search

spatialaudio / nbsphinx

:ledger: Sphinx source parser for Jupyter notebooks
https://nbsphinx.readthedocs.io/
MIT License
457 stars 130 forks source link

PDF Conversion failing in nbconvert docs builds with 0.7.0 #474

Closed MSeal closed 4 years ago

MSeal commented 4 years ago

I had to pin nconvert's docs build to 0.6.1 to get it to build again. There's an error with an invalid png (only header written in the file it appears?) being generated by latexmk based on converting the results of python3 setup.py build_sphinx -b latex. I tried a few things to correct for the issue but didn't find a root cause in the change. It appears to have begun failing off the master build of nbsphinx between the 7th and 18th of March (the builds used to use git master during doc gen). https://github.com/jupyter/nbconvert/blob/master/.circleci/config.yml has the execution commands that fail on the latexmk -pdf task using the latest release of nbsphinx. If you nuke the build directory and rerun with 0.6.1. it succeeds.

mgeier commented 4 years ago

This error was due to multi-line data in an <img src="data:..."> tag in a notebook:

https://github.com/jupyter/nbconvert/blob/e67aad48c293071149125aeaf0e8980b988ecf5e/docs/source/customizing.ipynb#L24691-L24693

I've created a fix in #475.

MSeal commented 4 years ago

Thanks! I'll unpin our doc builds once that fix is released. Appreciate you digging into the issue.

mgeier commented 4 years ago

@MSeal I've just released nbsphinx 0.7.1 which contains the fix.

MSeal commented 4 years ago

@mgeier Great, looks like it's working. I'll change the pin to >=0.7.1