Gluon initial states with beam statement cause crash

[fbach]

The following input

process proc = gl, gl -> t, T
compile
sqrts = 7 TeV
beams = p, p -> lhapdf
integrate (proc) { iterations = 1:1000 }
simulate (proc) { n_events = 10 }

produces

******************************************************************************
******************************************************************************
*** FATAL ERROR: Evaluator product: no matching states found
***                 -----------------------------------
***              Maybe you chose the wrong initial state
***              for the hard interaction, or there is a
***              mismatch in the structure functions.
******************************************************************************
******************************************************************************
| There were no errors and    1 warning(s).

see also the attached file.

[Juergen Reuter]

Daniel and JRR confirmed. JRR has a first look at it.

[dwiesler]

Note that if the simulate statement is left out, there is no fatal! Thus it must somehow be connected to event generation.

[Juergen Reuter]

Wow, with gfortran on the MAC, I even get NaNs? in the state matrix density. Something is screwed up here

[Juergen Reuter]

It seems, that the problem appears only for gg and qg initial state. So whenever, there is colour in the initial state.

[Juergen Reuter]

The bug has nothing to do with LHAPDF, it also appears when removing the -> LHAPDF statement, it appears for qg and gg in the initial state whenever using a beams statement.

[kilian]

Fixed in r1937. Either color_invert or color_canonicalize has to handle the case of gluons, where color/anticolor entries are swapped. Chose the former.

[sschmidt]

The gg->Tt is fixed but now I get the same error message Fabian got for the process

process proc = g, u -> A, u
beams = p, p -> lhapdf

[sschmidt]

example .sin file for the new error

[kilian]

I'll have another look.

[kilian]

done in r1949. This was an unrelated bug in the routine which calculates the color-index offset when combining the structure functions for both beams.

Please double-check.

[Juergen Reuter]

Milestone v2.0-rc3 deleted