Re: Tests for non-default configurations

On Tue, Nov 8, 2016 at 11:19 AM, James Graham <james@hoppipolla.co.uk>
wrote:

> We just had a situation where a gecko contributor wrote some
> web-platform-tests for a feature. So far, so good. However a number of
> these tests depended on disabling mixed content blocking, which is possible
> in Firefox through a pref. AIUI it's also possible in other at least some
> other browsers through a flag.
>

Blink (and WebKit, last I checked) exposes a test-only API to disable mixed
content
checking (`window.testRunner.setAllowRunningOfInsecureContent(...)`).


> At the present time I think we will take the opinion that such tests
> relating to features that are not on by default and will not become the
> default in the future are not worth upstreaming, and keep them in a
> mozilla-specific directory. Is this acceptable to everyone or are there
> situations where sharing such tests is valuable? If there are, how do you
> want to communicate the information that a specific setting is required?
>

That works for me. For this setting in particular, I'd be pretty surprised
if we ever turned it off by default in Blink; it's something we can assume
as foundational to the platform, even if some folks' clients are configured
strangely.

In other words, unless a setting is controlled via an API we ship as part
of the platform, I don't think the setting should be considered part of the
platform.

-mike