onefang | And the pros and cons of HTTPS is a #debianfork discussion. The support question was answered. | 00:00 |
---|---|---|
some_alex | onefang: got it, thanks | 00:01 |
systemdlete | Here's something kinda yeccchy: sh: 0: getcwd() failed: No such file or directory | 00:30 |
systemdlete | this happened while I was cd'd into a directory under /sys | 00:30 |
systemdlete | once I cd'd back out to, say, root home, the problem stopped. I was trying to figure out what is causing an error message at boot. | 00:31 |
systemdlete | The error I have been seeing at boot (even after a cold reboot) is usbhid 1-5.2:1.1: couldn't find an input interrupt endpoint | 00:31 |
systemdlete | It turns out the device is my mouse, which is wireless and connected via my keyboard, through a switchbox I paid a bit of money for. | 00:32 |
systemdlete | But I don't notice any problems with the mouse, or the keyboard, or even the switch (which also runs on USB, btw) | 00:33 |
systemdlete | This is getting... kinda frightening. | 00:33 |
systemdlete | This is devuan ascii, and I've updated/upgraded as of yesterday. | 00:35 |
systemdlete | correction: Seems that there is a new linux image and headers... but why didn't these upgrade? I had to manually install them, which is odd since they are already installed | 00:39 |
ullet | questions questions | 00:42 |
systemdlete | I know, I know... LOL | 00:42 |
systemdlete | (wanna buy some linux? Hot copies, so you have to promise not to tell anyone, ok?) | 00:43 |
systemdlete | I'll try rebooting into this new image and see if the error disappears... | 00:45 |
* ullet ponders | 00:46 | |
MinceR | are you pondering what i am pondering? | 00:46 |
DonkeyHotei | i think so, brain, but wouldn't the salmon just swim upstream anyway? | 00:48 |
MinceR | :) | 00:49 |
DonkeyHotei | narf | 00:53 |
systemdlete | Still getting that error. I'll check my logs and see if they've been showing prior to recently | 01:29 |
ullet | maybe drive is going bad? | 01:33 |
systemdlete | ullet: The hard drive is RAID1, so... | 01:59 |
systemdlete | I just rebooted a couple of times and I have not seen errors from fsck on the way up | 02:00 |
systemdlete | I guess I could force a fsck on root, in case that was missed | 02:00 |
fsmithred | to always get the newest kernel for the release, install linux-image-amd64 (or other arch) | 02:00 |
fsmithred | same for headers | 02:01 |
systemdlete | That is already the case, thanks | 02:01 |
systemdlete | so that is not it | 02:01 |
fsmithred | maybe it has to be dist-upgrade to get a newer version | 02:01 |
systemdlete | but anyway, I am more concerned about the usb problem. I can see from my system logs that the system was rebooted (by me) on Apr 27 and Apr 28. The dmesg log for the 27th and prior do NOT show this error (usb endpoint missing), but 28th forward all show it. | 02:02 |
systemdlete | No packages were added or upgraded in the time in between. | 02:02 |
systemdlete | The only thing I can think of is the usb kvm switch. I may have installed it right around that time, but I don't remember exactly when that was now. | 02:03 |
systemdlete | Maybe the introduction of that switch is generating the reference to the mouse... the information about this switch says it does not support wireless mouse. But I don't see how that can matter given that the mouse is running off the keyboard. But it does say that. Funny how I have not had any problems with *usb* until today (other than the usual usb weirdness that does no harm) | 02:04 |
systemdlete | sorry, it says | 02:05 |
systemdlete | does not support keyboard mouse... hmmm. | 02:05 |
systemdlete | I wonder if that is what they mean. Does this imply I must use a wired mouse to the keyboard? | 02:05 |
* systemdlete is asking a bit rhetorically here | 02:06 | |
systemdlete | https://www.amazon.com/dp/B075KBCPKZ?tag=duckduckgo-ffab-20&linkCode=osi&th=1&psc=1 | 02:07 |
systemdlete | "FJGEAR 4 Ports USB 2.0 HDMI KVM Switch Keyboard Mouse Switcher for PC, Windows, (Wireless mouse keyboard is not supported,No power)" | 02:07 |
