You are not logged in.
The problem occurs on two computers with different graphics cards and monitors by different manufacturers. It only appeared on the older nvidia machine when the 16x10 monitor was replaced by one that was 16/9. It appeared on the brand new machine with onboard Intel graphics immediately. The problem ONLY occurs using the old X11 "grey" background; that is the common feature, and I assume it is the problem. In both cases I did the setup, one eight years ago and the other just recently, by letting xorg do its hardware checking, and in the more recent case it runs without an xorg.conf file. The new 16/9 monitor displays beautiful detailed graphic output over the whole screen area for all applications, so I have had no reason do do any BIOS adjustment.
Sorry for the delay, as I have been busy handling the video again for the beowulf upgrade. The video card is GT440 1G from Nvidia and does not work properly without the Nvidia specific driver. That was a lot of trouble eight years ago when the machines were new, but the driver became a standard package from Debian, and now Devuan; it requires compilation of a kernel module each time there is a kernel upgrade. I didn't make any comment about inches (and as a Canadian I would prefer to use centimetres :-) The problem seems to depend on the screen aspect ratio, and showed up on one computer when its monitor gave up and was replaced with a newer one, having the wider 16x9 aspect ratio. The cables and connections are the standard that came with a usual desktop (tower) computer. I just got a new computer with Intel B360M on board graphics and it shows the same behaviour. The only problem is with the traditional X11 background, so-called grey but actually a checkerboard; all other graphics display perfectly. I suspect there is a problem with the X11 background, and maybe that is why they replaced it; it is no longer the default.
I have inadvertently installed the nvidia-driver 418.74-1 on a machine with a graphics card that is no longer supported. The problem can be corrected by installing nvidea-legacy-360xx-driver, and I would appreciate some advice on how to proceed. I assume I have to remove all the nvidia 418.74 packages and also xserver-xorg-video-nvidia 418.74-1 and replace them.
Which xserver package should be installed? Once that is done, I assume I can install the nvidia legacy meta-package and get X11 to function again. I also assume one can not, or should not try to install the replacement nividia packages over the existing ones. So, what is the best sequence of steps?
Congratulations to the good people who maintain Devuan. With just a little trepidation I did my upgrade on one machine which was using 917 packages. The whole upgrade process took less than an hour, and all that I had to deal with was the usual questions about keeping modified configuration files. It seems as if everything works properly ... except the graphics, but that is an Nvidea problem, and both my X11 log and the email report on the installation give explicit instructions on how to handle the problem. I have had a lot more fallout from distribution upgrades in the past.
As regards graphics, oddly I am in a no-man's-land as I have a new machine that will not work without the 4.19.0 kernel, and my older ones which are having trouble with Nvidea and the Beowulf upgrade. It keeps me busy, and confined to the house, which is not a bad idea these days.
I wonder if this is a general problem, or is related to specific monitors. I run X11 with the fvwm window manager, and have used the traditional "grey" X11 background on the root window; it is in fact a checkerboard mosaic using very small black and white squares. When I replaced a 16x10 display with a 16x9 replacement the far right area of the screen displayed an undulating background with wave-like horizontal lines. When a window was located in that area its contents displayed normally.
I am now setting up a new computer, and the 16x9 display exhibits exactly the same glitch. Anything displayed in that area is rendered properly, and if I use "xsetroot -solid <somecolour>" there is no problem. Since this happens with two monitors, of different vintage, I wonder if it is in fact an X11 problem, and there is something wrong with the rendering of the checkerboard pattern. I note that X11 changed the default background a couple of upgrades ago.
It is not of any great significance, as I have discovered that I prefer the solid background, and if it has to be grey, there are several shades available.
Anyone else have the same experience?
Just did a test run on a virtual machine, and it repeated what happened previously adding the 4.19.0 kernel to an ascii system, namely installing apparmor. That gave me trouble with the previous 4.9.0 kernel, specifically handling locales. My quick check on what apparmor is leads me to conclude that I don't want it, and that furthermore it might get in the way of running one's system without any unnecessary impediments.
It seems to recommended for the 4.19.0 kernel, not a dependency, so I don't think dist-upgrade should be insisting on it. I moved to Devuan because I did not want to be told how I ought to do things.
Those are the notes that I used. But I just checked the release notes for Debian buster and they recommend doing apt-get upgrade before the full upgrade. My comment was that Devuan do not make q similar suggestion.
