Almost there...

Tom Tromey
Mon Jul 22 20:06:00 GMT 2002

>>>>> "Bryce" == Bryce McKinlay <> writes:

Bryce> Ideally, GCJ should default to the same encoding that javac
Bryce> does on any given system. IIRC the reason why it doesnt is that
Bryce> the system's iconv implementation, used by GCJ, sometimes
Bryce> doesn't support the conversions we need.

There's another issue, too.  On Linux, the `C' locale is very strict
about checking the ASCII encoding.  So by default (assuming no locale
setting) a file with an umlaut is going to give an error with gcj.
However, on the same platform javac will let the file through, since
the JDK appears to use ISO-8859-1 as the default encoding.

The only way I've thought of to fix this is to special case the C
locale.  I think that's pretty ugly, though, so I've resisted the
temptation.  Instead I just reply to the bug reports.


More information about the Java mailing list