Move ChromiumMultiDex to BuildConfig.

build/config/android/rules.gni

Index: build/config/android/rules.gni
diff --git a/build/config/android/rules.gni b/build/config/android/rules.gni
index f9bfd3d43b6198d9cdb5a72c741b12303acdead5..4d3fbe94b7ce580f9cc25768eab425eb56dbf29b 100644
--- a/build/config/android/rules.gni
+++ b/build/config/android/rules.gni
@@ -1484,20 +1484,18 @@ template("android_apk") {
     _srcjar_deps += [ ":${_template_name}__native_libraries_java" ]
   }
 
-  if (!defined(invoker.apk_under_test)) {

[jbudorick, 2016/02/10 21:45:46]

Removing this condition makes me worry that we're going to hit duplicate
definition of ChromiumBuildConfig between APK under test + test APK. I saw this
manifest as a crash w/ "class resolved to unexpected dex" or something along
those lines appearing in the logcat.

[Peter Wen, 2016/02/10 21:56:48]

Makes sense. I'll check the logcat again after adding back this check.

Do we have a bot building multidex? Or is there a way for me to make sure I
don't break multidex locally?

-    java_cpp_template("${_template_name}__multidex_config_java") {
-      package_name = "org/chromium/base/multidex"
-      sources = [
-        "//base/android/java/templates/ChromiumMultiDex.template",
-      ]
+  java_cpp_template("${_template_name}__chromium_build_config_java") {
+    package_name = "org/chromium/base"
+    sources = [
+      "//base/android/java/templates/ChromiumBuildConfig.template",
+    ]
 
-      defines = []
-      if (enable_multidex) {
-        defines += [ "ENABLE_MULTIDEX" ]
-      }
+    defines = []
+    if (enable_multidex) {
+      defines += [ "ENABLE_MULTIDEX" ]
     }
-    _srcjar_deps += [ ":${_template_name}__multidex_config_java" ]
   }
+  _srcjar_deps += [ ":${_template_name}__chromium_build_config_java" ]
 
   java_target = "${_template_name}__java"
   java_library_impl(java_target) {