Add support for using ICU with a data file instead of compiling the data in....

Add support for using ICU with a data file instead of compiling the data in.

This might help the Mac bots as it means each unittest that depends on base
will be ~10meg smaller, helping link and also helping with the copy of
bits from the builder to the tester.

If Mac ever support a Universal binary (32bit and 64bit) it will also probably
want to use this to avoid carrying two copies of the ICU data.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=74439

[TVL, 2011-02-10 15:26:42]

Mark - Can you take a look.

Jungshik - fyi

[Mark Mentovai, 2011-02-10 16:08:11]

Is it then the user’s responsibility to initialize ICU by pointing it to
icudt42l.dat?

http://codereview.chromium.org/6482029/diff/1/third_party/icu/icu.gyp
File third_party/icu/icu.gyp (right):

http://codereview.chromium.org/6482029/diff/1/third_party/icu/icu.gyp#newcode8
third_party/icu/icu.gyp:8: 'icu_use_datafile%': 0,
data_file

http://codereview.chromium.org/6482029/diff/1/third_party/icu/icu.gyp#newcode72
third_party/icu/icu.gyp:72: # Make sure any binary depending on this gets the
datafile.
data file

[TVL, 2011-02-10 16:11:09]

On 2011/02/10 16:08:11, Mark Mentovai wrote:
> Is it then the user’s responsibility to initialize ICU by pointing it to
> icudt42l.dat?

Yup, chrome already has this bottlenecked because windows uses the data DLL.

Other changes done.

[Mark Mentovai, 2011-02-10 16:18:34]

LGTM

Can you also merge this patch to /trunk/deps/third_party/icu46?

[TVL, 2011-02-10 16:22:16]

On 2011/02/10 16:18:34, Mark Mentovai wrote:
> LGTM
> 
> Can you also merge this patch to /trunk/deps/third_party/icu46?

Yup, I'll pull and send out another cl in a little bit.