Mac: Part 1 of a fix to get OOP font loading working in the renderer on 10.6.6 .

Mac: Part 1 of a fix to get OOP font loading working in the renderer on 10.6.6 .

When the renderer can't load a font due to sandbox restrictions, it sends an IPC over to the browser to return a copy of the font file, which it then activates.  Before doing this, the renderer retrieves the ATSFontContainerRef corresponding to the font it couldn't load.  This value is unique for each font file and having this value allows the renderer to determine if the font in question has already been retrieved from the browser and activated.  Without doing this, the renderer would need to activate a font each time it was requested.

In 10.6.6 the function used to get the unique ID from the font in the renderer was changed so it fails in the sandbox (it now tries to access the on-disk font file).  In order to work around this, we need to get the unique id on the browser side and send it back over IPC.

This is the first part of said change.

BUG=72727
TEST=None

[jeremy, 2011-05-30 09:07:24]

brettw: owner approval, rubberstamp is fine.
nico, robert: only one of you need review.

[brettw, 2011-05-30 13:05:47]

rubberstamp LGTM

http://codereview.chromium.org/7080024/diff/1/content/renderer/renderer_webki...
File content/renderer/renderer_webkitclient_impl.cc (right):

http://codereview.chromium.org/7080024/diff/1/content/renderer/renderer_webki...
content/renderer/renderer_webkitclient_impl.cc:476: bool
RendererWebKitClientImpl::SandboxSupport::loadFont(NSFont* srcFont,
Style: the args should be aligned horizontally, so move the first arg down to
the next line (I'd probably put them on separate lines as well, but that's less
rigidly defined).

[Nico, 2011-05-30 13:19:37]

LGTM with the below addressed

http://codereview.chromium.org/7080024/diff/1/content/browser/renderer_host/r...
File content/browser/renderer_host/render_message_filter.cc (right):

http://codereview.chromium.org/7080024/diff/1/content/browser/renderer_host/r...
content/browser/renderer_host/render_message_filter.cc:542: if (!ok ||
font_data_size == 0 || font_id == 0) {
Do you mean *font_id == 0 here? If not, should you just CHECK(font_id)?

http://codereview.chromium.org/7080024/diff/1/content/browser/renderer_host/r...
content/browser/renderer_host/render_message_filter.cc:545: " font id=" <<
font_id;
*font_id here too

http://codereview.chromium.org/7080024/diff/1/content/browser/renderer_host/r...
content/browser/renderer_host/render_message_filter.cc:548: *font_id = 0;
You really mean *font_id above, else this would be a null write

http://codereview.chromium.org/7080024/diff/1/content/common/font_loader_mac.mm
File content/common/font_loader_mac.mm (right):

http://codereview.chromium.org/7080024/diff/1/content/common/font_loader_mac....
content/common/font_loader_mac.mm:21: CHECK(font_data && font_data_size &&
font_id);
I'd make this 3 separate CHECK statements, that gives you more information when
it fails

[commit-bot: I haz the power, 2011-05-31 08:04:30]

Change committed as 87282