Description of problem: I tested the Beta1 Plasma Live DVD hardware : processor intel 15-9400F NVidia GeForce GT 1030 1st display iiyama PL2474H 2nd display Sony TV connected by HDMI A) without the nonfree driver 1) In the beginning the displays are cloned 2) But when I get into the GUI interface to select language, accept licence, choose country time etc.. these windows are spreaded on the two displays (cut in their middle) and I have to go from a display to the other to make a choice and validate it 3) After that Plasma is launched and the external TV gets black 4) The TV doesn't exist in the plasma system settings when I try to clone it from the main display B) with the non-free driver Everything is OK 1) the displays are cloned even for choosing language etc... 2) The plasma system setting allows to clone the TV from the main monitor
Thank you for the report, which basically says that with the proprietary nVidia driver all works OK, but not with 'nouveau'. Please post the output of: $ inxi -Gxx to detail the graphics setup; for both the Nouveau & Nvidia drivers if possible.
CC: (none) => lewyssmith
Created attachment 13729 [details] result of inxi with nvidia nonfree driver result of inxi with nvidia nonfree driver
Created attachment 13730 [details] result of inxi with nvidia free driver (nouveau) result of inxi with nvidia free driver (nouveau)
Surprising !!! After having launched the Live DVD with nouveau I got the same problem than described in comment 0 then I used inxi in a console the screen becomes black for one second And a window appears above the window of the console This window proposes a choice of display : 1st monitor (with no 2nd monitor) or 2nd monitor (with no first monitor) or 2 cloned monitors or screen spreaded on the 2 monitors side by side I choose cloned monitors and then it's OK : the two monitors exist in the Plasma system setting and I can modify the display I remember that for Mageia8 the window proposing a choice of display appeared at the end of the launching of a LiveDVD... this step seems to be missing for Mageia9 : => no problem for the non free driver since the 2 monitors exist for the Plasma system setting ...) => but this is wrong for nouveau since it's not possible to use a the second monitor
When booting using the free (nouveau) driver, try editing the boot command line in the GRUB menu and removing the word "noxconf".
CC: (none) => mageia
Hi Martin I have edited the command line in the GRUB menu and removed "noxconf" 1) That solved the first part of the problem : Now the 2 monitors are cloned in the beginning and the four windows and the mouse cursor appear twice (once on each monitor) no problem to choose language, to accept licence, to choose country, local time and keyboard BUT 2) Only the first monitor displays something, the TV is black the display is clean and readable I have to launch again inxi -Gxx since in the Plasma setting only one monitor exists After that I don't get the window proposing the choice of display like in comment 4 Nevertheless, in the Plasma setting the two monitors then exist and I can choose to clone the second one (TV) on the first But after having used inxi -Gxx the polices in Plasma are fuzzy, whatever police I choose, and even if I increase their size (that was not the case with the method used in comment 4) It's quite unreadable (fortunately I know Mageia enough to guess what is written) the problem is the same on the two monitors... Is it now a Plasma problem ? or still a nouveau problem ? NB those fuzzy polices didn't appear before I use inxi -Gxx which induces this time something different than in comment 4 NB with NVidia non-free driver everything works well out of the box : - no need to remove noxconf in GRUB (I don't know if this word is present in GRUB for the non-free driver) - no need to use inxi -Gxx to have the two monitors present in Plasma setting Anything else to try for nouveau ?
Hi Philippe Could you try both removing "noxconf" and changing "xdriver=free" to "xdriver=modesetting" on the boot command line. Check with "inxi -G" - that should still show "nouveau" as the active kernel driver but should show "modesetting" as the active X11 driver. Could you also test the Xfce Live ISO, to see if it behaves the same.
Thank you both for attacking this. I do not think it is right that using 'inxi' changes the video behaviour; it is only a reporting tool.
Hi Martin Hi Lewis Sorry for lasting to answer (not home today) to Lewis It's strange but indeed using inxi created a strange video behaviour each time I used it with nouveau : It allowed to find the second monitor inside Plasma settings (which did not exist before !) but created garbage display after !!! same for XFCE to Martin I tested Xfce Live out of box : 1st surprise the two monitors are cloned in the beginning and the windows to choose language, accept licence, choose country, choose local time, and choose keyboard are not spreaded on a wide screen using the two monitors side by side (that's different from Plasma Live for which I had to navigate with the cursor from half a window on left monitor and the second half window on the right monitor ) But when XFCE is launched, I get a wide screen spreaded on the two monitors and no way to clone the TV on the primary monitor with the XFCE parameters I modified GRUB removing noxconf => nothing better But you found the trick in comment 7 :-) congratulations If I modify GRUB for Plasma Live and for XFCE Live : remove noxconf modify xdriver= from free to modesetting Everything works perfectly on Plasma and XFCE : I can clone my monitors and don't have fuzzy police For XFCE and Plasma inxi -G gives : nouveau as kernel driver and modesetting and v4l as active
NB modifying GRUB is necessary for nouveau but for NVidia non-free driver there's nothing to modify it works perfectly out of the box When the solution has been found (congratulations again to Martin) and when GRUB will have been modified this bug should be closed as resolved Unfortunately I won't test a real installation of Beta ISO on my computer and see if the problem with nouveau exists too Let someone else test a real install on a bare computer Thanks again for you wonderful work : except the nouveau driver problem everything else is quite perfect : for a Beta it's unexpected
Aw
Await Martin's response. Is this for assigning to Mageiatools, - is there anything we can do? - or better left as is with an ERRATA note? Noting that in case.
