ParaView fails to read composite data correctly in parallel
Steps to reproduce :
- launch
pvserver
in parallel (mpirun -n 2 ./bin/pvserver
or./bin/mpiexec -n 2 ./bin/pvserver
), - launch ParaView and connect to the local server,
- open a VTK XML multi-group data file (.vtm) (see attached file below).
The example I use is a MultiblockDataset containing 4 leafs (PolyData). By default by reading the file with 2 MPI process we have 2 blocks on each process (can be seen by coloring with vtkProcessId
array).
The parallel XML reader fails to read correctly the information data of the leafs of the 2nd process. On the information panel, the names of the 3rd and 4th blocks are wrong, the data type is unknown and no metadata about points, cells numbers or data arrays are shown. Furthermore, when we display the data on a spreadsheet view, we can see the data of the two first blocks is duplicated on the two other blocks (the data about the 3rd and the 4th block begin on the row n°48) :
Here is the valid data, obtained by opening the file in a single process. We can see on the spreadsheet that we have valid data for the 3rd and 4th block :
Doing an ExtractBlock
on this data, for example, fails. When gathering the data on the main thread, we got an error saying the structure of the composite data of each thread is different.
Test case : Test_Case.zip