The originally-proposed patch to add event_t was more restrictive, not even
allowing implicit conversions from 0 to event_t except for function
... but it looks like the restriction was not completely removed, and still
exists for explicit conversions?
The OpenCL specification is hopeless when it comes to defining the event_t
type, so I have no idea what the right behavior is. Perhaps someone with
more familiarity with OpenCL can guess what the standard intended?