Re: two bugs and a buggy patch

[Per Bothner <per at bothner dot com>]

Marco Vezzoli <marco.vezzoli@st.com> writes:

> This patch fix only class creation... the idea is to mangle the names
> when accessing instances too.

Please remember to send patches in diff -u or diff -c format.
The plain diffs are not reliable.

I would like to make it so that record-type-name also works for
non-Java names.  In the short run, I think it makes sense to
write an "unmangle" procedure.  However, mangling is not
guaranteed to be invertible, so you may not get back exactly the
same type and field names.  It is possible and desirable to get
the exact names, but that may not be worth it for the record feature,
which should be deprecated anyway.

You do currently have the alternative of using the module feature
for defining new class types, though that may be a bit cumbersome.
(You have to put each class in a separate .scm file.)

What do you think about the SRFI-9 Record Type feature
(http://srfi.schemers.org/srfi-9/srfi-9.html)?  I would like
Kawa to implement that.

Or perhaps getting define-class working is more important?

> Another problem with quasiquoting.
> 
> #|kawa:1|# ,a
> gnu.mapping.UnboundSymbol: Unbound symbol unquote

That is the correct behavior, I believe.  Perhaps you're
complaining that error message is not that information?
Remember that ,a is syntactic sugar for (unquote a),
according to R5RS.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/~per/