duplicate ROOT modules with "./bin/alienv enter O2/nightly-20180503-1"

[sawenzel]

Loading the o2 module from cvmfs/alice-nightlies.cern.ch gives (using module list):

 1) BASE/1.0                                   10) GEANT4/v10.3.3-3                           19) ZeroMQ/v4.1.5-6                            28) HepMC3/v3.0.0+git_d43693ce0e-1
  2) GSL/v1.16-26                               11) vgm/v4-4-12                                20) nanomsg/v1.0.0+git_c52f1bedca-1            29) O2HLTCATracking/hlt_o2_ca_tracking-v1.7-3
  3) FreeType/v2.6-35                           12) GEANT4_VMC/v3-5-12                         21) DDS/2.0-5                                  30) ApMon-CPP/v2.2.8-alice1-2
  4) libpng/v1.6.34-4                           13) protobuf/v3.0.2-6                          22) FairRoot/alice-dev-20180413-2              31) Monitoring/v1.6.0-1
  5) Python/v2.7.10-33                          14) pythia6/428-alice1-5                       23) Vc/1.3.3-4                                 32) Ppconsul/0.0.1-11
  6) Python-modules/1.0-16                      15) boost/v1.66.0-3                            24) libpng/v1.6.18-30                          33) Configuration/v1.4.1-1
  7) lzma/v5.2.3-1                              16) yaml-cpp/v0.5.2-26                         25) Python/v2.7.10-27                          34) ms_gsl/1-3
  8) ROOT/v6-12-04-12                           17) lhapdf/v6.2.1-alice1-1                     26) Python-modules/1.0-10                      35) arrow/v0.9.0-alice1-1
  9) GEANT3/v2-5-12                             18) pythia/v8230-10                            27) ROOT/v6-12-04-2                            36) O2/nightly-20180502-1

I notice that ROOT is loaded twice as ROOT/v6-12-04-12 and ROOT/v6-12-04-2 and one is shadowing the other. I believe there is mistake in the module file dependencies.

[sawenzel]

In fact, the same thing happens for a number of different packages.

[ktf]

Looks like HepMC3 is inconsistent with the rest. Not sure why at this point. This should simply not happen at build time, by construction. @dberzano was there any manual intervention there?