Added Android build target

This builds upon http://codereview.adblockplus.org/10228016/ but avoids having to configure toolchains manually by invoking ndk-build.

[Wladimir Palant, 2013-05-16 10:32:19]

http://codereview.adblockplus.org/10488022/diff/1/common.gypi
File common.gypi (right):

http://codereview.adblockplus.org/10488022/diff/1/common.gypi#newcode15
common.gypi:15: '-pedantic', '-std=gnu++0x', '-fexceptions', ],
-Wno-error flags were merely left-overs from back when we included
standalone.gypi everywhere. Adding -std=gnu++0x fixes a bunch of warnings that
we had in regular builds as well and I figured that -fexceptions won't hurt if
specified for all *ix OSes.

http://codereview.adblockplus.org/10488022/diff/1/libadblockplus.gyp
File libadblockplus.gyp (right):

http://codereview.adblockplus.org/10488022/diff/1/libadblockplus.gyp#newcode43
libadblockplus.gyp:43: 'third_party/v8/include',
Not sure why this only came up when compiling for Android - abpshell includes
v8.h indirectly (via JsEngine.h) but doesn't have the necessary include path
configured.

[Wladimir Palant, 2013-05-16 12:06:26]

http://codereview.adblockplus.org/10488022/diff/1/Makefile
File Makefile (right):

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode31
Makefile:31: APP_PLATFORM=android-9 \
This is pretty random - I simply chose it because V8 compiles for the android-9
platform. android-3 platform works fine as well but probably doesn't make too
much sense if we use V8 code.

[Felix Dahlke, 2013-05-23 13:35:43]

http://codereview.adblockplus.org/10488022/diff/1/Makefile
File Makefile (right):

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode20
Makefile:20: test -d third_party/v8/build/gyp || mkdir third_party/v8/build/gyp
How about "mkdir -p third_party/v8/build/gyp"?

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode21
Makefile:21: cp -f third_party/v8_gyp_launcher third_party/v8/build/gyp/gyp
How come this is only necessary for Android?

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode31
Makefile:31: APP_PLATFORM=android-9 \
On 2013/05/16 12:06:26, Wladimir Palant wrote:
> This is pretty random - I simply chose it because V8 compiles for the
android-9
> platform. android-3 platform works fine as well but probably doesn't make too
> much sense if we use V8 code.

I think this should be the same as our minSdkVersion, android-7. I think code
compiled for a new platform won't work on older Android versions, but I'm not
sure.

http://codereview.adblockplus.org/10488022/diff/1/common.gypi
File common.gypi (right):

http://codereview.adblockplus.org/10488022/diff/1/common.gypi#newcode15
common.gypi:15: '-pedantic', '-std=gnu++0x', '-fexceptions', ],
On 2013/05/16 10:32:19, Wladimir Palant wrote:
> -Wno-error flags were merely left-overs from back when we included
> standalone.gypi everywhere. Adding -std=gnu++0x fixes a bunch of warnings that
> we had in regular builds as well and I figured that -fexceptions won't hurt if
> specified for all *ix OSes.

We should use -std=c++0x if possible, -std=gnu++0x enables some GNU-specific
stuff we probably don't want.

Note that -std=c++0x will give us a bunch of features, like nullptr, lambdas and
variadic templates - depending on the compiler version. We should probably
define somewhere which subset of C++0x we can use, in order support slightly
older compilers. Previously, we only had access to C++0x features in the
standard library, no new language features.

http://codereview.adblockplus.org/10488022/diff/1/libadblockplus.gyp
File libadblockplus.gyp (right):

http://codereview.adblockplus.org/10488022/diff/1/libadblockplus.gyp#newcode3
libadblockplus.gyp:3: # We don't want to use curl on Windows, skip the check
there
How about "... on Windows or Android ..."?

[Wladimir Palant, 2013-05-24 08:41:56]

This is meant to be applied on top of the changes in
http://codereview.adblockplus.org/10727002/ now meaning that v8 include
directory is no longer necessary in the dependent targets. However, I had to
remove dependencies on AdblockPlus::Utils from tests because Utils.h depends on
v8.h.

[Wladimir Palant, 2013-05-24 08:47:36]

http://codereview.adblockplus.org/10488022/diff/1/Makefile
File Makefile (right):

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode20
Makefile:20: test -d third_party/v8/build/gyp || mkdir third_party/v8/build/gyp
On 2013/05/23 13:35:44, Felix H. Dahlke wrote:
> How about "mkdir -p third_party/v8/build/gyp"?

Heh, very obvious... I actually looked at man mkdir before implementing it this
way.

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode21
Makefile:21: cp -f third_party/v8_gyp_launcher third_party/v8/build/gyp/gyp
On 2013/05/23 13:35:44, Felix H. Dahlke wrote:
> How come this is only necessary for Android?

Normally we run our own gyp instance, we don't run V8's make files that expect
to find gyp at a fixed location.

http://codereview.adblockplus.org/10488022/diff/1/Makefile#newcode31
Makefile:31: APP_PLATFORM=android-9 \
On 2013/05/23 13:35:44, Felix H. Dahlke wrote:
> I think this should be the same as our minSdkVersion, android-7. I think code
> compiled for a new platform won't work on older Android versions, but I'm not
> sure.

Given that V8 is compiled with android-9 this would be rather pointless. Also,
I'm pretty certain that the binary parts are very much backwards compatible -
they don't use any Java-based APIs. Our app works fine on Android 2.1 despite
using V8.

[Felix Dahlke, 2013-05-24 10:40:28]

LGTM