add customized printing setting for headless

headless/BUILD.gn

Index: headless/BUILD.gn
diff --git a/headless/BUILD.gn b/headless/BUILD.gn
index 7b3788b2c5f4beb9e68d59385205e5194464a5d7..a9362e6100c77842ba889b6bd2d635217cc3469a 100644
--- a/headless/BUILD.gn
+++ b/headless/BUILD.gn
@@ -29,6 +29,7 @@ repack("pak") {
   sources = [
     "$root_gen_dir/blink/public/resources/blink_image_resources_100_percent.pak",
     "$root_gen_dir/blink/public/resources/blink_resources.pak",
+    "$root_gen_dir/components/components_resources.pak",
     "$root_gen_dir/components/strings/components_strings_en-US.pak",
     "$root_gen_dir/content/app/resources/content_resources_100_percent.pak",
     "$root_gen_dir/content/app/strings/content_strings_en-US.pak",
@@ -51,6 +52,7 @@ repack("pak") {
 
   deps = [
     ":resources",
+    "//components/resources:components_resources",
     "//components/strings",
     "//content:resources",
     "//content/app/resources",
@@ -316,6 +318,10 @@ static_library("headless_lib") {
     "//net",
   ]
 
+  if (enable_basic_printing) {
+    public_deps += [ "//skia" ]

[Sami, 2017/05/05 16:54:12]

This is probably fine if it's needed to fix a build error, but I'll just note
that embedders of headless shouldn't depend on skia to avoid build breaks due to
API churn.

[Wei Li, 2017/05/05 17:18:28]

drive-by: Should this be a deps instead? I don't see why headless lib needs to
expose skia?
btw, is it better to merge sections at line307, this one and line352 all
together?

[Lei Zhang, 2017/05/05 21:39:28]

It's not a deps entry because it is no the headless_lib target that needs it.
It's any dependents that #includes headless/lib/browser/headless_print_manager.h
that has problems.

As an example, components/constrained_window does the same thing, even though
the directory does not have a single skia #include.

[blundell, 2017/05/09 15:34:04]

I've been following this discussion, and I still don't really understand it. If
the target doesn't need to directly depend on or expose //skia, then it also
shouldn't need to have //skia as a public dependency. If the issue is that this
target is exposing its dependents to some intermediate target that then exposes
//skia, I believe that the following structure should have the same effect and
is cleaner from an encapsulation POV:
- the intermediate target lists //skia as a public dep
- this target lists the intermediate target as a public dep

Am I missing something here?

[jzfeng, 2017/05/10 06:55:09]

headless_lib need to expose //skia to make the target depend on it to build.
Among the current deps of headless_lib, several expose //skia. They are:
content/public/app:both
content/public/browser
content/public/child:child
content/public/common
ui/base
ui/display
ui/events/devices
So it wouldn't make much sense to expose one of them from headless_lib.

+  }
+
   deps = [
     ":gen_devtools_client_api",
     ":tab_socket",
@@ -388,6 +394,10 @@ test("headless_unittests") {
     "public/util/testing/fake_managed_dispatch_url_request_job.h",
   ]
 
+  if (enable_basic_printing) {
+    sources += [ "lib/browser/headless_printing_unittest.cc" ]
+  }
+
   deps = [
     ":headless_lib",
     "//base/test:run_all_unittests",
@@ -490,6 +500,10 @@ test("headless_browsertests") {
     "//testing/gmock",
     "//testing/gtest",
   ]
+
+  if (enable_basic_printing) {
+    deps += [ "//pdf" ]
+  }
 }
 
 static_library("headless_shell_lib") {