firefox 4: problem with NPN_GetURL target=_self

[Deepak Bhole]

* Andriy Gapon <avg at icyb.net.ua> [2011-04-19 13:35]:
> on 19/04/2011 20:20 Deepak Bhole said the following:
> > 
> > That'd be great, thanks!
> 
> OK, just got it - geturl has to be initiated from the applet itself.
> Here is the updated example:
> http://people.freebsd.org/~avg/java_test/Test.html
> (http://people.freebsd.org/~avg/java_test/ is browsable)
> 
> Please compare behavior of:
> 1. pressing the html button
> 2. pressing first the html button then navigating back and then the applet button
> 3. pressing only the applet button (e.g. after force reloading the page)
> 

There appears to be an odd race condition happening in Firefox code. If
I do #3 normally, it freezes. If I open the browser via gdb and try it
though, it works fine. Anyway...

Though a bug in Firefox, this is a shortcoming in the plugin as well.
Unfortunately it is not one that can be easily mitigated with the
current plugin design. 

The LiveConnect engine (that handles JS <-> Java communications) can
handle recursive calls. But the part that handles other (non
LiveConnect) requests (geturl, setstatus, etc.) is not designed to
handle recursions as there should never be a need for it.

The GetJavaObject request being made by NPN_GetURL is rather odd as the
latter is a non-liveconnect functionality, yet it is making a
liveconnect related call (GetJavaObject). I will look into recursion
handling, but it is unlikely to happen soon as it will be a fair bit of
change. In the mean time, I believe it would be worthwhile to pursue
this on Mozilla side as well, since it's code is definitely doing
something odd.

Cheers,
Deepak

> Thanks!
> -- 
> Andriy Gapon