HXT Meshing stops with error 11 (Segmentation failed) (try 2)
I have simple, but massive geometrical model: big box (fragmented into 8 equal parts) with inserted thousands of randomly placed small boxes (then BooleanFragments performed). If the number of small boxes is more than some value (5k-20k or more) then the meshing usually stops with "Signal: Segmentation fault (11)".
This issue is reproduceable only in multithreaded mode (-nt > 1). Single thread HXT stopped normally. If "Mesh.MeshSizeMax" is very small (about 0.25...0.5 of small box edge size), then multithread meshing usually stops normally (~8 of 10 times, related onto random factor of small boxes placing; but number of tetrahedra is crazy).
Info : Final tet. mesh contains 3147497 tetrahedra
Info : Final tet. mesh contains 510508 vertices
Info : tEmptyMesh = 52.300
Info : tVerifyBnd = 8.801
Info : tBndRecovery = 21.359
Info : tConvertMesh = 27.710
Info : tRefine = 204.129
Info : tOptimize = 2145.993
Info : Mesh generated
Debug : Start Hxt2Gmsh
*** Process received signal ***
Signal: Segmentation fault (11)
Signal code: Address not mapped (1)
Failing at address: 0x45f8a378c
[ 0] /lib/x86_64-linux-gnu/libc.so.6(+0x42520)[0x87d1520]
[ 1] gmsh(+0x9342bb)[0xa3c2bb]
[ 2] /lib/x86_64-linux-gnu/libgomp.so.1(+0x1db9e)[0x8742b9e]
[ 3] /lib/x86_64-linux-gnu/libc.so.6(+0x94b43)[0x8823b43]
[ 4] /lib/x86_64-linux-gnu/libc.so.6(clone+0x44)[0x88b4bb4]
*** End of error message ***
Issue reproduced with prebuild Gmsh and with manually built Gmsh onto various hardware (FX-6300 + 32GiB RAM, Ryzen 3600 + 64 GiB, i7-870 + 12 GiB; all machines with Ubuntu 22.04).
Here unrolled geometry to reproduce (big box with 10k small boxes; 10k - before fragmentation): t16v10_debug.geo_unrolled
And here Valgrind log ( valgrind --verbose --log-file=valgrind2.txt gmsh t16v10_debug.geo_unrolled -3 -algo hxt -nt 6 -v 99
): valgrind2.txt
I hope this information will be useful.