| holst | hi! | 13:43 |
|---|---|---|
| holst | i am trying to get nvidia and ati restricted drivers work from the installer | 13:43 |
| holst | I want the auto-detection to choose any of those two, and then and only then fall back to the free one | 13:44 |
| holst | Now I have this: | 13:45 |
| holst | xserver-xorg xserver-xorg/config/device/driver select nvidia | 13:45 |
| holst | that will probably not work so good with any other restricted driver, like ati :) | 13:45 |
| stgraber | holst: hmm, I guess you will need to create a script and have it called during the install | 14:12 |
| stgraber | because you can only install one of the two drivers (both cary libGL and only one can be installed at once) | 14:13 |
| stgraber | and then update the xorg.conf | 14:13 |
| stgraber | so basically have a script looking at lspci (best would be to use the same code as jockey to detect the best driver for the hardware), then installing the right driver and finally update xorg.conf | 14:14 |
| holst | hmm, not optimal... | 14:17 |
| holst | would rather give a "priority list" | 14:17 |
| cjwatson | this is all up to the X packaging rather than the installer as such; you'd be better off asking its maintainers | 14:19 |
| holst | how do the big companies do this? | 19:46 |
| holst | how do they install ubuntu/whatever linux efficiently on heterogeneous computers? | 19:46 |
| hardwire | well paid admins | 20:06 |
| hardwire | :P | 20:06 |
| holst | that go around installing by hand? :) | 20:10 |
| holst | i generally think that peoplare are lazy | 20:10 |
| holst | and generally, stupid | 20:10 |
| holst | thats what I am anyway :) | 20:10 |
| hardwire | I'd guess most that can do the job are experienced enough to do it efficiently, whatever that solution may be. | 20:11 |
| holst | yes, but they solution is mostly ugly | 20:12 |
| hardwire | I'm not sure what we are arguing about anymore. | 20:12 |
| holst | I am saying: there must be a good way to do it | 20:12 |
| hardwire | and there probably is | 20:12 |
| hardwire | read up. | 20:12 |
| cjwatson | hardwire: um, that isn't terribly helpful | 20:18 |
| cjwatson | if you aren't prepared to understand users' problems, please let somebody else answer | 20:18 |
| cjwatson | I suspect, that said, that big companies doing this kind of thing either (a) buy hardware in bulk and therefore they're not all that heterogeneous, or (b) divide their installations into classes, each of which have distinct preseed files | 20:20 |
| cjwatson | there are a number of hooks available to preseed files that let you run arbitrary scripts during the installation | 20:21 |
| cjwatson | the best one to use here is probably dropping scripts into /usr/lib/post-base-installer.d/ - those are run after the base system has been installed, but before X is installed | 20:22 |
| cjwatson | so you can chroot into the target system and use facilities from there | 20:22 |
| cjwatson | it's not as neat as just specifying a priority list, but it would make it feasible | 20:22 |
| hardwire | I find myself identifying unique information between base installs, then imaging everything after creating a one-off script to set up that system as a unique machine. | 20:24 |
| cjwatson | so you could key your driver selection based on PCI IDs | 20:24 |
| cjwatson | the nvidia and ati vendor ids are well-known, and that's probably all you need | 20:24 |
| holst | can you use those scripts to feed more preseeing data as well? | 20:25 |
| cjwatson | yes | 20:25 |
| cjwatson | as long as they run before whatever's going to consume it, obviously | 20:25 |
| holst | hmm, that can actually solve both the xorg problem as well as the partitioning problem | 20:25 |
| cjwatson | the clumsy but simple way is to use the debconf-set utility; the neater but slightly longer way is to do '. /usr/share/debconf/confmodule' at the top of the script and then use db_set | 20:25 |
| cjwatson | (and possibly db_fset if you want to mark the question seen as well) | 20:26 |
| cjwatson | or you can feed preseed-file-syntax input to debconf-set-selections | 20:26 |
| cjwatson | partitioning is fiddly in hardy, unfortunately | 20:26 |
| cjwatson | you can do it, it's just awkward | 20:26 |
| cjwatson | you have to have a preseed/early_command script that writes out a second script in /lib/partman/display.d/ and makes it executable | 20:27 |
| cjwatson | in intrepid, there'll be a direct partman/early_command preseed to short-cut that | 20:27 |
| cjwatson | I might actually try to come up with something more generic | 20:27 |
| holst | yes, I will probably not be done with this before intrepid is released | 20:27 |
| cjwatson | main-menu/hook/<udeb-name> | 20:27 |
| cjwatson | (hypothetical) | 20:28 |
| holst | thanks very much for the advices; I have kept them in my /lastlog | 20:39 |
| holst | I think that using debconf-set-selections in a hook is the way I will go | 20:39 |
| holst | then I can make some fancy scripts that prove for a /NOBACKUP partition as well (my other problem) | 20:40 |
| holst | probe* | 20:40 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!