svn commit: r420966 - head/security/openvpn

[Alexey Dokuchaev]

On Sat, Aug 27, 2016 at 07:58:08PM +1000, Kubilay Kocak wrote:
> On 27/08/2016 7:32 PM, Matthias Andree wrote:
> > New Revision: 420966
> > URL: https://svnweb.freebsd.org/changeset/ports/420966
> > 
> > Log:
> >   Make self-test the TEST option, support make test. Enabled by default.
> 
> >   NB: This is a critical port with many users, and the test is low on
> >   resources, it takes two minutes idling, waiting for timers to expire.
> 
> I'm a HUGE proponent of additional tests/QA/quality in ports/packaging.
> 
> However, users can run make test at any time to verify QA in their
> environments. What is the actual rationale *for* running tests
> post-build in all/any environments (including packaging) by default? Is
> the aim to have packaging fail if it doesn't pass?
> 
> Should all our 'critical' ports do the same thing? Should we instead
> mark 'critical' ports in some manner and have the framework run these
> tests-post-build by default?

+1.  We do have proper support for tests since CHANGES at 20150928.  Ideally,
those TEST options should be gone, and on-demand tests/QA logic implemented
in standard terms that are usable by package builders.  Ports users do not
need to opt-out of tests (by deselecting an OPTION or even patching port's
Makefile); those who need it should just run `make test'.

./danfe