan 2, 2023 at 3:04 AM Andrew Dunstan <andrew@dunslane.net> wrote:
>>>> On 2022-12-19 Mo 11:16, Andrew Dunstan wrote:
>>>>> There is currently no test for the use of ldapbindpasswd in the
>>>>> pg_hba.conf file. This patch, mostly the work of John Naylor, remedies that.
>>>>>
>>>>>
>>>> This currently has failures on the cfbot for meson builds on FBSD13 and
>>>> Debian Bullseye, but it's not at all clear why. In both cases it fails
>>>> where the ldap server is started.
>>> I think it's failing when using meson. I guess it fails to fail on
>>> macOS only because you need to add a new path for Homebrew/ARM like
>>> commit 14d63dd2, so it's skipping (it'd be nice if we didn't need
>>> another copy of all that logic). Trying locally... it looks like
>>> slapd is failing silently, and with some tracing I can see it's
>>> sending an error message to my syslog daemon, which logged:
>>>
>>> 2023-01-02T07:50:20.853019+13:00 x1 slapd[153599]: main: TLS init def
>>> ctx failed: -1
>>>
>>> Ah, it looks like this test is relying on "slapd-certs", which doesn't exist:
>>>
>>> tmunro@x1:~/projects/postgresql/build$ ls testrun/ldap/001_auth/data/
>>> ldap.conf ldappassword openldap-data portlock slapd-certs slapd.conf
>>> tmunro@x1:~/projects/postgresql/build$ ls testrun/ldap/002_bindpasswd/data/
>>> portlock slapd.conf
>>>
>>> I didn't look closely, but apparently there is something wrong in the
>>> part that copies certs from the ssl test? Not sure why it works for
>>> autoconf...
>>
>> Let's see how we fare with this patch.
>>
>>
> Not so well :-(. This version tries to make the tests totally
> independent, as they should be. That's an attempt to get the cfbot to go
> green, but I am intending to refactor this code substantially so the
> common bits are in a module each test file will load.
>
>
This version factors out the creation of the LDAP server into a separate
perl Module. That makes both the existing test script and the new test
script a lot shorter, and will be useful for the nearby patch for a hook
for the ldapbindpassword.
cheers
andrew
--
Andrew Dunstan
EDB: https://www.enterprisedb.com
