[Rivet] Python, Rivet 1.20, CERN AFS, LXBATCH

[Terrance Figy]

Hello Andy,
I completely agree this is a problem with the software setup at CERN  
TH. This is why I decided to test the software setup. I wanted to make
sure that it actually works. I will make sure the person in charge  
with the software setup is aware of this problem. Thanks for your help.
Cheers,
Terrance

On 26 Mar 2010, at 18:05, Andy Buckley wrote:

> On 26/03/10 16:34, Frank Siegert wrote:
>> Andy Buckley, Friday 26 March 2010:
>>> But the real problem is that we aren't bundling the generated source
>>> files anymore, so when they *are* generated, they are generated with
>>> the system copy that isn't GCC4.4-compatible. I wasn't aware that  
>>> we'd
>>> stopped bundling these files... we used to have problems with HepMC
>>> version details being written into these files, which caused
>>> compatibility problems, but now that is all done in C++ and I think
>>> it's safe for SWIG-generated files from my Ubuntu 9.10 machine (with
>>> up-to-date SWIG) to be bundled in the tarball for use on SL  
>>> machines.
>>> At least, it finished compiling just fine when I copied the attached
>>> files into the rivet/pyext directory: can you try that, Terrance?
>>>
>>> Riveteers: Any opposition to me re-adding the rivet_wrap.cc and
>>> rivet.py bundling, if this works for Terrance?
>>
>> We stopped bundling them because it wasn't necessary anymore:  
>> Without all
>> the HepMC complications, SWIG worked just fine even on SL4. Of  
>> course I
>> wasn't aware of the gcc 4.4 non-compatibility of that SWIG  
>> version. If
>> somebody provides/uses a gcc 4.4 SW environment, shouldn't this also
>> include an updated SWIG that would then be used? So maybe that  
>> problem
>> should rather be fixed in CERN TH's software setup.
>>
>> I don't exactly remember why it is a bad idea to bundle them...  
>> are they
>> really as system agnostic as you describe?
>
> They *should* be: they're just meant to be generic standard C++  
> wrapper
> functions, connecting to whatever headers were included for the thing
> being wrapped.
>
> I have a feeling that we removed them because a) there was  
> (apparently)
> no longer a need and b) rivet_wrap.cc is pretty large, at ~500k.
>
> Terrance, can you request via the TH group that the gcc44 setups  
> need an
> updated copy of SWIG, as Frank has suggested, since the system copy
> doesn't produce gcc44-compatible code? That is the preferable  
> solution,
> I think.
>
> Andy
>
> -- 
> Dr Andy Buckley
> SUPA Advanced Research Fellow
> Particle Physics Experiment Group, University of Edinburgh