From: Gordon Watts [gwatts@phys.washington.edu]
Sent: Friday, October 31, 2003 10:28 PM
To: daviesbg@fnal.gov
Cc: Harry Melanson
Subject: RE: Local rcp's not picked up when referenced from official one

Hi,
  Thanks for the suggestion, that is an excellent idea. Harry -- if you
are still putting together your hits file, this sis a technique that
might get included.

   Cheers, 
      Gordon 


-----Original Message-----
From: daviesbg@fnal.gov [mailto:daviesbg@fnal.gov] 
Sent: Thursday, October 30, 2003 3:05 AM
To: Drew Alton
Cc: Alan Jonckheere; Protopopescu, Serban D; 'Gustaaf Brooijmans ';
Gordon Watts
Subject: Re: Local rcp's not picked up when referenced from offical one

I believe dave evans made an "RCP Tree" maker using python for runjob...
Is this the kind of thing you meant. ( I have no idea what happened to
it).
Brian 

----- Original Message -----
From: Drew Alton <dalton@fnal.gov>
Date: Wednesday, October 29, 2003 9:29 pm
Subject: Re: Local rcp's not picked up when referenced from offical one

> Hi all,
>   I know this isnt the best way...but the way I've always dealt with
> this was:
> 1) Intentionally corrupt the rcp I am interested in. (I add the word
> 'break')
> 2) Modify the chain of rcp's I think is necissary
> 3) try running...if it doesnt die quickly with an rcp error
> 4) go back and rethink 2 to find what I missed.
> 
> Not the most elegant method.
> 
> 
> 
> On Tue, 28 Oct 2003, Alan Jonckheere wrote:
> 
> > Even better would be some way to trace the chain so they'd know what
> > they had to include!
> >
> >
> >
> > Protopopescu, Serban D wrote:
> >
> > >  Yes, it is the documented beahviour, but people get bitten
> > > by this over and over. It would be nice if users could
> > > be made aware of this when they  modify an rcp file.
> > > I don't know how to do it but maybe someone can come
> > > up with a way of warning users what is happening.
> > >
> > >     Serban
> > >
> > > -----Original Message-----
> > > From: Alan Jonckheere
> > > To: Gustaaf Brooijmans
> > > Cc: Gordon Watts; d0rug@fnal.gov
> > > Sent: 10/28/2003 10:26 AM
> > > Subject: Re: Local rcp's not picked up when referenced from 
> offical one
> > >
> > > This is the documented, designed, deliberate, etc etc etc. 
> behavior.> >
> > > The reason is illustrated by the following:
> > >    1) You use an rcp which includes another which includes etc
> > >    2) You store that first rcpid in your data
> > >    3) You readback the data and want to lookup the rcps used to 
> create> > the data.
> > >
> > > You don't want it to find an rcp that happens to be in your 
> area. You
> > > want the original ones used.
> > >
> > > Physically this is enforced by storing the actual rcpids of rcps
> > > referenced in the db. Thus, once you've gotten into the 
> official db in
> > > an rcpid lookup, you can't get out of it. All references in 
> that chain
> > > will be satisfied by rcps in the official db. It's actually a 
> bit more
> > > complicated than that, but loosely speaking that's a good way to
> > > remember this behavior.
> > >
> > > Alan
> > >
> > > Gustaaf Brooijmans wrote:
> > >
> > >>Hi Gordon,
> > >>
> > >>This is the documented behavior.  You need to have the full rcp 
> tree> >>locally when changing one of the pointed at rcps.  I 
> understand this
> > >
> > > is
> > >
> > >>not the intuitively expected behavior, but many compromises 
> apparently> >>needed to be made to save us from copying every 
> single needed rcp to
> > >
> > > the
> > >
> > >>running directory (which was definitely worse).
> > >>
> > >>Gustaaf
> > >>
> > >>
> > >>On Mon, 27 Oct 2003, Gordon Watts wrote:
> > >>
> > >>
> > >>
> > >>>Hi,
> > >>> I've gotten myself confused over the RCP resolution strategy 
> (to the
> > >>>tune of two hours).
> > >>>
> > >>> I'm using rund0exe, and p15.05.00. I have my own copy of the
> > >
> > > framework
> > >
> > >>>(from the head of p15-br).
> > >>>
> > >>> I'm building trigsim, and that includes the reco_trig.rcp from
> > >
> > > d0reco,
> > >
> > >>>which then includes the l1l2_reco.rcp from the l1l2_reco 
> package. The
> > >>>d0trigsim and 1ll2_reco packages are in my local release, but 
> d0reco> >
> > > was
> > >
> > >>>not. The executable would pick up the l1l2_reco.rcp that was 
> in the
> > >>>official database, not my local database. If I then did "addpkg
> > >
> > > d0reco"
> > >
> > >>>and ran again, it would pick up the local version of the 
> l1l2_reco.rcp> >>>file.
> > >>>
> > >>> Is this correct behavior? Or is there something more subtle 
> going> >
> > > on?
> > >
> > >>>You can look at what I've got in
> > >>>/work/gasworks-clued0/gwatts/higgs/trigger/p15.05.00
> > >>>
> > >>>     Cheers,
> > >>>           Gordon.
> > >>>
> > >>
> > >>
> > >>---------------------------------------------------------
> > >>Gustaaf Brooijmans
> > >>Fermilab, MS # 357              Tel: +1 (630) 840 4269
> > >>P.O. Box 500                    Fax: +1 (630) 840 8481
> > >>Batavia, IL 60510
> > >>---------------------------------------------------------
> >
>