Crash with index error and floating-point exception

Running and Runtime Errors
1

Dear FLUKA experts,

I have several very similar heavy ion jobs that keep failing in different ways and I cannot trace the root of the problem. First there is the file with 2.7 GeV/u Uranium that results in an error
At line 63 of file preeq/xinpro.f Fortran runtime error: Index '113' of dimension 1 of array 'amssst' above upper bound of 100

second, there is the same file, but at 2.0 GeV/u, which fails with
Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.

I have a suspicion that the errors are related to the corrections made in FLUKA 4.0.1 as in 4.0 the 2.7 GeV/u job failed with a floating point exception as well.
Note, as this is a heavy ion job, the custom source file was linked via ldpmqmd, not lfluka.

Best regards
Roman

SIS100_Dump_2000_Uranium.inp (6.2 KB) SIS100_Dump_2000_Uranium001.err (43.0 KB) SIS100_Dump_2000_Uranium001.log (449 Bytes) SIS100_Dump_2000_Uranium001.out (155.6 KB) SIS100_Dump_2700_Uranium.inp (6.2 KB) SIS100_Dump_2700_Uranium001.err (22.8 KB) SIS100_Dump_2700_Uranium001.log (638 Bytes) SIS100_Dump_2700_Uranium001.out (141.7 KB) source_realbeam.f (9.6 KB)

2

Thanks @rmartin for reporting. Since the failure is not instantaneous (I have been running both your inputs with no crash for a while), could you please upload also the random files of the crashing jobs? Namely ranSIS100_Dump_2700_Uranium001, ranSIS100_Dump_2000_Uranium001, ranSIS100_Dump_2700_Uranium002, and ranSIS100_Dump_2000_Uranium002, with the *002 ones as normally available in the respective fluka_# directory.
For upload purposes, you have to rename them with the extension .ran

3

here are the respective files
ranSIS100_Dump_2000_Uranium001.ran (1.6 KB) ranSIS100_Dump_2000_Uranium002.ran (1.6 KB) ranSIS100_Dump_2000_Uranium001.ran (1.6 KB) ranSIS100_Dump_2700_Uranium002.ran (1.6 KB)

4

Thanks again @rmartin. You met two actual (rare) bugs.

The second, i.e.
Program received signal SIGFPE: Floating-point exception - erroneous arithmetic operation.
was due to a not initialized variable in a photonuclear reaction by … a relatively low energy muon (!),

while the first, i.e.
At line 63 of file preeq/xinpro.f Fortran runtime error: Index '113' of dimension 1 of array 'amssst' above upper bound of 100
comes from the pre-equilibrium de-excitation of a too heavy nucleus produced in a U ion reaction on tungsten.

The respective fixes, concerning only users who experienced these crashes, will be released with FLUKA 4.0.2 later this autumn (let us know at fluka.teamATcern.ch if the crash occurrence prevents you from waiting until then). Note also that, contrary to your suspicion, both issues were already affecting FLUKA 4.0.0 and the fact that you observed one of them only with FLUKA 4.0.1 is just a matter of random history.