Conflict between fontspec and geometry

Question

Using the option lines=x of geometry does not give the expected result when the fontspec package is used in the same document. For example, declaring lines=40 does not result in 40 lines per page but 29:

\documentclass[a4paper,11pt]{article}
\usepackage{fontspec}
\usepackage{geometry}
\geometry{lines=40}
\usepackage{lipsum,lineno}

\begin{document}
\linenumbers
\lipsum
\end{document}

This hapens with standard classes (book, article), but not with the KOMA classes nor with memoir.

Loading geometry before fontspec and using newgeometry instead of geometry produces the expected result, but I would like to know what the problem is and how to resolve it.

Answer 1

The problem is that \baselineskip is "wrong" (10pt instead of 13.4pt) when geometry calculates the lines as an \DeclareErrorFont from the euenc package changes it.

It is a bit unclear if it is a latex "bug" that \DeclareErrorFont resets the \baselineskip but even without this change geometry can err. E.g. This example

\documentclass[a4paper,11pt]{article}
\usepackage{setspace}
\onehalfspacing
\usepackage{geometry}
\geometry{lines=40}
\usepackage{lipsum,lineno}
\begin{document}
\linenumbers
\lipsum
\end{document}

gives a different layout then this one

\documentclass[a4paper,11pt]{article}
\usepackage{setspace}
\usepackage{geometry}
\geometry{lines=40}
\usepackage{lipsum,lineno}
\onehalfspacing
\begin{document}
\linenumbers
\lipsum
\end{document}

I can imagine other szenarios and imho geometry can't catch all possible cases, so if you want to use the lines option (and heighrounded) you will sometimes have to help it a bit to get the correct/wanted line height. In your cases by issuing \normalsize before \geometry{lines=40}. Or by setting the option with \newgeometry only at the begin of the document.