WHIZARD event generation stuck in endless loop

[Juergen Reuter]

The attached SINDARIN file where top decays are simulated leads to an endless loop in the event generation.

[Juergen Reuter]

SINDARIN file triggering the problem

[Juergen Reuter]

This has been reported by Marcus Rammes. But I get some events in the event files, but only two for roughly 16 hours runtime.

[sschmidt]

I played around with it a bit, and I found a sin file that sometimes works, sometimes runs into an (semi?-)endless loop. Running the sinfile with a WWHIZARD compiled with "nagfor -C=all -nan" gave a crash with the backtrace:

Runtime Error: cascades.f90, line 238: Subscript 1 of REDUCED (value 2) is out of range (1:1)
Program terminated by fatal error
cascades.f90, line 238: Error occurred in CASCADES:CASCADE_WRITE_FILE_FORMAT:REDUCED
cascades.f90, line 197: Called by CASCADES:CASCADE_WRITE_FILE_FORMAT
cascades.f90, line 800: Called by CASCADES:CASCADE_SET_WRITE_FILE_FORMAT
processes.f90, line 2621: Called by PROCESSES:PROCESS_SETUP_PHASE_SPACE
integrations.f90, line 303: Called by INTEGRATIONS:INTEGRATION_SETUP_PHASE_SPACE
integrations.f90, line 661: Called by INTEGRATIONS:INTEGRATION_INIT0
integrations.f90, line 840: Called by INTEGRATIONS:INTEGRATE_PROCESS0
integrations.f90, line 860: Called by INTEGRATIONS:INTEGRATE_PROCESS1
integrations.f90, line 931: Called by INTEGRATIONS:INTEGRATE_MISSING_PROCESSES

[ohl]

I have a similar report from Ben O'Leary. Here adding a new channel, e.g. letting a W decay into muons and electrons instead of only into electrons sends WHIZARD in an endless loop. His example is very complicated, I will try to simplify it.

[kilian]

There were two issues, actually:

(1) For the decays, the process objects were copied but the cut expressions contained a pointer to the original. Therefore the momenta in the cascade decays were all zero.

(2) The cut expressions for the decays were evaluated in the lab frame. They should have been boosted to the decaying particles' respective rest frames first.

These issues are solved in r3593. The test now does generate events.

However, there is an issue with the test file: Cuts on the particle decays must be specified in an invariant form. The pT cuts listed there make no sense when applied to boosted/rotated frames. This should be mentioned, once cascades are properly described in the manual.