| ham5urg | I tried 'apt install qutebrowser -o Dpkg::Options::=--root=/pkgs/qutebrowser' with the directory already created. Got a bunch of configure-errors from apt. | 00:12 |
|---|---|---|
| gnarface | i'd guess you probably have to also specify other things besides the root, since the defaults for those things would probably be relative to root so it might be expecting a whole functional install at /pkgs/qutebrowser | 00:14 |
| gnarface | i don't actually know though | 00:15 |
| gnarface | if you're building the packages you would | 00:15 |
| gnarface | maybe it doesn't need the whole thing though, maybe it just needs /var/lib/dpkg, inferring from the dpkg man page | 00:16 |
| gnarface | or rather in your case, /pkgs/qutebrowser/var/lib/dpkg.. | 00:16 |
| gnarface | i mean /pkgs/qutebrowser/var/lib/dpkg | 00:16 |
| ham5urg | I could add a symlink from /var/lib/dpkg to /pkgs/qutebrowser/var/lib/dpkg | 00:17 |
| ham5urg | or linking just var | 00:17 |
| gnarface | i'm not even sure that is safe but you could certainly try it | 00:17 |
| ham5urg | But the installed dependencies wont be inside PATH so the package should break anyway | 00:19 |
| gnarface | you might need to symlink other stuff too | 00:19 |
| ham5urg | And even if in PATH, there is no guarantee that a pkg will not use the absolute path of a dependency | 00:20 |
| ham5urg | Better to stay away from such things. | 00:20 |
| gnarface | it would be safer to build the package over | 00:20 |
| gnarface | seems rebuilding the whole package is lot of work just to change the install location but the change itself is relatively simple | 00:22 |
| ham5urg | yes. I was glimpsing into guix and was impressed by their package management. | 00:23 |
| debdog | FTR, there's also "dpkg -i [ -root=$DIR | --instdir=$DIR ]" | 07:21 |
| gnarface | hmm, i wonder if the instdir part would have fixed this issue | 07:30 |
| libredev2 | do i can get any non free blob even after expert install. | 16:38 |
| libredev2 | or the system will be fully free | 16:39 |
| gnarface | if non-free is not in your /etc/apt/sources.list then the system should remain fully free unless you download and install something manually | 16:40 |
| libredev2 | ok | 16:43 |
| libredev2 | great | 16:43 |
| fifiopenbsd | hello gnarface, do you remember me, filipdevuan_? | 16:45 |
| fifiopenbsd | you gave me lots of valuable hints in the past | 16:45 |
| fifiopenbsd | glad you're still alive | 16:45 |
| golinux | libredev2: "Non-free firmware packages are available on all install media. These packages will only be installed if required by your wifi adapter. To avoid the automatic installation and loading of needed non-free firmware, choose the Expert install option during the installation process." | 16:53 |
| golinux | https://www.devuan.org/os/install | 16:53 |
| libredev2 | i was just confirming. | 17:00 |
| libredev2 | thanks | 17:00 |
| Kitty | is there a way to specify which device driver is used for a certain pci device? | 17:19 |
| Kitty | I have an ethernet device I want to use the vfio-pci driver, rather than i40e | 17:19 |
| alphalpha | Hi | 19:06 |
| debdog | Kitty: I have no clue, but have you tried just blacklisting i40e? | 20:50 |
| debdog | (assuming both drivers are compatible with your device) | 20:52 |
| used____ | to avoid tinkering with bl, rename i40e.ko to i40e.ko.moved, rmmod, then see if it works when inserting the other module. | 21:33 |
| used____ | if you reboot the moved module will be loaded, from ramdisk. To make it work across reboots, restore old .ko name and use blacklist | 21:35 |
| Kitty | debdog: no, cos I have two physical devices, and I want to use i40e on one, and vfio-pci on the other | 21:49 |
| used____ | Aiee. | 21:52 |
| used____ | That will require some deeper eudev rule poking. I hope your devices have serial numbers, distinct. Look in dmesg for them Kitty | 21:52 |
| Kitty | 0000:44:0.1 and 0000:44:0.0 | 21:53 |
| used____ | That does not sound like a s/n. MAC addresses perhaps. | 21:53 |
| Kitty | pci id | 21:54 |
| used____ | lspci -vvv or such should show serial numbers. Not PCI id. | 21:54 |
| Kitty | will have to look on Tuesday. I have a 4 day weekend now | 21:54 |
| Kitty | but thank you for the suggestion | 21:55 |
| used____ | https://wiki.archlinux.org/title/Udev these people have nice docs. | 21:55 |
| used____ | https://wiki.archlinux.org/title/Udev#Identifying_a_disk_by_its_serial a hint on how to write a serial number based rule from ^ | 21:56 |
| used____ | see differences udev and eudev. devuan uses eudev | 21:57 |
| used____ | fwiw this NVIDIA Ethernet controller does not expose MAC address or Serial no. to lspci -vvv | 22:01 |
| bb|hcb | used____: this has nothing to do with udev | 22:02 |
| bb|hcb | kernel modules for nics, e.g. i40e do not have any option to specify which set of devices to handle and which not | 22:02 |
| used____ | oh | 22:02 |
| used____ | lshw does list serial and mac Kitty | 22:03 |
| bb|hcb | but here is a way to achieve it - load both modules, then i40e would claim all devices and then use bind/unbind | 22:03 |
| used____ | bb|hcb: bind/unbind? | 22:03 |
| Kitty | I did it manually with unbind/bind | 22:03 |
| Kitty | but I need to make it persist on reboots | 22:04 |
| used____ | Right. | 22:04 |
| Kitty | worst case a shell script that is run on boot... | 22:04 |
| bb|hcb | like: echo -n 0000:02:00.0 > /sys/bus/pci/drivers/i40e/unbind | 22:04 |
| bb|hcb | and: echo -n 0000:02:00.0 > /sys/bus/pci/drivers/other_driver/bind | 22:04 |
| used____ | yes /etc/rc.local | 22:04 |
| used____ | bb|hcb: nice, did not know that trick, only via script which probably does that | 22:05 |
| bb|hcb | rc.local is also not a good idea, because it runs after e/n/i; maybe some pre-up in e/n/i calling the shell script? | 22:05 |
| used____ | yes, better. | 22:06 |
| bb|hcb | i'd do pre-up on lo ;) :P | 22:06 |
| used____ | on grounds that runs 1st? | 22:06 |
| bb|hcb | yes, and also, depending on other settings devices may change names under the different driver | 22:08 |
| Kitty | ooh funky idea | 22:08 |
| bb|hcb | Kitty: in case you are lucky, the mobo will assign the same pci id on every boot, but i have seen ones that shuffle the pci ids randomly; if you hit that, you need to script detecting the current id based on e.g. the mac address | 22:10 |
| bb|hcb | HTH :) | 22:14 |
| Kitty | oh fun | 22:14 |
| Kitty | thanks for the heads up | 22:14 |
| used____ | so lshw does show mac addresses. | 22:15 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!