Re: java.lang.RuntimeException: invoke: no method named `$000a' in class gnu.expr.PrimProcedure

[Per Bothner <per@bothner.com>]

On 2/28/22 15:10, Panicz Maciej Godek via Kawa wrote:
> I have a new problem. When I invoke a function, I get the following error:
> 
> java.lang.RuntimeException: invoke: no method named `$000a' in class
> gnu.expr.PrimProcedure

This is an exception thrown by the 'invoke' procedure:

https://www.gnu.org/software/kawa/Method-operations.html#idm45230715888224

This is a way to explicitly call a Java method. 'invoke' can appear in user-code, and
sometimes the compiler generates call to it.  In this case it looks like the
compiler has translated some form to the equivalent of (invoke OBJECT '$000a ARGS ...).
With appropriate type information about OBJECT and the ARGS, in further steps
the compiler can translate this to an inline JVM invokexxx instruction,
but when it doesn't have the needed type information it will invoke at run-time.
The call to invoke should should up in the --debug-print-final-expr output.
Probably for some reason the method $000a isn't being generated or it
is generated in the wrong class.

> I suppose there could be some bug in the treatment of continuations when
> call/cc appears in the body of a method (but it's only a hypothesis).

Could well be.  Kawa will try to optimze call/cc when invoking and handling the
continuation are in the same method, but because Kawa does a lot of
inlining and other optimizations it could well have gotten confused.

> Anyway, I wonder if Kawa provides a way of performing an early return from
> a method (other than call/cc)?

Well you could use exceptions: primitive-throw with try-catch.  call/cc is
basically compiled into that anyway, except when it can be optimized to a goto.
-- 
	--Per Bothner
per@bothner.com   http://per.bothner.com/