fsmithred | _abc_, I just got home | 00:48 |
---|---|---|
fsmithred | to install grub to partition, you select Partition instead of MBR. That comes up before you select partitions for installation. | 00:50 |
fsmithred | if you didn't see that window, it's because the wrong grub was installed, and you get a change to install the grub packages later | 00:51 |
fsmithred | and I didn't make Partition a choice for that part | 00:51 |
fsmithred | s/change/chance | 00:52 |
fsmithred | I'll be back in a few minutes | 00:52 |
_abc_ | fsmithred: thanks for answering, I was away, it's almost 3AM here | 01:24 |
_abc_ | fsmithred: re: it's meant to work in a live session: yes, except I used the persistence partition also to store backups and it was rather large. So when the install started it started pulling in the backups. Morale: remove any stuff NOT to be installed on the system from the persistence volume before starting the install. | 01:26 |
_abc_ | It basically tried to copy things it should not have imo, since it is confused about what's in the persistence/rw mounted with / union and other things on that volume which are visible in the running system under /media/devuan/persistence. | 01:27 |
_abc_ | It worked out eventually but it was not nice | 01:27 |
_abc_ | There should be a way to make the rsync copy not see that dir, it's a cli option. | 01:28 |
_abc_ | Signing off for now. | 01:28 |
fsmithred | DocScrutinizer05, is it possible to leave a message for someone when they aren't here? | 01:35 |
golinux | That's what the znc bouncer is for. You get everything that happened while you were gone. | 01:37 |
fsmithred | yeah, but I want to leave a message for abc | 01:37 |
fsmithred | we keep missing each other. | 01:38 |
Uberius | fsmithred, if abc is a registered nick then simply perform: /msg memoserv send abc blablabla | 01:38 |
fsmithred | thanks | 01:38 |
Uberius | you're welcome | 01:38 |
golinux | I noticed. You haven't been around much lately. | 01:39 |
fsmithred | been working | 01:39 |
fsmithred | so I haven't been home much. And pretty beat when I do get home. | 01:40 |
fsmithred | been painting | 01:40 |
DocScrutinizer05 | fsmithred: yes, there's a message feature in one of greenode's service bots. But it's very inconventient to use and not guaranteed to notice the info message sent by service when the addressee logs in | 01:55 |
DocScrutinizer05 | freenode* | 01:55 |
fsmithred | thanks. Then I hope I see him again soon. | 01:56 |
DocScrutinizer05 | what works best for me is to post something to a pastebin and ask channel to inform the addressee about the URL of that pastebin | 01:58 |
fsmithred | that's more reliable? | 01:59 |
DocScrutinizer05 | you also can look up channel logs and even refer to a point in log with a URL | 01:59 |
fsmithred | oh, cool. I've seen your logs. Thanks for that. | 01:59 |
DocScrutinizer05 | yesm that's way more reliable since messageserv's notifications are basically invisible, the way they get posted during addressee's irc client connecting and authenticating even before it has joined any channels | 02:00 |
DocScrutinizer05 | also I'm not sure but possibly memoserv only works for addressees with a user account | 02:02 |
DocScrutinizer05 | yeah, exactly : Notice] -MemoServ- MemoServ allows users to send memos to registered users. | 02:03 |
DocScrutinizer05 | fsmithred: infobot knows an onjoin featire that makes the bot say something as soon as a particular iser joins a particular channel. It's locked since it is too complicated to manage | 02:10 |
aitor | good morning | 07:57 |
aitor | i've been reading one of the links about runit written by steve litt: | 07:59 |
aitor | 170) se ha unido | 07:59 |
aitor | * Vall se ha marchado (Ping timeout: 25 | 07:59 |
aitor | sorry: http://troubleshooters.com/linux/init/manjaro_experiments.htm#12232014_pure_runit_init_proof_of_concept | 07:59 |
aitor | i'm testing my devuan jessie with runit and vdev | 08:01 |
