(j3.2006) J3/15-244

[John Reid]

Malcolm Cohen wrote:
> <<<
> This is not so. The first sentence of 8.2 of the TS says "An atomic
> subroutine is an intrinsic subroutine that performs an action on its
> ATOM argument or the count of its EVENT argument atomically."
>>>>
>
> And the rest of the description of what an atomic subroutine does simply
> does not work with EVENT_QUERY ***AT ALL***.  The description of the atomics
> has moved on since 2010.
>
> Just read the Atomic subroutines section in the standard.  It is rather
> obviously completely inapplicable to how EVENT_QUERY works.  Sprinkling "or
> EVENT argument" and "or execution of EVENT POST statements" throughout that
> subclause would make it work, but the end result would be significantly
> harder to understand than just describing how EVENT_QUERY works (it has NO
> interaction with the other atomic subroutines!!!) separately.
>
>> I have to disagree here. Changing an atomic subroutine to be nonatomic
>> is surely a big change.
>
> Describing it correctly is not a change at all.
>
>> It is like making ATOMIC_REF nonatomic.
>
> No it is not.
>
> There are some changes needed to 244 (at least I think it needs to say it
> operates atomically w.r.t. EVENT POST), but blathering on about how we are
> destroying the TS is not shedding any light on the matter.

Yes, words saying that it operates atomically w.r.t. EVENT POST and 
EVENT WAIT would meet my objection.

Cheers,

John.