systemdlete | my keyboard is wired. The mouse is not. The mouse dongle is plugged into the keyboard. | 02:07 |
systemdlete | So I am not sure if I am in violation or not. | 02:08 |
systemdlete | It depends on how "wireless mouse keyboard" is defined. | 02:08 |
systemdlete | The thing is... it works. Had not noticed the boot error until today | 02:09 |
systemdlete | Was investigating a different problem that was occurring in a VM, which may be related to Zoom. | 02:10 |
ullet | try swapping mouse | 02:10 |
systemdlete | The VM, btw, is running Ascii also, so maybe you can help me there, or at least give a listen. | 02:10 |
systemdlete | ullet: ooh ooh. Good idea! | 02:10 |
systemdlete | (but I'll have to reboot) | 02:10 |
systemdlete | +1 | 02:11 |
systemdlete | I have another wireless mouse around here I think | 02:11 |
systemdlete | And a wired mouse also | 02:11 |
fsmithred | can you plug the dongle directly into the computer instead of into the keyboard? | 02:12 |
systemdlete | anyway, Zoom was crashing badly and kern.log shows stack traces | 02:12 |
systemdlete | Oh, the mouse works | 02:12 |
ullet | zoom appears to be a new thing | 02:12 |
systemdlete | and it works through the switch. What are you trying to have me test, fsmithred? | 02:12 |
ullet | therefore it is bad | 02:12 |
systemdlete | LOL. You are probably right, ullet. | 02:13 |
ullet | :) | 02:13 |
fsmithred | test to see if it makes the warning go away | 02:13 |
systemdlete | It *is* proprietary, so... | 02:13 |
systemdlete | ok, fsmithred. I can try that also when I reboot. I'll be sure to take notes on all these tests. | 02:14 |
systemdlete | But if I solve the riddle, then I deserve to help solve the Covid problem. | 02:14 |
systemdlete | because it seems to be almost as difficult. | 02:14 |
ullet | that is in people's heads | 02:14 |
systemdlete | What is it about lsusb? The tree option doesn't show the device descriptions. That would be helpful right now. | 02:15 |
systemdlete | and usbtree no longer works. | 02:15 |
systemdlete | ullet: More in the chest and lungs, actually. | 02:15 |
onefang | Some of these switches fake info, so that when you switch your mouse and keyboard to another computer, the original still thinks it has them plugged in. | 02:17 |
systemdlete | onefang: Exactly, yes. That is so the quiescent boxes won't think they've lost the head. | 02:18 |
* systemdlete thinks deeply about wireless mouse attached to keyboard, via the switch... | 02:20 | |
systemdlete | This actually makes some sense, now that I think about it. Linux tries to find an endpoint for the mouse, but the actual mouse is in front of the switch; I am thinking maybe it cannot query it -- it will query the "fake" kb and mouse. | 02:21 |
systemdlete | Howeer, it is worth noting that lsusb does show the wireless mouse and correctly identifies it as a logitech, etc | 02:21 |
systemdlete | well, I'll know more after these tests | 02:22 |
fsmithred | so the fake info contains real content | 02:24 |
systemdlete | apparently copied from the real "content" | 02:26 |
systemdlete | but Linux does not squawk about the keyboard | 02:27 |
systemdlete | but first, lunch | 02:27 |
systemdlete | or dinner | 02:27 |
systemdlete | then more testing | 02:27 |
systemdlete | thanks for all the suggestions -- that's very helpful | 02:28 |
systemdlete | So... I tried a number of different combinations of different mice connected to the kb, the pc, and so forth, and one thing I noticed was that, no matter which combination, I ALWAYS got the boot error about missing endpoint for the mouse. But how could it be just any mouse, wired, wireless, directly connected, etc... so then I got to thinking | 04:24 |
systemdlete | I booted the pc without ANY mice attached whatsoever! | 04:24 |
systemdlete | And STILL the error message appears. | 04:24 |
gnarface | usb mice? | 04:24 |
systemdlete | Soooooo. I think I conclude that the switchbox itself is the issue, though apparently not fatal. | 04:24 |
