[rfc][icedtea-web] Make 'getvalue' nonblocking to reduce hanging chance & improve responsivity

[Pavel Tisnovsky]

Hi Adam,

in overall - this change looks ok.

Have you tried additional tests you mentioned in the original message?
> > I have done testing on the patch and it looks OK, but I will still test
> > it more.

Cheers,
Pavel

----- Adam Domurad <adomurad at redhat.com> wrote:
> On 06/07/2013 02:11 PM, Adam Domurad wrote:
> > Hi all. I am very happy to present this patch!
> >
> > All Firefox hanging bugs had a common theme, Firefox waits for the
> > result of some function to finish from the plugin, and IcedTea-Web
> > plugin-side waits as well for the java-side.
> >
> > This patch should reduce the chance of Firefox outright hanging, by
> > making the most frequent site of these occurences nonblocking. This will
> > also improve responsivity, eg Firefox no longer 'hangs' while waiting
> > for you to accept or deny a confirmation message.
> >
> > The patch delays waiting for initialization of the applet until its
> > absolutely needed - which is only when the applet is communicated with
> > using liveconnect.
> >
> > This especially benefits the majority of applets that do not use
> > liveconnect. Eg the applets at
> > http://www.walter-fendt.de/ph14e/resultant.htm no longer presents a
> > blank page for a while, for me.
> >
> > The patch introduces (with tests) a helper class, NPObjectRef. This was
> > done to make this patch easier to reason about. It should be far easier
> > to know now when to retain/release an object (almost always automatic,
> > only call raw_retain() when returning raw NPObject* to browser).
> >
> > It also introduces a wrapper around g_strdup_printf with a (very) simple
> > test.
> >
> > I have done testing on the patch and it looks OK, but I will still test
> > it more.
> >
> > Cheers,
> > -Adam
> 
> Ping ?
> -Adam