Re: [Gnash-commit] [SCM] Gnash branch, master, updated. release_0_8_9_final-1634-g404343a

[Sandro Santilli]

On Fri, Jul 05, 2013 at 01:05:00PM +0200, Benjamin Wolsey wrote:
> > commit 404343a510a88352b465ae37026102cd1df705e0
> > Author: Sandro Santilli <address@hidden>
> > Date:   Wed Jul 3 09:10:04 2013 +0200
> > 
> >     Initialize NetStream prototype only once
> > 
> > diff --git a/libcore/asobj/NetStream_as.cpp b/libcore/asobj/NetStream_as.cpp
> > index b915df9..034e85c 100644
> > --- a/libcore/asobj/NetStream_as.cpp
> > +++ b/libcore/asobj/NetStream_as.cpp
> > @@ -1590,11 +1590,13 @@ netstream_new(const fn_call& fn)
> >          if (isNativeType(toObject(fn.arg(0), getVM(fn)), nc)) {
> >              ns->setNetCon(nc);
> >              if ( nc->isConnected() ) {
> > -                // TODO: initialize only once ?
> > -                // Test it (but keep re-entrancy, in case)
> > -                as_object* proto = obj->get_prototype();
> > -                if ( proto ) {
> > -                    attachPrototypeProperties(*proto);
> > +                static as_object* proto = NULL;
> 
> This is rather a shame; it took a long time to get rid of all the
> statics so that multiple or successive VMs would work. Even some kind of
> isNetstreamInitialized flag in the VM would be better than that.

Hey, I knew it was a step backward. And indeed I felt ashamed,
but getting you to join the mailing list traffic was a good reason,
wasn't it ?  :D

So it looks like we're missing a testcase to ensure the reentrancy
work won't be destroyed by ashamed people like me.
Until then I guess I can revert this one.

--strk;

 ~~ http://strk.keybit.net