[webkit-dev] Following up on removing -khtml- and -apple- CSS vendor prefixes

[Adam Barth]

On Thu, May 3, 2012 at 7:57 PM, Ryosuke Niwa <rniwa at webkit.org> wrote:
> Resent from the right email address.
>
> ---------- Forwarded message ----------
> From: Ryosuke Niwa <rniwa at berkeley.edu>
> Date: Thu, May 3, 2012 at 7:49 PM
> Subject: Re: [webkit-dev] Following up on removing -khtml- and -apple- CSS
> vendor prefixes
> To: Adam Barth <abarth at webkit.org>
> Cc: Alexey Proskuryakov <ap at webkit.org>, webkit-dev at lists.webkit.org
>
>
> On Fri, Apr 6, 2012 at 4:27 PM, Adam Barth <abarth at webkit.org> wrote:
>>
>> 2012/4/5 Alexey Proskuryakov <ap at webkit.org>:
>> > 05.04.2012, в 01:08, Adam Barth написал(а):
>> >
>> >> Based on this information, I've posted a patch that limits the -apple-
>> >> and -khtml- vendor prefixes to ENABLE(DASHBOARD_SUPPORT):
>> >>
>> >> https://bugs.webkit.org/show_bug.cgi?id=83256
>> >>
>> >> This approach lets ports that wish to remain compatible with Dashboard
>> >> widgets continue to support these prefixes while also letting ports
>> >> that aren't constrained by dashboard widgets disable them, hopefully
>> >> lessening their use on the broader web.
>> >
>> > Guarding on ENABLE(DASHBOARD_SUPPORT) will make it so that these names
>> > will be available in Safari on Mac, in every Mac application that uses
>> > WebKit, but not in Safari on Windows or iOS, and not in other WebKit based
>> > browsers such as Chrome. I don't think that this level of fragmentation is
>> > good.
>> >
>> > We normally guard Dashboard-only quirks with
>> > settings->usesDashboardBackwardCompatibilityMode().
>>
>> I attempted to implement this approach in
>> <https://bugs.webkit.org/attachment.cgi?id=136093&action=prettypatch>,
>> but I ran into a problem because cssPropertyID is a free function.
>> Requiring all the callers to pass in a Settings object seems
>> undesirable.  Note: there are already a number of uses of
>> ENABLE(DASHBOARD_SUPPORT) in CSSParser.cpp, and none of them check
>> usesDashboardBackwardCompatibilityMode().
>
> I don't think you need to do that. Looking at your latest work-in-progress
> patch, it's probably sufficient to add a guard in CSSParser.cpp. It's okay
> to have those free functions (in fact we DO need them)
> when ENABLE(DASHBOARD_SUPPORT) is true
> and usesDashboardBackwardCompatibilityMode() is false since they will never
> be called if the parser rejects values.

Ok.  I'll update the patch.

Thanks!

Adam