2013-07-01, 15:56
Greetings everyone. This might be more related to X than XBMC itself; but hopefully some of you would also like to run XBMC in a multiseat config on a small form-factor system with a built-in Intel GPU. This is for a config where DVI: main display & HDMI: projector in a different room. These ports show up as HDMI1 & HDMI2 in XBMC, respectively. (The system is a Shuttle XH61V, btw.)
Synopsis:
So I'm trying to get XBMC running on two separate (independent) displays with Intel HD4000 (i.e, ":0.0" & ":0.1" or ":0" / ":1") I've been banging my head against the wall for the better part of a week without getting any closer to a working solution; no matter how I've tried to force the creation of two separate Screens (Screen #0 & Screen #1) in xorg.conf; X _always_ configures one screen on the primary display (when looking at the config with xdpyinfo), despite Xorg.0.log indicating that a Screen #0 & Screen #1 has indeed been created. Obviously, this causes XBMC to only see the primary display when launching and one has to override the X config with xrandr to get the second port to be detected. But this again makes the main session unusable, as XBMC always grabs the pointer (even when using -nocursor and the wmctrl workaround posted elsewhere.)
I've tried launching XBMC from the console (with an xinit script), with a separate/dedicated XBMC ServerLayout section, and with a combined xorg.conf (for both the LCD & PJ) and running it in a separate MDM session; without any luck. XBMC works, but one can only use one display at a time (which isn't what I am looking for)–I want XBMC to run independently 24/7 on the HDMI port while using the DVI port for regular (basic) Linux usage. I've also added the relevant ModeLines to xorg.conf and confirm that these are being used when starting the X server.
I've also not been able to create an X-config which isn't tied to a specific VT and which does not force a VT change (using the -sharevts & -noswitchvt options when starting (noticed when trying to start separate X servers for each of the output ports), i.e one server on ':0' and another on ':1'.
Somehow XBMC (or, most likely, X) reverts to the primary display and seems to ignore the fact that HDMI1 is tied to the LCD (via the ConnectedMonitor directive in xorg.conf) and HDMI2 is the PJ–i-e, selecting either HDMI1 or HDMI2 from within the XBMC UI results in it displaying on the same screen.
I know the ConnectedMonitor setting is being taken into account from looking at the Xorg.0.log, but this apparently gets ignored/overridden) So what happens is that XBMC launches with the resolution and settings of the correct port (HDMI2 in this case) on the wrong output– HDMI1.
I've thought about running XBMC from the console via EGL/DirectFB to get around any X interference, but I guess that will not solve the problem of being able to use the system as a desktop while XBMC is running on a different display.
If anyone has any xorg.conf files to share with a working independent screen setup (i.e, not a 'big desktop'), I would appreciate it as my searches for how to enable this have ended up being fruitless, and my own X hacking has proved to be mediocre.
On a related note:
When I do get XBMC running it is only using a spanning display (3480x1920) set up with xrandr, or launched individually from via a XBMC desktop session or SysVinit script. However, running it in fullscreen and 24Hz the video playback is very ugly (tearing in the middle of the picture). This is with the Xvba port of XBMC (which was the only one that was able to detect more than one screen, due to the limitation in SDL mentioned elsewhere on this forum)
I'm running the latest Intel drivers from 01.org, btw. (Which, at the time of writing are at the following versions)–
o Mesa - 9.1
o xf86-video-intel - 2.21.3
o Libdrm - 2.4.42
o Libva - 1.1.1
o vaapi intel-driver - 1.0.20
o Cairo - 1.12.14
o Xserver - 1.13.2.902
On kernel 3.8.0-25-generic (Ubuntu Raring), distro is Linux Mint 15. Appreciate any input from others who might also have tried this (or know more about the intricacies of X than I do!)
Synopsis:
So I'm trying to get XBMC running on two separate (independent) displays with Intel HD4000 (i.e, ":0.0" & ":0.1" or ":0" / ":1") I've been banging my head against the wall for the better part of a week without getting any closer to a working solution; no matter how I've tried to force the creation of two separate Screens (Screen #0 & Screen #1) in xorg.conf; X _always_ configures one screen on the primary display (when looking at the config with xdpyinfo), despite Xorg.0.log indicating that a Screen #0 & Screen #1 has indeed been created. Obviously, this causes XBMC to only see the primary display when launching and one has to override the X config with xrandr to get the second port to be detected. But this again makes the main session unusable, as XBMC always grabs the pointer (even when using -nocursor and the wmctrl workaround posted elsewhere.)
I've tried launching XBMC from the console (with an xinit script), with a separate/dedicated XBMC ServerLayout section, and with a combined xorg.conf (for both the LCD & PJ) and running it in a separate MDM session; without any luck. XBMC works, but one can only use one display at a time (which isn't what I am looking for)–I want XBMC to run independently 24/7 on the HDMI port while using the DVI port for regular (basic) Linux usage. I've also added the relevant ModeLines to xorg.conf and confirm that these are being used when starting the X server.
I've also not been able to create an X-config which isn't tied to a specific VT and which does not force a VT change (using the -sharevts & -noswitchvt options when starting (noticed when trying to start separate X servers for each of the output ports), i.e one server on ':0' and another on ':1'.
Somehow XBMC (or, most likely, X) reverts to the primary display and seems to ignore the fact that HDMI1 is tied to the LCD (via the ConnectedMonitor directive in xorg.conf) and HDMI2 is the PJ–i-e, selecting either HDMI1 or HDMI2 from within the XBMC UI results in it displaying on the same screen.
I know the ConnectedMonitor setting is being taken into account from looking at the Xorg.0.log, but this apparently gets ignored/overridden) So what happens is that XBMC launches with the resolution and settings of the correct port (HDMI2 in this case) on the wrong output– HDMI1.
I've thought about running XBMC from the console via EGL/DirectFB to get around any X interference, but I guess that will not solve the problem of being able to use the system as a desktop while XBMC is running on a different display.
If anyone has any xorg.conf files to share with a working independent screen setup (i.e, not a 'big desktop'), I would appreciate it as my searches for how to enable this have ended up being fruitless, and my own X hacking has proved to be mediocre.
On a related note:
When I do get XBMC running it is only using a spanning display (3480x1920) set up with xrandr, or launched individually from via a XBMC desktop session or SysVinit script. However, running it in fullscreen and 24Hz the video playback is very ugly (tearing in the middle of the picture). This is with the Xvba port of XBMC (which was the only one that was able to detect more than one screen, due to the limitation in SDL mentioned elsewhere on this forum)
I'm running the latest Intel drivers from 01.org, btw. (Which, at the time of writing are at the following versions)–
o Mesa - 9.1
o xf86-video-intel - 2.21.3
o Libdrm - 2.4.42
o Libva - 1.1.1
o vaapi intel-driver - 1.0.20
o Cairo - 1.12.14
o Xserver - 1.13.2.902
On kernel 3.8.0-25-generic (Ubuntu Raring), distro is Linux Mint 15. Appreciate any input from others who might also have tried this (or know more about the intricacies of X than I do!)
