H5py version in Makefile.CP2 contains severe bug


I’d been experiencing sporadic problems with accessing .h5 files, and I think it may be due to a bad version of the h5py package. In testing I would occasionally get a segmentation fault (see http://code.google.com/p/h5py/issues/detail?id=167, which was discovered in h5py version 1.4.0, but was likely included in 1.3.1 that is part of the CP2 install from source.

Updating the Makefile.CP2 version of h5py to the latest release (2.0.1, accessible from h5py.googlecode.com/files/h5py-2.0.1.tar.gz) fixes the issue, and I no longer see any segfaults. Given the extent to which CP2 now relies on hdf5, I think this change should be incorporated into the Makefile.CP2.



Hi Lee,

Thanks for pointing is out. Once all the HDF5 changes are fully vetted, the makefile will be updated accordingly.

One thing to note is that Makefile.CP2 is not updated very frequently. It is probably best to use it with the same SVN version as it was last updated at, which is 10946. It’s probably fine to use it with the last “released” version of CP 10997, as well. This is all prior to HDF5 output, which is still undergoing development.

And to anyone finding this somewhat old thread – please refer to the updated Linux installation notes here (github.com/CellProfiler/CellPro … -for-Linux) or even easier, our rpm repository instructions here