[rhel7/master] Don't call BusyCursor before Gdk is setup (#1078868)

Vratislav Podzimek vpodzime at redhat.com
Fri Sep 19 05:51:42 UTC 2014


On Thu, 2014-09-18 at 13:15 -0700, Brian C. Lane wrote:
> On Thu, Sep 18, 2014 at 10:50:44AM +0200, Martin Kolman wrote:
> > On Wed, 2014-09-17 at 16:06 -0700, Brian C. Lane wrote:
> > > The setup method is too early, the Gdk.get_default_root_window call
> > > fails. So Remove the call to busyCursor, but just in case also silently
> > > return if busyCursor or unbusyCursor doesn't get a window.
> > What about logging it as warning/error ? Not having a window in this
> > case is always an error - either we are calling it too early or GTK is
> > blowing up.
> 
> It actually gets printed to the console, I'm not sure how useful logging
> it would really be -- it isn't an essential operation if it does fail.
> 
> FYI, this was pushed and built into 19.31.91-1 last night and it at
> least solved the immediate issue of not being able to start anaconda.
> 
> I still see this message on the console:
> 
> ** (anaconda: 1090): WARNING **: Could not open X display
> 
> I'm not sure where that comes from.
> 
> While debugging this I also threw in a big delay after metacity started
> to see if it was timing, and it would still fail to get a window. I also
> tried looping to see if repeated calls would eventually work, and they
> did not.
That's what I did too and what get me to a notion that there's something
going really wrong.

> 
> I expected this to be related to some gdk/gtk change, but vpodzime's
> report that 7.0 + updates.img also does it contradicts that. So I really
> have no solid idea what's going on.
If I find enough time, I'll try to bisect the issue.

-- 
Vratislav Podzimek

Anaconda Rider | Red Hat, Inc. | Brno - Czech Republic



More information about the anaconda-patches mailing list