Add COMPONENT_BUILD global define.

Add COMPONENT_BUILD global define. 

This avoids the need to define FOO_DLL macros for each project that we wish to
optionally build as a DLL (when component=="shared_library").  This in turn
means that we do not need direct_dependent_settings to define FOO_DLL, and that
means that we don't need to update projects to convert transitive dependencies
into explicit dependencies.  This makes the component build more consistent
with the static build.

An alternative would be to use all_dependent_settings, but I feel that the
global approach is simpler as it creates less repetition in each target
definition for components.

A side-effect of this change is that I needed to make base_nacl_win64 be a
shared_library in the component build.

R=rvargas,bradnelson,evan
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=92409

[darin (slow to review), 2011-07-12 22:25:30]

http://codereview.chromium.org/7344022/diff/4001/base/base.gypi
File base/base.gypi (right):

http://codereview.chromium.org/7344022/diff/4001/base/base.gypi#newcode634
base/base.gypi:634: 'target_name': 'base_i18n_nacl_win64',
This target is required since icu_util::Initialize is not an exported function. 
We cannot just include its implementation with base_nacl_win64.  We need a
separate static lib to mirror the base_i18n static lib.

http://codereview.chromium.org/7344022/diff/4001/net/base/ssl_false_start_bla...
File net/base/ssl_false_start_blacklist.h (right):

http://codereview.chromium.org/7344022/diff/4001/net/base/ssl_false_start_bla...
net/base/ssl_false_start_blacklist.h:17: class SSLFalseStartBlacklist {
This change was required for the ssl_false_start_blacklist_process.exe target. 
It just wants to include this header and call these methods without linking
against ssl_false_start_blacklist.cc.  It does not use the IsMember method.

http://codereview.chromium.org/7344022/diff/4001/ui/base/resource/resource_bu...
File ui/base/resource/resource_bundle_dummy.cc (right):

http://codereview.chromium.org/7344022/diff/4001/ui/base/resource/resource_bu...
ui/base/resource/resource_bundle_dummy.cc:16: class Font {};
This change does not strictly need to be part of this CL, but since it is fairly
trivial, I'm adding it.

[rvargas (doing something else), 2011-07-12 22:45:54]

We should also update crypto.gyp

http://codereview.chromium.org/7344022/diff/4001/net/net.gyp
File net/net.gyp (right):

http://codereview.chromium.org/7344022/diff/4001/net/net.gyp#newcode827
net/net.gyp:827: 'conditions': [
We can remove this block.

[darin (slow to review), 2011-07-12 22:56:36]

PTAL

[rvargas (doing something else), 2011-07-12 22:58:09]

LGTM

[bradn, 2011-07-12 23:47:22]

LGTM
Do keep an eye on the nacl_integration stage on try jobs before you commit.

[commit-bot: I haz the power, 2011-07-13 07:12:37]

Change committed as 92325

[darin (slow to review), 2011-07-13 17:52:45]

On 2011/07/13 07:12:37, I haz the power (commit-bot) wrote:
> Change committed as 92325

Was reverted as 92329

[darin (slow to review), 2011-07-13 20:15:34]

I made one change in the latest CL:

In base/logging.h, I now have this patch:

 // MSVC doesn't like complex extern templates and DLLs.
-#if !defined(COMPILER_MSVC) && defined(BASE_DLL)
+#if !defined(COMPILER_MSVC) && !defined(COMPONENT_BUILD)

In the earlier versions of the patch, I just changed BASE_DLL to
COMPONENT_BUILD, but this broke the Linux build, which was never actually
defining BASE_DLL for direct dependents.  Using !defined(COMPONENT_BUILD)
actually preserves that old behavior.

[darin (slow to review), 2011-07-13 20:15:59]

Also, I tested locally that a linux component build compiles.

[bradn, 2011-07-13 20:22:56]

LGTM
2nd time's the charm :-)

[Evan Martin, 2011-07-13 20:39:29]

base LGTM (though I don't uflly understand the gyp changes)