I have been building CVS Paraview against my own HDF5 1.8.1 install by defining H5_USE_16_API with no problems. After a recent CVS update, my compile failed due to an ABI change in the (undocumented) struct H5FD_class_t used in Utilities/Xdmf2/libsrc/XdmfH5Driver.cxx. I don't know why Xdmf2 needs to use this undocumented struct, but the patch below is a simple way to make things work with 1.8 again. It would be better for CMake to define a variable for the real version of HDF5 rather than branch on H5_USE_16_API.
Index: Utilities/Xdmf2/libsrc/XdmfH5Driver.cxx =================================================================== RCS file: /cvsroot/ParaView3/ParaView3/Utilities/Xdmf2/libsrc/XdmfH5Driver.cxx,v retrieving revision 1.5 diff -r1.5 XdmfH5Driver.cxx 134a135,139 > #ifdef H5_USE_16_API > static haddr_t H5FD_dsm_get_eoa(const H5FD_t *_file, H5FD_mem_t); > static herr_t H5FD_dsm_set_eoa(H5FD_t *_file, H5FD_mem_t, haddr_t addr); > static haddr_t H5FD_dsm_get_eof(const H5FD_t *_file); > #else 137a143,144 > #endif > 596a604,606 > #ifdef H5_USE_16_API > H5FD_dsm_get_eoa(const H5FD_t *_file, H5FD_mem_t) > #else 597a608 > #endif 624a636,638 > #ifdef H5_USE_16_API > H5FD_dsm_set_eoa(H5FD_t *_file, H5FD_mem_t, haddr_t addr) > #else 625a640 > #endif 665a681,683 > #ifdef H5_USE_16_API > H5FD_dsm_get_eof(const H5FD_t *_file) > #else 666a685 > #endif When I build with testing, I get an error because these directories don't exist. Index: VTK/Examples/Infovis/Cxx/CMakeLists.txt =================================================================== RCS file: /cvsroot/ParaView3/ParaView3/VTK/Examples/Infovis/Cxx/CMakeLists.txt,v retrieving revision 1.6 diff -r1.6 CMakeLists.txt 27,33d26 < # If Qt is enabled include the EasyView example < IF(VTK_USE_QVTK) < SUBDIRS(ChartView) < SUBDIRS(EasyView) < SUBDIRS(StatsView) < ENDIF(VTK_USE_QVTK) < Jed
pgpGRMUld5BAi.pgp
Description: PGP signature
_______________________________________________ ParaView mailing list ParaView@paraview.org http://www.paraview.org/mailman/listinfo/paraview