[regression] Mir gives up too easily - std::exception::what: Failed to find the current VT

[Daniel van Vugt]

I have always run Mir from an ssh login, but also made sure I have a VT1 login on the console too. This has worked well until recently. Now we seem to have introduced a restriction that mir can only be started from the physical console...

$ bin/mir
ERROR: /home/dan/bzr/mir/trunk/src/server/graphics/gbm/linux_virtual_terminal.cpp(70): Throw in function int {anonymous}::find_active_vt_number()
Dynamic exception type: boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >
std::exception::what: Failed to find the current VT

[Alexandros Frantzis (afrantzis)]

This should work when running as root. I can add some more logic, so that if the user that is trying to run mir is logged in any VT, then mir will be able to find the current VT, regardless of whether it is the VT the user is logged in. I think this should allow you to work as before (of course, proper VT switching will still not work if your run as a normal user).

[Daniel van Vugt (vanvugt)]

Confirmed that root can run mir from ssh. Only non-root can't.

[Daniel van Vugt (vanvugt)]

I've been playing with Mir in a chroot on an armhf chromebook recently and this error is presently a blocking issue. I'm pretty sure my system does not even have VTs, and while it does use the mesa-kms driver, the inability to find or control VTs should not be considered an error.

[Daniel van Vugt (vanvugt)]

No longer a critical priority for the new device bring-up. Might finish it at a later date.

[Daniel van Vugt (vanvugt)]

Still happening...

[Michał Sawicz (saviq)]

Syncing task from Mir.