0> In article <4A1063F8.8080209@free.fr>, 0> frmas <URL:mailto:frmas@free.fr> ("Frmas") wrote: Frmas> <?xml version='1.0' standalone='no'?> Okay, so the XML is UTF-8 (since no coding is specified).
/-------- | sed -n -e '148424{' -e 'p' -e 'q' -e '}' cartes/paris.osm | od -t x1c \--------
to view the problematic part.
Frmas> 0000120 6a 6f 73 e9 2c 6a 6f 73 c3 a9 6d 2c 45 73 70 Frmas> j o s 351 , j o s 303 251 m , E s p That's the problem: E9 2C isn't a valid UTF-8 sequence - it looks like that part has been encoded as Latin-1. The later sequence C3 A9 is okay: a valid UTF-8 2-byte sequence. So mkgmap is right to reject the file; now the question is, how did that malformed sequence get in there in the first place? Could your script have written it, or has it come from (say) the planet.osm file (or a subset cut from it)?