I have been having trouble with https://bugzilla.redhat.com/show_bug.cgi?id=1167103 and decided to have a look at fixing it. Basically, if you let dracut DHCP to pick up your kickstart file, and in your ks file you set up a static network config and use --activate, _and_ you attempt to do a VNC install then: 1) The VNC window opens 2) The system IP address changes. 3) You're stuck with a useless grey VNC window connected to an IP which doesn't exist. Fedora 20 just left the old IP in place as well as configuring the new one. This worked, though the wrong hostname appeared on the VNC window. F21 doesn't keep the old IP around. This all happens because setupDisplay() is called before initializeNetwork(). So I moved the network setup up above the display setup, and it pretty much works. However, there are two caveats: When the interface runs, hostname resolution isn't yet completely configured. (The DNS server is defined, I assume from DHCP, but no domain search list is setup.) That appears to get done when the network spoke calls write_network_config. Since the spokes run concurrently and the network setup appears to take the longest, SourceSpoke runs first, and cannot resolve the unqualified hostname I use for my local fedora mirror. I can fix that by using a fully qualified hostname, but that's still a bug to run down. If you just click the installation source button and immediately hit done, things work. I snuck in a progressively lengthening sleep in _getTreeInfo where it's calling urlgrab and this helps (though it takes nearly ten seconds of sleep for me). If I fix my repo hostname so the interface doesn't stall, the interface often either doesn't initialize (I just get a black VNC window) or I get an interface where everything is squished into a small space. It's as if the installer doesn't wait for the display to get properly setup and the interface to appear before continuing. It's OK if I add a short sleep in the code before calling setupDisplay, but obviously that's a hack. I've no idea what's going on here. Is it possible to make SourceSpoke wait for the network spoke? I could see a solution where if the repo setup fails, it waits for the network spoke to finish and then retries. Any hints about where I should look next to dig into this further would be appreciated. - J< _______________________________________________ Anaconda-devel-list mailing list Anaconda-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/anaconda-devel-list
