[webkit-dev] Experimental and (new) Internal feature flags
Ryosuke Niwa
rniwa at webkit.org
Wed Sep 12 15:58:27 PDT 2018
On Wed, Sep 12, 2018 at 3:53 PM Dean Jackson <dino at apple.com> wrote:
> On 13 Sep 2018, at 08:50, Ryosuke Niwa <rniwa at webkit.org> wrote:
>
>
> On Wed, Sep 12, 2018 at 3:09 PM Dean Jackson <dino at apple.com> wrote:
>
>> On 13 Sep 2018, at 08:05, Ali Juma <ajuma at chromium.org> wrote:
>>
>> Is there a plan for how to handle web platform tests (where we can't add
>> such a header)?
>>
>> Currently, WebKitTestRunner enables all experimental features in tests
>> (including those disabled by default), so adding an experimental feature
>> (even one that's disabled by default) is a convenient way to get web
>> platform test coverage for the feature. If WebKitTestRunner will no longer
>> do that, we'll need some other way to keep these tests running.
>>
>>
>> I thought about this. I see three options:
>>
>> 1. Continue to enable all experimental features in testing. The downside
>> is that we're testing something different to what we ship.
>>
>> 2. Have the default value only apply to testing, not to shipping. For
>> Internal Features, instead of enabling them all, I had them revert to their
>> default value.
>>
>> 3. Have some way to add the headers after importing the WPT test.
>>
>> For now, I added headers to the WPT code, because there already were a
>> bunch of tests that had such headers.
>>
>
> Is that automated? We don't want get into a state where we'd have to
> manually add those headers whenever we re-import tests.
>
>
> It isn't.
>
> I think maybe we should go for option 2 at the moment.
>
> Or, ultimately, have a per-directory configuration file that could enable
> features.
>
Okay. I can think of two options for (3):
- Have a per-directory/per-test setting to enable/disable features
- Have a per-directory/per-test importation rule to add headers.
I think the second option is better because implicitly changing WTR/DRT's
behavior based on directory names would be quite subtle for people who
don't directly work on the test infrastructure. e.g. moving an imported
test from one place to another would change the test's behavior.
- R. Niwa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20180912/13675691/attachment.html>
More information about the webkit-dev
mailing list