Color-flow mismatch in LHAPDF event generation

[Juergen Reuter]

Whizard produces segfaults of the former type (backtraceable to flavor_get_pdg), when started without the 'beams = xx' parameter in the SINDARIN file. If the beams parameter is set (and event generation is requested), there are still segfaults with quarks in the initialstate e.g. drell yan, backtraceable to state_matrices.f90:

This is probably linked with the flavor problem, since the LHEF output of W2 misses information:

For all listed events, the incoming flavor pdg is always zero (only for the HEPEUP event blocks, the HEPRUP init common block gives the correct pdg code). The cross section and the color information is also zero. I have compared the results with the LHEF output of W1.94.

__state_matrices_MOD_state_iterator_get_qn_multi () at state_matrices.f90:693
693         node => it%node
(gdb) bt
#0  __state_matrices_MOD_state_iterator_get_qn_multi () at state_matrices.f90:693
#1  0xb7e41f3e in __state_matrices_MOD_state_matrix_factorize () at state_matrices.f90:1074
#2  0xb7e46495 in __interactions_MOD_interaction_factorize () at interactions.f90:584
#3  0xb7e68fe0 in __particles_MOD_particle_set_init_interaction () at particles.f90:444
#4  0xb7ef8ccd in __events_MOD_event_factorize_process () at events.f90:182
#5  0xb7ef8d69 in __events_MOD_event_generate_unweighted () at events.f90:157
#6  0xb7f35c2d in __commands_MOD_cmd_simulate_execute () at commands.f90:4386
#7  0xb7f2fc57 in __commands_MOD_command_execute () at commands.f90:1579
#8  0x080c7c38 in ?? ()
#9  0xb7faa040 in known.8357 () from /localscratch/dwiesler/w2/lib/whizard/libwhizard.so.0
#10 0x080ac830 in ?? ()
#11 0xb7faa040 in known.8357 () from /localscratch/dwiesler/w2/lib/whizard/libwhizard.so.0
#12 0xbfc94668 in ?? ()
#13 0xb7fd5090 in _dl_rtld_di_serinfo () from /lib/ld-linux.so.2
#14 0xb7f2fcd2 in __commands_MOD_command_list_execute () at commands.f90:5380

[kilian]

Should be fixed in r1582. Please check.

[Juergen Reuter]

The comment from Daniel Wiesler is: The missing beam parameter does not produce a seg fault any more, but still he cannot generate LHEF events, as the second seg fault problem still persists as described by the ticket.

[kilian]

Thanks for the input file; doing this with nagfor shows that the problem is unrelated to #147 and unrelated to LHEF. There is a mismatch in color codes when building up complete events. Investigating further ..

[kilian]

The color-flow handling in the convolution of structure function + hard ME was buggy. The problem appears only for event generation.

In r1590, two bugs are fixed. There is a remaining problem (which also occurs here) that the in-state is a color singlet. I still have to handle that case.

[sschmidt]

I have two input files that crash, they look similar to the cases above. I'll just attach these here.

[kilian]

With r1600, this should be finally fixed. The segfaults have disappeared, since now color-singlet processes can be handled when convoluting beam and hard interaction.

[Juergen Reuter]

Milestone v2.0.0final deleted