Description of problem: I'm not sure if this should be assigned to firstboot or systemd... During graphical firstboot, after enabling NTP and synchronizing the clock, a new X server w/ GDM starts and becomes the active console. You have to know that firstboot is not done, and manually switch console back with ctl-alt-F1 in order to see the hardware profile screen and complete the firstboot session. Version-Release number of selected component (if applicable): firstboot-1.117-2.fc15.x86_64 systemd-22-1.fc15.x86_64 How reproducible: Not sure Steps to Reproduce: 1. boot using the pxeboot kernel and initrd.img installer 2. install a default graphics workstation, using the network 3. reboot into firstboot 4. advance and click to enable NTP 5. advance and it shows a modal dialog about synchronizing the clock Actual results: firstboot disappears before completing, as described above. Expected results: firstboot continues, and user completes tasks before GDM becomes visible Additional info: This was from a fresh network install today, using pxeboot installer via GRUB (on a laptop without a CD drive). I also enabled the "encrypt system" option. I don't know if that affects the firstboot sequence in any significant way.
Yes, I found about this myself yesterday, so I did already some investigations. The gdm is run during the execution of code in scdMainWindow.py apply() method from system-config-date. I don't know which part exactly causes systemd to run gdm, even though firstboot is not finished yet.
Obviously s-c-date doesn't do anything with gdm, I imagine that gdm is started as a side-effect of starting or stopping ntpd in the apply() method. Here are the relevant parts, I've added comments describing the resulting /sbin/service or /sbin/chkconfig calls: class scdMainWindow(object): [...] def apply (self): [...] if ntpEnabled == False: # We're not using ntp, so stop the service self.dateBackend.stopNtpService () # --> /sbin/service ntpd stop [...] self.dateBackend.chkconfigOff () # --> /sbin/chkconfig --level 2345 ntpd off elif ntpEnabled == True: # We want to use NTP [...] pid = os.fork () if pid == 0: [...] retval = self.dateBackend.startNtpService (None) # --> /sbin/service ntpd (re)start [...] os._exit (0) [...] self.dateBackend.chkconfigOn () # --> /sbin/chkconfig --level 2345 ntpd on [...] I'll change the component to ntp for further examination and add Lennart to Cc. Lennart, do you have an idea how starting/stopping ntpd this way can cause gdm to start prematurely?
It's a bug in systemd. Nothing wrong in firstboot. *** This bug has been marked as a duplicate of bug 694079 ***