Re: Generating manifests for the test runner

Thanks!

On Wed, Jul 23, 2014 at 11:28 AM, James Graham <james@hoppipolla.co.uk> wrote:
> On 23/07/14 19:08, Brad Hill wrote:
>> So, what in-test-file data does it use?  I was already asking
>> previously about what metadata needed to be in tests, and the
>> consensus seemed to be pretty much "none".
>
> The correct answer is "pretty much none".
>
>> How does the manifest generator distinguish between support files,
>> READMEs, and real test cases, for example?
>
> Real testharness.js testcases are html/XML/SVG files that contain a
> <script src="/resources/testharness.js">. Real reftests either have a
> <link rel="[mis]match" href="/url/of/ref.html"> or a file in the same
> directory with the same name and a -ref suffix. The idea is that in the
> commonest case of a testharness.js test the author doesn't have to
> supply any metadata at all; things Just Work.
>
> I notice that some of the documentation around reftests in
> testthewebforward.org differs from the web-platform-tests reality. That
> sucks. Unfortunately the documentation setup at the moment makes it
> unnecessarily difficult to keep up to date. I have a plan to change
> that, but it requires some time to rework the system for generating the
> site so it can pull in documentation from the various source projects.
>
> Also there are some differences between CSS tests and
> web-platform-tests, which makes for difficulties. Frustratingly, efforts
> to consolidate these testsuites have stalled.
>
> In good news, I have spent some time today rewriting the manifest
> generator to work in a way that seems much more friendly to including
> local changes by default, whilst still allowing for incremental updates.
> It isn't tested enough to call it done yet, but hopefully I can finish
> it off soon.
>

Received on Wednesday, 23 July 2014 19:44:38 UTC