aitor | the device management works, the audio works, the network connection works and the respawn and virtual terminals work... | 08:02 |
aitor | i have also another system with udev instead vdev, but the virtual terminals don't work | 08:03 |
aitor | running update-grub2 (a symlink to update-grub) with vdev the grub.cfg file loses its entry for runit; so, some changes are needed in the /usr/sbin/grub-mkconfig | 08:05 |
aitor | running update-grub in the other computer (with udev), update-grub works correctly | 08:05 |
aitor | need to go (time to work), have a nice day :) | 08:06 |
msiism | how can i find out which task in tasksel contains which packages? | 20:16 |
gnu_srs | apt-cache show task-xfce-desktop | 20:17 |
gnu_srs | Look at the Depends: and Recommends: entries | 20:18 |
msiism | gnu_srs: ok, that's nice. but you would need a running devuan system to do it (which i happen to have). | 20:19 |
muep | you could choose to just not install any of the tasks and then install them once you have the running (but minimal) system | 20:20 |
msiism | muep: ok, well i'm trying to create sort of a devuan pure blend with a customized selection of packages (but using the standard base system you'd get from the installer without selecting anything in tasksel). | 20:22 |
msiism | but ok, i can always simulate installing a package with apt (or rather apt-get). it's just that i'm trying to avoid meta-packages. | 20:24 |
msiism | ok, so there are about 220 task-* packages, most of them language/locate-related. but due to the naming scheme, everything is pretty mixed up. | 20:29 |
KatolaZ | msiism: what do you need to know? | 20:34 |
msiism | KatolaZ: what's the name of the basic system utilities task? | 20:34 |
msiism | i can't find it. | 20:34 |
KatolaZ | what do you mean by basic system utilities? | 20:35 |
msiism | KatolaZ: i don't really remember what it's called in english. i think it's the last thing you can select in tasksel. (i don't mean console-productivity.) | 20:37 |
msiism | KatolaZ: standard system utilities | 20:38 |
KatolaZ | ok, so yu don't mean the base system | 20:38 |
msiism | no | 20:38 |
KatolaZ | msiism: that package installs anything that is Priority: standard | 20:38 |
msiism | ok, so it's different from the other tasks in how it works, right? | 20:39 |
KatolaZ | it is | 20:39 |
msiism | i see | 20:40 |
KatolaZ | msiism: I guess it makes also sure that all the packages markes |"d |""iummportant" are installed | 20:47 |
KatolaZ | but I am not sure if that is already done in the base install | 20:47 |
msiism | KatolaZ: having trouble reading the above. some characters are not being displayed as they should | 20:48 |
KatolaZ | oops sorry | 20:51 |
KatolaZ | msiism: I guess it makes also sure that all the packages markes |"d "important" are installed as well | 20:52 |
KatolaZ | oh FFS | 20:52 |
KatolaZ | I am not sure if all the packages marked "important" are installed too | 20:52 |
golinux | You can find all that info a click away with synaptic. | 20:53 |
KatolaZ | no golinux you can't | 20:53 |
KatolaZ | in this specific case synaptic is useless | 20:53 |
KatolaZ | (unless synaptic can list by Priority) | 20:54 |
KatolaZ | msiism: cat /var/lib/apt/lists/pkgmaster.devuan.org_merged_dists_ascii_main_binary-amd64_Packages | grep -i -B 20 "Priority: standard" | grep -i "Package: " | 20:55 |
golinux | It lists dependencies, recommends and suggests for every package | 20:55 |
KatolaZ | golinux: this is not a package :) | 20:55 |
KatolaZ | it's a special task | 20:55 |
golinux | OK. I guess I missed domething. | 20:55 |
msiism | KatolaZ: ok, i was gonna try dpkg-query. | 20:58 |
KatolaZ | msiism: it's not accurate though | 20:58 |
msiism | i've found `dpkg-query -W --showformat='${Package} ${Priority}\n' | grep "required$"`. that works. | 21:00 |
KatolaZ | msiism: OK, but you should use "standard" instead of "required" | 21:06 |
KatolaZ | IIRC, "required" and "essential" are part of the base install | 21:07 |
KatolaZ | (whatever is installed before tasksel is started) | 21:07 |
