Error in decays.f90

[sschmidt]

See the attached sin file, e+e- => tTH with subsequent factorized decays for the t and the W.

Executing it leads to a crash with the backtrace:

Runtime Error: decays.f90, line 430: Subscript 1 of IGNORE_HEL (value 1) is out of range (1:0)
Program terminated by fatal error
decays.f90, line 430: Error occurred in DECAYS:DECAY_INIT
decays.f90, line 663: Called by DECAYS:DECAY_TREE_GENERATE_EVENT:DECAY_NODE_GENERATE_EVENT
decays.f90, line 632: Called by DECAYS:DECAY_TREE_GENERATE_EVENT
events.f90, line 207: Called by EVENTS:EVENT_GENERATE
simulations.f90, line 978: Called by SIMULATIONS:SIMULATION_GENERATE_EVENT
simulations.f90, line 1240: Called by SIMULATIONS:SIMULATION_EVENT
commands.f90, line 5158: Called by COMMANDS:CMD_SIMULATE_EXECUTE
commands.f90, line 1021: Called by COMMANDS:COMMAND_EXECUTE
commands.f90, line 6050: Called by COMMANDS:COMMAND_LIST_EXECUTE
whizard.f90, line 326: Called by WHIZARD:WHIZARD_PROCESS_STREAM
whizard.f90, line 304: Called by WHIZARD:WHIZARD_PROCESS_FILE
main.f90, line 353: Called by MAIN

during event generation. Compiled with gfortran, a similar error message occurs:

At line 430 of file decays.f90
Fortran runtime error: Index '1' of dimension 1 of array 'ignore_hel' above upper bound of 0

Reported by Jan Strube.

[Juergen Reuter]

This is really rather urgent....

[kilian]

The problem is that the decay routine tries to use a process record that is completely un-initialized (garbage). This happens only if 'integrate' for the subprocess is issued after 'unstable'. If 'integrate' is done before 'unstable', or 'integrate' is omitted completely, everything's ok.

So the first workaround is to write 'unstable' after all integrations.

Investigating further ...

[kilian]

In r3664, I catch this as an error condition.

The user should write 'unstable' after all relevant integrations, since all decay channels have to be set up again if an integral changes.

Of course, this could be done automatically, so I keep the ticket. Either the automatic recheck may be implemented, or the required order is documented in the manual for the next release.

[Juergen Reuter]

So hard it may seem: please try to work around this, and prepare a bug report.... The bug is still there in 4.6.2, actually. Could someone check 4.7.0., maybe Sebastian !?

[kilian]

@JR: Your last comment belongs to #460, I guess? Anyway, that is fixed. I'll file the bug report tomorrow.

[Juergen Reuter]

Oops, yes indeed, I'm not as multi-threaded as I thought...

[Juergen Reuter]

Any update on that one here?...

[Juergen Reuter]

This is now the main open issue for 2.0.7...

[kilian]

The decays are now updated, where applicable, after each integration command, cf. r3746. The example works as intended.