Fix coverage build.

webrtc/build/common.gypi

Index: webrtc/build/common.gypi
diff --git a/webrtc/build/common.gypi b/webrtc/build/common.gypi
index f4ecc314ca15b9fcf4cf2dcdb1bb91d76c09f633..f8446d9f91bce9e16aa52cd9dfc5df9ae293007e 100644
--- a/webrtc/build/common.gypi
+++ b/webrtc/build/common.gypi
@@ -372,6 +372,7 @@
       ['coverage==1 and OS=="linux"', {

[kjellander_webrtc, 2016/04/11 19:53:08]

I think we should add clang==0 here to make it clear that it's GCC's coverage.

[Michael Achenbach, 2016/04/12 08:58:34]

The cflags are valid for both the clang and the gcc coverage builds.

         'cflags': [ '-ftest-coverage',
                     '-fprofile-arcs' ],
+        'ldflags': [ '--coverage' ],

[kjellander_webrtc, 2016/04/11 19:53:08]

Adding this, which seems to be Clang specific, just confuses the user.
We should probably adapt something similar to V8's
https://chromium.googlesource.com/v8/v8/+/master/build/standalone.gypi instead.

See https://bugs.chromium.org/p/chromium/issues/detail?id=568949 for the great
work machenbach@ has done for V8's coverage effort.

[Michael Achenbach, 2016/04/12 08:58:34]

This seems to fix the clang case as described e.g. here:
http://logan.tw/posts/2015/04/28/check-code-coverage-with-clang-and-lcov/

GCC needs 'ldflags': [ '-fprofile-arcs']. If you want to support both, you need
to add clang=1 or clang=0 to the conditions.

[kjellander_webrtc, 2016/04/19 05:31:51]

OK. If it needs more work when/if we decide to setup Clang coverage that's fine
with me.

         'link_settings': { 'libraries': [ '-lgcov' ] },
       }],
       ['os_posix==1', {