systemdlete | yes, all usb mice | 04:25 |
gnarface | oh, and a usb switch? mabye | 04:25 |
systemdlete | the problem is on a usb switchbox, so yes, usb mice. all. | 04:25 |
systemdlete | usb switch yes | 04:25 |
systemdlete | gnarface: On the amazon web page for it, it even says does not work with wireless mouse keyboard, though I am not sure what that means precisely | 04:26 |
systemdlete | wireless mouse attached to kb, or wireless mouse+kb | 04:26 |
systemdlete | but either way, I am guessing I can look forward to this issue for the near future. | 04:26 |
systemdlete | Not sure if kernel devs can handle this or not. It could be an actual hw flaw. | 04:27 |
systemdlete | But I feel a little more sane now that I've narrowed it down to the switch and/or kernel | 04:27 |
systemdlete | I think I will write to the mfr and see if they can offer some insight or confirm this | 04:28 |
gnarface | systemdlete: i do recall hearing of some type of fakey cheap usb kvm switches that were on the market that needed specifically a windows-only feature to work right, but i don't recall much specifically about it | 04:36 |
gnarface | it wouldn't fail completely, it would just critically fail at the one feature that made it useful | 04:37 |
systemdlete | Well, this one was recommended by a fellow Linux user and it was not cheap. | 04:37 |
gnarface | oh hmm. well that is maybe not the issue then | 04:37 |
systemdlete | OTOH, I also figured something out | 04:37 |
systemdlete | the ad says "no power" -- so I think what they are saying is that since the switch is not powered, it can't talk to a wireless kb/mouse | 04:37 |
gnarface | oh is this ascii? maybe it's a kernel version issue, maybe there's quirks and they've been added to a newer kernel but not the ascii one | 04:37 |
systemdlete | So I think my hardware config is fine -- it must be, since it does work... | 04:38 |
gnarface | it *might* just mean that since it's unpowered some wireless devices might not work? | 04:38 |
systemdlete | the mouse supplies its own power by its battery. | 04:38 |
systemdlete | right | 04:38 |
systemdlete | there is some risidual power in usb that can power a keyboard I think | 04:39 |
gnarface | yea, i don't know the numbers, but it's probably not a good idea to try to use it to charge anything | 04:40 |
onefang | If the switch is unpowered, then it gets it's power from the computer, and it can ask for more power and distribute that downstream to devices. Assuming the computer can supply extra power. | 04:40 |
onefang | It might not do that if it's just assuming typical mouse and keyboard. | 04:41 |
gnarface | i would worry about frying stuff accidentally | 04:41 |
gnarface | it should be safe, one would think, but it seems risky | 04:41 |
gnarface | like, if it tried to charge a iphone would it trickle or would the wires get hot? | 04:42 |
gnarface | i would assume the former but also not risk it | 04:42 |
onefang | Modern USB stuff sorta negotiates power levels, but it's tricky with all the various versions, and cables may or may not support some of that. | 04:42 |
onefang | My keyboard isn't powered, but has a small USB hub in the back to support two other devices, like a mouse and graphics tablet. Should be plenty of juice for that sort of thing. | 04:44 |
systemdlete | Yeah, I have a Happy Hacker (my 2nd one after 20 years with the first!) and this new one has 2 USB ports. I attach the wireless mouse to the kb, since the mouse is powered by a battery. | 06:18 |
systemdlete | The KVM does have a (5mV?) power port, but the adapter does not come with it. However, I have not needed additional power so far, and hopefully I won't. | 06:19 |
systemdlete | I found a matching adapter I have here, just in case I need it. | 06:20 |
onefang | Mines a Happy Hacker Lite 2. B-) | 06:21 |
systemdlete | mine too! | 06:21 |
systemdlete | $200 | 06:21 |
systemdlete | I got 20 years out of the first one. The letters on the keys had become worn, and the keys themselves had sort of melted a bit... | 06:22 |
