Re: PR java/25429: can't see private static final int CREATE = 1

[Andrew Haley <aph at redhat dot com>]

Tom Tromey writes:
 > >>>>> "Andrew" == Andrew Haley <aph@redhat.com> writes:
 > 
 > I was just writing a comment in the RH PR ...
 > 
 > Andrew> +             if (nested_member_access_p (current_class, decl)
 > Andrew> +                 && ! (JDECL_P (decl) && CLASS_FINAL_VARIABLE_P (decl)
 > Andrew> +                       && DECL_INITIAL (decl) != NULL_TREE))
 > 
 > I'm not sure about this patch.  It seems to me that the test has to be
 > for whether the field's value is inlineable, i.e., the DECL_INITIAL is
 > a constant expression.

Well, it's the exact same test that is used to do the same thing
elsewhere in the compiler.  

 > For instance, won't this skip the accessor if we see:
 > 
 >     private static final Object FIELD = new Object();
 > 
 > ... but that would be wrong.

No, because when we see more complex initialization we don't set the
DECL_INITIAL.  Instead, we generate a synthetic method that contains
all the initializations.  We only set DECL_INITIAL if the
initialization requires no code to be executed at runtime.

Andrew.