msiism | KatolaZ: ok, but i deselected everything in takssel. | 21:07 |
msiism | but, interestingly, why i run the above command without the grep part, i can see that the installation gave me a wild mix of "required", "important", "standard", "optional" and "extra" packages. | 21:09 |
msiism | s/why/when | 21:09 |
KatolaZ | msiism: those might be deps | 21:10 |
msiism | ok. i was thinking that hese are like more or less separate spheres. | 21:10 |
KatolaZ | msiism: there is no real separation there, AFAICT | 21:11 |
msiism | KatolaZ: ok, but doesn't that make it sort of chaotic? | 21:12 |
KatolaZ | msiism: the world is chaotic :) | 21:13 |
KatolaZ | think that Debian is the work of hundred people | 21:13 |
KatolaZ | ~hundreds of | 21:13 |
KatolaZ | and still it works | 21:13 |
KatolaZ | and is the base of 80% of the current distributions | 21:13 |
KatolaZ | :) | 21:13 |
KatolaZ | msiism: anyway, the base system should be around 200 packages | 21:14 |
KatolaZ | you could try: | 21:14 |
KatolaZ | dpkg-query -W --showformat='${Priority}\n' | sort | uniq -c | 21:14 |
KatolaZ | msiism: ^^^ | 21:14 |
msiism | KatolaZ: ok, give me a sec to understand that. | 21:15 |
KatolaZ | (that's just counting how many packages for each Priority: level are installed) | 21:15 |
msiism | KatolaZ: ok, just had a look at the man page of uniq. unfortunately, it's not on my new minimal install. but i can change that. | 21:16 |
KatolaZ | uh? | 21:16 |
KatolaZ | uniq is not in the minimal install? | 21:16 |
KatolaZ | it's part of coreutils | 21:17 |
msiism | KatolaZ: yes, there's no coreutils in the minimal install (i mean "minimal" as in "deselcted everything in tasksel") | 21:17 |
KatolaZ | uh? | 21:17 |
KatolaZ | :D | 21:17 |
msiism | this is the ascii rc netinstall | 21:18 |
msiism | i first noticed that less isn't more if you install like that :) | 21:18 |
KatolaZ | less is in a separate package | 21:19 |
KatolaZ | but you still should have coreutils | 21:19 |
KatolaZ | without that you have no kill | 21:19 |
KatolaZ | no cat | 21:19 |
KatolaZ | no cp | 21:19 |
msiism | i have that. maybe i mistyped sth | 21:19 |
msiism | ok, typo. coreutils are there. sorry... | 21:20 |
KatolaZ | :D | 21:20 |
KatolaZ | np | 21:20 |
KatolaZ | if coreutils is there, uniq is there | 21:20 |
KatolaZ | and sort as well | 21:20 |
msiism | right. i'm seeing that now | 21:20 |
msiism | now, that is a nice output | 21:21 |
msiism | 62 required, 54 important, 19 standard, 97 optional, 17 extra. | 21:22 |
msiism | about this being chaotic: i recently installed slackware on another machine and was sort of blown away by its grouping all packages into collections that you can work through in a menu (like when configuring a kernel). but then, slackware doesn't do any dependency tracking. | 21:24 |
muep | I'm quite happy that most programs I use are using something else than a build configuration system similar to what the kernel has | 21:26 |
msiism | muep: i'm a prospective choice-fanatic (in a certain way) and control-freak (to use the worst possible descriptions). but atm, i still lack the knowledge to be too serious about that. | 21:29 |
muep | If I'm going to go through all those thousands of build options, I'm a control freak too | 21:30 |
muep | so it bothers me quite a lot that I can't find a documented way of setting options in a non-interactive way | 21:31 |
msiism | muep: ok, i see. well, for the kernel, you could edit the config in a text file, or possibly even have that done by some program, i guess. | 21:32 |
muep | there is nothing like, "I'd want these specific items, please use them alongside some sensible defaults and give me an error if I requested something that is not available" | 21:32 |
msiism | muep: well, i think gentoo's USE flags facilitate that (but that's a bit off-topic...) | 21:34 |
muep | editing kernel .config manually is very tedious because the build will just happily override things that conflict with the dependencies of options | 21:35 |
