(j3.2006) Use of STAT_FAILED_IMAGE in TS 18508 subclause A.2.1

[John Reid]

Van Snyder wrote:
> What is the use of STAT_FAILED_IMAGE in TS 18508 subclause A.2.1?
>
> The STATUS argument from EVENT_QUERY is compared to STAT_FAILED_IMAGE at
> [43:41-46].
>
> The description of the STATUS argument of EVENT_QUERY in subclause
> 7.4.11 merely says "it is assigned the value 0 if no error condition
> occurs and a processor-defined positive value if an error condition
> occurs."

Yes, this example is broken. There needs to be an invocation of 
FAILED_IMAGES at the start of the master DO loop.

Cheers,

John.

>
> Under what conditions is it assigned the value STAT_FAILED_IMAGE (or
> STAT_STOPPED_IMAGE)?  If it is ever assigned one of those values, which
> image is it that failed or stopped?
>
> Presumably, since no image number argument is present, EVENT_QUERY is
> either looking at its own EVENT, or, if EVENT is a coindexed variable,
> it's looking at the specified one.  If it's looking at its own EVENT the
> image presumably is not stopped or failed, else the call didn't get
> executed and the STATUS cannot be examined.  Since the <event-variable>
> in an EVENT WAIT statement cannot be coindexed, there seems to be no
> point to sending a coindexed EVENT argument to EVENT_QUERY -- why would
> an image care about another image's event count?  Maybe that ought to be
> prohibited.  If EVENT is not coindexed, the STATUS certainly will not be
> STAT_FAILED_IMAGE or STAT_STOPPED_IMAGE.
>
>
> _______________________________________________
> J3 mailing list
> J3 at mailman.j3-fortran.org
> http://mailman.j3-fortran.org/mailman/listinfo/j3
>