ee -> jets MEPS@NLO example crashes

[Enrico Bothmann]

Running Examples/Jets_at_LeptonColliders/LEP_Jets/Run.NLO.dat crashes on the rel-2-2-4 branch. The only modification I did was to set the loop generator placeholders to

  LOOPGEN1:=BlackHat;
  LOOPGEN2:=BlackHat;
  LOOPMGEN:=OpenLoops;

However, the crash occurs when the BVI integration for ee -> jj starts, which should be taken of by the internal ME due to the default LOOPGEN0:=Internal; in the run card.

The binary search window might be quite large, because the same crash happens for rel-2-2-0 (sic!).

The crash happens in BBar_Emission_Generator::Active, because p_active is a null pointer at that time.

I've attached the print-out for the argument bviproc being passed to Active.

[Enrico Bothmann]

At least in rel-2-2-4, the crash seems to be non-deterministic, i.e. every 10 or so attempts, it actually starts integrating. This happens both for MacOS/clang and for GNU/Linux/gcc environments.

[Enrico Bothmann]

I've tracked down the crash to r24465 on trunk, which merges several commits from the unlops branch. One of those is r24459. If one does not include this unlops commit when merging, the crash does not appear.

r24459 initialises thresholds in ISR_Channels to p_psh->Cuts()->Smin(), instead of leaving them at 0.0. If one reverts this in later revisions, the crash disappears. I've tested this on rel-2-2-1 and rel-2-2-4. Somehow, the initialisation seems to be wrong for lepton collider set-ups, but I have not digged deeper so far.

[Enrico Bothmann]

a patch for rel-2-2-4 (and other more recent revisions)

[anonymous]

Fixed with r31007 on trunk and with r31009 on rel-2-2-4.

Thanks to Stefan for helping me with the fix!