On Mon, Feb 20, 2017 at 10:16 AM, Joe Orton <jor...@redhat.com> wrote: > > FYI, since I've seen people thinking seriously about test suites, this > kind of issue is impossible to test comprehensively with the current > test framework. We really need to spin up httpd multiple times with > different configurations, to test different combinations of global to > vhost merges with default vs non-default options. > > (If there is some way to do that with the current test suite, I'd be > interested to hear anyway.)
The headache is that testing anything more complex than loopback implies you kick off a remote with -server and then pound on that remote it from the client/framework. Perhaps if we spun up parallel instances in the same port-ordered manner, but those instances were truly distinct configs? (These need distinct var/ directories/pid files etc, of course.)