Do not inherit TaskPriority in TaskTraits.

base/task_scheduler/task_traits.cc

Index: base/task_scheduler/task_traits.cc
diff --git a/base/task_scheduler/task_traits.cc b/base/task_scheduler/task_traits.cc
index 6acf3244f59e475c51ee21a1741ba94259fbfdf3..67a0eeb6169ec1d982ab34ef7209bce47b95597b 100644
--- a/base/task_scheduler/task_traits.cc
+++ b/base/task_scheduler/task_traits.cc
@@ -19,6 +19,7 @@ namespace base {
 TaskTraits::TaskTraits()
     : may_block_(false),
       with_base_sync_primitives_(false),
+      priority_set_explicitly_(false),

[gab, 2017/04/25 17:16:41]

Should be set to true in .WithPriority()?

[fdoray, 2017/04/25 17:57:48]

Sorry bad merge.

       priority_(internal::GetTaskPriorityForCurrentThread()),

[gab, 2017/04/25 17:16:41]

This doesn't match CL desc.

[fdoray, 2017/04/25 17:57:48]

Sorry bad merge.

       shutdown_behavior_(TaskShutdownBehavior::SKIP_ON_SHUTDOWN) {}

[gab, 2017/04/25 17:16:41]

Use inline member initialization in header for all traits

[fdoray, 2017/04/25 17:57:48]

We have to choose between initialization in header of bitfield.
http://stackoverflow.com/questions/29286203/default-values-for-bit-fields-in-c11

I'm leaning towards bitfield since member initialization won't be needed with
constexpr constructor (all members will be initialized with a call to a template
function in the constructor https://codereview.chromium.org/2829083002/) 

[gab, 2017/04/25 18:01:21]

Fast-access matters more than bit savings IMO (i.e. extra codegen to extract bit
is undesired).

Member initialization OTOH is about readability IMO and I prefer that.