Re: Class.h and Class.java both define zero-arg constructor

[Adam Megacz <gcj at lists dot megacz dot com>]

Tom Tromey <tromey@redhat.com> writes:
> Adam> I'm doing a static build for mingw; perhaps on other platforms
> Adam> the constructor gets inlined or otherwise stripped.

> Yes, it should be inlined.  I think it shouldn't be emitted in
> natClass.o.  Do you know why it is happening?

I don't know... could have something to do with the mingw backend, or
perhaps the fact that I'm doing a "-static" compile (although I've
successfully done that on linux targets). Maybe a combination of the
two.

Is it safe to assume that that method will get inlined?

What does the #pragma implementation "Class.h" in natClass.cc mean?


> Less ideally, we could add a dummy argument to the constructor in
> Class.h.

We may end up having to do that... Where does that constructor get
called from? There's a comment pointing to prims.cc, but I don't see a
"new Class()" anywhere in there.

  - a