Bug 1950053

Created attachment 1772222 [details]
screenshot

This looks like yet another case of GTK (which is not thread safe) being broken by access from a non-main thread.

The Connect to Red Hat screen is initialized in a thread as it needs to wait for some other threads to finish:



    threadMgr.add(AnacondaThread(name=THREAD_SUBSCRIPTION_SPOKE_INIT,
                                 target=self._initialize))


    def _initialize(self):
        # wait for subscription thread to finish (if any)
        threadMgr.wait(THREAD_SUBSCRIPTION)
        # also wait for the payload thread, which migh still be processing
        # a CDROM source, to avoid the Subscription being mandatory by mistake
        # due to CDN still being default at the time of evaulation
        threadMgr.wait(THREAD_PAYLOAD)

        # update overall state
        # - the update methods manipulate GTK and this method is called in a non-main thread,
        #   so make sure the calls are serialized to the main thread by using the appropriate
        #   wrapper functions
        self._update_registration_state()
        self._update_subscription_state()

        # Send ready signal to main event loop
        hubQ.send_ready(self.__class__.__name__, False)

        # report that we are done
        self.initialize_done()


Once the two threads it waits on terminate, it calls the _update_registration_state() and _update_subscription_state() methods, notifies the hub and then terminates as well. Calling the two _update* methods is most likely where things go wrong, as they update methods manipulate GTK widgets while being called from a non-main thread, which can result in GTK hevaing in undefined ways. Quite likely including the spoke status not being set correctly.

This behavior would be also consistent with the spoke status being fixed by visiting a different spoke and returning to the hub - the _initialize() method tha "breaks GTK" is called ony once during Anacoonda startup and return from a spole triggers a refresh that likely puts the broken GTK back in order.

To fix the issue we need to make sure all the calls in the Coneect to Red Hat spoke _initialize() method that could interact with GTK are properly synchronized into the main thread. Thankfully, we have a set of wrapper we can use for this, making the change quite easy.

I've opened a (draft) PR thart wrappes the problematic calls:
https://github.com/rhinstaller/anaconda/pull/3293

Would it be possible to inject it to the testing environment to check if it fixes the issue ?

RHEL-9 clone: bug 1959383.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory (anaconda bug fix and enhancement update), and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHBA-2021:4254