muep | and then the menu often has things marked enabled with no obvious explanation. the non-obvious explanation will be that something else somewhere entirely else required it to be enabled | 21:35 |
msiism | ok, i didn't know that. i've experienced the same sort of problems trying to build customized versions of software from debian source packages. (but what i did wrong there if probably documented.) | 21:36 |
msiism | s/if/is | 21:36 |
muep | I used to keep a personal kernel configuration in version control and there were so many hoops to go through | 21:37 |
muep | nothing like how I usually could just have kept a script file with all the desired options to configure, cmake or whatever | 21:37 |
msiism | i see. i haven't built a my own kernel in quite a while | 21:37 |
batmore | anyone on who knows if there is a clear path from Jessie or Ascii to Beowulf? I've installed Ascii RC, then changed sources to beowulf and update && dist-upgrade... and that'd fine until a reboot... seems to be stuck at Slim login prompt... just cycles when you enter userid/pswd, even root/pswd | 21:38 |
muep | I still think the kernel is an amazing piece of software and especially an exceptional software project, but the build configuration system is one of the main things that I would hope to not get used as an inspiration for other things | 21:38 |
msiism | muep: ok. good point. | 21:40 |
msiism | batmore: no idea, but can you login on the tty? | 21:41 |
batmore | away from machine, but hmm... from GUI login I don't have a choice that accepts alt-f2, etc... is there an easy way? | 21:43 |
nacelle | ctrl-alt-f2 ? | 21:44 |
batmore | nacelle: oh? sh*t that easy? | 21:44 |
nacelle | i've no idea what you're trying to do, but sure | 21:44 |
nacelle | that will net you a real tty instead of X | 21:44 |
nacelle | _usually_ | 21:45 |
RandomDamage | That'll get you to the text consoles | 21:45 |
golinux | batmore: There hasn't been much work done yet on Beowulf. I've heard server should work but DE's are likely to have problems | 21:45 |
nacelle | and then you can login as usual and diagnose why yer *dm login thingahoozit is brokenated™ | 21:45 |
batmore | nacelle: ok, then I'll give that a try when back at machine. Golinux: oh, rats... ok, then I'll stick with server for now. | 21:46 |
msiism | what's "server"? | 21:46 |
batmore | missim: just an install without the DE or wm | 21:46 |
batmore | msiism: sorry, I mangled your nik | 21:47 |
nacelle | batmore: cool. you can also usually kill an X server with ctrl-alt-backspace if its not wired out (which, sadly, its not on most boxes)... which may restart things enough on yer *dm that it'll work | 21:47 |
golinux | Most of the issues are with DE. WM might still work. | 21:47 |
nacelle | but not likely. | 21:47 |
msiism | batmore: np | 21:47 |
golinux | heads uis now using beowulf. | 21:48 |
batmore | golinux: interesting, I didn't know that... maybe that'll work then. | 21:48 |
batmore | all: ok, thanks all. I've got enough info for now. Cheers. | 21:49 |
golinux | See ys | 21:49 |
golinux | ya | 21:49 |
RandomDamage | muep: I think you want "make oldconfig" for your kernel configurations. | 22:01 |
muep | RandomDamage: I don't because it hides all the old state from me | 22:02 |
muep | I can easily be building tons of stuff that I don't use or that I used to use but not anymore | 22:02 |
KatolaZ | muep: there is no other way than going through the options | 22:03 |
KatolaZ | and the corresponding documentation | 22:03 |
KatolaZ | as you say, the kernel is a big piece of code | 22:03 |
muep | yes but imo the config system should help be at least as much as a package manager would | 22:03 |
KatolaZ | the best way to have something minimal is to go through the whole thing | 22:04 |
RandomDamage | muep: You can always edit the config file directly. I've found that easier in a lot of cases. But then I end up going back to menuconfig... | 22:04 |
KatolaZ | muep: the config system is *awesome* IMHO | 22:04 |