systemdlete | I think I got my moneys worth from it. | 06:22 |
systemdlete | heheheh | 06:22 |
systemdlete | I only paid $100 for the first one | 06:22 |
systemdlete | I cleaned the old one about 3x over the years, and it still worked like new. Just couldn't see the keys anymore | 06:23 |
systemdlete | letters | 06:23 |
nemo | hmmm I'm on a beowulf devuan, and I was trying to build chafa | 17:32 |
nemo | so I installed libmagick++-d | 17:32 |
nemo | ev | 17:32 |
nemo | everything built and linked correctly | 17:32 |
nemo | but, on run it claims the libs are missing | 17:34 |
nemo | checking in symaptic I see there's a couple of variants of libmagick++ | 17:34 |
slvr | maybe build and execute are using different paths | 17:35 |
slvr | I think you can run ldd on the bin and see the paths to all linked libs, and if they are missing. | 17:35 |
nemo | I appear to be using libmagick++-6.q16-8 | 17:35 |
nemo | the linker expects libMagickWand-6.Q16.so.3 | 17:35 |
nemo | q16-8 provides: | 17:36 |
nemo | /usr/lib/x86_64-linux-gnu/libMagick++-6.Q16.so.8 | 17:36 |
nemo | /usr/lib/x86_64-linux-gnu/libMagick++-6.Q16.so.8.0.0 | 17:36 |
nemo | I see there's a q16-7 (probably provides so.7) | 17:36 |
nemo | aaaand a libmagickcore-6.q16-3 which might have something to do with 3 | 17:37 |
nemo | slvr: I was just puzzled how I got into this situation in first place. normally debian/devuan ensures libs I'm linking to actually exist | 17:37 |
nemo | eh. lemme do a purge and rebuild | 17:37 |
nemo | maybe some stale cache | 17:37 |
nemo | yep. guess it was that. nevermind | 17:42 |
slvr | woot | 17:43 |
ullet | it seems more fun to move to ceres, for me | 17:51 |
mason | For the life of me I can't get LightDM to show a background, either per-user or global. The config is straightforward, but the convoluted pile is fighting me. | 20:45 |
fsmithred | what convoluted pile? | 20:51 |
fsmithred | can't you just set the bg image in the config file? | 20:51 |
mason | fsmithred: I tried with a couple syntax examples I found. | 20:51 |
fsmithred | are you messing with the alternatives system? | 20:52 |
mason | Evidently it can/does/? depend on an AccountsService thing I'd never heard of before today. | 20:52 |
fsmithred | that name sounds familiar | 20:52 |
mason | No, I've left all the package-installed stuff untouched. | 20:52 |
fsmithred | cool | 20:52 |
mason | I found this that seemed applicable: https://github.com/linuxmint/slick-greeter/issues/94 | 20:52 |
mason | (LightDM/slick-greeter I should say.) | 20:53 |
mason | I'll try again with the gtk-greeter and see if that does any better. | 20:53 |
fsmithred | I've never done per user bg. Don't see how that would work. | 20:54 |
fsmithred | how would it know which user is going to log in before they log in? | 20:54 |
mason | According to one creation myth, there's a per-user file in /var/lib/AccountsService/users/ that is populated with this, but I've not dug into what yet, largely because even with it set properly it doesn't display an image. | 20:55 |
mason | fsmithred: So, with slick-greeter, the user list is all there in a list, and whichever one is selected is the one whose background shows. It's slick when it works - Ubuntu does it. | 20:55 |
mason | It's not so slick when it doesn't work. :P | 20:55 |
mason | For myself, I use xdm, and my background pattern is the standard grey X stipple as God intended, but the kids like more bling. | 20:56 |
fsmithred | for lightdm, I've always changed it in the config file | 20:56 |
fsmithred | there's a line for background_image | 20:57 |
fsmithred | probably in the greeter config | 20:57 |
mason | Yeah. I'm trying again with lightdm-gtk-greeter and we'll see if that obeys. | 20:57 |
mason | It won't show the user list, but that's okay. | 20:57 |
fsmithred | there's a setting to show or not show user list | 20:57 |
mason | My four year old isn't reading just yet, so the theory was that seeing her background would help her figure it out. Starting with her typing in a simple password will be task enough. :) | 20:58 |
