Fixes for the build of Python and Lhapdf for the analysis release

Python was installing incorrectly until now, breaking the SLC6 build of AnalysisBase.

The fix for the Lhapdf installation should just make the project a bit more robust. Allowing it to install everything else correctly if/when the build of Lhapdf breaks.

So... This is a first merge request.

@graemes , @leggett , I think we should start modifying the repo like this from now on. I'll talk today once again to Christian about possibly setting up a CI test for these merge requests as well.

Reassigned to @cgumpert

Jenkins Build FAILURE

Results available at: Jenkins [AE-MERGE-REQUEST #1]

@akraszna now atlasbot also wants to be a developer of your atlasexternals fork

Jenkins Build FAILURE

Results available at: Jenkins [AE-MERGE-REQUEST #2]

Jenkins Build SUCCESS

Results available at: Jenkins [AE-MERGE-REQUEST #3]

Jenkins Build FAILURE

Results available at: Jenkins [AE-MERGE-REQUEST #4]

Ok, the good news are: it works in principle. The build failed because I set a timeout at 30min which is probably to aggressive. atlasexternals will always be built on aibuild092 in /build2/atlasexternals in case you want to check some local logfiles. I will increase the timeout and restart the build.

Cool. In the meanwhile I'm actually playing with doing things in a slightly more orthogonal way. But I'll announce it if I have something that actually works. (And pushes its results to CDash in a way that I now hope.)

The current build looks good so far. I just noticed that AnalysisBaseExternals is not happy with the default python version:

http://aibuild080.cern.ch:8080/job/AE-cmake/5/PROJECT=AnalysisBaseExternals,label=externals/console

I believe that AthenaExternals build does not care because it is taking python from LCG, right?

Jenkins Build SUCCESS

Results available at: Jenkins [AE-MERGE-REQUEST #5]

Maybe I should add some messages to the configuration to make people less worried. The code not finding Python 2.7 is not an error. It's not even a warning. The code just checks whether Python 2.7 is available on the system to decide how to build stuff. Like:

https://gitlab.cern.ch/atlas/atlasexternals/blob/master/External/ROOT/CMakeLists.txt#L57

I guess I should just add a QUIET keyword to the find_package calls not to make people worried.

Here I'm just waiting with https://gitlab.cern.ch/atlas/athena/merge_requests/227 to be merged in before I would accept this MR. Just not to affect any of the Athena merge requests. (Even though this change should not affect them, we're not in a too big hurry with this.)

Resolved all discussions

Status changed to merged

Mentioned in commit b4639e56