Hi, Anton. Fons did describe how the UNIX version TTimer works.
Under Windows NT asynch TTimer is served by the separate thread
and doesn't interrupt the main thread but simply suspend it for a while
and resume it later. (NT has no "good" signal mechanism like UNIX)
This way you will get your 30 ms if your Notify function may complete
(for me it must complete anyway) its activity for that time.
I think we should not go ahead with creating a separate thread for each
TTimer
the user created since it makes ROOT fragile.
think one single asynch TTimer is just enough (on either system) to allow
you to start
your OWN (not ROOT) thread('s) (or event process) and pass it whatever you
want.
Just my 2 cents.
Valery
> Hi Fons,
>
> thanx for the reply. It seems to be what I expected. Let me ask a more
> concrete question. Imagine I have two event processors based on
> TTimer.Notify(). Now I want to be sure that the first one will be
> triggered
> every say 30 ms, otherwise I may lose data. Now as soon as the second one
> is triggered, the first one should wait until the second one Notify() is
> completed. And that's not good. Can this situation be resolved somehow? Of
> course I can (as you suggest) use Notify of the second processor to set a
> flag and then check for this flag in the very body of the
> program, but that
> doesn't seem to be an elegant solution especially if several
> processes with
> different priorities should be run "in parallel". (well, if you have only
> one process which must actually do something every 30 ms or so and let all
> the other job to be done "when convenient time comes" it is OK, but if you
> have two processes with certain scheduling ... ). I m still voting for a
> Java-like solution :))
>
> Always,
>
> Dr. Anton Fokin
>
> Division of Cosmic and Subatomic Physics
> Lund University
> Sweden
>
> ----------
> > From: Fons Rademakers <Fons.Rademakers@cern.ch>
> > To: Anton Fokin <kosu_fokin@garbo.lucas.lu.se>
> > Cc: roottalk@hpsalo.cern.ch
> > Subject: Re: TTimer
> > Date: den 12 March 1999 16:12
> >
> > Hi Anton,
> >
> > timers never trigger at the same time or when in a TTimer::Notify.
> > Little explanation:
> >
> > - there are two types of timers synchronous and a-synchronous.
> > - synchronous timers are only handled via the ROOT eventloop
> > (see TUnixSystem::DispatchOneEvent()). If there are no
> mouse/keyboard
> > events then the synchronous timer queue is checked. So if the
> processing
> > of a mouse/keyboard event takes a long time synchronous timers are
> not
> > called for a while. To prevent this from happening one can call in
> long
> > procedures gSystem->ProcessEvents(). The system schedules only the
> > next timer in the queue when the current one's Notify() has
> finished.
> > - a-synchronous timers are triggered via SIGALARM, i.e. the
> program is
> > interupted and execution jumps to the Notify() function. When the
> > notify is finished the next a-sync timer is scheduled and
> the system
> > resumes from the place where it was initially interrupted. One of
> the
> > things to remember when using a-sync timers is don't make any
> graphics
> > calls in them. X11 is not re-entrant and it might be that the
> SIGALARM
> > signal interrupted the system while being in X11. A-sync timers are
> best
> > used to set flags that you can test at a convenient and controlled
> > time.
> >
> > Hope this helps.
> >
> >
> > Cheers, Fons.
> >
> >
> >
> > Anton Fokin wrote:
> > >
> > > Hi there,
> > >
> > > well, I am curious what happens if one sets up two TTimer derived
> objects
> > > and FirstTimer.Notify() call comes while SecondTimer.Notify() is
> executing.
> > > I would also like to know 1. if there is any way to protect a portion
> of
> > > code in a way it is done in Java (synchronized()) 2. If I can set any
> > > "priorities" for timer calls. (Well, to get this discussion useful for
> all
> > > rooters let's assume that I am a lamer in Unix/whatsoever system
> > > programming and do not know much about system calls
> (semaphores, etc.))
> :)
> > >
> > > Always yours,
> > >
> > > Dr. Anton Fokin
> > >
> > > Division of Cosmic and Subatomic Physics
> > > Lund University
> > > Sweden
> >
> > --
> > Org: CERN, European Laboratory for Particle Physics.
> > Mail: 1211 Geneve 23, Switzerland
> > E-Mail: Fons.Rademakers@cern.ch Phone: +41 22 7679248
> > WWW: http://root.cern.ch/~rdm/ Fax: +41 22 7677910
>
This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 00:43:30 MET