GROMACS easyblock updated for version 5#513
Conversation
|
Test PASSed. |
|
This gromacs.py did not succeed in installing GROMACS 5.0.4 on SL 6.6. |
|
@sbecuwe do you mean that those files are directly under |
|
It is installed directly under lib64. (In our 4.6.1 tree, files were directly installed under lib.) |
|
Okay, it seems to do the same thing on one of my test machines. It's bizar that the install dir changed on my system on some point. Might have something to do with plumed. I'm gonna drop this PR for the time being. |
|
OK, closed a bit too soon. The situation is:
A check on Linux distro could fix this. Or figuring out why autoconf/libtool want to use a platform name subdir. The latter is probably the preferred but hardest option. If I find the time, I will have a look. Don't merge this for the moment. |
|
Test PASSed. |
|
I tried this one and is working great for gromacs 5.0.4. Thanks @wpoely86 !! |
|
@pescobar on what distro did you try it? |
There was a problem hiding this comment.
GROMACS 5.x includes a gmx binary that should be used as main program; in 5.0 these are symlinks, in 5.1 the symlinks are gone, so just check for gmx in 5.x
There was a problem hiding this comment.
feedback provided by @rapostolo, he had more suggestions for improvements, more info later...
|
Regarding the library installation directory have a look at Gromacs commit gromacs/gromacs@d5c5abd. There we started using the cmake GNUInstallDirs to set paths on multiarch systems. |
|
@rapostolo: with respect to GROMACS 5.x support, we're seeing something weird... For me, on Scientific Linux 6, the GROMACS installation is producing files like For @wpoely86 on Debian on the other hand, the library is going to Do you have any idea why the library is one level deeper on Debian? This makes sanity checking, and including the right |
|
@boegel : the different paths are likely due to the GNUInstallDirs cmake module (http://www.cmake.org/cmake/help/v3.0/module/GNUInstallDirs.html) for settings paths on multiarch systems. Probably Debian reports x86_64-linux-.., while SL6 reports lib64 |
|
@rapostolo: thanks for the feedback; another reason to despise CMake... ;) |
|
There might be something in python that works the same way as GNUInstallDirs?? |
|
@rapostolo: I'll let the easyblock just figure out where |
detect library subdirectory in GROMACS easyblock (+ sync with develop)
|
Refer to this link for build results (access rights to CI server needed): |
fix libname in lib detection performed after installation
|
Refer to this link for build results (access rights to CI server needed): |
loosen up check used to determine lib subdir
|
Refer to this link for build results (access rights to CI server needed): |
fix definition of self.lib_subdir in GROMACS install_step
|
Refer to this link for build results (access rights to CI server needed): |
|
well-tested with easybuilders/easybuild-easyconfigs#1221 and easybuilders/easybuild-easyconfigs#1599, both on Debian & RedHat systems (which matters a lot for this easyblock) all existing GROMACS easyconfigs were retested, see https://gist.github.com/4db020f79ec7ef3f1651 going in, thanks @wpoely86 for kicking this off! |
GROMACS easyblock updated for version 5
GROMACS easyblock updated for version 5
Small fix in sanity check but it doesn't work everywhere:
get_platform_name()doesn't always return the correct string...