You are not logged in.
I am not really angry with systemd, eventually it is just a program, but with everything that surround it, I also find disheartening the lack of leadership and orientation that pervades Debian. From the leading to follow whatever the big siblings of Linux decide.
Honestly the thing that drove me away from Debian was the constant pandering to whatever overengineered crap GNOME and freedesktop/redhat was peddling on any given day.
Systemd is fine for those who want to use it, but because GNOME depends hard on it at compile-time it's been given special status in Debian... Where special means an unavoidable, unremovable, low-level please-link-everything-against-me bloat-kit.
I'd have much preferred that they simply stuck with XFCE and dumped GNOME3 from the repos until the relevant developers got over their NIH, one-true-way, ewontfix corporate nu-linux mentality. Then we could have kept our freedom to choose our low-level system tools, and those who want a good whacking with the UX-braindamage bat could just use fedora...
But that would have required balls, which apparently the Debian leadership lacks.
Debian has never been primarily a desktop distro, and railroading everyone running it on servers for the sake of leg-humping a DE project they don't want was a right slap in the face.
Devuan did the sensible thing and kept XFCE as the default desktop, because unlike certain others it doesn't try to dictate your choice of init system.
Hi folks,
Anyway sysv stinks hence I am thinking to use OpenRC...
I would hesitate to say stinks, but it can be a bit clunky, especially with all of Debian's mods (insserv etc) on top.
Really though, sysv is just a bit old and a bit crusty. It still does what it was meant to do, and does it tolerably well.
I haven't tried openrc on Devuan yet, but I can say it's pretty nice with Gentoo. Small, simple, cleaner configuration than sysv, and it just does what an init system should without getting in the way.
If and when Devuan ships native openrc init scripts instead of relying on sysv compatibility, it'll likely be as good as it is on Gentoo. Running openrc with sysvinit as pid1 and a bunch of sysv scripts makes no sense to me, I might as well run sysv.
I am still making my tests... I am still wondering if it is worth encrypt the disks for a domestic computer, even if it is a laptop. Complex scheme partitions, encryption, lvm make difficult clone your disk setup or repair your computer to recover your data if something get wrong.
I for one absolutely take the KISS approach to storage. Encrypted LVM does what it says on the tin, but it also adds a couple extra layers between you and what's on your disks - complicating, as you say, data recovery when (not if!) a drive dies.
As for full-disk encryption in general, I've never really seen the point. If I have sensitive data to protect, I'll encrypt that filesystem only and skip the overhead for the rest of the system.
TBH I really don't care if someone stealing my machine can read /usr/lib or not, it's irrelevant. Just, you know, don't store the keys to your encrypted /home or whatever on the unencrypted /.
Very strange.
Indeed.
Once you get in (or if you image the disk before reinstalling), it would be interesting to see the apt policy and debsums for those packages... Libraries don't change themselves (this is stable, not ceres, right?), barring some proper odd variety of disk corruption.
I'm not sure exactly what you mean by "remote dpkg install" here, but all you should need for a look around is a working mount command, the minimal live image would do nicely.
Incidentally, if that's some kind of thin terminal and you have a bunch of identical installs, you might like clonezilla's bare-metal PXEboot image restore features. It sure beats lugging HDDs around IME.
That error regarding libmount.so.1 is almost certainly the culprit, and such a library version mismatch should never happen unless a n upgrade was forcibly interrupted or one has mixed releases and/or otherwise messed about with sources.list.
You might be able to fix it by installing the correct versions of the relevant packages (mount & libmount1) from a livecd chroot, but YMMV. Depending on how the system ended up in this messed up state any number of other things may be borked as well.
I'd like to give you my reason why I would prefer the graphical installer if available: I can copy/paste passwords and encryption keys to make sure they are the same.
Well, the netinstall image doesn't include any mouse support, which I guess is a shame (but hey, it's tiny)... So some extra keystrokes and/or gratuitous use of cat might be needed.
Both the graphical and minimal live images can install the system with refractainstaller though, and they both have mouse-driven copy-paste, so there's a viable option if you run full disk encryption or something (what else?) that needs input of long strings to the installer.
golinux wrote:Note that this option is no longer available in Beowulf and probably beyond.
