In today's Fedora Rawhide compose (Fedora-Rawhide-20230512.n.0), anaconda's help doesn't work. Clicking a Help button in anaconda doesn't seem to do anything. Reproducible: Always Steps to Reproduce: 1. Download https://kojipkgs.fedoraproject.org/compose/rawhide/Fedora-Rawhide-20230512.n.0/compose/Everything/x86_64/iso/Fedora-Everything-netinst-x86_64-Rawhide-20230512.n.0.iso 2. Boot it 3. Hit the Help button Actual Results: Nothing appears Expected Results: The help should appear (in yelp) In the journal, this error appears when the help button is clicked: at-spi-bus-laun[pid]: g_dbus_method_invocation_return_dbus_error: assertion 'error_mesage != NULL' failed and at-spi2-core changed in the affected compose (from at-spi2-core-2.48.0-1.fc39 to at-spi2-core-2.48.1-1.fc39), so I'm filing against at-spi2-core initially. anaconda and yelp did not change in the affected compose. CCing the anaconda devs for info. Proposing as an F39 Final blocker per "Any element in the installer interface(s) which is clearly intended to display 'help' text must do so correctly when activated." - https://fedoraproject.org/wiki/Fedora_39_Final_Release_Criteria#Installer_help
Oh, this is only broken in traditional installer images - netinst, server DVD, silverblue installer. It works OK on live images (KDE and GNOME).
Whoops - turns out I just didn't wait long enough. yelp does *eventually* launch, after 56 seconds.
*** Bug 2203446 has been marked as a duplicate of this bug. ***
So, I think the blivet-gui issue is probably the same thing, and there's (at least) a third affected case: nm-connection-editor (which can be launched from the Network & Host Name spoke. All three do eventually run, but take much longer to start than before. So for now I'm gonna figure these are all the same case, and it's *maybe* caused by at-spi2-core. I'll see if I can spin up a Rawhide image with at-spi2-core downgraded and confirm if that's the cause.
Ah - this may well be https://gitlab.gnome.org/GNOME/at-spi2-core/-/issues/115 , which should be fixed in 2.48.2. We've built 2.48.2 already but it didn't make the compose. I'll grab the image openQA built with that at-spi2-core in it and see if this is fixed.
Hah, yes, that *does* fix it. Sorry for the noise. So this'll be fine in tomorrow's compose.