search

labscript-suite-temp-2 / lyse

lyse is an analysis framework. It coordinates the running of python analysis scripts on experiment data as it becomes availiable, updating plots in real time.
BSD 2-Clause "Simplified" License
0 stars 0 forks source link

HDF5 library version mismatched error #38

Closed philipstarkey closed 6 years ago

philipstarkey commented 6 years ago

Original report (archived issue) by Lars Kohfahl (Bitbucket: lkohfahl, GitHub: lakodarmstadt).

The original report had attachments: hdf5_h5py_update_Windows_py2.txt, hdf5_h5py_update_Windows_py3.txt

Hey, my name is Lars Kohfahl and I am currently keeping labscript in Darmstadt, Germany up to date after Jan Werkmann finished his masters thesis. I now came across a problem where I do not know if it is an error from my side or some bug: When updating hdf5 (to 1.10.2) or h5py (to 2.8.0) with conda and starting lyse afterwards, a window pops up saying "python does not work anymore" and an error message is displayed saying that there is a "HDF5 library version mismatched error" saying that the versions of the headers and the library do not match. I can reproduce this error in Python 2 and Python 3. I have attached the error messages for Py2 and Py3. Note that in Py2 I first updated HDF5 and then h5py, in Py3 I did it the other way around. Can anyone help my how to solve it?

philipstarkey commented 6 years ago

Original comment by Philip Starkey (Bitbucket: philipstarkey, GitHub: philipstarkey).

Hi Lars, good to hear from you!

This sounds like you might have conflicting HDF5 DLLs in your system. How are you installing HDF5 before h5py? I didn't think that step was necessary, so possibly uninstalling everything and then just using conda to install h5py might do the trick.

Edit: I should add that if you have a reason for needing that specific HDF5 version, and not say 1.10.1) then you'll need to compile h5py from source I think rather than having conda install the pre-built version (there may be other simpler ways to do it but I'm not all that familiar them)

philipstarkey commented 6 years ago

Original comment by Lars Kohfahl (Bitbucket: lkohfahl, GitHub: lakodarmstadt).

Thanks for the advice. For the Py3 environment this did the trick, for the py2 it didn't. Finally I found the reason: conda and pip had both installed h5py. If the error comes up, both versions were different. So I first tried to just remove the pip h5py version which resulted in new errors. When installing the latest h5py in pip and conda, it works just fine.

I am not quite happy with a pip and a conda version, but okay.

philipstarkey commented 6 years ago

Original comment by David Meyer (Bitbucket: dihm, GitHub: dihm).

Pip/conda issues can be quite annoying. Pip installing dependencies through pip instead of conda is where I always get tagged. It gets worse because the only way to remove the pip package is to uninstall the conda package first, then the pip package, then reinstall the conda package.

If you do that you should only need the one package.

philipstarkey commented 6 years ago

Original comment by Lars Kohfahl (Bitbucket: lkohfahl, GitHub: lakodarmstadt).

Thanks for that advice. I did it by chance on a different computer and there it worked, so I will use it everywhere now.

philipstarkey commented 6 years ago

Original comment by Lars Kohfahl (Bitbucket: lkohfahl, GitHub: lakodarmstadt).

Wasn't a bug