I think this is problem one can see using Cyrillic locales; also German, French, Turkish etc where letters can have diacritics - umlauts etc.
In my case, replacement characters appeared in panels when I compiled PDF with knitr
and tried to find an unsearchable thing.
I have a cp-1251
character encoding by default.
If you want to see this problem by yourself, you can try to compile an example.
\documentclass [russian]{article}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{babel}
\begin{document}
\title{minimal knittr example in R}
\author {John Doe}
\maketitle
Давайте писать по-русски
<<>>=
set.seed(99)
k = 1 * rnorm(200,3,2)
mean(k); var(k)
@
the first two of x are \Sexpr{k[1:2]}
\end{document}
For the demonstration of this problem, see pic:
How to fix it?
I tried a lot of things, but a solution was easier, than i could imagine.
Sys.setenv("LANGUAGE"="EN")
and you can see: