[Rivet] Minor bug in rivet-bootstrap [Re: Rivet 1.5.0 released]

[Andy Buckley]

Hmm, annoying. Can you do one last check... are the Rivet C++ libraries
installed by automake/libtool (libRivet.so, etc.) also located in the
lib64 directory or are they in .../lib/? I just tried installing on
another 64 bit SLC5 system (the Durham Grid UI machine), with Python 2.4
and everything was installed into .../lib/! Can you post the configure
command, config.log, and Python version that you used?

As far as I know, autotools will always install into $prefix/lib, and
the lib64 is actually coming from distutils (in your case)... so rather
than not being aware of the lib64 convention, your distutils *seems* to
use it when installing but not when being queried about what the install
location will be.

The Makefile.am where the Python install script is called has no mention
of the automake $libdir at all:
$(PYTHON) setup.py install --prefix=$(DESTDIR)$(prefix)

Why are the simple things always the most troublesome?!

Andy


On 15/03/11 21:57, Ben Waugh wrote:
> waugh> python
> Python 2.4.3 (#1, Jan 14 2011, 22:42:01)
> [GCC 4.1.2 20080704 (Red Hat 4.1.2-48)] on linux2
> Type "help", "copyright", "credits" or "license" for more information.
>>>> import distutils.sysconfig
>>>> print distutils.sysconfig.get_python_lib(prefix='/foo')
> /foo/lib/python2.4/site-packages
> 
> If I understand correctly after a few minutes with Goole, this is a
> long-standing bug in distutils. See e.g.
> http://mail.python.org/pipermail/distutils-sig/2005-April/004466.html
> 
> Cheers,
> Ben
> 
> On 15/03/11 18:11, Andy Buckley wrote:
>> On 15/03/11 13:14, Ben Waugh wrote:
>>> On 15/03/11 12:51, Andy Buckley wrote:
>>>> On 15/03/11 12:38, Ben Waugh wrote:
>>>>> There is only one minor bug that I have discovered so far. The
>>>>> generated
>>>>> rivetenv.sh sets PYTHONPATH to
>>>>>    $INSTALLDIR/lib/python2.4/site-packages
>>>>> but the Python libraries are actually in
>>>>>    $INSTALLDIR/lib64/python2.4/site-packages
>>>>>
>>>>> So just running Rivet out of the box gives the infamous "No module
>>>>> named
>>>>> rivet" error.
>>>>
>>>> Damn, so close! Hmm, I *thought* that I was now taking the generated
>>>> rivetenv scripts from the Rivet build, in which case the installation
>>>> directory is actually generated (in configure.ac) by Python's own
>>>> installation system! I'll have to check the bootstrap script again.
>>>>
>>>> In the meantime, can you take a look at the generated rivetenv.sh which
>>>> should be in the Rivet build directory on your system (i.e. not
>>>> necessarily the one that rivet-bootstrap tells you to use) and see
>>>> if it
>>>> also specifies lib rather than lib64? It would certainly be strange if
>>>> Python reports that it'll be installing into one directory and then
>>>> actually copies the files into another...
>>>
>>> Unfortunately...
>>>
>>> waugh>  grep PYTHONPATH build/rivet/rivetenv.sh
>>> export
>>> PYTHONPATH="/tmp/waugh/rivet/local/lib/python2.4/site-packages:$PYTHONPATH"
>>>
>>
>> Hmmm! Can you now try running a python prompt on the machine with which
>> you did the installation, run this:
>>
>> import distutils.sysconfig
>> print distutils.sysconfig.get_python_lib(prefix='/foo')
>>
>> and post the output? This should be the method used by Python to do the
>> module installation, and is the method used to generate that PYTHONPATH
>> entry in the rivetenv.sh script, so it *really* should be returning a
>> value with "lib64" in it if that's the dir name that Python is choosing
>> to install the libraries into!
>>
>> Cheers,
>> Andy
>>
> 


-- 
Dr Andy Buckley
SUPA Advanced Research Fellow
Particle Physics Experiment Group, University of Edinburgh

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.