Re: Disabling src/test/[ssl|ldap] when not building with SSL/LDAP support

On Wed, Feb 28, 2018 at 10:02:37AM -0500, Peter Eisentraut wrote:
> On 2/24/18 18:29, Michael Paquier wrote:
>> Sure. But then I think that it would be nice to show up on screen the
>> reason why the test failed if possible. As of now if SSL is missing the
>> whole run shows in red without providing much useful information.
>> Instead of 0001 as shaped previously, why not using as well diag to show
>> the failure on the screen?
>>
>> For example the following block at the top of each test:
>> if (!check_pg_config("#define USE_OPENSSL 1"))
>> {
>> diag "SSL tests not supported without support in build";
>> die;
>> }
>
> I think BAIL_OUT() is intended for this.

That's a better idea. I have reworked that in 0001. What do you think?
This has the same effect as diag(), which shows information directly to
the screen, and that's the behavior I was looking for.

>>> What I had in mind would consist of something like this in
>>> src/test/Makefile:
>>>
>>> ifeq ($(with_ldap),yes)
>>> ifneq (,$(filter ldap,$(YOUR_VARIABLE_HERE)))
>>> SUBDIRS += ldap
>>> endif
>>> endif
>>
>> OK. So let's call it PG_TEST_EXTRA_INSECURE or PG_TEST_EXTRA, which can
>> take 'ldap', 'ssl' or 'ldap,ssl' as valid values. Seeing that
>> documented is really necessary in my opinion. Any idea for a better
>> name?
>
> I don't have a great idea about the name. The value should be
> space-separated to work better with make functions.

I have concluded about using the most simple name: PG_TEST_EXTRA. A
documentation attempt is added as well. This is unfortunately close to
EXTRA_TESTS. It would be tempting to directly reuse EXTRA_TESTS as
well, however this is used only for the core regression tests now, so a
separated variable seems adapted to me.

Thoughts and reviews are welcome.
--
Michael