muep | so I could tell it to get me things and it will tell me the extra required things | 22:04 |
KatolaZ | if you think how many thousands modules and options you have | 22:04 |
muep | or I could tell it I do not want something and it tells me what other stuff needs to get removed | 22:04 |
KatolaZ | muep: dependencies are satisfied automatically | 22:04 |
muep | only to one direction | 22:05 |
KatolaZ | yes | 22:05 |
KatolaZ | but the config system cannot know if you need something that is needed by something you just removed or not... | 22:05 |
KatolaZ | that's on you | 22:05 |
RandomDamage | muep: I haven't seen many basic config tools that undo things included for dependencies. | 22:05 |
KatolaZ | if it had to remove all the reverse deps, it should remove everything.... | 22:05 |
muep | it does know if because all the config options have that metadata | 22:05 |
muep | but it does not tell me. it just gives me options in the menu that can not be cleared | 22:06 |
RandomDamage | apt-get autoremove is remarkable for existing | 22:06 |
KatolaZ | RandomDamage: that's easier | 22:06 |
muep | like, if I tell apt to remove gtk2 it will tell me that ok, but let's remove also these things that require gtk2 | 22:06 |
KatolaZ | mmmhhh | 22:07 |
muep | if apt was like the kernel configurator, it would just say that it is not possible to remove gtk2 | 22:07 |
KatolaZ | muep the kernel config system already does that | 22:07 |
KatolaZ | if you remove networking, you won't have any ethernet card driver | 22:07 |
RandomDamage | KatolaZ: would that be in the xconfig target? | 22:07 |
KatolaZ | RandomDamage: ? | 22:07 |
muep | some specific items work like that, especially if they are nested | 22:07 |
muep | but a network driver might require e.g. some less common bus or suchlike | 22:08 |
muep | those just get enabled and I'm not told why | 22:08 |
KatolaZ | err...because the driver requires them? :D | 22:08 |
muep | yes the information is in the config metadata but not presented to me | 22:08 |
muep | and if I go find the whatever required thing it is in the menu, I often can't clear it | 22:09 |
muep | I just need to go through all my enabled items to find out which one requires it | 22:09 |
KatolaZ | muep: do you have a specific example? | 22:09 |
muep | not right now because it is quite a long time since I last was doing this | 22:09 |
muep | it may even have gotten improved | 22:09 |
KatolaZ | all the required deps are explicitly listed in the documentation file accompanying each item | 22:09 |
RandomDamage | KatolaZ: I see what you mean, but groups are handled differently than item dependencies. It's a legit gripe, but a non-trivial problem to solve | 22:10 |
KatolaZ | RandomDamage: see above ^^^ | 22:10 |
KatolaZ | the deps are listed in the info file | 22:10 |
KatolaZ | IIRC | 22:10 |
RandomDamage | KatolaZ: Yes, all very well documented, but I can see why not having a tool that handles things more gracefully is an issue. | 22:11 |
KatolaZ | RandomDamage: you don't compile a kernel every other day :) | 22:11 |
KatolaZ | and if you do, you know what to touch (or you learn it anyway, the hard way) | 22:12 |
RandomDamage | I lived in the kernel for a few years, and I did recompile at least twice a week when I did. I'm just remembering how much I had to keep track of when I did. | 22:12 |
KatolaZ | yep RandomDamage :) | 22:12 |
RandomDamage | A smarter tool would have been noce | 22:12 |
RandomDamage | *nice | 22:13 |
muep | IIRC there was also the mode that some feature is not shown at all because some required feature entirely elsewhere is disabled | 22:14 |
RandomDamage | muep: that happens with some config groups in menuconfig and xconfig | 22:16 |
muep | and then there is actually something I found I could use. point to a partial kernel config file with the KCONFIG_ALLCONFIG env variable while running make allnoconfig | 22:17 |
muep | but it took me hours to find all those settings that "unlock" the ones that I actually knew I wanted to have | 22:17 |
