GNOME Bugzilla – Bug 344902
/usr/bin/dogtail-run-headless - fails on FC5 - make the script distro-aware or just document?
Last modified: 2006-07-13 18:25:55 UTC
Please describe the problem: /usr/bin/dogtail-run-headless - fails on FC5 - make the script distro-aware or just document? In order to run in headless mode on FC5, this line in dogtail-run-headless: xinit -- /usr/X11R6/bin/Xvfb :1 -screen 0 1024x768x24 -fbdir /tmp :1 Must be changed to: xinit -- /usr/bin/Xvfb :1 -screen 0 1024x768x24 -ac -noreset -shmem Also - although it seems obvious in retrospect, headless mode must be run from the console, not from within a GNOME session. If you do try to run it from within a GNOME session, this error results: 'I've detected a panel already running and will now exit.' Steps to reproduce: 1. 2. 3. Actual results: Expected results: Does this happen every time? Other information:
The change in path of Xvfb is a sideeffect of the switch to the modular X server. Can we expect this stuff to simply be in $PATH ?
Doesn't really matter, since if it's not in $PATH we've got a couple good guesses anyway. If it's in /usr/local/bin/haha/try/to/find/me, we can just bail.
And - Zack actually just pointed out a better way to run scripts in headless mode - create another user account, su into that account and then run the headless script. This will get around the problem of having a GNOME session as the new user will not have a GNOME session.
I think it should be possible to run a headless dogtail session without having to do things like have root access, or create custom accounts. Normal users should be able to run the scripts.
It certainly is possible. You just can't have more than one GNOME session running, and that's gnome-session's fault, not dogtail's.
(In reply to comment #5) > It certainly is possible. You just can't have more than one GNOME session > running, and that's gnome-session's fault, not dogtail's. > So the in-ability to run multiple headless dogtail sessions, or just a dogtail session simultaneous to a gnome desktop is not available due to a limitations in gnome. Could this restriction be lifted by, for instance, telling dogtail to use an entirely virtual or constructed GNOME session? Perhaps a bug against GNOME could be filed? Testing, while clearly important, is often the most difficult activity to encourage. Keeping that in mind, I think it is really important that we keep all potential barriers to entry minimized, and to that end having a UI test framework that has zero cornered limitations is very important. If we don't to that we'll be faced with the constant and perenial assertion of << testing is too hard >>.
Actually, the situation seems to have improved since I last took a detailed look. It looks like we can work around this, at least in many cases, by simply not starting gnome-session or gnome-settings-daemon. That can't be the default, though. I guess dogtail-run-headless needs to be more flexible in terms of how it sets up its environment.
What's the status of this? Any progress? This is impacting Frysk testing in a big way. thanks
(In reply to comment #8) > What's the status of this? Any progress? This is impacting Frysk testing in a > big way. As you (and most other people don't) know, I've been having to dedicate all of my time to RHEL update testing for the last... long time. During the few spare moments I've gotten, I rewrote dogtail-run-headless, in Python, with more options, far better error handling, and ... a usage statement! It now finds an available X display all by itself. And if you want to run multiple scripts as the same user, at the same time, use the 'none'-type session. It also works on FC5.