Proposed Target Macro for Java Front-end: MODIFY_JNI_METHOD_CALL

[Andrew Haley]

Ranjit Mathew writes:
 > Hi,
 > 
 >      I would like to propose a new target macro 'MODIFY_JNI_METHOD_CALL'
 > to allow targets like Cygwin and MinGW to modify the call generated
 > by the GCC Java front-end to invoke a JNI (Java Native Interface)
 > method.
 > 
 > For background information on the need for such a macro, please
 > see the thread starting from this message:
 > 
 >      http://gcc.gnu.org/ml/gcc/2002-11/msg00734.html
 > 
 > The following proposed patch to "gcc/doc/tm.texi" explains this
 > macro in more detail:
 > ------------------------------- 8< -----------------------------------
 > --- gcc/doc/tm.texi	2002-12-05 05:00:30.000000000 +0530
 > +++ gcc/doc/tm.texi	2002-12-05 07:03:16.000000000 +0530
 > @@ -9141,4 +9141,20 @@
 >   lists.
 > 
 > +@findex MODIFY_JNI_METHOD_CALL
 > +@item MODIFY_JNI_METHOD_CALL (@var{mdecl})
 > +Define this macro to a C expression representing a variant of
 > +the method call @var{mdecl}, if Java Native Interface (JNI) methods
 > +must be invoked differently than

differently from

 > other methods on your target
 > +machine. For example, on 32-bit Windows, JNI methods must be invoked
 > +using the @code{stdcall} calling convention and this macro
 > +could then be defined as this expression:

I suggest "is defined as"

 > +
 > +@smallexample
 > +build_type_attribute_variant (@var{mdecl},
 > +                              build_tree_list
 > +                              (get_identifier ("stdcall"),
 > +                               NULL))
 > +@end smallexample
 > +
 >   @end table
 > ------------------------------- 8< -----------------------------------
 > 
 > The following proposed patch would make the Java front-end use this
 > macro:
 > ------------------------------- 8< -----------------------------------
 > --- gcc/java/expr.c	2002-12-05 07:12:31.000000000 +0530
 > +++ gcc/java/expr.c	2002-12-05 07:15:34.000000000 +0530
 > @@ -2266,7 +2266,11 @@
 >   		 tree_cons (NULL_TREE, build_utf8_ref (tem), lookup_arg));
 > 
 > -  jni_func_type
 > -    = build_pointer_type (build_function_type (TREE_TYPE (TREE_TYPE 
 > (method)),
 > -					       arg_types));
 > +  tem = build_function_type (TREE_TYPE (TREE_TYPE (method)), arg_types);
 > +
 > +  #ifdef MODIFY_JNI_METHOD_CALL
 > +  tem = MODIFY_JNI_METHOD_CALL (tem);
 > +  #endif
 > +
 > +  jni_func_type = build_pointer_type (tem);
 > 
 >     jnifunc = build (COND_EXPR, ptr_type_node,
 > ------------------------------- 8< -----------------------------------
 > 
 > Finally, the following proposed patch will make the Cygwin/MinGW targets
 > define this macro appropriately (mingw32.h #include-s cygwin.h):
 > ------------------------------- 8< -----------------------------------
 > --- gcc/config/i386/cygwin.h	2002-12-05 07:19:36.000000000 +0530
 > +++ gcc/config/i386/cygwin.h	2002-12-05 07:23:10.000000000 +0530
 > @@ -398,4 +398,12 @@
 >        }
 > 
 > +/* Java Native Interface (JNI) methods on Win32 are invoked using
 > +   the stdcall calling convention */
 > +#undef MODIFY_JNI_METHOD_CALL
 > +#define MODIFY_JNI_METHOD_CALL(MDECL) \
 > +  build_type_attribute_variant ((MDECL), build_tree_list \
 > +                                         (get_identifier ("stdcall"), \
 > +                                          NULL))
 > +
 >   /* External function declarations.  */
 > ------------------------------- 8< -----------------------------------

The Java parts of this are OK, but you'll need to get approval from
someone with appropriate priviliges for the i386 and genereric part.

Andrew.