mason | In the gtk greeter, that works, but it controls a pop-up menu, not a list with everyone there on-screen at once. | 20:58 |
fsmithred | oh | 20:58 |
mason | fsmithred: Does setting background= work for you? It seems not to work here, regardless of whether I have it set in /etc/lightdm/lightdm-gtk-greeter.conf or /etc/lightdm/lightdm.conf.d/foo | 21:01 |
fsmithred | trying to find a VM that has lightdm | 21:01 |
fsmithred | in the past, it has always worked for me to just list the path to the bg image | 21:02 |
mason | This is on Beowulf, FWIW, as I expect they're different versions. | 21:03 |
mason | Hm, even if I install lightdm-gtk-greeter-settings and have it set the image via the gui, no love. | 21:06 |
fsmithred | installing lightdm now | 21:12 |
mason | Ah, cool - I'll be interested in seeing how it goes. | 21:13 |
fsmithred | ok, after install it uses the desktop-base bg without my doing anything other than selecting lightdm as default dm | 21:16 |
fsmithred | and now it's stuck and I have to start over | 21:17 |
fsmithred | don't restart dm in qemu | 21:17 |
fsmithred | use init 1 instead | 21:17 |
fsmithred | yeah, it works. I got a different bg | 21:21 |
fsmithred | my guess is you changed it near the top of the file instead of in the section that starts with [greeter] | 21:21 |
mason | doh | 21:26 |
mason | Which file? | 21:26 |
mason | trying lightdm.conf itself | 21:27 |
mason | Even putting it there, at the end, no luck. I'll purge and try again with a fresh config, just that one line added. | 21:41 |
mason | fsmithred: Which config file did you amend? | 21:41 |
eyalroz | Hello devuaners, | 21:44 |
eyalroz | How do I get rid of the message "udevd: Error running install command for nvidia" which I get on every boot of Beowulf? | 21:44 |
eyalroz | I've been led to believe I might need to disable something in /etc/modprobe.d/nvidia.conf, but I'm not sure what line. Or - perhaps I need to do something else entirely? | 21:44 |
ullet | idk. maybe a setup script | 21:46 |
ullet | a kernel module generally doesn't handle installing | 21:46 |
eyalroz | ullet: I should mention my nvidia card works fine (although I don't use it as a display adapter) | 21:47 |
eyalroz | I'm just annoyed by the error message. | 21:47 |
gnarface | eyalroz: best guess is that nouveau and nvidia are both fighting over the card. you have to pick one then blacklist the other. | 22:00 |
eyalroz | I have a "blacklist nouveau" in another .conf file in modprobe.d ... | 22:01 |
eyalroz | Also - none of that stuff was done manually by me, nor by a manual nVIDIA installer... | 22:01 |
eyalroz | I think those files came with the distro, | 22:01 |
eyalroz | ... so it's not exactly about what I have to do personally. IIANM. | 22:02 |
fsmithred | mason /etc/lightdm/lightdm-gtk-greeter.conf | 22:02 |
gnarface | eyalroz: oh, hmmm.... well the nvidia official drivers are in non-free, so if you didn't include non-free you wouldn't have got them even by accident. but if you're blacklisting nouveau already then my hypothesis is probably wrong. | 22:02 |
gnarface | eyalroz: (to be clear, my hypothesis was that nouveau was already loaded and then it was blocking nvidia's attempt to load after that) | 22:03 |
gnarface | eyalroz: is it a very old card? perhaps the error is about using the new drivers with a legacy card. there's also a legacy nvidia driver package in there for older cards | 22:03 |
Venker | hi people | 22:03 |
gnarface | eyalroz: or, alternately, if you're on ascii and it's a very new card, you might need to get the drivers&kernel from backports | 22:04 |
gnarface | Venker: hi, just ask | 22:05 |
ullet | my vote for next release name is PETSCII | 22:05 |
Venker | I was coding a pull request or something like that for https://git.devuan.org/devuan-packages/openstack-devuan-images | 22:06 |
Venker | but I can't get it to work | 22:06 |
Venker | it complains about extlinux package not found, but it is | 22:06 |
