Fixes for Wayland on FreeBSD #116
Conversation
Otherwise, keyboard input would touch the console, and e.g. pressing Enter would crash Weston
valpackett
force-pushed
the
fbsd-wayland
branch
from
a56c98f
to
ea4a569
Compare
src/ck-sysdeps-unix.c
Outdated
| @@ -308,6 +308,13 @@ ck_get_a_console_fd (void) | |||
| } | |||
| #endif | |||
|
|
|||
| #if defined(__FreeBSD__) | |||
| fd = ck_open_a_console ("/dev/ttyv0"); | |||
There was a problem hiding this comment.
What's rationale for this change? Doesn't /dev/console work?
There was a problem hiding this comment.
The function uses /dev/console as fallback, but prefers a VT specifically (on other platforms already, FreeBSD just was missing).
CK is usually spawned by the D-Bus daemon for the system bus, who would guarantee that /dev/console is a VT?
There was a problem hiding this comment.
So, ck_get_a_console_fd should return a fd for a virtual terminal?
In this case, why don't we try all possible ttyv's? When writing this code 1 I found out that if no one is logged in at /dev/ttyv0, you can't open() it.
There was a problem hiding this comment.
Doesn't seem necessary, it just works like this so far.
There was a problem hiding this comment.
Well, I'm debugging https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221452#c36 it seems related. Maybe you have some insights?
There was a problem hiding this comment.
I have no idea how the x11 support works and I haven't tried any display managers yet.
With just two wayfires the switching works, the currently active session is always active e.g.
Session1:
unix-user = '1001'
realname = 'Greg'
seat = 'Seat1'
session-type = 'unspecified'
session-class = 'user'
session-state = 'active'
active = TRUE
x11-display = ''
x11-display-device = ''
display-device = '/dev/ttyv0'
remote-host-name = ''
is-local = TRUE
on-since = '2021-10-22T10:00:23.671559Z'
login-session-id = ''
XDG_RUNTIME_DIR = '/var/run/user/1001'
VTNr = '1'
Session2:
unix-user = '1001'
realname = 'Greg'
seat = 'Seat1'
session-type = 'unspecified'
session-class = 'user'
session-state = 'online'
active = FALSE
x11-display = ''
x11-display-device = ''
display-device = '/dev/ttyv1'
remote-host-name = ''
is-local = TRUE
on-since = '2021-10-22T12:56:56.714321Z'
login-session-id = ''
XDG_RUNTIME_DIR = '/var/run/user/1001'
VTNr = '2'
There was a problem hiding this comment.
According to this PR, using /dev/console seems more correct: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=221452#c39
Should I commit the change proposed in this PR?
There was a problem hiding this comment.
Yes, everything works fine, I have removed the commit from this PR.
Only got a NULL here on FreeBSD/aarch64, for some reason
valpackett
force-pushed
the
fbsd-wayland
branch
from
42b8c93
to
5debe34
Compare
|
@rnagy @ajacoutot We're running these patches on FreeBSD for years now. Can this PR be merged in, please? |
| @@ -1482,6 +1486,16 @@ ck_session_setup_vt_signal (CkSession *session, | |||
| return; | |||
| } | |||
|
|
|||
| #if defined(__FreeBSD__) || defined(__FreeBSD_kernel__) | |||
There was a problem hiding this comment.
Could this be made feature dependent maybe instead of OS dependent?
/dev/dri/card0is a symlink to../drm/0on FreeBSD. When CK2 tries to determine the device type, the path gets resolved, and so it's notcard0. Also the subsystem is not set by libudev-devd./dev/*device doesn't die until udev consumers have detached it, because/devis controlled by udev — in our case, it's controlled bydevfsin the kernel) Let's check if the device died on re-attaching.