This is the mail archive of the mailing list for the Cygwin project. See the Cygwin home page for more information.
Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: jni,egcs1.1.2 compiling problem

John McNally <> writes:
> The problem I was having seems to have been solved by using 
> the -jni option on javah.  Sorry, I don't know much about what I'm
> trying to do, so I tend to make a lot of mistakes.

I love problems that solve themselves.

> What is your opinion on the likelihood of success on trying to call a
> FORTRAN subroutine thru jni?
> I know Sun only supports C and C++, but was hoping the integration of
> g77 as a front end to gcc might make it possible.  I am currently
> getting a java.lang.UnSatisfiedLinkError when I call the method.
> Unless I've made some other stupid mistake, I am guessing the symbol
> names in the DLL file are not compatible.
> Do you think I might be able to do something with the dllinit.c (or some
> other foo.c file) as a go-between the java and FORTRAN?
> Or is this out of the league of a neophyte programmer, not prepared to
> do some compiler programming?

One issue here is that you need to compile the F77 functions and
subroutines to use "pascal" calling convention as opposed to the usual
"cdecl" calling conventions (that's what JNICALL expands to). Since
you can't use attributes in f77 code currently, you can just compile
the whole file with -mrtd option to g77 and you should be all set.

The other issue is that since you're not using __declspec(dllexport)
(which is JNICALL expands to), you'll need to manually export the
f77 function/subroutine yourself or use --export-all --add-stdcall-alias
option to dllwrap.

If you send me a simple f77 JNI example, I'll set it up and add it to
the jni-examples package.


Want to unsubscribe from this list?
Send a message to