fix: clear[] crashed after a fire whose front had stopped

[yanissrairi]

When a fire's front stops everywhere (e.g. an ignition in non-burnable fuel), the next clear[] segfaults.

getUpcomingEvent took the last event out but left head pointing at it. The stopped node's event has an infinite time, so insert() deleted it, and clear() deleted it again through head. The fix sets head = nullptr in that branch.

Reproduction, from tests/runff:

include[params.ff]
loadData[data.nc;2025-02-10T17:35:54Z]
startFire[loc=(38305.0,28205.0,0);t=0]
goTo[t=600]
clear[]
loadData[data.nc;2025-02-10T17:35:54Z]
startFire[loc=(35881.873264425,28699.674854985,0);t=0]
goTo[t=600]

Added tests/unit/test_timetable.cpp, which crashes without the fix. ctest and runff pass.

[filippi]

thanks @yanissrairi the clear[] hasn't been much tested yet. My problem is testing it into a coupled fire/atmospher context with MPI and Fortan bindings, which i have no time to do yet.. very sorry. I'll be back on these around Oct 20.

[antonio-leblanc]

Thanks @yanissrairi! Reproduced the crash on dev and the fix here; real_case output is byte-identical before and after. I merged dev in to fix the CHANGELOG conflict.

[antonio-leblanc]

Tested the coupled case with this branch vs dev:

• Ubuntu 24.04 (WSL2), GCC/gfortran 13.3, Open MPI 4.1.6
• Meso-NH 6.0.1 (Zenodo release), built as LXgfortran-R8I4-MPIAUTO-O2
• ForeFire built with -DFOREFIRE_ENABLE_MPI=ON, linked as $SRC_MESONH/exe/libForeFire.so: dev at f4b3fd8, this PR at 1b011b7
• Case tests/mnh_real_nested, mpiexec -np 4, with LHORELAX_SVC1R3, NLBLX and NLBLY removed from the two .des
• 4 runs: dev, PR, dev, PR

Result: 107 of 119 output files (ForeFire/Outputs, MODEL1, MODEL2, RUN12.*.FIRE.*.nc) are bit-identical across all 4 runs. The other 12 (the air_density dumps and the NetCDF history timestamp) also differ between the two dev runs, so it's run-to-run noise, not this change.