Re: GCJ 3.2 for Win32: Updated Snapshot

[Andrew Haley <aph at redhat dot com>]

Jeff Sturm writes:
 > On Fri, 13 Dec 2002, Andrew Haley wrote:
 > > Jeff Sturm writes:
 > >  > On 12 Dec 2002, Tom Tromey wrote:
 > >  > > Jeff> That is what -fno-assume-compiled was for, as I understand it.
 > >  > > Jeff> However it doesn't work at this time.
 > >  > >
 > >  > > I think it never worked.
 > >  >
 > >  > I looked at the code.  I think it can be useful; see the example below.
 > >  > Would you be interested in patches to make it work?
 > >
 > > I have some patches to make -fno-assume-compiled work.  I certainly
 > > could send them along if there's interest.
 > 
 > As do I.  I've rebuilt most of libgcj and several applications with
 > -fno-assume-compiled for testing.
 > 
 > One sticky problem is the unwind info.  The Utf8Const+1 addressing gcj
 > uses to store its catch_type doesn't work well with PIC and the
 > PE_indirect encoding.  Did you find a solution?

No, I didn't.  Most of the work was just finding the places in libgcj
that didn't exepct a constant pool index and fixing them.

It sound like maybe you got further than I did.

Andrew.