[Moonlight-list] Firefox removing XPCOM for FF 3.6a1

[Chris Toshok]

This thoroughly screws any sort of interop story between C# and JS.  
there's simply no nice way to hook up a C# delegate to an html object on 
the page without access to XPCOM.

hrm..  perhaps there is an incredibly hacky way to do it, involving JS + 
eval.  add a method to the plugin, say, function _internalProxyDomEvent 
(id, args, ...).  when hooking up a delegate to an html object we create 
a proxy object wrapping the delegate and do:

eval ('$dom_element.addEventListener ("$event_name", function (...) { 
$plugin._internalProxyDomEvent ($proxy->GetId(), ...); })')

like I said.  no nice way.

nice going, mozilla.

That said, we can still have 1 plugin.  We'll just have a different 
bridge for ff 3.5 and .6.

Chris

On 10/05/2009 01:29 PM, Brad Taylor wrote:
> Hey,
>
> Recently, I started encountering segfaults in the moon-unit tests.
> Originally, I attributed the crashes to instability caused by the
> mono-2-6 move, but I finally got a chance to find the cause today.
>
> The segfault I was getting was at http://paste2.org/p/448706.  It seems
> that in ff3-dom.cpp:182, Firefox is setting item to nsnull, causing the
> actual segfault later on.
>
> As I've been (epically) working on a Firefox patch myself, I've been
> tracking the latest FF 3.6 branch.  It seems that in FF 3.6a1, they have
> deprecated the XPCOM interface, causing any calls to NPN_GetValue with
> the DOM window to return nsnull.  Instead they recommend replacing the
> code with npruntime API calls.  A discussion board post to
> mozilla.dev.tech.plugins yields more information[1].
>
> Anyway, just thought I'd give you guys a heads up, and brighten your
> days a bit.  I know my day got brighter hearing that we'll have to ship
> two plugins for FF 3.5 and 3.6.  Yay!
>
> Best,
>
> -Brad
>
> --
> [1]
> http://groups.google.com/group/mozilla.dev.tech.plugins/browse_thread/thread/87b5dc9d4383755e
>
>
> _______________________________________________
> Moonlight-list mailing list
> Moonlight-list at lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/moonlight-list
>