=== himcesjf_ is now known as him-cesjf [10:52] Hi, LocutusOfBorg asked me to address a current issue here: the kernel now has the modules for a Virtualbox guest included by default. What is missing, is the udev rule which is part of the virtualbox-guest-dkms package. I recommended to put it into the virtualbox-guest-x11 or virtualbox-guest-utils package. What's your opinion on that? [10:54] Whoopie: it doesn't sound like something for the -x11 package, that's GUI/X-server specific isn't it? [10:54] yes [10:54] there're the libraries for 3D acceleration and some config files. [10:55] TJ-: in the virtualbox-guest-utils package, there's also /sbin/mount.vboxsf, which doesn't work properly without the udev rules. [10:56] Whoopie: that seems to seal the decision then :) [10:56] or put the udev rule into the kernel package? [10:57] I don't think so - the linux-image-* packages don't carry any udev rules currently [10:58] problem is the dependency chain [10:58] dpkg depends on utils, not the opposite [10:59] is it for /lib/udev/rules.d/60-virtualbox.rules ? [10:59] TJ-: /lib/udev/rules.d/60-virtualbox-guest-dkms.rules [10:59] LocutusOfBorg: but -x11 depends on -utils, so it's always installed. [11:01] dpkg depends on it? [11:02] people might have the utils without x11 and without dkms [11:03] oh, DKMS not dpkg you meant? [11:03] yep [11:03] virtualbox-guest-dkms [11:04] LocutusOfBorg: then it doesn't hurt at all. [11:06] hang on, if the kernel module is now in the linux-image then is there any need for virtualboc-guest-dkms at all? ? [11:08] TJ-, probably none, except for udev [11:10] I don't see anything in the virtualbox-guest-dkms other than the udev rule that needs relocating, so that package could become virtual/meta(?) if the udev rule moved to virtualbox-guest-utils. [11:11] well, I don't usually trust the module on the kernel, I mean, kernel updates might break stuff, I prefer to have the possibility to choose my version in case of issues [11:11] the kernel modules are not mainline, they are updated from time to time [11:12] My observations - probably needs someone more aware of the intricacies of package reorg than me, but I don't think the udev rule belongs in the linuximage package - certainly no others are currently [11:12] LocutusOfBorg: ahh, i see, so just move the udev rule to virtualbox-guest-utils and keep guest-dkms as is otherwise [11:15] if v-g-d gets installed on it's own with no other VB guest packages that's just the same as having linux-image-* installed; the udev rule is only required when of the v-g-{utils,x11} packages is installed, and -x1ll depends on -utils, so that sounds correct [11:15] s/when of/when one of/ [11:16] so x11 depends on utils, utils depends on dkms/source/virtual meta package [11:16] and the dkmd itself doesn't depend on anything else [11:16] *dkms [11:16] because also guest-source might provide kernel modules [11:17] "virtualbox-guest-modules" is provided by virtualbox-guest-dkms, virtualbox-guest-source -> (the deb created after the build), and the one provided by src:linux [11:17] how will it work if the -dkms is installed - the kernel will have 2 identical modules, does modprobe know to prefer the /lib/modules/*/updates/dkms/ path to /lib/modules/*/kernel/ ? [11:17] yes the kernel will prefer the updates [11:18] I also patched dkms to not error out in case there is already one installed [11:18] https://launchpad.net/ubuntu/+source/dkms/2.3-3ubuntu5 [11:19] how does depmod handle that? there'll be 2 modules with identical symbols/aliases won't there? [11:20] the updates version is preferred, depmod knows that newer kernel modules might appear [11:20] we hacked some versioning inside the module, to make them differ in version [11:20] that dkms upload was to fix exactly that versioning check [11:25] yeah, just been playing about with it to check [11:26] http://debomatic-amd64.debian.net/distribution#unstable/virtualbox/5.2.8-dfsg-7/buildlog [11:26] let me see [12:40] Whoopie, TJ- uploaded in unstable, will sync in Ubuntu if nobody complains [12:40] thanks for your help! [12:40] https://anonscm.debian.org/cgit/pkg-virtualbox/virtualbox.git/commit/?id=a58ce37a83a281baa85db238d28e6bc82b49114c [12:40] this is the diff [13:02] fortunately the udev rules has packagename in filename, so the mangling avoids me to add breaks+replaces [13:05] apw, ^^ so now we should have fixed that issue in the kernel [13:05] thanks to Whoopie and TJ- :) [13:05] LocutusOfBorg, sounds good, i think [14:16] LocutusOfBorg: thank you!!! [14:18] LocutusOfBorg: "Make the guest-utils depends on dkms not vice-versa" -> so -utils always installs -dkms? [14:23] nope, utils depends on -dkms | -source | virtual version provided by dkms or source or kernel [14:23] virtualbox-guest-dkms (= ${source:Version}) | virtualbox-guest-source (= ${source:Version}) | virtualbox-guest-modules, [14:23] thi is the dependency [14:24] actually I could even remove and just depend on "virtualbox-guest-modules", because it is already provided by dkms [14:24] the problem is that installing guest-source is not sufficient to have a kernel module [14:24] you have to install guest-source, use module-assistant to build a custom .deb kernel module file, and then dpkg -i it (and this custom deb has the "Provide: virtualbox-guest-modules" ) [14:25] so, depending on guest-source is theoretically wrong [14:25] ok, thanks for the explanation. [14:25] but I couldn't find a better [15:18] jsalisbury, tworevert is bad === Elimin8r is now known as Elimin8er [22:49] if bionic will be on kernel 4.15, why wasn't a longterm kernel chosen? [23:13] phil42: because the mainline release cadence doesn't match Ubuntu releases. [23:17] * trippeh_ mumbles something about people wanting support for the latest radeons and stuff [23:45] it just seems to me like letting the kroah-hartman et al backport the fixes would be better [23:52] The kernel team already does that and a lot more