[Rivet] Warning: change to analysis loader!

[Andy Buckley]

Hendrik Hoeth wrote:
> Thus spake Frank Siegert (frank.siegert at durham.ac.uk):
> 
>> I have two reasons, why I would argue against splitting the directories:
>>
>> - Our bin/rivet-mkanalysis script isn't able to put the analysis skeleton 
>> into the correct directory anymore now (using the --srcroot option).
>> - Machines have more and more cores now, making the "make -j" option quite 
>> useful (especially if you use distcc ;)). So having fewer source files in 
>> one directory is slowing down the build unnecessarily, because make won't 
>> be able to "multithread" across directories.
> 
> I agree with Frank here, especially with the second issue (is anybody
> actually using rivet-mkanalysis?!?)

 From having just been through all the analyses as part of the 
conversion process, I think the answer is "yes" ;) It's pretty trivial, 
but I think quite useful for newbies (if we remember to mention it in 
the wiki getting started guide, manual, etc., which we currently don't)

> We could put the *.lo files into
> some build directory to make the output of "ls" shorter, but I don't
> really see any need for that either.

No, I don't think that's useful. I considered the parallelisation thing 
but thought the neatness might win out. I guess I mainly like it because 
I'm the person who hacks the infrastructure code that forces repetitive 
large rebuilds, so it's nice to be able to trivially disable rebuilds in 
subdirs... but it seems that the weight of opinion favours the all in 
one directory approach. Fine --- we can pretend for now that Rivet is a 
democracy ;) The flat form also echos the anainfo file structure and the 
fact that they get installed into one dir, and will make generating the 
docs easier... so even I'm convinced! I'll move them back tomorrow.

Is there any opposition to keeping the separate plugin libs? That I'm 
more attached to, and I don't think they block any distcc speedup, do they?

Andy

-- 
Dr Andy Buckley
Particle Physics Experiment Group, University of Edinburgh