mason | fsmithred: Alright, that did work, so now I need to pick apart how I was doing it to see what was breaking. Difference this time, I purged all the related packages and put them back in pristine, freshly installed state. Thank you. | 22:08 |
fsmithred | yw | 22:09 |
fsmithred | you saw what I meant about the right place to edit the file? | 22:09 |
fsmithred | it's a common mistake | 22:09 |
mason | Yes. | 22:09 |
eyalroz | gnarface: I do include non-free. But - maybe nouveau is being blacklisted "too late"? | 22:12 |
eyalroz | gnarface: I'm on Beowulf and it's a GTX 1050 Ti, so not that new. | 22:13 |
eyalroz | ... but not that old either. | 22:13 |
Venker | https://gist.github.com/pgmartinez/c5c8ae2ecf7d21aff53fdaadef4b1679 | 22:16 |
eyalroz | gnarface: perhaps I should try an unload command for nouveau in the same file as the nvidia load commands, on an earlier line? | 22:30 |
gnarface | eyalroz: uh... well maybe. it's easy to check by looking in the output of lsmod. afaik only one or the other can be loaded, though it's possible you've managed to end up with neither working right. fyi the 1050 *is* too old though for the stock ascii versions, i'm pretty sure | 22:30 |
eyalroz | gnarface: The nvidia drivers are what actually get loaded. | 22:31 |
gnarface | eyalroz: sorry i mean that the 1050 *is* too new for the stock ascii versions | 22:31 |
eyalroz | gnarface: But why would the stock ascii version matter, if I'm running Beowulf? ... and with CUDA 10.1 being installed? | 22:32 |
eyalroz | Also, GTX 1050 Ti was released in late 2016. | 22:33 |
gnarface | eyalroz: oh, sorry i missed that you said that. and actually if you're just using this for cuda and not using it as a display, i don't think it matters | 22:33 |
eyalroz | Using it just for computation. | 22:33 |
ullet | are you good with ze NN's eyalroz | 22:33 |
ullet | i'd like to upgrade old videos but it's OT here. sorry. | 22:34 |
gnarface | eyalroz: but fyi according to this, 1050 support wasn't added to the linux version of the driver until version 375.20, in november of 2016: https://www.nvidia.com/Download/driverResults.aspx/111596/en-us | 22:35 |
gnarface | eyalroz: (and the ascii version is frozen at 340.something, which is why i thought that was relevant) | 22:36 |
eyalroz | ullet: I'm actually exploring other aspects of GPU computing. NN's are very fashionable, I think they're not very challening in terms of putting the GPU to new uses. They work too well without you having to break a sweat, design-wise. | 22:36 |
ullet | i'd be happy if i could operate them properly. oh well. i can write shadertoys. | 22:37 |
eyalroz | gnarface: Luckily, beowulf has CUDA 10.1 with driver version 430.50 ... | 22:37 |
gnarface | eyalroz: did you get CUDA from the beowulf repo or is that something you have to install from nvidia's shell script thingy directly? | 22:38 |
eyalroz | ullet: I can't do NN work myself. I just know that it's not a challenge getting them to run efficiently on GPUs. Analytic databases, however, are a different story... | 22:38 |
eyalroz | gnarface: beowulf repo - I'm 95% certain. | 22:39 |
gnarface | eyalroz: beowulf is at 418.something, mabye you have some of two versions installed and that's the conflict? | 22:39 |
gnarface | eyalroz: do a "dpkg -l |grep nvidia -i" and see if you see 418.xx cruft mixed in there | 22:39 |
eyalroz | I see the two different versions in the apt package cache | 22:40 |
gnarface | we only care about installed versions | 22:41 |
eyalroz | the installed version is 430.50-1 | 22:41 |
eyalroz | Status: install ok installed | 22:41 |
gnarface | eyalroz: i'm not seeing any sign of version 430 in the devuan repos | 22:42 |
eyalroz | Maybe it's another sign of me having a somewhat messed up apt state again :-( | 22:42 |
mason | fsmithred: Ah, found one issue with accountsservice - it only provides a systemd service unit. | 22:42 |
gnarface | eyalroz: well, or mabye you got some stuff directly from nvidia and forgot? | 22:42 |