Keywords: (none) => FOR_ERRATA9
Unfortunately there isn't one single configuration that works for every hardware setup. The current boot command line options were chosen as the best compromise. There were some notes in the Mageia 7 Errata about changing the defaults for the X drivers, but it seems they got removed from the Mageia 8 Errata. I also notice that some other Live-specific details got moved inappropriately. According to https://bugs.freedesktop.org/show_bug.cgi?id=94844#c3, the modesetting DDX driver is preferred over the nouveau DDX driver for G80 and later Nvidia GPUs. CC'ing tmb for his opinion on whether we should change ldetect-lst accordingly.
CC: (none) => tmb
(In reply to Martin Whitaker from comment #13) > There were some notes in the Mageia 7 Errata about changing > the defaults for the X drivers, but it seems they got removed from the > Mageia 8 Errata. I also notice that some other Live-specific details got > moved inappropriately. Just about ERRATA for M9: Your comment suggests that M8 Errata lacked carried-forward items from M7 which were still valid. And still are for M9 ? Unsure what you mean by "moved inappropriately": removed? Tell me privately.
(In reply to Lewis Smith from comment #14) > (In reply to Martin Whitaker from comment #13) > > There were some notes in the Mageia 7 Errata about changing > > the defaults for the X drivers, but it seems they got removed from the > > Mageia 8 Errata. I also notice that some other Live-specific details got > > moved inappropriately. > Just about ERRATA for M9: > Your comment suggests that M8 Errata lacked carried-forward items from M7 > which were still valid. And still are for M9 ? Unsure what you mean by > "moved inappropriately": removed? Tell me privately. here is the errata from Mageia7 that disappeared for Mageia8 << Live Images : Non-working graphics Some graphics hardware does not work correctly when using the xorg.conf file that gets created by harddrake when booting the Live system. If you have no graphical display at the end of the boot cycle, try adding noxorgconf to the boot options in the initial boot menu (press e to edit the boot options). Some NVIDIA graphics cards don't work at all with the free video driver (nouveau). In this case, try replacing xdriver=free with xdriver=modesetting in the boot options in the initial boot menu, or alternatively, try using the non-free (proprietary) drivers by selection directly in the boot menu. >> The solution for the problem with nouveau was here !!!
CC: (none) => friKeywords: (none) => FOR_ERRATA8
I tested again with my old Mageia8 Live ISO (final release) (I never have done this : always testing NVidia non-free drivers) The problem is exactly the same with nouveau for Mageia8 and Mageia9 as described in comment 1 in plasma settings monitor 2 doesn't exist launching inxi -Gxx In a console monitor 2 appears in the result After that I tried again plasma settings and the second monitor exists if I try to clone from monitor 1 I get garbage in monitor 2 Editing GRUB the same way as for Mageia9 solves the problem for Mageia8 This errata from Mageia7 shouldn't have disappear for Mageia8 and is needed for Mageia9 unless there is a way to have GRUB conditionally adapted after the detection of the graphic card (does the launcher of the Live ISO use ldetect or hardrake to adapt GRUB ?)
Thank you Philippe. I updated Errata 8 & 9. I also copied two other sections regarding Live from 7, which AFAIK are still valid. (For the errata part on Gnome I dont know, do anyone reading this have?)
Keywords: FOR_ERRATA8, FOR_ERRATA9 => IN_ERRATA8, IN_ERRATA9
(In reply to Morgan Leijström from comment #17) > (For the errata part on Gnome I dont know, do anyone reading this have?) Ah. It was in the wrong place in mga7 errata, it is in correct place in mga8 & 9 errata (not specific to Live)
(In reply to Morgan Leijström from comment #18) > (In reply to Morgan Leijström from comment #17) > > (For the errata part on Gnome I dont know, do anyone reading this have?) > > Ah. It was in the wrong place in mga7 errata, it is in correct place in mga8 > & 9 errata (not specific to Live) No, the "nowayland" option is specific to Live. I've added that back, also adjusted the instructions for non-working graphics, for which there was a change between 7 and 8.
Thank you Martin Shouldn't we move the Gnome section about "nowayland" https://wiki.mageia.org/en/Mageia_9_Errata#GNOME to up under the Live header then?
(In reply to Morgan Leijström from comment #20) > Shouldn't we move the Gnome section about "nowayland" > https://wiki.mageia.org/en/Mageia_9_Errata#GNOME > to up under the Live header then? I did, but sturmvogel has undone my changes.
OK we need to communicate. (sorry about keeping hijacking this bug) So the "nowayland" option is specific to Live. This is a medicine. But the described problem with Gnome and Wayland, does it not also appear in a system installed by classic installer?
CC: (none) => sturm-fr
(In reply to Morgan Leijström from comment #22) > But the described problem with Gnome and Wayland, does it not also appear in > a system installed by classic installer? If it does occur the user can edit /etc/X11/gdm/custom.conf and set WaylandEnable=false. The "nowayland" option on the Live ISOs just does that for you, because it's a pain to have to do it manually every time you boot the Live ISO. This may not be needed any more - Wayland is more mature now - but I felt I might as well keep the option (and document it) just in case.
Good. I added that info, while yet cut down text length, all in one place: https://wiki.mageia.org/en/Mageia_9_Errata#GNOME_and_Wayland (to be backported to mga8 errata if OK)
As it is documented (thanks always to Morgan), can we close this? It is (comment 13) not something we can fix universally.
(In reply to Lewis Smith from comment #25) > As it is documented (thanks always to Morgan), can we close this? > It is (comment 13) not something we can fix universally. This kind of bugs remain open to check if it is still valid when new version is available, I see that at less with my weird bugs
OK, but it would be nice to have a special assignee for bugs like this with no feasible fix, which are basically ERRATA with work-around advice. Leaving with bugsquad for now.