We observe core dumps of ParaView reading some large ICON data with v6.1. The same works well with v6.0. The problem occurs when reading the full 3D mesh, i.e. ``` reader = CDIReader(...) reader.Show3DSurface=1 ``` We have traced the problem in Debug mode with the following backtrace ``` #0 0x000040001555efa0 in __memcpy_generic () from /lib64/libc.so.6 #1 0x00004002b1a4cd4c in H5VM_memcpyvv () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #2 0x00004002b17a877c in H5D__compact_readvv () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #3 0x00004002b17d1f6c in H5D__select_io () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #4 0x00004002b17d2f64 in H5D__select_read () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #5 0x00004002b17a0be4 in H5D__chunk_read () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #6 0x00004002b17bf214 in H5D__read () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #7 0x00004002b1a424ec in H5VL__native_dataset_read () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #8 0x00004002b1a2da48 in H5VL_dataset_read () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #9 0x00004002b178e310 in H5D__read_api_common.constprop.0 () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #10 0x00004002b1790740 in H5Dread () from /user-environment/linux-neoverse_v2/hdf5-1.14.6-cobrby6yjq7vyf4x7m6wy7aoembrddul/lib/libhdf5.so.310 #11 0x00004002b14e8fa4 in NC4_get_vars () from /user-environment/linux-neoverse_v2/netcdf-c-4.9.2-vnbwjuaxwbokvwnvvvuel7kk6npgtwge/lib/libnetcdf.so.19 #12 0x00004002b149e064 in NC_get_vara () from /user-environment/linux-neoverse_v2/netcdf-c-4.9.2-vnbwjuaxwbokvwnvvvuel7kk6npgtwge/lib/libnetcdf.so.19 #13 0x00004002b13b5b60 in cdf_get_vara_float (ncid=65536, varid=varid@entry=11, start=start@entry=0xfffffe259450, count=count@entry=0xfffffe259478, fp=fp@entry=0x400a64020010) at cdf_int.c:508 #14 0x00004002b13baa70 in cdfReadVarSPPart (streamptr=0x19f81a70, varID=0, varType=<optimized out>, startpoint=<optimized out>, length=59354129, data=0x400a64020010, numMissVals=0xfffffe259508) at cdf_read.c:1107 #15 0x00004002b13bad8c in cdiStreamReadVarPart (numMissVals=0xfffffe259558, data=0x400a64020010, memtype=2, size=59354129, start=0, varType=2, varID=0, streamID=<optimized out>) at cdf_read.c:1140 #16 streamReadVarPart (streamID=<optimized out>, varID=0, varType=2, start=0, size=59354129, data=0x400a64020010, numMissVals=0xfffffe259554, memtype=2) at cdf_read.c:1171 #17 0x00004002b134154c in cdi_tools::cdi_get_part_unstruct<float> (cdiVar=0x1a7430c0, start=0, size=59354129, buffer=0x400a64020010, nlevels=3) at /capstor/scratch/cscs/jfavre/ParaView/ParaView-v6.1.0/Plugins/CDIReader/Reader/cdi_tools.h:122 #18 0x00004002b13391bc in cdi_tools::cdi_get_part<float> (cdiVar=0x1a7430c0, start=0, size=59354129, buffer=0x400a64020010, nlevels=3, grib=false) at /capstor/scratch/cscs/jfavre/ParaView/ParaView-v6.1.0/Plugins/CDIReader/Reader/cdi_tools.h:66 #19 0x00004002b1321cbc in vtkCDIReader::CheckForMaskData (this=0x19b20170) at /capstor/scratch/cscs/jfavre/ParaView/ParaView-v6.1.0/Plugins/CDIReader/Reader/vtkCDIReader.cxx:1834 #20 0x00004002b131fe5c in vtkCDIReader::AllocSphereGeometry (this=0x19b20170) at /capstor/scratch/cscs/jfavre/ParaView/ParaView-v6.1.0/Plugins/CDIReader/Reader/vtkCDIReader.cxx:1609 #21 0x00004002b131c7dc in vtkCDIReader::ReadAndOutputGrid (this=0x19b20170, init=true) at /capstor/scratch/cscs/jfavre/ParaView/ParaView-v6.1.0/Plugins/CDIReader/Reader/vtkCDIReader.cxx:1235 #22 0x00004002b131432c in vtkCDIReader::RequestData (this=0x19b20170, outVector=0x19204d10) ``` I have bisected the problem to https://gitlab.kitware.com/paraview/paraview/-/commit/22aea3dcbf090431317c58b0468fe88c56826a19 Can @nils.dreier revisit the topic? Our data is hard to share 55GB per timestep (full Earth oceanographic data on an R2B11 grid (1.2 km resolution)), but Nils most likely has similar data. I have reverted that commit into the v6.1 source code and I can read the data without crashing.