Hi Jens et al, On 12/07/2012 10:57, Jens Timmerman wrote: > If you don't need anything in the toolkit but the compiler you can > create a new toolkit with just GCC and use this to build your packages. > > Feel free to create your own.
Sure :) I was just trying to understand the argument behind putting an explicit hard dependency on a gcc version for a language like Python (I can grasp things like arguments in paper "Reflections on trusting trust") but, I wanted to hear a war-story why this is important to keep strict... > Are you using the latest version here? > This typo was fixed 2 months ago, and the file actually has a different > name. > https://github.com/hpcugent/easybuild/commit/0e09e3210ff79879519f494746da29bfb0819ee3#easybuild/easyconfigs/l/Libint/Libint-1.1.4-goalf-1.1.0-no-OFED.eb Mea culpa; It may be an effect of me doing multiple checkouts and copy-around; It's a good moment for another clean start for me apparently! > Apparently something is missing here, ./autogen.sh needs to be run > before configure > Since BEAGLE has no source by default, our tests didn't pick up on this. > Thanks for reporting this. Throw to me some more and I can test some more :) btw. I also have some comments as regards cross-distro support: "tcsh" is never satisfied as "rpm dependency" in our local Debian HPC world. What I do now as a workaround is, to manually remove such dependencies especially for the items that are visible here: (yeap, libib* should be there) https://hpc-tracker.uni.lu/projects/hpc/wiki/Software_Development_Environment Can we find a more elegant solution to this? (eg. some "escape mechanism") thanks, Fotis -- echo "sysadmin know better bash than english" | sed s/min/mins/ \ | sed 's/better bash/bash better/' # Yelling in a CERN forum
