I don't know how it breaks Cisco AnyConnect, all I know is that reverting this back to webkit2gtk-2.28.0-8 brings back the correct behavior. If this proceeds to stable I'll have to open up a bug on it here.
Ok, how do I found out what changed between webkit2gtk-2.28.0-8 and webkit2gtk-2.28.0-9? I'm willing to put in the work to put in a proper bug request to the appropriate place(s), but allowing this to go to stable will mean everyone that uses Cisco AnyConnect will be broken - surely that is not a desired state?
In a bug report, you could describe more precisely what issue you are seeing. (I'm uncertain how a web engine could possibly be related to VPN connectivity, unless you are using some sort of custom VPN app instead of the system AnyConnect NetworkManager integration.) But presumably there is some problem with the WPE renderer. Accelerated compositing in -8 is severely broken, so we have to update away from that ASAP one way or another.
I submitted this bug here https://bugzilla.redhat.com/show_bug.cgi?id=1823421. I hope there is enough detail here.
I was using the NetworkManager integration up until last year, when our VPN provider switched to requiring an MFA token in addition to a username/password to authenticate, which I never could get working (I did submit bugs upstream for that, AFAIK they are still working on it). The Cisco AnyConnect application works by spawning a WebKit instance through which it asks you for these 3 pieces of information, and then closes the WebKit instance passing the information back to the application to authenticate and set up the VPN. What is happening is the WebKit instance gets created, briefly, (you see an empty white window for less than a second), and then the window disappears and the app complains your credentials are invalid. In the log you see the WebKit instance being spawned, and then it immediately segfaults. Could be renderer, as you do see the window before any content is drawn, but you never get to see the window with the content actually drawn complete.
This update has been submitted for testing by catanzaro.
This update's test gating status has been changed to 'waiting'.
This update's test gating status has been changed to 'ignored'.
This update has been pushed to testing.
Epiphany works well with this
This update can be pushed to stable now if the maintainer wishes
This update has been submitted for stable by catanzaro.
This breaks Cisco AnyConnect!
Please report a bug (ideally upstream on WebKit Bugzilla) for Cisco AnyConnect.
I don't know how it breaks Cisco AnyConnect, all I know is that reverting this back to webkit2gtk-2.28.0-8 brings back the correct behavior. If this proceeds to stable I'll have to open up a bug on it here.
If you want it fixed, you definitely need to report a bug.
Ok, how do I found out what changed between webkit2gtk-2.28.0-8 and webkit2gtk-2.28.0-9? I'm willing to put in the work to put in a proper bug request to the appropriate place(s), but allowing this to go to stable will mean everyone that uses Cisco AnyConnect will be broken - surely that is not a desired state?
You can see here: https://src.fedoraproject.org/rpms/webkit2gtk3/tree/f32
In a bug report, you could describe more precisely what issue you are seeing. (I'm uncertain how a web engine could possibly be related to VPN connectivity, unless you are using some sort of custom VPN app instead of the system AnyConnect NetworkManager integration.) But presumably there is some problem with the WPE renderer. Accelerated compositing in -8 is severely broken, so we have to update away from that ASAP one way or another.
I submitted this bug here https://bugzilla.redhat.com/show_bug.cgi?id=1823421. I hope there is enough detail here. I was using the NetworkManager integration up until last year, when our VPN provider switched to requiring an MFA token in addition to a username/password to authenticate, which I never could get working (I did submit bugs upstream for that, AFAIK they are still working on it). The Cisco AnyConnect application works by spawning a WebKit instance through which it asks you for these 3 pieces of information, and then closes the WebKit instance passing the information back to the application to authenticate and set up the VPN. What is happening is the WebKit instance gets created, briefly, (you see an empty white window for less than a second), and then the window disappears and the app complains your credentials are invalid. In the log you see the WebKit instance being spawned, and then it immediately segfaults. Could be renderer, as you do see the window before any content is drawn, but you never get to see the window with the content actually drawn complete.
This update has been unpushed.