[Rivet] Pluggable analyses branch

[Andy Buckley]

Lars Sonnenschein wrote:

> I would like to implement the CDF jet shape analysis.
> To start with I could use the Salam Cacciari fast jet package.
> But if it is not installed the analysis has to be protected against
> linking since undefined symbols would appear. How would I do this in a 
> clean way (the analysis would be pluggable and somewhere I would have to 
> check if the fastjet package is available).

I'm not sure if this is a real issue - analyses should be implemented in 
terms of projections rather than direct calls to external algorithms, so 
if you were able to compile and link an analysis against Rivet's FastJet 
projection, then the FastJet library is presumably available.

If developing a new analysis and not yet using a projection, then the 
analysis will be directly linked against FastJet, so again it's 
presumably there.

The possible complications involve use of rpath to ensure that library 
chaining happens automatically - this is a robustness requirement on any 
official analyses - and what happens if a library is moved between 
linking and running. The first has a simple, standard engineering 
solution; anyone who invokes the second deserves all they get, if you 
ask me!

Anyway, I don't think this is a show-stopper, but please have a go at 
implementing the analysis and we'll find any problems soon enough!

Andy