I am simply curious about how the upgrade process works. I assumed packages are upgraded continuously, I certainly do enough in between major releases. I suppose changes are held back until a major release change if they involve new versions and repackaging.
I did not have to upgrade from Devuan 2.0 to 2.1 because routine maintenance got me there seamlessly. Clearly the step to 3.0 involves just about everything.
A question on the upgrade notes. As I recall from previous upgrades on Debian, instructions are to do an upgrade before the dist-upgrade; presumably to upgrade packages that do not require reorganization and changes in dependencies. Is that not a good idea?
I did comment elsewhere on the fact that essentially all packages were upgraded when I made the changeover, and my recollection was that previously something like half of them were unchanged if regular checking had been done on upgrades with aptitude. I am normally upgrading packages weekly, and I had assumed this takes many of them to the level of the next release. In that case I thought the dist-upgrade dealt only with reorganized packages and changes in versions.
I have a new machine which I am in the process of setting up and tailoring. It had a minimal installation set up from debootstrap, and it seemed like a good idea to test upgrading it to Beowulf before doing any further software installation. Following the upgrade notes, I discovered at "apt-get dist-upgrade" that the process was going to upgrade ALL 327 packages, remove 2 and add 97 new.
I have been doing these distribution upgrades since Debian 2.2 (potato) and it seems unusual for everything to be updated, and there are also normally only new packages because of repackaging. So I have two questions: (1) is it to be expected in this case that all packages need an upgrade? (2) Why are new packages being installed?
I am interested in the answers as I have two machines presently running ascii on about 1000 packages, and they will be upgraded soon.
It is possible (1) occurred because I missed a step and did not do a second "apt-get update" after installing devuan-keyring?
Did (2) happen because there is a list of "standard packages"? I notice one of them was exim4 (which of course I will want eventually, but it looks as though I am being told what packages I ought to have. No complaints, as it is probably good advice.
In any case the upgraded system works well, and the upgrade process is fast and straightforward. I am very happy being in the present company.
It is indeed the problem! Adding a line "needs_root_rights=yes" to Xwrapper.config fixes things so that user roger can start X11. I am pleased with this outcome, not just because an annoying problem has been solved, but because I understand the cause and it makes sense. The only difference between the machines running my script and one that wouldn't was the video card, and apparently that makes the decision about rights when the auto setting is used.
Thanks for the quick and helpful response Ralph.
That may be the problem, as there is a default setting "needs_root_rights=auto" and what auto decides depends on the video card. I experimented with the setting "no", but I think that is the reverse of what I want. I'll experiment more later. I am optimistic about this because all the machines have identical software and settings, so the video card is a likely source of difference.
From my research on the web this appears to be a classic problem, but the suggested solutions are less than ideal, or do not work for me. I am setting up a new desktop machine as a potential backup or replacement for two others that have been running continuously for eight years or so. I am mainly interested in programming, and write my own software for some video applications and record keeping. The typical desktop environments do not appeal to me as they all seem to be attempts to make Linux machines look and behave like windows, and if it's GNOME even preventing the user from being able to take control. So, I run fvwm and xterms with most things available on drop-down menus. My procedure on other machines has been, for many years, to let myself, not as root, be the principal operator, and the person that starts an .xinitrc script that all other processes hang from. If I want to be root, or any other user, I set up xterms for them.
The problem with the new machine is errors such as "parse_vt_settings: Cannot open /dev/tty0 (Permission denied)"; making myself a member of the tty, input, and video groups does not help. The suggestions frequent on the web are to change the file permissions for all the /dev/tty? files; that is somewhat drastic, and does not seem to work, and in any case permissions get changed depending on which user does the first login. There is a suggestion on the gentoo website that seems a likely cause of my problem. Apparently the xorg people changed a default involving sudo in their generic device driver, and apparently users with nvidia and intel drivers avoid the problem. I have tried to make an entry in xorg.config.d for the intel driver, but that doesn't seem to work; maybe I have to download a driver from Intel and not use the one supplied by xorg.
I would appreciate any insight members of this forum could give me. In the meantime (reluctantly) I can continue with my testing and tailoring using sudo. That complicates things as it messes up paths and directory references.