Heh, I didn't even notice!
Neither did I. If I need an installer I'll pick the TUI one every time, and frankly I don't understand what a GUI provides in that context besides a bigger install image and more to go wrong. Starting an X server and all the baggage that entails just to have a pretty installer nuts.
On the OP at hand:
As the patient does not belong to the tap-and-drool selfie-stick demographic, I suspect this is simply a case of classic CLI-o-phobia. For the unfamiliar, a disorder brought on when suppressed memories of CP/M, DOS, the Windows command prompt and other shells designed to torment their users are brought to light, often upon encountering a similar text-interface aesthetic years later.
For this I usually prescribe the LFS treatment, followed by a light regime of Slackware or Gentoo use. Symptoms generally resolve in short order given sufficient exposure to bash/zsh, the GNU utilities and a proper POSIX environment.
The Debian Ncurses installer is better laid out than the Slackware Ncurses installer, making it just a little easier to navigate, but both are fast, functional, and fine even in 2020.
Indeed. Hell, the Gentoo "installer" is perfectly fine in 2020.
Ascii did have a nice gui installer, cut and paste options with the mouse.
Why would you need copy-paste in the installer anyway?
I mean you can use GPM at the console, but why?
It seems like the only thing holding it up is because they think that using a systemd tool to manage what looks like one temp file is a good idea.
Pulling in over a million lines of code to create a single directory. ![]()
Do we need a better example of the install-bloat and complexity-creep problems systemd creates?
I heard about something called opentmpfiles that might be able to manage this. All that would be needed then is a dummy systemd-tmpfiles packages to allow the Debian package to be installed, if it's truly a drop-in replacement.
As a stopgap until this is patched out in the devuan repos, you can whip up such a package with equivs and a symlink. Or just use the POC helpfully provided here.
It's quicker than patching and rebuilding php-fpm locally anyway ![]()
Opentmpfiles should be a drop-in, so long as systemd and the freedesktop crowd actually adhere to their own standards...
From his response to the various bug reports on this, and his and flat refusal to consider such a trivial change even when provided a patch, I guess Ondřej Surý is going on my personal "do not bother contacting, has irreconcilable attitude problems" list.
So long and thanks for all the fish, deb.sury.org. It was convenient while it lasted.
I think that it is somehow related with the systemd issue.
Indeed it is, but that would be off topic for this thread. ![]()
Then tried to add to /etc/modules: the entry vboxdrv was a game changer, the error when starting my Win7 VM changed. I saw that I needed to enter also vboxnetadp and vboxnetflt, and now it works.
To be fair you shouldn't have to, something must be going wrong in the byzantine mess that is oracles init script...
The init script that I didn't see on cursory examination of the package because it's copied from /usr/lib/virtualbox by the byzantine postinst-common script, which is in turn called from another postinst script...
The init script that takes 20KB of shell to load 3 frickin modules, and clearly doesn't manage to do it reliably.
Sheesh, commercial vendors and their overengineering.
You could try debugging it, or you could just stick with those 3 lines in /etc/modules... I know which one I'd choose.
Can't remember that I had to do this before, e.g. under Squeeze, Wheezy, or ASCII. Anyhow, I think I have got a solution now.
VirtualBox was in the repos back then (It was dropped for buster due to oracle being "uncooperative" IIRC), so presumably you were using those packages rather than the ones from oracle, no?
All the nice hints found do not fix the issue.
What hints? The very useful one that VirtualBox provides when you try to start a VM without the kernel module loaded?
When I follow the instruction I can work with VBox
By "the instruction" you mean
modprobe vboxdrv?
If that works then the kernel module has been built just fine, it simply needed to be loaded.
the game starts next day when I have shut down the PC the evening before and restarted or past any reboot.
What is going on here?
Unless you have configured the system to load that module at startup (i.e. in /etc/modules or some init script), you will need to load it when you want to use it.
What's unexpected about that?
ED. The repo at
https://people.debian.org/~lucas/virtualbox-buster/is probably a better option than Oracle's .debs. It looks like it includes virtualbox-dkms for automatic rebuilds against new kernel versions, as well as an init script to load the modules at system startup.
I don't see anything in there that would make those Debian buster packages problematic on Devuan beowulf, but as always, YMMV.
