Segmentation fault in shower code

[Juergen Reuter]

With gfortran 4.7.4, WHIZARD produces a segmentation fault when run with MoXin?'s NMSSM SINDARIN script. The backtrace is

Program received signal SIGSEGV, Segmentation fault.
lorentz::add_vector4 (p=<error reading variable: Cannot access memory at address 0x11>, q=...) at lorentz.f90:638
638	    r%p = p%p + q%p
(gdb) bt
#0  lorentz::add_vector4 (p=<error reading variable: Cannot access memory at address 0x11>, q=...) at lorentz.f90:638
#1  0x00007ffff62858d8 in shower_core::shower_add_interaction_2ton_ckkw (shower=..., partons=..., ckkw_pseudo_weights=...) at shower_core.f90:694
#2  0x00007ffff6275443 in shower_core::shower_add_interaction_2ton (shower=..., partons=...) at shower_core.f90:602
#3  0x00007ffff627fbcc in shower_core::shower_generate_emissions (shower=..., particle_set=..., model=..., model_hadrons=..., os_data=..., matching_settings=..., data=..., valid=.TRUE., vetoed=.FALSE., 
    number_of_emissions=<error reading variable: Cannot access memory at address 0x0>) at shower_core.f90:302
#4  0x00007ffff61fe486 in shower::evt_shower_generate_weighted (evt=..., probability=6.95335580692039e-310) at shower.f90:300
#5  0x00007ffff6193f99 in event_transforms::evt_generate_unweighted (evt=...) at event_transforms.f90:198
#6  0x00007ffff62029f8 in events::event_evaluate_transforms (event=..., r=...) at events.f90:433
#7  0x00007ffff620346f in events::event_generate (event=..., i_mci=<optimized out>, r=...) at events.f90:619
#8  0x00007ffff60a865e in simulations::simulation_generate (simulation=..., n=50, es_array=...) at simulations.f90:1232
#9  0x00007ffff613c5f6 in commands::cmd_simulate_execute (cmd=..., global=...) at commands.f90:4253
#10 0x00007ffff6125f9e in commands::command_list_execute (cmd_list=..., global=...) at commands.f90:5577
#11 0x00007ffff6184062 in whizard::whizard_process_stream (whizard=..., stream=..., lexer=..., quit=.FALSE., quit_code=0) at whizard.f90:449
#12 0x00007ffff6183cc7 in whizard::whizard_process_file (whizard=..., file=..., quit=.FALSE., quit_code=0) at whizard.f90:424
#13 0x00007ffff7d91872 in MAIN__ () at main.f90:401
#14 0x00007ffff7d9275f in main (argc=<optimized out>, argv=<optimized out>) at main.f90:35
#15 0x000000396761ed5d in __libc_start_main () from /lib64/libc.so.6
#16 0x0000000000400789 in _start ()

The input file is attached.

[Juergen Reuter]

Interestingly this happens in a line in an if clause that is only entered when isr_is_possible_and_allowed = true (which should not be the case here) (?)

[Juergen Reuter]

When I set all debug flags in the shower to true, then with gfortran 6.0.0 I get:

| Events: event normalization mode '1'
 Copy incoming partons from particle_set to partons
 Copy outgoing partons from particle_set to partons
 Insert partons in shower
 Add interaction2toN
 isr_is_possible_and_allowed =     T
 D: shower_add_parent: for parton            1
 D: shower_enlarge_partons_array
 D: shower_add_parent: for parton            2
 D: integral_over_z_simple: t =   -13564.485149557437     
 mother =            0
******************************************************************************
******************************************************************************
*** FATAL ERROR: Shower: pdf only implemented for (anti-)proton
******************************************************************************
******************************************************************************
WHIZARD run aborted.

[Juergen Reuter]

This is the output from the nagfor 6:

Runtime Error: shower_core.f90, line 263: Subscript 1 of HADRONS (value 0) is out of range (1:2)
Program terminated by fatal error
shower_core.f90, line 263: Error occurred in SHOWER_CORE:SHOWER_GENERATE_EMISSIONS
shower.f90, line 298: Called by SHOWER:EVT_SHOWER_GENERATE_WEIGHTED
event_transforms.f90, line 198: Called by EVENT_TRANSFORMS:EVT_GENERATE_UNWEIGHTED
events.f90, line 433: Called by EVENTS:EVENT_EVALUATE_TRANSFORMS
events.f90, line 619: Called by EVENTS:EVENT_GENERATE
simulations.f90, line 1232: Called by SIMULATIONS:SIMULATION_GENERATE
commands.f90, line 4253: Called by COMMANDS:CMD_SIMULATE_EXECUTE
commands.f90, line 5577: Called by COMMANDS:COMMAND_LIST_EXECUTE
whizard.f90, line 449: Called by WHIZARD:WHIZARD_PROCESS_STREAM
whizard.f90, line 424: Called by WHIZARD:WHIZARD_PROCESS_FILE
main.f90, line 401: Called by MAIN
Aborted

I think the best would be to veto against ?ps_isr_active = true when using leptonic beams.

[Bijan Chokoufe Nejad]

r6926 should remove the segfault. It is a left-over of SSs notion that beam particles are always hadrons. But the veto would be a good idea, I think. The program should make clear that ps_isr_active is an option for hadron colliders (at least for now).

[Juergen Reuter]

Confirmed, the segfault or the stuck (depending on the compiler) is gone. In principle, we can close the ticket. Or shall we really fatal against using ISR QCD shower with lepton initial beams?

[Juergen Reuter]

The only open issue is the warning/fatal error when someone tries to use QCD ISR when having both beams leptons. Noted by BACN, will come with his shower/matching refactoring. Closing this one for now.