Non-Unicode characters before n are OK; other formats are OK.
GAUSS [label="Carl Gaußn1799"];
[quarl] It works fine normally, at least for the latin1 subset, to simply encode it in UTF-8. I worked around the Gaussn problem by adding a space between ss and n -- I didn't mention this in the bug report because I thought you knew Unicode worked most of the time.
From looking at the postscript output it seems like a bug in
parsing n, because everywhere else n is split into multiple PS
commands; here it was encoded as one PS line with "\n".