Hi Steve
...
Well that all looks fine. Rats.
Does the root account actually have a password?
I was a bit of confused that I should work from now on with "su -".
It's just another upstream solution in search of a problem. ![]()
'su' no longer adds /sbin and /usr/sbin to your $PATH, so you will get "command not found" for any binaries that live there.
'su -' or 'su --login' will read root's profile, so you get root's $PATH, which does include /sbin and /usr/sbin.
Stick alias su='su -' in your ~/.bash_aliases and forget about it.
I have no idea how to check up why I get that "ash" output.
Any suggestions?
Sounds like some cruft in root's .profile / .bashrc etc?
What does env say after you su to root, and what's in root's .bashrc, .profile, .bash_profile and .bash_aliases (if they exist)?
Being able as normal user to get a root-shell by just typing su "--login" without need for a password makes me a little nervous.
Not you?
Smells like a pam screwup to me (assuming it's not some new and wonderful default).
What do /etc/pam.d/su and /etc/pam.d/su-l contain?
You should not be able to su of any kind without a password unless you are already root, or you have something like "auth sufficient pam_<something>.so trust" in your pam su config, which would be very silly.
Unfortunately allmost every gui to day are using pulseaudio, and thereby reducing latency for sound (it makes the sound worse). I know only two current gui programs that works with alsa. Volumicon-alsa for gtk (wich I use myself) and as far as I know it qasmixer for qt.
Fortunately almost every GUI today (with the notable exception of anything related to GNOME) works just fine with plain ALSA... So long as you don't compile it against pulseaudio, which unfortunately Devuan has, and it doesn't assume pulseaudio is available because libpulse0 was installed, which unfortunately Devuan does.
That will probably happen eventually. We just have to be patient. The team have a lot on their plate.
EDIT: Or maybe not... https://dev1galaxy.org/viewtopic.php?id=2934
I'm going to go with "not".
Pretty much every sound-related application in the repos which can be linked to pulseaudio has been...
Because apparently upstream Debian just loves needless bloat these days, and dubious "features" like lobotomising your ability to configure your system properly by burying everything behind a GUI and a wall of XML or JSON (then adding a daemon or two and a dbus interface for good measure) is progress.
This is one of the reasons I'm not running Devuan (or any "modern" binary distro for that matter) on my desktops.
A minimal server install is still quite feasible, but as soon as you want a usable web browser or a DE, the tendency for distro maintainers to compile in everything including the kitchen sink bites you in the ass.
Devuan:
apt install <pretty much anything>
...
I see you're installing $software, would you like bloat with that? Too bad, we compiled everything we could find with '--install-lennart --enable-bloat --enforce-more-latency --everyone-uses-laptops --bug-policy=redhat'
The following 900 packages you don't need or want will be installed...
Gentoo:
USE="-pulseaudio alsa" emerge --newuse @world && emerge --depclean
...
Bask in the complete and satisfying eradication of libpulse, Firefox without nasty LD_PRELOAD hacks, XFCE with full ALSA support, etc. etc. ![]()
Because the primary target is average Joe and they readily suck up marketing and empty promises.
I kinda hoped most Devuan users were smarter than that. Aren't we here to get away from marketing and empty promises?
If I saw this request over on the *buntu forum I'd understand, but someone caring enough about systemd, KISS/UNIX principles and/or the redhat/corporate encroachment to be running Devuan ought to be well wary of such hype no?
Even with Firefox, the target is windows/android/apple.
I actually have some faith in Firefox and the Mozilla foundation, if only because they're the only independent browser vendor without a glaring conflict of interest... That and Netscape/Mozilla was once the only real option on GNU/Linux.
We're not the priority, sure. But they gave us a usable modern browser back in ~2002, and it's still all that.
We're obviously never going to agree.
We certainly won't if your only interaction with this community is to drop in and say "Devuan is too hard, I'm going to use Mint."
Feel free to use whatever distro you want. If no-reading one-click installs are your thing, Devuan probably isn't for you.
If you have questions about Devuan, we're happy to provide answers.
If you have a problem to solve, we'll probably enjoy helping you.
If you have constructive ideas on how to make Devuan better (ideally with some code to share), we'll gladly discuss them.
