Summary: | hostnamed should emit PropertyChanged for Hostname property | ||
---|---|---|---|
Product: | systemd | Reporter: | Marius Vollmer <marius.vollmer> |
Component: | general | Assignee: | systemd-bugs |
Status: | RESOLVED FIXED | QA Contact: | systemd-bugs |
Severity: | enhancement | ||
Priority: | medium | CC: | dh.herrmann |
Version: | unspecified | ||
Hardware: | Other | ||
OS: | All | ||
Whiteboard: | |||
i915 platform: | i915 features: |
Description
Marius Vollmer
2014-09-16 13:39:29 UTC
hostnamed sets the kernel hostname itself, so there is little reason in tracking it. We don't support other processes setting the kernel hostname directly (and there is little reason for it), but only via hostnamed APIs. Currently, hostnamed sends dbus notifications for *any* hostname change. That is, if you're interested in a specific hostname, watch those notifications and then either parse them are read the hostname via the gethostname(2) syscall. Btw., I just noticed that we have the code to send notifications for "Hostname" changes, but we lacked a vtable annotation. I fixed this [1], so you should now get notifications for "Hostname" changes, too. [1] http://cgit.freedesktop.org/systemd/systemd/commit/?id=caffbef636ec48958dcb22e4b4140a9889a6a769 (In reply to comment #1) > Currently, hostnamed sends dbus notifications for *any* hostname change. > That is, if you're interested in a specific hostname, watch those > notifications and then either parse them are read the hostname via the > gethostname(2) syscall. I only now noticed that gethostname will not always correspond to the Hostname property, at least since c779a442 (released with v213)... Should hostnamed maybe have one more property EffectiveHostname that is guaranteed to have the same value as gethostname? How can things like the hostname be so complicated? :-) > Btw., I just noticed that we have the code to send notifications for > "Hostname" changes, but we lacked a vtable annotation. I fixed this [1], so > you should now get notifications for "Hostname" changes, too. Great, I am happy to hear that this was just a bug, and not some fundamental limitation, as I had feared. Thanks! (In reply to comment #2) > I only now noticed that gethostname will not always correspond to the > Hostname property, at least since c779a442 (released with v213)... Should > hostnamed maybe have one more property EffectiveHostname that is guaranteed > to have the same value as gethostname? We never expose systemd/bus APIs if there is already an API to retrieve the given value. That is, our recommendation is to *always* use gethostname() if you require the effective hostname of your system. We enforce this policy in quite a lot places, and so far there hasn't been any reason to change this. > How can things like the hostname be so complicated? :-) Yeah.. auto-configuration via DHCP makes stuff awfully complicated. What can we do.. (In reply to comment #3) > Yeah.. auto-configuration via DHCP makes stuff awfully complicated. What can > we do.. Also, NetworkManager seems to have its own ideas and calls sethostname itself. To watch the system hostname please use poll() on /proc/sys/kernel/hostname. You'll get a wakeup each time the hostname changes and you can reread the new hostname. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.