Subpixel text 3/8 of a pixel too far to the right.

[bungeman, 2012-01-03 16:34:33]

The pre-add for rounding should be 1/2 the sampling frequency. Without subpixel
this means adding 1/2; with subpixel this means adding 1/8.

[caryclark1, 2012-01-03 19:15:07]

LGTM; it looks like some of the changes are formatting fixes, and some change
actual code. It's helpful to me if the two are separate CLs so I don't have to
grok which is which, although if it's the group style to do both at once it's OK
with me.

[bungeman, 2012-01-12 21:10:14]

With these changes, and defining 
SK_DRAW_POS_TEXT_IGNORE_SUBPIXEL_LEFT_ALIGN_FIX for 'mac' in the chromium
skia.gyp, there are no new layout tests failures. Two GMs, fontscaler and
lcdtext, will need to be rebaselined on Mac and Linux (Windows does not
currently support subpixel glyphs/positioning).

Without SK_DRAW_POS_TEXT_IGNORE_SUBPIXEL_LEFT_ALIGN_FIX there are around 400
layout failures on Mac, which are mostly controls, justified text, and complex
layout text.

[epoger, 2012-01-13 15:29:49]

Is this going to require massive rebaselines in Chrome?  And if so, is it worth
the trouble?  (Is there a bug out there where people said "yes, let's fix
this"?)

[epoger, 2012-01-13 15:31:50]

On 2012/01/13 15:29:49, epoger wrote:
> Is this going to require massive rebaselines in Chrome?  And if so, is it
worth
> the trouble?  (Is there a bug out there where people said "yes, let's fix
> this"?)

Possibly answering my own question... LCD text rendering is turned off for the
webkit layout tests, so hopefully this will not require any rebaselining in
webkit.

[bungeman, 2012-01-13 20:44:27]

Committed revision 3037.
Committed rebaselines revision 3037.