eyalroz | gnarface: Nope, the package maintainer is "Debian NVIDIA Maintainers" | 22:43 |
gnarface | hmm, weird | 22:43 |
eyalroz | and when I get stuff directly from nvidia, I get CUDA+drivers together and install CUDA to /usr/local/cuda-X.Y ; yet CUDA 10.1 is installed under /usr | 22:43 |
gnarface | eyalroz: but pkginfo.devuan.org says you should have cuda 9.2.x, i think. can you confirm? https://pkginfo.devuan.org/cgi-bin/d1pkgweb-query?search=cuda&release=beowulf | 22:46 |
gnarface | eyalroz: oh, 10.1 may have come from beowulf-backports... | 22:46 |
gnarface | eyalroz: i'm seeing it here: https://packages.debian.org/search?keywords=cuda&searchon=names§ion=all&suite=buster-backports | 22:46 |
eyalroz | gnarface: Very possibly. | 22:46 |
gnarface | eyalroz: but nvidia-driver in beowulf backports should be 440 now not 430, so maybe your backports are just out of date | 22:46 |
gnarface | eyalroz: remember to add "-t beowulf-backports" or it will ignore backported packages during upgrades | 22:47 |
eyalroz | My backports are out of date, since I commented them out at some point to resolve some apt issues you helped me with :-) | 22:47 |
eyalroz | Ok, apt-get upgrade time | 22:47 |
gnarface | eyalroz: yea that happens to often for me to remember the specific issue, sorry, but i would have definitely recommended to NOT just "apt-get -t beowulf-backports upgrade" | 22:48 |
eyalroz | gnarface: Sorry I took you on a bit of a wilf goose chase again | 22:48 |
gnarface | eyalroz: no worries, just don't apt-get upgrade all of backports. uncomment it, then upgrade just the nvidia packages (with -t beowulf backports), then comment it out again | 22:48 |
gnarface | eyalroz: (mabye add a comment about this in the sources.list so you remember next time) | 22:49 |
gnarface | eyalroz: what kernel do you have? you might need that too, not sure. the backports one should be at 5.4 and the stock one is 4.19 | 22:49 |
eyalroz | 5.2.0-3 | 22:50 |
gnarface | yea, so that's definitely also an out-of-date backports kernel | 22:50 |
eyalroz | gnarface: Turns out I also have a libc6 from backports. Ouch. | 22:53 |
gnarface | eyalroz: ouch. yea, that's specifically why i would have said not to just blindly apt-get upgrade the whole thing from beowulf and only just get the kernel and nvidia packages specifically (and any related firmware if relevant) | 22:53 |
gnarface | eyalroz: from beowulf-backports* | 22:54 |
eyalroz | gnarface: I had enabled backports when I first installed beowulf - a mistake on my part. Well, it looks like I'm going to have to lie in this bed after I've made it. | 22:56 |
fsmithred | you could pin beowulf-backports to -1 to downgrade | 22:58 |
fsmithred | dpkg -l |grep bpo | 22:58 |
fsmithred | to see what's from backports (plus a few false positives) | 22:58 |
eyalroz | fsmithred: Can I actually downgrade everything from backports to the non-backports versions? When that includes downgrading libc6 and all of its dependent packages? | 23:01 |
gnarface | eyalroz: according to Debian, downgrades aren't officially supported, however you can probably make it happen | 23:03 |
eyalroz | gnarface: Not sure it's worth the trouble at this point, but it's good to know. | 23:04 |
eyalroz | fsmithred: How to I do the pinning, exactly? | 23:05 |
fsmithred | eyalroz, I don't know what happens if libc6 is from backports | 23:15 |
fsmithred | I'll get you an example | 23:15 |
fsmithred | three lines in a file in /etc/apt/preferences.d | 23:15 |
fsmithred | Package: * | 23:16 |
fsmithred | Pin: release a=beowulf-backports | 23:16 |
fsmithred | Pin-Priority: -1 | 23:16 |
fsmithred | oh | 23:16 |
fsmithred | no | 23:16 |
eyalroz | no? | 23:16 |
fsmithred | pin beowulf to 1001 | 23:16 |
fsmithred | sorry - got a fever and toothache right now | 23:16 |
eyalroz | oh... I'm sorry to hear that! | 23:17 |
fsmithred | anything I say could be from another dimension | 23:17 |
fsmithred | first line is same - all packages | 23:17 |
fsmithred | Pin: release n=beowulf | 23:17 |
fsmithred | Pin-Priority: 1001 | 23:17 |
