Closed jayaddison closed 1 week ago
The removal of the module causes our LaTeX writer tests to fail, as evident in a recent GitHub Actions docutils HEAD workflow failure.
Strictly speaking this failure only occurs unless the roman
Python package is available on the system already, since we will use that as a fallback when available: https://github.com/sphinx-doc/sphinx/blob/a1510de4777eaa2e569435f95b05f6f3293d7035/sphinx/writers/latex.py#L29-L31
Yep, we discussed this on the DU tracker recently. I'll prepare a fix.
A
Closed in #13131 (a pleasing PR number).
A
Describe the bug
With a recent modification in the upstream
docutils
codebase (git-svn-id: http://svn.code.sf.net/p/docutils/code/trunk@9977 929543f6-e4f2-0310-98a6-ba3bd3dd1d04
), thedocutils.roman
module has been removed, superseded by an alternative implementation indocutils.utils._roman_numerals
.We rely on the former to provide an implementation of
toRoman
in the Sphinxsphinx.writers.latex
module, with theroman
package -- a non-vendored implementation of the code found indocutils.roman
, with some updated releases since the version(s) used indocutils
-- as a fallback.The removal of the module causes our LaTeX writer tests to fail, as evident in a recent GitHub Actions
docutils
HEAD workflow failure.How to Reproduce
[test]
extras into a virtualenv.docutils
from source and checkout the relevant commit (d6e9dd44523c00ae175442b795f68542653a7b78).docutils
package (e.g.pip install ../docutils/docutils
, or similar).pytest tests/test_writers/test_writer_latex.py
Environment Information
Sphinx extensions
Additional context
N/A
Edit: clarification about the relationship between
roman
and the implementation that exists/existed indocutils.roman
.