Progress report: I planned to install a basic version of ascii on an empty disk, using debootstrap on a new computer, as a first step to setting up Devuan. The vendor had installed Ubuntu simply to check whether everything was working. As reported I was unable to use netinstall.iso because an ethernet card was not found. I have had some fun using a mixture of tty2 on the Ubuntu disk and my own efforts on the second disk to set up Devuan, with the initial result that there was no ethernet connection. Using very helpful suggestions from members of this forum, I have now upgraded the 4.9.0 ascii kernel to beowulf's 4.19.0 and the ethernet connection appears and works properly.
So: the problem was, as suggested by others above, that the device driver in kernel 4.9.0 was not able to handle the updated hardware of Intel's B360M chipset. The e1000e driver for Intel has been around, in name at least, for the eight year life of my existing desktop machines, but it has been significantly changed, as is apparent from its increased size. Research on the web suggests that kernel 4.15 or so is necessary for the latest Intel ethernet hardware.
I appreciate the helpful comments and advice from members of this forum. Thanks.
Thanks very much. What I was missing was that one can control which parts of sources.list are used with entries in an
/etc/apt/preferences file, and that those preferences can be overridden by a -t flag to apt-get. Of course one also has to run apt-get update to get access to sections added to sources.list
I can now install selected items from the beowulf release into a system running ascii. In principle I am happy to stay with the current stable release, but I am planning for a possible problem with outdated device drivers installing Devuan onto a new computer. I have had problems like that before with ethernet and video. That was several releases ago, and Linux seems much better these days in being up to date with hardware.
Forgive my ignorance, but I am having trouble persuading dpkg or apt to download an updated kernel for testing. I simply want to get the beowulf version without changing my sources.list and having apt then want to upgrade everything else. There must be a simple command to download a specific *.deb file from the repository. What is it?
My computer usage is mainly to do programming running compilers and making relatively simple graphics applications (by gaming standards) - so I am happy to stay with whatever the current stable release of Devuan is, and I also appreciate not having to be continually involved in upgrades.
So: I am running the 4.9.0-12 kernel, and my computers have a file
/lib/modules/4.9.0-12-amd64/kernel/drivers/net/ethernet/intel/e1000e/e1000e.ko
The driver is definitely is there. My problem is that running netinst.iso on a new machine with the intel chipset does not find the ethernet card. I conclude the driver is not in netinst.iso and so my best course of action is to get debootstrap into the machine currently running Ubuntu and use the untar procedure described in instructions. That is probably preferable to using apt to install it, mixing it up with Ubuntu sources.
I'll report progress. It is probably useful to know that the Debian wiki for debootstrap describes the correct procedure for Devuan. I have done a quick check of a test installation of debootstrap on a Devuan virtual machine and it definitely has some Devuan specific features.
The advice is appreciated, and is what I follow since I only use the Devuan sources. But I did find debootstrap listed when I run aptitude using my exclusively Devuan sources list; the description given is "Bootstrap a basic Devuan system." Obviously it would handle the init package properly as that is essential to the basic operating system. The idea is that once a basic Devuan system in available, it can be used to install kernel packages, which would provide any necessary modules.
Apropos the e1000e ethernet driver module, I assume that is not included in the netinst.iso download. I do not know enough about the vast varieties of hardware available to know how common the B360M-C motherboard is among Linux users.
A follow-up to the above: web research indicates that my problem is likely that devuan_ascii_2.1_amd64_netinst.iso does not include a driver for the ethernet section of the Intel B360M-C motherboard used by Asus. The computer provided by the vendor does contain a working installation of Ubuntu, and a check with lsmod shows that a module e1000e is installed, as a component of the 4.15 kernel.
In one of my existing machines currently running Devuan, the e1000e module is present in /lib/modules so it would presumably be available after kernel installation. The problem is then to get the driver into the machine in which Devuan is to be installed, and it now seems that it would be a good idea to follow the advice I was given in my previous thread and use debootstrap to install Devuan in the new machine currently running Ubuntu. It seems an even better idea given that the machine has two hard drives (in preparation for using mdadm to set up RAID1). Devuan can presumably be installed on /dev/hdb, using debootstrap, without messing up the existing Ubuntu installation on /dev/sda; once Devuan is working, Ubuntu can be removed, and RAID1 set up.
It sounds like a fun project; precisely why I got into working with Linux.
