Add U_PF_NATIVE_CLIENT

Add U_PF_NATIVE_CLIENT

NaCl has to be identified by '__native_client__'. Currently, we assume that
NaCl has a signature of Linux (e.g. __linux___), but that is not the case.

Add a new platform (U_PF_NATIVE_CLIENT) to platform.h and treat it as
Linux-based.

Upstream bug : http://bugs.icu-project.org/trac/ticket/11033

BUG=NONE
TEST=NaCl builds fine when ICU is rolled to this revision.
R=ncbray@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=285387

[jungshik at Google, 2014-07-24 20:59:19]

Hi Nick,

Can you take a look?  I've added U_PF_NATIVE_CLIENT and assign 4050 to it.
(U_PF_LINUX=4000, U_PF_ANDROID=4100). 

In a part of the file (platform.h) not shown in the file,
U_PLATFORM_IS_LINUX_BASED is set in the following way: With this CL, putilimpl.h
does not need to be changed. Are you ok with this approach? 

#ifdef U_PLATFORM_IS_LINUX_BASED
    /* Use the predefined value. */
#elif U_PF_LINUX <= U_PLATFORM && U_PLATFORM <= U_PF_ANDROID
#   define U_PLATFORM_IS_LINUX_BASED 1
#else
#   define U_PLATFORM_IS_LINUX_BASED 0
#endif

[jungshik at Google, 2014-07-24 21:07:37]

It's 'riskier' than your patch. (
https://cs.corp.google.com/#search/&sq=package:%5Epiper$%20-file:experimental...
).   

A 3rd way may be almost identical to your patch except that U_PF_NATIVE_CLIENT
is defined with such a value that it won't be treated as LINUX_BASED.  In that
case, a check for __native_client__ would be replaced by a check for
U_PF_NATIVE_CLIENT in putilimpl.h

[Nick Bray (chromium), 2014-07-24 21:54:16]

On 2014/07/24 21:07:37, Jungshik Shin wrote:
> It's 'riskier' than your patch. (
>
https://cs.corp.google.com/#search/&sq=package:%5Epiper$%20-file:experimental...
> ).   
> 
> A 3rd way may be almost identical to your patch except that U_PF_NATIVE_CLIENT
> is defined with such a value that it won't be treated as LINUX_BASED.  In that
> case, a check for __native_client__ would be replaced by a check for
> U_PF_NATIVE_CLIENT in putilimpl.h

The search you linked to doesn't work for me.
I tried it against the patch that stop defining __linux__ for NaCl, and
everything builds fine.
It looks reasonable from the NaCl perspective.
LGTM, although you may want to have someone familiar with the ICU code review
it.

[jungshik at Google, 2014-07-24 22:14:22]

Committed patchset #1 manually as r285387 (presubmit successful).