This method of installation uses a custom binary distribution of
Windows libraries that the previous pyhdf maintainer provided. The
binary distribution contained hdf4 and all it's dependencies (zlib,
libjpeg, and szip). These libraries are most likely outdated at this
point and there is no 64-bit version. I have no plans on updating them.
You can still use the custom binary distribution if you really wanted
by setting the environment variables LIBRARY_DIRS and INCLUDE_DIRS
appropriately, but I don't recommend using them anymore.
The "quick install" for Windows right now is to use conda-forge pyhdf
build.
This method of installation uses a custom binary distribution of Windows libraries that the previous pyhdf maintainer provided. The binary distribution contained hdf4 and all it's dependencies (zlib, libjpeg, and szip). These libraries are most likely outdated at this point and there is no 64-bit version. I have no plans on updating them.
This removes use of the environment variable
HDF4, which must be set to the directory containing the above mentioned binary distribution. conda-build sets it to the HDF4 version4.2, which makes builds fail. (see https://github.com/conda-forge/pyhdf-feedstock/pull/20#discussion_r264302801)You can still use the custom binary distribution if you really wanted by setting the environment variables LIBRARY_DIRS and INCLUDE_DIRS appropriately, but I don't recommend using them anymore.
The "quick install" for Windows right now is to use conda-forge pyhdf build.