Hi, On Jun 26, 2014, at 9:17 AM, Kenneth Hoste wrote: > buildininstalldir is not being used there, it's using the MakeCp easyblock > and just copies what is really needed. > > Please test that PR, and let us know (in the PR) whether it worked for you.
oops! buildininstalldir vs makecp! this thread highlights an issue with scientific software that is delivered in such way: A) is it the intention of the upstream developer to really build in the source dir? => Potential problem: too much noise (or unneeded files) in the build dir B) is it the intention of the upstream developer indeed to only copy binaries/doc? => Potential problem: directories of future releases may be left behind Either way, upstream must make clear what's the case, when there is build process involved. IMHO, the only way to get around this dilemma is, to ask upstream developers about it; or, check documentation, if that is indeed specified there. Not all software to deliver necessarily faces the above dilemma, fi. here is a script of mine that goes for tarball extraction w/out build: http://fotis.web.cern.ch/fotis/QTOP/ Or, do you fiercely object to that? read http://en.wikipedia.org/wiki/Occam's_razor Let the flamewars begin, now :) I have a dream; that one day we will be able to provide a quality report on software; here is a criteria top10 collected during an effort of last year: https://github.com/fgeorgatos/easybuild.experimental/wiki/HPC-Software-Quality#criteria I'd be more than happy to pool effort with others on this, especially if you think alike that with proper communication with upstream, things can be improved seriously much. Sure, we cannot improve *all* software, but having scientific sw on HPC platforms described in a consistent way, for basic features, is a lot more tractable target! best, Fotis -- echo "sysadmin know better bash than english" | sed s/min/mins/ \ | sed 's/better bash/bash better/' # Yelling in a CERN forum
