Mir/unity-system-compositor crashed with std::exception::what: Couldn't clear output ... (drmModeSetCrtc = -13)

Status changed to 'Confirmed' because the bug affects multiple users.

I also used live image and after install of unity8-desktop-session-mir and login to Unity 8 ended up with black screen.

Then I switched to TTY1 and restarted lightdm and when it restarted I was on Unity 8 greeter. The session started but it was totally unusable. The Scopes displayed only webbrowser-app , system settings, checkbox and app store.

When I tried to start system settings it only displayed a small window for a very, very short time and disappeard. The same with webbrowser. Also the fonts display was terrible. Mouse cursor was very slow.

Hmm, for one, I'm not how well supported just installing unity8-desktop-session-mir works on a LiveCD. I seem to remember there were special hacks in place when the Desktop guys were creating a Unity8 based LiveCD ISO (which sadly has been deprecated).

Also, looks like Mir may be having some issue with the display based on this:

ERROR: /build/mir-UeyFew/mir-0.19.0+16.04.20160128/src/common/fatal/fatal.cpp(55): Throw in function void mir::fatal_error_except(const char*, ...)
Dynamic exception type: boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >
std::exception::what: Couldn't clear output eDP-1 (drmModeSetCrtc = -13)

As a stab in the dark, I'll add the Mir project to this bug to see if they can weigh on on this error.

Error 13 is "Permission denied". That could mean unity-system-compositor is being shut down from a VT other than the one it is running on. So we might expect to see that error.

I'm tempted to say if this only happens on shutdown then we just need to make it a non-fatal error. However there's only one location where I can see the offending function getting called:

mgm::Display::clear_connected_unused_outputs()

and even then it's only called when:

        if (conf_output.connected &&
            (!conf_output.used || (conf_output.power_mode != mir_power_mode_on)))

So how is your eDP-1 (laptop screen) considered connected and unused? Is something turning the screen off prior to shutdown?!

@Daniel, let me know if there's something I can test for you (or a different log I can scrape) to check if something is turning the screen off when I try to enter the Mir session.

I tried reproducing this yesterday with the Mir demos, but no luck yet. I suspect we'll have to modify some code to more easily reproduce the problem. I don't think you need to do anything more right now, thanks.

I tried this yesterday with the latest daily 16.04 image and just wanted to note that the user-facing behavior is the same. I didn't check the logs to see if the same messages were appearing but could do more if it would help.

Another report of the same crash came in today.

Perhaps we need a fix like that of bug 1579630.

Lucky me. My yakkety desktop experienced this crash today.