I think I can solve this issue.
Could you assign this issue to me?

I've replaced ASSERT with DCHECK without ASSERT(isGraphOwner()) first as followed.

https://codereview.chromium.org/2159403002/

In my investigation, RecursiveMutex do not have locked() API anymore.
Hence we can't remove ENABLE(ASSERT) for isGraphOwner() without solving 
RecursiveMutex API issue.

I didn't follow your comment on isGraphOwner and RecursiveMutex.  These ASSERTs are enabled when doing a debug build.

I got compilation error as following when I remove remove ENABLE(ASSERT) for isGraphOwner().

../../third_party/WebKit/Source/modules/webaudio/DeferredTaskHandler.cpp:74:32: error: no member named 'locked' in 'WTF::RecursiveMutex'; did you mean 'lock'?
    return m_contextGraphMutex.locked();

This is strange. These are supposed to be enabled for a debug build and I have no problems building a debug build.  This configuration is tested by the trybots as well.

I just tried this on my linux machine.  Debug builds work fine.

hyungwook.lee@

What platform are you using for the development? OSX works fine with the removal of ENABLE(ASSERT).

I'm using gcc version 4.9.3 (Ubuntu 4.9.3-8ubuntu2~14.04) on Ubuntu 14.04

My first investigation seems wrong, I can see MutexBase has locked() API.

class WTF_EXPORT MutexBase {
    WTF_MAKE_NONCOPYABLE(MutexBase); USING_FAST_MALLOC(MutexBase);
public:
#if ENABLE(ASSERT)
    bool locked() { return m_mutex.m_recursionCount > 0; }
#endif

I'm testing it on release build not debug build.

Ah, that explains it.  I think you should ask on the blink-dev mailing list on whether it's ok to remove the #if ENABLE(ASSERT) there. That should fix the problem.

How is this issue different from the  issue 637509 ? Perhaps we can mark this as a dupe and close?

No,  issue 637509  is for replacing ASSERT with DCHECK, except for the isGraphOwner case, which is what this is about.

Better summary.