[mythtv] [mythtv-commits] Ticket #8512: Mythwelcome spawned frontend crashes when using internal web browser

[Michael T. Dean]

On 06/05/2010 03:41 AM, Mike Parker wrote:
> I can confirm that removing the OpenJDK plugin resolves (or works around) the issue - webpages load successfully in both a MythBrowser session from a standalone frontend and from a frontend spawned by MythWelcome.
>
> Having replaced OpenJDK with the Sun/Oracle JRE, I no longer have a problem.
>
> However, I can't help but think this is a workaround for a genuine MythWelcome/MythFrontend/MythBrowser issue. There are definite differences in the two MythBrowser sessions - for instance, the loaded URL is displayed in the "progress bar" at the foot of the screen in a "standalone" MythBrowser session whilst it's not in a spawned browser session. I suspect that differing behaviour wrt. the OpenJDK plugin is another such difference.
>
> *Something* is causing the two frontends (standalone and spawned) to differ when, AFAIK, they should be identical invocations of /usr/bin/mythfrontend (or similar). It's probably beyond my somewhat megre knowledge to determine what this difference is, but I'm more than willing to help someone who can.
>
> Thanks for the OpenJDK pointer,
>    

Mike,

Thanks for reporting back.  I'm going to go ahead and close the ticket.  
Since Otto said he was getting the crash with OpenJDK's IcedTeaPlugin 
without using mythwelcome, mythwelcome doesn't seem to be a factor 
here.  I do understand your concern, but at this point, it's very much 
more likely that the difference in behavior between 
mythfrontend/mythbrowser run by mythwelcome and run at the command line 
is due to some other difference in your configuration.  For example, 
with a different HOME specified, it's possible that only one of the 2 
instances actually uses the IcedTeaPlugin (if the plugin is installed in 
the HOME).  Or it could be some other difference in the environments.

Futhermore, there's not really anything MythTV does to use plugins--we 
simply tell Qt's WebKit to enable them and then everything else about 
them is Qt/WebKit.  So, I really think if there's a bug, it's a Qt bug 
or an OpenJDK bug.  I'm currently leaning toward it being a Qt 
bug--simply because Adobe Flash 10.1 beta for GNU/Linux x86_64 (possibly 
also 32-bit) also crashes WebKit.  That said, it's possible it's fixed 
in a newer version of Qt (perhaps in the Qt 4.7 snapshot).

I'll also try to find some way that we can log a message indicating what 
plugins are enabled in webkit to help users diagnose issues like this.

Thanks,
Mike