|
[Rivet] auxiliary scripts in several generators.Andy Buckley andy.buckley at ed.ac.ukTue Apr 9 13:19:14 BST 2013
Hi Pere et al, I notice that ROOT has a root-config script: does that somehow survive Grid/CVMFS relocation? I have to admit that I've only used it locally. In my local, privately made build of ROOT I see lines like bindir=/usr/local/bin in this script, (I installed into prefix=/usr/local) so I suspect that it suffers from the same problem. If so, there are *-config script problems closer to the heart of SFT/LCG than the myriad external generator packages! But if there's a clever trick in the there that makes it "just work", then I'd be happy to try to put it into the packages that I have control over. Cheers, Andy On 08/04/13 23:47, David Grellscheid wrote: > Hi Pere, > >> I understand that nothing will be done. > > In Herwig++ that is. I can't decide by myself for the others. > >> So I will not insist but you are in my opinion wrong. > > Don't misunderstand me. I'd like to be convinced, but I really cannot > see at the moment how it can be done in Herwig. And dumping the design > flaws of the grid lower and lower down the chain until you stop at the > original authors of various tools is not an efficient way to solve the > problem. If Genser comes up with a neat way of how we can work around > this Genser-created reloc problem centrally, I'd be happy to look at it > for Herwig++. > >> The fact is that many packages (more than 100) are being re-installed >> at various locations (shared files systems, CVMFS, etc.) > > Re-installing always works anyway. It's only the raw copying of binaries > that's problematic. Also, don't both your examples have a fixed path > structure, so the DESTDIR approach works perfectly fine? > >> making these "config'" scripts useless. > > I fully agree, *-config scripts are useless other than for fairly > mainstream end users, who need to link one or two self-written things to > some library. I would actually say that ./configure setups that rely on > *-config scripts for their checks are broken, and the discussion we've > had shows why. To ease relocation, all other lookup aspects can be > controlled with environment variables (PATH, LD_LIBRARY_PATH, > PYTHONPATH, MANPATH), but typically *-config scripts ignore these values > completely. > >> Aiming for relocatable packages is not violating any GNU-style >> commandment and it can be done. Indeed the GNU community is providing >> a module to support relocation > > Sure. There's also Binreloc. It works fine if all you have is binary > executables and libs with hardcoded paths that need rewriting, but not > easily with *-config scripts or packages like Hw++ with its own separate > tree of generator supplementary files and their cross-relations. > > As far as I see, the most straightforward solution is still a small > number of separate builds/installations for 1) AFS, 2) CVMFS, (which > covers all the Grid issues right there) and 3)..n) the shared file > systems you mentioned. Rather than copying the final result over, why > not make the build process portable, so deploying on a new shared FS is > a button push. > > David > -- Dr Andy Buckley, Royal Society University Research Fellow Particle Physics Expt Group, University of Edinburgh The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
More information about the Rivet mailing list |