waitpid event-loop enabled

waitpid event-loop enabled

[Andrew Cagney]

FYI,

I've turned on the new event loop.

The one outstanding issue - stray waitpid events - was tracked down to 
the linux kernel and how it provides two waitpid notifications for a 
termination - one for the parent and one for the attached process.

Andrew

Re: waitpid event-loop enabled

[Phil Muldoon]

Andrew Cagney wrote:
> FYI,
>
> I've turned on the new event loop.

Some questions:

How does this differ from the old eventloop?

and


Does this affect any existing usage of eventloops? Two scenarios here:

1) The Frysk UI asks for eventloop refreshes every 3 second in a 
different Java thread

2) LinuxCoreFileHost takes an eventloop as a parameter and schedules 
(among other things) a refresh inside that eventloop?

Regards

Phil

Re: waitpid event-loop enabled

[Andrew Cagney]

Phil,

Good questions.  Functionally, the new event-loop does not differ from 
the old, they both implement the public methods defined in 
frysk.event.EventLoop which includes support for timers and the 
execution of requests.  Testing shows no difference in results between 
the two.  The code you mention would use those interfaces.

Internally, the new code is very different though.  Instead of using 
select for timeout and SIGCHILD to detect a waitpid event, it uses 
SIGALRM for timeouts and waitpid to detect a waitpid event.  This means 
that the new code is 100% insulated from SIGCHLD/WAITPID changes.

Andrew

Phil Muldoon wrote:
> Andrew Cagney wrote:
>> FYI,
>>
>> I've turned on the new event loop.
>
> Some questions:
>
> How does this differ from the old eventloop?


>
> and
>
>
> Does this affect any existing usage of eventloops? Two scenarios here:
>
> 1) The Frysk UI asks for eventloop refreshes every 3 second in a 
> different Java thread
>
> 2) LinuxCoreFileHost takes an eventloop as a parameter and schedules 
> (among other things) a refresh inside that eventloop?
>
> Regards
>
> Phil