I have separated this from my previous post which just concerned matters of curiosity. My plan is to do a Devuan install from an iso USB stick on a computer which has a working Ubuntu system installed by the vendor. I have run into a problem early in the install procedure because the installation program reports that no ethernet card is detected. It does present a list of drivers, from which a selection can be made, but there is no way to exit from that screen and continue the process.
Ethernet and video are included in the Intel B360M-C motherboard used by Asus in this machine. Running Ubuntu I can use F2 to get a tty and run regular unix commands; lsmod tells me that a module called e1000e is loaded, which turns out to be a driver for Intel ethernet hardware. e1000e is also included on the list of drivers presented by the Devuan install process, but selecting it does not make a "continue" button available.
I assume that if the driver is listed by the install process it is available in the iso file used to set up a basic system. Maybe that is a wrong assumption and I have to download and install the driver, but it appears that there is a hardware detection problem with the installer. I would appreciate any assistance that anyone could give me for this problem.
I would also welcome any comment about whether this topic belongs under "hardware" rather than "installation".
My private network has two desktop machines running Devuan, and they have been running virtually continuously for eight years. I thought it would be a good idea to get a new machine in order to be ready for a failure, and also to keep my hand in with the installation process. The new machine has an Asus B360M motherboard, which I believe will handle Linux, and to my surprise when I first turned it on, I was immediately dumped into a Gnome desktop. So obviously the hardware handles X11 graphics which is good news.
But it is filled with Gnome bloatware and garbage, and I am hampered in investigating what is there by not knowing the password set up by the vendor, and Ubuntu's ridiculous attitude about root accounts. I had intended to install Devuan in an empty machine, using an iso file on a USB stick, and of course it is still possible to do that; however, I am interested in finding our whether one can make the iso file accessible to the existing Ubuntu system, and install Devuan by overwriting from there. Has anyone any ideas on that?
In passing, I must comment that if I wanted a desktop system filled with bloatware, and a Gnome menu devoid of xterms, I might as well stay with Windows. Linux does of course allow access to tty's using function keys, but that isn't any use without a password to log in. I can get the password, either from the vendor, or by diverting the boot process, and would like to investigate what Ubuntu has provided, before I erase it and set up my familiar Devuan.
I have similar experience, and had the same question, but I routinely use aptitude (once a week or so) and my system transparently changed from 2.0 to 2.1 -- my only minor problem was updating the mechanism for reporting the release.
I think there is only more effort necessary when the major release number changes, e.g 9 to 10 in Debian; I assume that will happen with Devuan when 3.0 comes on line. Debian does hold back upgrades from stable until the announcement of a minor release change.
Thanks. It does work indeed, and the re-installed lsb_release identifies the system as Devuan 2.1
It's one of those situations where apt-get works better than aptitude which can be confused on occasion. I think the problem was version 9.20161125 was identified as being higher than 4.1+devuan2 so aptitude saw no reason to upgrade. Once 9.20161125 was gone, aptitude found the correct version from the Devuan sources.
A question I raised under news and announcements, but the follow-up is probably more appropriate under a different topic. I have Devuan running on a virtual machine in a Windows laptop where it was installed from scratch:
lsb_release -sir
Devuan
2.0
correctly reported, as I have not updated the system. But on regular desktop machies, which were transitioned from Debian:
lsb_release -sir
Devuan
9
These machines are continually updated, and as a very happy Devuan user I would like them to be convinced that they are running Devuan 2.1 and to report correctly.
The problem appears to be that the package lsb-release will not update from version 9.20161125 to version 4.1+devuan2 because it requires a package distro-info-data without which it claims to be broken. I think distro-info-data should be removed, and lsb-release reinstalled to fix the problem, but I would appreciate any comments. The package distro-info-data is specific to Debian and not included on a clean install of Devuan.
This is obviously not vitally important as I have a fully functional Devuan system, but I would like it to identify itself properly.
Thanks, I didn't know that command. But it gives me
Devuan
9
I suspect there is a remnant from Debian that should be removed.
As a relative newcomer to Devuan I have a question. Having received the notice about the 2.1 release, I note that many packages I have installed are at later versions than the ones listed in the announcement. I have updated regularly, and do so at least once a week; does that mean that I in fact have the 2.1 version of Devuan?
If that is the case, might I suggest that the file /etc/devuan_version be updated to indicate the current level of the system.
To conclude on a lighter note, typing this response prompts another question. Why does the Galaxy Forum spell checker query the word "Devuan" -- I like it!