Do not call CGFontSetShouldUseMulticache. Apple says to not.

Do not call CGFontSetShouldUseMulticache. Apple says to not.

BUG=255125
R=thakis@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=153513

[Robert Sesek, 2013-06-27 20:57:40]

See bug for details.

[Nico, 2013-06-27 21:04:59]

Does "does nothing" mean "does nothing all the way back to 10.6"? Should we
still call this on 10.7 and and 10.6?

[Robert Sesek, 2013-06-27 21:14:58]

On 2013/06/27 21:04:59, Nico wrote:
> Does "does nothing" mean "does nothing all the way back to 10.6"? Should we
> still call this on 10.7 and and 10.6?

This sets a thread-local preference (_use_two_level_cache) on 10.6-10.8. I don't
know on which OSes they check/respect it. The only difference would be
performance, which my machine is not set up to test. If there's any performance
regression, I'd just revert this and do it conditionally.

[Nico, 2013-06-27 21:18:50]

On 2013/06/27 21:14:58, rsesek wrote:
> On 2013/06/27 21:04:59, Nico wrote:
> > Does "does nothing" mean "does nothing all the way back to 10.6"? Should we
> > still call this on 10.7 and and 10.6?
> 
> This sets a thread-local preference (_use_two_level_cache) on 10.6-10.8. I
don't
> know on which OSes they check/respect it. The only difference would be
> performance, which my machine is not set up to test. If there's any
performance
> regression, I'd just revert this and do it conditionally.

Do we run perf tests on interesting OS X versions?
http://build.chromium.org/f/chromium/perf/dashboard/perf.html lists 10.5 and
10.6, which doesn't make me very confident :-/

I'd keep this call for <= 10.8, but up to you. lgtm either way.

[Robert Sesek, 2013-07-01 17:45:22]

On 2013/06/27 21:18:50, Nico wrote:
> On 2013/06/27 21:14:58, rsesek wrote:
> > On 2013/06/27 21:04:59, Nico wrote:
> > > Does "does nothing" mean "does nothing all the way back to 10.6"? Should
we
> > > still call this on 10.7 and and 10.6?
> > 
> > This sets a thread-local preference (_use_two_level_cache) on 10.6-10.8. I
> don't
> > know on which OSes they check/respect it. The only difference would be
> > performance, which my machine is not set up to test. If there's any
> performance
> > regression, I'd just revert this and do it conditionally.
> 
> Do we run perf tests on interesting OS X versions?
> http://build.chromium.org/f/chromium/perf/dashboard/perf.html lists 10.5 and
> 10.6, which doesn't make me very confident :-/
> 
> I'd keep this call for <= 10.8, but up to you. lgtm either way.

I think if 10.6 shows no regression, then this should be acceptable on future
versions.

[Robert Sesek, 2013-07-03 16:41:25]

Committed patchset #1 manually as r153513 (presubmit successful).