Consolidate most simple vs. complex Font methods

Consolidate most simple vs. complex Font methods

This CL encapsulates the simple/complex logic in a low level glyphBuffer builder function, and removes all paint-related codepath-specific methods. Everything above buildGlyphBuffer() now operates on a codepath-agnostic data structure (glyphBuffer).

To pull this off, the CL unifies GlyphBuffer & GlyphBufferWithOffsets by adding offset support to the former. The approach has no downside beyond a slightly increased stack frame size for the simple path, and is aligned with plans to eventually store offsets for simple text also (and possibly drop advances altogether).

Also, most text blob caching logic is relocated down into Font drawText() - which allows us to trivially add support for complex text blobs (currently still off due to the lack of offsets support in buildTextBlob). The advance return value (which doesn't play well with blob caching) is dropped, and the only existing client for it is converted to use a new dedicated, blob-cache-bypassing drawUncachedText() method.

R=jbroman@chromium.org,eae@chromium.org,dominik.rottsches@intel.com,leviw@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=183620

[f(malita), 2014-10-13 17:44:52]

fmalita@chromium.org changed reviewers:
+ dominik.rottsches@intel.com, eae@chromium.org, jbroman@chromium.org

[f(malita), 2014-10-13 17:45:48]

fmalita@chromium.org changed reviewers:
+ leviw@chromium.org

[f(malita), 2014-10-13 17:49:55]

At -71 net LoC I think this is not a bad cleanup, but to sweeten the deal I'll
also throw in a free 2.5% record time reduction for the regular/non-textblob
path (Wikipedia rasterize_and_record_micro, 3 vs.3 runs):
http://i.imgur.com/5TOw6Z4.png

[eae, 2014-10-13 17:57:09]

Very nice! LGTM

[eae, 2014-10-13 18:02:20]

The CQ bit was checked by eae@chromium.org

[commit-bot: I haz the power, 2014-10-13 18:03:27]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/645363003/340001

[commit-bot: I haz the power, 2014-10-13 18:23:56]

Committed patchset #4 (id:340001) as 183620

[Dominik Röttsches, 2014-10-14 06:43:30]

Thanks! - that makes it again more readable.

One comment - From your description: "...and possibly drop advances altogether"
- I don't think we want to do that. We basically just refactored the GlyphBuffer
to keep advances + offsets, since it avoids issues with baseline shifts when
having Runs with mixed FontData, and also makes placing the initial glyph more
sane. It's also closer to the data we have in the font files itself. This change
allows us to get rid of the initialAdance value which basically belongs into
GlyphBuffer and not as a separate passed-around value. I think we can remove the
initialAdvance return that's still in Font.cpp in a separate patch.