Re: [java/c++/rfc] avoid plt references to cni methods

[Andrew Haley <aph at redhat dot com>]

Richard Henderson writes:
 > The idea is that CNI code will be in the same DSO as its corresponding
 > Java code.  Indeed, after this patch it MUST be in the same DSO.

I think that's a reasonable requirement.  I guess we should document
it.

 > We arrange for the C++ front end to emit hidden aliases for all of
 > the functions that we need, and we use those in the _Jv_Method
 > table.

OK.

 > On the C++ side, I need to wait until we've resolved clones.  Since
 > this is done during cgraph_analyze_function, which makes it difficult
 > to do this except for a separate pass over the generated functions.
 > 
 > I've somewhat arbitrarily chosen to extend the mangling grammar:
 > 
 >   <special_name> := HA <encoding>
 > 
 > where "HA" stands for "hidden alias".  It's not inconcievable that
 > such aliases could be useful in other cases.  In any event, it seemed
 > safer to inject myself into a logical point in the grammar like this
 > than invent my own private prefix/suffix which could just as easily
 > conflict with other symbols.

I guess that's OK, but it seems a little arbitrary.  I wonder if there
ought to be some well-defined langauge-specific way to do special
mangling: we're going to need to extend it again for Java methods that
only differe by return type.

 > On the Java side, I've rearranged things to use the name mangling 
 > langhook.  This reduces the number of places that need to touch
 > DECL_ASSEMBLER_NAME to the builtins and the mangler itself.  This
 > ensures that we don't set DECL_ASSEMBLER_NAME before it's absolutely
 > necessary, which in turn allows us time to call java_mark_class_local.
 > Which will in turn set bits that will in fact *change* the mangling
 > for CNI methods.

I see.

This looks good.

Andrew.