Remove the dynamic initialization of WebKitSystemInterface. Just call into the library directly.

Remove the dynamic initialization of WebKitSystemInterface. Just call into the library directly.

BUG=101156
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=148861

[abarth-chromium, 2013-04-22 19:00:39]

LGTM

Why did the code previously do this runtime initialization?

[Robert Sesek, 2013-04-22 19:01:50]

On 2013/04/22 19:00:39, abarth wrote:
> LGTM
> 
> Why did the code previously do this runtime initialization?

They had #if BUILDING_ON_MAC_CAT conditionals, since certain functions didn't
exist on certain OSes. I got rid of all that in the last CL, though.

[eseidel, 2013-04-22 19:02:24]

It's not immediately clear to me why Apple did it this way?  I assume because
they would build a single copy of WebKit which would need to work on multiple
versions of the OS?

This is simply abstraction we don't need since we statically link Blink into
Chrome?

There was a bunch of hubbub when we open sourced WebKit around the creation of
this library.  IIRC it was originally designed as a way to insulate the rest of
Apple from WebKit's SPI dependancies.  It also served as a burn-down list of
SPIs which needed some sort of API equivalent.  Obviously that list was never
completely burned. :)  Although we use remarkably few SPIs in Blink these days
it seems.

I'm OK with this change, I would just like to better understand the
history/consequences before approving.

[eseidel, 2013-04-22 19:03:25]

So you believe now that the #ifdefs are gone, this runtime initialization serves
no purpose?

[abarth-chromium, 2013-04-22 19:06:00]

On 2013/04/22 19:03:25, Eric Seidel (Google) wrote:
> So you believe now that the #ifdefs are gone, this runtime initialization
serves
> no purpose?

From reading the CL, it doesn't seem to serve any purpose anymore.  The function
pointers are just written can called unconditionally.

[Robert Sesek, 2013-04-22 19:07:31]

On 2013/04/22 19:03:25, Eric Seidel (Google) wrote:
> So you believe now that the #ifdefs are gone, this runtime initialization
serves
> no purpose?

Correct. There's no reason to indirect through function pointers, since we only
ever built with the Leopard version of this library. In the future, I'm going to
try and remove the static library entirely and just call the SPIs directly. I
was working in WebKit to do that, too.

[eseidel, 2013-04-22 19:09:12]

I'm very interested to hear what sort of resistance/lack-there-of you find from
Apple-peeps who have more a more current understanding of this than I do.

lgtm.

[Nico, 2013-04-22 19:15:58]

lgtm as well

[Robert Sesek, 2013-04-22 19:57:44]

On 2013/04/22 19:09:12, Eric Seidel (Google) wrote:
> I'm very interested to hear what sort of resistance/lack-there-of you find
from
> Apple-peeps who have more a more current understanding of this than I do.
> 
> lgtm.

They were receptive to changes like this upstream. E.g.,
https://bugs.webkit.org/show_bug.cgi?id=102405.

Maciej in particular said he didn't think WKSI turned out the way he wanted, and
would support removing it in favor of calling SPI directly.

[Robert Sesek, 2013-04-22 22:10:06]

Committed patchset #2 manually as r148861 (presubmit successful).