If you're just here to (passive aggressively) complain and make pointless comparisons to $newb_friendly_distro though, I'm pretty sure this conversation is over.
Perhaps I'm missing something but if you don't opt in to those schemes then it seems to me there's not a lot of difference to Firefox etc.
Indeed. AFAICT it's nothing more exciting than YACB (Yet Another Chromium Build). So why all the yammering about it, like it's the hottest new thing?
What is with all the noise about this "brave" browser recently anyway? Is it demonstrably better than Firefox/Iceweasel or a degoogled chromium build?
Personally I'd be highly suspicious of a new browser that is advertised as heavily as this one appears to be...
Interesting article about Brave linked on lobste.rs:
https://practicaltypography.com/the-cow … brave.html
Ed. Yup, that's pretty much what I expected. "Have our ads and tracking instead of theirs". Free cryptocurrency woo and virtue signalling included.
No thanks, I'll stick to my adblockers and javascript defangers.
Double click install.
Oh dear, this is much too complicated for me.
The Devuan installer is almost identical to the Debian installer, which is thouroghly documented and has served well for many years.
Comparing Devuan to Mint/Ubuntu is disingenuous, they have very different intended audiences.
One focuses on being beginner friendly, the other on flexibility and reliability. This is why they have an easy-mode installer, while ours has lots of options. Making the install process like Ubuntu would exclude a large number of possible configurations.
If the flexible and configurable installer is "too complicated" for you, perhaps you should read the manual?
If that doesn't answer your questions, there's always the Debian install guide, which is 99% applicable to Devuan.
I'm not a newcomer having started out with Ubuntu 6.06
If Ubuntu/Mint is all you have used since then, you're still a "newcomer" to Devuan.
Both of those distros are noob-friendly Debian derivatives targeted squarely at those who are allergic to the command-line, and they exist almost entirely to make Debian (and thus Devuan) less complicated.
Gentoo
Your plan is to upgrade to AMD YD195XA8AEWOF Ryzen Threadripper 1950X ?
Nah, got down on a used Supermicro X9DRL-iF + 2x E5-2650 for cheap. Just waiting on some more RAM (also cheap) and non-obnoxious coolers. Damn virus slowing down shipping and all that.
Ryzen would be nice and all, but the price was right in this particular case, and I want the dual LAN, IPMI & 10x SATA ports more than I want another AMD machine. Haven't owned one of those since 2001. ![]()
For this box, where raw performance is less important than reliability and expansion options, I'm fine with running a couple of generations behind the cutting edge. Used server parts are readily available and IME far more reliable than even new consumer-grade stuff.
Then of course there's the small matter of scoring a board, 16GB of RAM & 2x CPUs for ~1/3 the price of that shiny AMD CPU alone...
My first Linux kernel compilation was on a 486 with 4 MB of RAM.
Yeah, that'll take a while. ![]()
Dunno if a modern kernel would build on that, it has grown some. Guess I try it next time I have a free week to watch paint dry...
Now that I've finally upgraded the old under-desk battleaxe to Beowulf:
..,,;;;::;,.. steve@redacted
`':ddd;:,. ---------------
`'dPPd:,. OS: Devuan GNU/Linux 3 (beowulf) x86_64
`:b$$b`. Model: X7DCL-3
'P$$$d` Kernel: 4.19.0-9-amd64
.$$$$$` Uptime: 11 days, 19 hours, 45 minutes
;$$$$$P Packages: 1537 (dpkg)
.:P$$$$$$` Shell: bash 5.0.3
.,:b$$$$$$$;' Terminal: /dev/pts/3
.,:dP$$$$$$$$b:' CPU: Intel Xeon X5460 (8) @ 3.167GHz
.,:;db$$$$$$$$$$Pd'` GPU: XGI Technology Inc. eXtreme Graphics Innovation)
,db$$$$$$$$$$$$$$b:'` Memory: 40164MiB / 48296MiB
:$$$$$$$$$$$$b:'`
`$$$$$bd:''`
`'''`The best bit: 32TB storage ![]()
Soon to get a hardware update to 16 cores & 64GB. Those old dual Harpertowns are a power hog.
I played DOOM first time in about 1994 too on a i386 with 8GB of RAM.
