Re: Java: fixes for -fno-assume-compiled

[Andrew Haley <aph at redhat dot com>]

Jeff Sturm writes:
 > On Sun, 15 Dec 2002, Andrew Haley wrote:
 > > Could you explain this next hunk a bit more?  It seems to defeat an
 > > optimization.  Perhaps the optimization didn't work, or didn't help?
 > 
 > Didn't work.  Under some circumstances (inner classes), class constants
 > were getting compiled into the wrong class's cpool.  I suspect
 > current_constant_pool_data_ref needed to be cleared somewhere where
 > current_class changes, but I didn't look, since...
 > 
 > >  > -      TYPE_CPOOL_DATA_REF (current_class) = current_constant_pool_data_ref
 > >  > +      TYPE_CPOOL_DATA_REF (current_class) = cpool_data_ref
 > >  >  	= build1 (ADDR_EXPR, ptr_type_node, decl);
 > 
 > ...it appears TYPE_CPOOL_DATA_REF also saves this value.  If it is already
 > saved in a per-class structure, why bother with a global variable?

I see.  I guess we should delete the global variable altogether.

The patch is OK.

Andrew.