[PATCH] Fix a race condition with kickstarts (#868834)

[Jesse Keating]

We were simluating the continue click before consuming all the messages
in the queue, which would cause the next hub to be in use, but we were
attempting to process messages for the previous hub.  It would cause a
crash of anaconda, but not a backtrace that could be caught with our
autoreporter.  The solution is to only continue when the queue is empty.

This also makes use of a property to see if we are ready to continue
instead of duplicating code.
---
 pyanaconda/ui/gui/hubs/__init__.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/pyanaconda/ui/gui/hubs/__init__.py b/pyanaconda/ui/gui/hubs/__init__.py
index 13f6192..3a88cc3 100644
--- a/pyanaconda/ui/gui/hubs/__init__.py
+++ b/pyanaconda/ui/gui/hubs/__init__.py
@@ -278,11 +278,12 @@ class Hub(GUIObject, common.Hub):
                     if not args[1]:
                         spoke.execute()

-                    if len(self._incompleteSpokes) == 0 and len(self._notReadySpokes) == 0:
+                    if self.continuePossible:
                         if self._inSpoke:
                             self._autoContinue = True
                         else:
-                            self.continueButton.emit("clicked")
+                            if q.empty():
+                                self.continueButton.emit("clicked")
             elif code == communication.HUB_CODE_MESSAGE:
                 spoke.selector.set_property("status", args[1])

[Vratislav Podzimek]

On Thu, 2012-10-25 at 20:40 -0700, Jesse Keating wrote:
> We were simluating the continue click before consuming all the messages
> in the queue, which would cause the next hub to be in use, but we were
> attempting to process messages for the previous hub.  It would cause a
> crash of anaconda, but not a backtrace that could be caught with our
> autoreporter.  The solution is to only continue when the queue is empty.

It is not a crash, it just hits the place in
pyanaconda/ui/gui/__init__.py:GraphicalUserInterface._on_continue_clicked which calls sys.exit(0). See my patch from Tuesday which I think should also be applied together with this one. However this patch looks to me as a nice solution for an awkward problem. Just please see one comment below.
> This also makes use of a property to see if we are ready to continue
> instead of duplicating code.
>
> pyanaconda/ui/gui/hubs/__init__.py | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/pyanaconda/ui/gui/hubs/__init__.py b/pyanaconda/ui/gui/hubs/__init__.py
> index 13f6192..3a88cc3 100644
> --- a/pyanaconda/ui/gui/hubs/__init__.py
> +++ b/pyanaconda/ui/gui/hubs/__init__.py
> @@ -278,11 +278,12 @@ class Hub(GUIObject, common.Hub):
>                      if not args[1]:
>                          spoke.execute()
>                if len(self._incompleteSpokes) == 0 and len(self._notReadySpokes) == 0:
>                if self.continuePossible:
>                    if self._inSpoke:
>                        self._autoContinue = True
>                    else:
>                        self.continueButton.emit("clicked")
>                        if q.empty():
>                            self.continueButton.emit("clicked")

The else-if could be replaced with:

elif q.empty():
    self.continueButton.emit("clicked")

[Jesse Keating]

On 10/26/2012 01:25 AM, Vratislav Podzimek wrote:
> >                        if q.empty():
> > >                            self.continueButton.emit("clicked")
>
> The else-if could be replaced with:
>
> elif q.empty():
>      self.continueButton.emit("clicked")

Fixed that and pushed.  Thanks!