Add font-equivalent mapping for a few Chinese/Japanese fonts for CrOS....

Add font-equivalent mapping for a few Chinese/Japanese fonts for CrOS.

The following aliases are added to be recognized in addition to what we have.

The first four are applicable to Chrome on Linux and ChromeOS. The last 
two are for Chrome OS.

IPAPMincho -> MS P Mincho
IPAMincho -> MS Mincho
IPAPGothic -> MS P Gothic
IPAGothic -> MS Gothic

Song ASC -> Simsun
N Song ASC -> NSimsun

BUG=65382, chromium-os:10182, chromium-os:8757
TEST=Install IPA fonts on Linux and make sure that Windows Japanese fonts are not installed on your machine. Also, add what's added in http://codereview.chromium.org/5695005/ to your copy of /etc/fonts/local.conf. And, load cjfontalias.html attached to http://crosbug.com/10182 to make sure that all three columns look identical (each rows should be different). 


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=71295

[jungshik at Google, 2010-12-11 01:10:50]

http://codereview.chromium.org/5578008/diff/6001/skia/ext/SkFontHost_fontconf...
File skia/ext/SkFontHost_fontconfig_direct.cpp (right):

http://codereview.chromium.org/5578008/diff/6001/skia/ext/SkFontHost_fontconf...
skia/ext/SkFontHost_fontconfig_direct.cpp:113: for (size_t i = 0; i <
kFontCount; ++i) {
Evan, do you think I'd better make a (hash)map once (with locking) and use that
instead of looping through the list everytime? This can be a hot spot, but I'm
not sure which would be better. One way is to land the CL and see if there's any
page cycler perf regression. Or, I can run page cycler tests locally with two
alternatives.

[jungshik at Google, 2010-12-14 17:49:40]

Evan, can you take a look?

[Evan Martin, 2010-12-14 18:04:20]

http://codereview.chromium.org/5578008/diff/6001/skia/ext/SkFontHost_fontconf...
File skia/ext/SkFontHost_fontconfig_direct.cpp (right):

http://codereview.chromium.org/5578008/diff/6001/skia/ext/SkFontHost_fontconf...
skia/ext/SkFontHost_fontconfig_direct.cpp:86: { PMINCHO, "ＭＳ Ｐ明朝" },
I worry that people's editors will eat these UTF-8 strings.
If the font name really is in UTF-8, maybe it'd be better to use
"\xA3\x12" etc in the quotes and then add a comment with the real UTF-8.

http://codereview.chromium.org/5578008/diff/6001/skia/ext/SkFontHost_fontconf...
skia/ext/SkFontHost_fontconfig_direct.cpp:113: for (size_t i = 0; i <
kFontCount; ++i) {
On 2010/12/11 01:10:50, Jungshik Shin wrote:
> Evan, do you think I'd better make a (hash)map once (with locking) and use
that
> instead of looping through the list everytime? This can be a hot spot, but I'm
> not sure which would be better. One way is to land the CL and see if there's
any
> page cycler perf regression. Or, I can run page cycler tests locally with two
> alternatives. 

I think whichever implementation is simpler would be better.  I agree this loop
is getting too large.

[jungshik at Google, 2010-12-23 02:01:26]

Oops. I thought I asked you to take another look after updating the CL, but
didn't send it away. Sorry for the delay. Can you take a look if you're not yet
gone for holiday :-). Thank you.



Anyway, I used \xXX for non-ASCII literals with a comment in UTF-8. I guess it'd
be ok even if somebody garbles up strings in comments. 

As for loop vs (hash)map, I left it alone out of 'laziness' to run page-cycler
tests locally. After landing the CL, I'll monitor the buildbot for the perf
regression. If there's a regression, I'll revert and use (hash)map.

[Evan Martin, 2010-12-23 21:28:29]

LGTM, I guess we should just profile a page before making performance
optimizations

[jungshik at Google, 2011-01-13 18:21:39]

Thank you for the review. I wanted to land this when few people are around, but
kept forgetting. Anyway, it's landed last night at r=71295 and all the page
cycler tests (on Linux and Linux LowMem) didn't slow down (the numbers fluctuate
about the same value before and after the CL landing).