Ok, I verified that with the MOVIECLIP as_value reverted and sprite_instance::set_member actually looking in the displaylist (the latter was experimentally disabled some time ago) the gravity_embed test works. Anyway, this makes VarAndCharTest testcases fail, as we want variables to take precedence over characters when there is a clash.
--strk; On Thu, Dec 21, 2006 at 08:03:46AM +0100, strk wrote: > On Thu, Dec 21, 2006 at 09:48:59AM +0800, zou lunkai wrote: > > > Maybe only Gnash creates two things: a variable 'container' and a > > character 'container'. To my opinion, that's not quite correct. > > Firstly, > > a chareacter 'container' is created. And then, there is a DefineLocal(0x3c) > > which trys to define an variable also named 'container'. But at last > > the variable > > will not be created, because DefineLocal will find the 'container' already > > exist. "0x3C" will just set the container again. So, there is always > > ONE 'container' > > and maybe no culprit. > > Good point. The only reason for this to fail in Gnash is then the fact > we 'descend' to upper "local variable frames" in getVariable(), as one > of my commits handed the DefineLocal for timeline scopes to NOT > use the set_local but directly the set_variable function (threating > the DefineLocal as it was a SetVariable). > > I guess a testcase for this will help, in particular trying to *first* > declare the local and *then* adding the character seems to be in interesting > one... > > --strk; -- _______________________________________________ Gnash-dev mailing list [email protected] http://lists.gnu.org/mailman/listinfo/gnash-dev
