Problem with multibyte characters

[MHellmund]

I am testing a quarto project of 'book' type with the new julia engine. Three times in about 50 pdf output pages it happened that a german umlaut ä,ü gets replaced by '��' (two copies of the unicode replacement character U+fffd).

How to reproduce: Run quarto render test3.qmd --to html with the attached file.

• It doesn't happen with jupter/IJulia.
• It doesn't happen with QuartoNotebookRunner.run!(server, "test3.qmd"; output = "test3.ipynb")
• Running quarto with keep-md:true, the '��' are already in the test3.html.md file which is fed to pandoc.

My guess is that somewhere a string/buffer is cut in the middle of an utf8 multibyte character and then both parts are themselves interpreted as utf8 strings. So the invalid utf8 bytes get replaced by U+fffd.

test3.qmd.txt test3.html.md.txt