[master/f20-branch][PATCH] Handle packaging exits in a better way

Vratislav Podzimek vpodzime at redhat.com
Wed Oct 16 08:17:29 UTC 2013


On Tue, 2013-10-15 at 13:58 -0700, Brian C. Lane wrote:
> On Tue, Oct 15, 2013 at 12:56:02PM +0200, Vratislav Podzimek wrote:
> > For various reasons we shouldn't call sys.exit() from a non-main thread. But
> > since we need some packaging threads to stop right in the middle and deep in the
> > stack, the only way that works the same as sys.exit() is raising an exception.
> > A special exception that is caught and "translated" into a sys.exit() call from
> > the main thread.
> > 
> > This way, packaging doesn't need to have a clue about our queues and exit
> > handling.
> 
> I really like this, have you done some error simulating to test it,
> especially with tui mode where we were having problems with the error
> handler working right?
I've tested it and it seems there is no difference with or without this
patch. That doesn't mean everything works as it should(*), but that's a
matter of a different patch (set).

(*) If software selection is checked and fails when leaving TUI Software
spoke, the quit/continue question dialog is mangled with the hub prompt
and if user wants to continue, the hub never gets the following input
and the installation hangs. However, the only case this could happen is
if a nonexisting package was listed in comps as part of some
environment.

-- 
Vratislav Podzimek

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



More information about the anaconda-patches mailing list