Decoupling rtc_base_approved from Obj-C code

webrtc/rtc_base/BUILD.gn

Index: webrtc/rtc_base/BUILD.gn
diff --git a/webrtc/rtc_base/BUILD.gn b/webrtc/rtc_base/BUILD.gn
index 60d1f57583cf9f86f0d74ac6c63484d30a5f39c6..7d133a7a8aeac5fd66263b4e4e1b4d33b1886c52 100644
--- a/webrtc/rtc_base/BUILD.gn
+++ b/webrtc/rtc_base/BUILD.gn
@@ -31,12 +31,6 @@ group("base") {
   }
 }
 
-config("rtc_base_approved_all_dependent_config") {
-  if (is_mac && !build_with_chromium) {
-    libs = [ "Foundation.framework" ]  # needed for logging_mac.mm
-  }
-}
-
 config("rtc_base_chromium_config") {
   defines = [ "NO_MAIN_THREAD_WRAPPING" ]
 }
@@ -86,16 +80,31 @@ source_set("compile_assert_c") {
   ]
 }
 
-# The subset of rtc_base approved for use outside of libjingle.
 rtc_static_library("rtc_base_approved") {

[kwiberg-webrtc, 2017/08/31 20:57:02]

If you make this a source set, can you drop the no-op file?

[mbonadei, 2017/09/01 09:46:38]

Yes, done.

+  sources = [
+    "noop.cc",
+  ]
+  if (is_mac && !build_with_chromium) {
+    public_deps = [
+      ":rtc_base_approved_objc",
+    ]
+  } else {
+    public_deps = [
+      ":rtc_base_approved_generic",
+    ]

[kwiberg-webrtc, 2017/08/31 20:57:02]

Can you make this public_dep unconditional, and turn the public_dep on line 313
into a regular dep? It's annoying to see what exports what already with one
layer of public_deps, and I'll be two layers are worse...

[mbonadei, 2017/09/01 09:46:38]

Done.

+  }
+}
+
+# The subset of rtc_base approved for use outside of libjingle.
+rtc_static_library("rtc_base_approved_generic") {
   # TODO(kjellander): Remove (bugs.webrtc.org/7480)
   # Enabling GN check triggers a cyclic dependency caused by rate_limiter.cc:
   #   :rtc_base_approved -> ../system_wrappers -> :rtc_base_approved
+  visibility = [ ":*" ]

[kwiberg-webrtc, 2017/08/31 20:57:03]

Could you tighten the visibility restrictions here, and the two below on lines
303 and 308? In your other CL, you had very strict visibility limits, and that
made it easy to see who was supposed to be depending on what.

[mbonadei, 2017/09/01 09:46:38]

Yes, I remember I talked about this with kjellander@ and we decided to use ":*"
(but I don't remember why) so probably we have to change the other CL.

Henrik: do you remember why we prefer ":*"?

[kjellander_webrtc, 2017/09/01 10:11:20]

It's just laziness, I wouldn't say we prefer one or the other. You could just
list the targets explicitly instead, and the more I think of it the more I
prefer that.
:* is just a little easier since it allows everything in this file, which is
usually sufficient (but not as clear).

   check_includes = false
   defines = []
   libs = []
   deps = []
-  all_dependent_configs = [ ":rtc_base_approved_all_dependent_config" ]
 
   sources = [
     "array_view.h",
@@ -210,7 +219,6 @@ rtc_static_library("rtc_base_approved") {
     sources += [
       "logging.cc",
       "logging.h",
-      "logging_mac.mm",
     ]
   }
   if (is_component_build && is_win) {
@@ -290,6 +298,24 @@ rtc_static_library("rtc_base_approved") {
   }
 }
 
+if (is_mac && !build_with_chromium) {
+  config("rtc_base_approved_objc_all_dependent_config") {
+    visibility = [ ":*" ]
+    libs = [ "Foundation.framework" ]  # needed for logging_mac.mm
+  }
+
+  rtc_static_library("rtc_base_approved_objc") {
+    visibility = [ ":*" ]
+    all_dependent_configs = [ ":rtc_base_approved_objc_all_dependent_config" ]
+    sources = [
+      "logging_mac.mm",
+    ]
+    public_deps = [
+      ":rtc_base_approved_generic",
+    ]
+  }
+}
+
 config("enable_libevent_config") {
   defines = [ "WEBRTC_BUILD_LIBEVENT" ]
 }