eyalroz | fsmithred: shouldn't I also pin beowulf-updates and beowulf-security to something higher then? | 23:17 |
fsmithred | probably | 23:17 |
fsmithred | yeah, that makes sense | 23:17 |
fsmithred | use n=beowulf, a=beowulf-whatever | 23:18 |
fsmithred | I think | 23:18 |
eyalroz | So, a 3-line entry for each repo? | 23:18 |
fsmithred | yes | 23:18 |
fsmithred | can all be in same file | 23:18 |
fsmithred | not sure if you need the -1 on bpo | 23:19 |
fsmithred | I'd try without that first | 23:19 |
fsmithred | apt -s upgrade to see what will happen | 23:19 |
fsmithred | how many backports packages did you end up with? | 23:20 |
eyalroz | fsmithred: Apparently, nothing happens when I apt-s upgrade | 23:20 |
eyalroz | oh, wait | 23:20 |
eyalroz | for the beowulf repo, should it be "n=beowulf, a=beowulf" ? | 23:21 |
fsmithred | n for name | 23:21 |
fsmithred | a for almost a name? | 23:22 |
eyalroz | fsmithred: I don't follow. Do I need an n= and an a= for each of three repos? | 23:22 |
fsmithred | no | 23:22 |
fsmithred | plain beowulf gets 'Pin: release n=beowulf' | 23:23 |
fsmithred | anything that's not exactly a codename gets a= | 23:23 |
fsmithred | that would be the other three | 23:23 |
fsmithred | -backports, -security, -updates | 23:24 |
eyalroz | Ah, ok. | 23:24 |
eyalroz | So, apt -s upgrade gives: | 23:25 |
eyalroz | calc-common : Breaks: apcalc-common (< 2.12.7.2-3~) but 2.12.7.2-2 is to be installed | 23:25 |
eyalroz | libogdi-dev : Breaks: libogdi3.2-dev (< 4.0.0) but 3.2.1+ds-4 is to be installed | 23:25 |
eyalroz | libogdi4.1 : Breaks: libogdi3.2 (< 4.0.0) but 3.2.1+ds-4 is to be installed | 23:25 |
eyalroz | E: Broken packages | 23:25 |
eyalroz | I could just remove these, maybe? | 23:26 |
fsmithred | maybe | 23:26 |
gnarface | it might work | 23:26 |
fsmithred | another thing that might work is to use aptitude | 23:26 |
fsmithred | it might give you alternatives | 23:26 |
gnarface | yea, it's made for figuring out this type of thing | 23:26 |
eyalroz | when I apt -s upgrade after the removal it comes up with another broken package | 23:27 |
fsmithred | that could go on for some time | 23:27 |
eyalroz | oh, but after removing that, apt -s upgrade has a super-long output | 23:27 |
eyalroz | with a zillion downgrades. success! | 23:28 |
eyalroz | Let's do this thing. | 23:28 |
fsmithred | cool | 23:28 |
eyalroz | and hope I don't get hosed. | 23:28 |
gnarface | make sure you have a live image ready on usb or cd | 23:28 |
fsmithred | yeah, good idea | 23:28 |
Venker | seeU | 23:29 |
eyalroz | gnarface: I think I have one somewhere; plus I have a laptop if push comes to shove. Now just need to wait for some 4.5 GB of downgrade package downloads. Thank you both (gnarface and fsmithred). | 23:29 |
gnarface | Venker: did anyone address your concern? i don't know the answer, sorry. maybe ask in #devuan-dev | 23:30 |
Venker | gnarface: I will, thanks. I think I did some progress regarding that code | 23:30 |
systemdlete | So I installed tlp, and that brought with it network-manager, which I do NOT want; I use wicd. So I uninstalled network-manager. But /etc/init.d/network-manager lingers. ??? | 23:44 |
systemdlete | It's not enabled, but still, why is it still there, even after I removed it with the apt tools? | 23:45 |
systemdlete | I tried re-installing it and uninstalling it. No dice. | 23:45 |
systemdlete | and why does installing tlp, which is power-related, drag a tool like network-mangler along with it? | 23:46 |
mason | systemdlete: When you delete things, you can purge them as well... apt purge foo, or apt --purge autoremove, etc. | 23:55 |
systemdlete | oh yes. I forgot about that. But why does it tell me that /etc/init.d/network-manager is owned by network-manager? Normally, if a package is not installed, I don't expect it to know anything about files owned by the package. | 23:57 |
systemdlete | but to my primary question: Why network-manager for tlp? | 23:59 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!