I’ve recently noticed PyROOT often taking a very long time (20-30s, sometimes more) to do the first operation, e.g. a simple from ROOT import TROOT in the environment I was working with (ATLAS software stack).
I’ve found that the amount of time the operation takes correlates with how many paths are in the LD_LIBRARY_PATH env var. When I stripped this down to just two paths (ROOT’s lib dir and the gcc lib dir) the import runs significantly faster (4-5 seconds max).
The startup time of root (cling) doesn’t seem to be affected in the same way.
Is there something in PyROOT’s initialization that is involving it repeatedly looking for libraries in the library paths or otherwise interacting with the libraries in all these paths?
Sorry it was bad of me not to say which version it was, and hence assume it affected latest versions. It is indeed 6.28/10 … hopefully we get an LCG version of 6.28/12 soon that the stack can then moved to. I’ll report back if the problems persist after the upgrade.