I think you mean 8MB, LOL. My 386 had 4, and it sure couldn't run Doom. 16MHz CPU, hercules monochrome graphics, and a spacious 80MB HDD.
Another dumpster special, and the first PC I could call "mine".
I still play (GZ)Doom from time to time, and it's still awesome. Fairly sure I have enough WADs to last me until the heat death of the universe too:
~ $ du -ch `find Games/Doom/Mods -iname '*.wad'` | tail -1
3.4G totalAside, I highly recommend timidity++ & the soundfonts from arachnosoft for full appreciation of Dooms epic soundtrack. ![]()
When we're happy with the installation media, we'll officially call it Stable and release it. Nothing else in beowulf needs to be changed.
Good to know. I didn't see this info anywhere official, so thanks for the heads-up.
May I know, why did you try to use that old hardware?
Because at the time (~2000 - 2003), a hopped-up 486 (120Mhz CPU, unofficial 40MHz bus, 2(!) VLB cards, 32MB RAM) junk-bin special was all I had.
The more recent install was simply a case of finding I still had enough old parts to assemble a retro box, and doing so for the hell of it.
IBM branded SFF 486 desktop, circa 1994. All the right bits, and worth a surprising amount nowadays as they're starting to become "vintage". Runs a stripped-down Gentoo, dual booting PC-DOS for DooM and Duke Nukem.
Ed. One of these. Top spec 100MHz model.
Is there any difference between i486 vs i586 in terms of the purpose why you tried this except Pentium having much more RAM available?
Plenty. The pentium brought many new instructions, as well as widespread PCI bus support.
Multimedia was especially painful on 486-class boxes, if you wanted to play MP3s at realtime you pretty much had to compile mpg123 with special optimisations. MMX changed all that.
Unless you're talking about really specialised embedded boards, there's no reason to use a 486 for anything but a nostalgia trip. i586 is better in every way... Well, except for that embarrasing FPU bug in the early pentium chips of course. ![]()
i486 is something about 16-64 Mb of RAM and on some boards a few MBs more?
Most late-model 486 boards supported 64MB, but finding compatible SIMMs wasn't particularly easy. 32MB was a lot more common.
I believe there were some server boards that supported 128MB, but I've never actually seen one.
16MB was considered a standard desktop for most of the era, and if you had more than that you usually had to map an address-space hole for the video memory... So although mine had 32MB installed, I could only use 30MB due to my 2MB graphics card. Had to bugger about with kernel patches to get that working too IIRC.
Then they are both good for me only to play with them in VM/chroot.
Personally I don't mind a rolling release as the primary (and only) OS on my desktop, and Gentoo has served me well in that capacity.
The customisation potential isn't just for special-purpose deployments or ricing either, it's being able to have the features you want, rather than what the distro maintainers thought would suit everyone. It's like having your own personal distro, and IMO that's a grand thing on a personal desktop.
FWIW there are two Gentoo branches, and the "stable" one sees very little breakage... Well, less than I remember from Arch anyway.
Each to their own though, I certainly understand the desire for a stable release and I sure wouldn't give a Gentoo box to my gran or deploy it in an office.
I would use Devuan for those, but frankly being a year behind Debian is too stable for me. It just makes too many compatibility problems.
In Arch I can find something new to try, test, check how it works, and then try to build on Devuan or wait when it is done by someone else.
In Gentoo I see some github repo I like the look of, whip up an ebuild in a few minutes, and add it to my local portage tree.
Or just add someone else's overlay.
Gentoo seems good for me to build something small and rare like hardened kernel by @anthrax for currently relatively rare architecture like i586.
Indeed it is. I actually installed current (as of a couple years ago) Gentoo on a 486 DX4-100 not so long ago. It's one of the very few distros that can easily be built with i486 (or i586 for that matter) specific optimisations, LFS + custom scripts being the other common choice. Architecture-specific optimisations make more difference than one might expect on old hardware.
Then I can try to build binary packages in Gentoo chroot and install them to physical old hardware?
Or set up the build box as a network binhost, or use distcc. I used to use distcc way back when my main desktop was a 486, with a mighty Pentium 2 downstairs pitching in for compiles. ![]()