[qmtest] Idle state of targets

[Mark Mitchell]

--On Wednesday, December 18, 2002 06:38:19 PM +0100 Sven Lundblad 
<svlu at enea.se> wrote:

> Hi,
>
> QMTest today assumes that the idle state of a target is an QMTest
> internal state, i.e. whenever it is idle or not only depends on the tests
> scheduled to it by the execution engine. However we have embedded targets
> that we intend to share among many user who all can run QMTest therefor
> when someone starts QMTest some targets may be occupied by somebody else,
> this can be detected by our target class implementation, which reserves
> the target, but there is no way to let the execution engine know about
> that, i.e. the target idle state is actually external to QMTest.
>
> I would like to implement an API to QMTest (and submit that as a patch)
> that does one of two things: (1) Asynchronously notifies the execution
> engine that the idle state of the target has changed.

We used to have something that was not too far from that, but it was very
hard to avoid deadlocks, so we went for something simpler.

It would be nice to have the facility you propose.

One key requirement is that we avoid spinning in the ExecutionEngine;
if there are no idle targets we want to be asleep so that we don't load
the machine.

Thanks,

-- 
Mark Mitchell                mark at codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com