RFR(xs): 8152440: Zero build fails after JDK-8146801

[Severin Gehwolf]

Hi Aleksey,

Thanks for the quick review!

On Tue, 2016-03-22 at 19:49 +0300, Aleksey Shipilev wrote:
> On 03/22/2016 07:25 PM, Severin Gehwolf wrote:
> > 
> > The enhancement of JDK-8146801 introduced a new
> > flag, InitArrayShortSize, which seems a C1/C2 only flag. It does not
> > make much sense to materialize this flag for, say, the Zero
> > interpreter. This causes a Zero build failure because an attempt is
> > made to materialize the macro pd_InitArrayShortSize without declaring
> > it via DECLARE_PD_DEVELOPER_FLAG. DECLARE_PD_DEVELOPER_FLAG is only
> > been declared conditionally which never gets included for a Zero build.
> I see at least ShareVtableStubs is defined as develop_pd in globals.hpp,
> and it is set with define_pd_global in globals_zero.hpp. Why would the
> same fail for InitArrayShortSize, if you add this to globals_zero.hpp?
> 
>  define_pd_global(intx,  InitArrayShortSize,     -1); // not used

Updated webrev does this:
http://cr.openjdk.java.net/~sgehwolf/webrevs/JDK-8152440/webrev.02/

More thoughts?

Cheers,
Severin

> > 
> > I'm proposing to move the macro expansion of developer_pd() for
> > InitArrayShortSize to c1_globals.hpp instead. Thoughts?
> It seems obscure to move C1/C2 flag to c1_globals.hpp. Therefore, I'd
> like us to look for a better fix.
> 
> 
> Thanks,
> -Aleksey
>