muep | if it worked more like a proper dependency solver (which it IMO should because the dependency information is meticulously provided by the devs), it would just automatically enable the things that are needed by the things I requested | 22:19 |
muep | and now I could keep this partial config in version control and document for myself why I want each of those enabled things | 22:20 |
RandomDamage | That would be a handy tool | 22:20 |
gnarface | google defconfig | 22:21 |
gnarface | "When using one of those configuration interfaces, it is always a good idea to start from a reasonable default configuration. The kernel provides such configurations in arch/arch/configs/*_defconfig and you can put your selected configuration in place with a command like make x86_64_defconfig (in the case of a 64-bit PC) or make i386_defconfig (in the case of a 32-bit PC)." | 22:21 |
gnarface | the kernel is shipped with many pre-made minimal configs as you have described making by hand | 22:21 |
gnarface | most the time you can just load up the one for your arch and add whatever is missing to that | 22:21 |
muep | I did not want someone else's minimal config. I wanted one that has things that I need | 22:22 |
gnarface | trust me, this will ONLY have the stuff you need | 22:22 |
gnarface | you'll probably still have to turn on other stuff just to get it to work | 22:22 |
KatolaZ | muep: there is a simple reason why the tool you have in mind does not exist: tracking reverse deps is not as easy as you imagine, IMHO | 22:22 |
muep | it made the difference between having kernel build take 5 minutes and having it take 30 minutes | 22:22 |
gnarface | or just spend a day doing it yourself by hand. that's fine too, but in the end you're not gonna find any difference in accuracy between the two approaches; just a lot more time spent on one than the other turning off page after page of stuff you don't need | 22:23 |
KatolaZ | if you remove the last network driver, should you remove also "networking" altogether? | 22:23 |
KatolaZ | (no you shouldn't, but a "proper" dependency system would) | 22:23 |
muep | I am not asking for that | 22:23 |
muep | I am asking that if I specify that I want a driver for my network interface, I also get networking without having to specify it manually | 22:24 |
muep | gnarface: my impression is that defconfig is more close to what a distribution could ship | 22:24 |
RandomDamage | It's definitely not a tool I would want to put in the kernel tree. There's too much stuff in there already. But a proper full dependency config tool would be pretty awesome. | 22:25 |
KatolaZ | RandomDamage: believe me, you don't want it | 22:25 |
KatolaZ | there must be a reason if nobody has thought about it in 25 years... | 22:25 |
muep | imo if they are happy to have a Qt interface to kernel configuration, I'd expect them to have a dependency solver too | 22:26 |
KatolaZ | and the reason is that it would introduces more problems than it "seems" to be able to solve | 22:26 |
gnarface | muep: that was not my impression even remotely, but i admit i never had to bother to resorting to them before i started building kernels for ARM/ARM64 devices | 22:26 |
KatolaZ | muep: there is one way to see if it works... | 22:26 |
KatolaZ | :) | 22:26 |
RandomDamage | Because to do it right means building a tool that could never be included in the source tree. It's an idea that *has* been floating around for 25 years. | 22:26 |
gnarface | muep: for whatever it's worth, the actual config that THIS distro DOES ship is easy to rip from the running kernel and stuff in another of approximately similar versioning... | 22:27 |
KatolaZ | RandomDamage: many things have been floating around :) | 22:27 |
muep | hm, indeed it seems that defconfig is not a distro-like configuration | 22:27 |
muep | but it does not set up the HW support I need eitehr | 22:28 |
KatolaZ | muep: manu defconfigs are very minimal | 22:28 |
KatolaZ | ~many | 22:28 |
RandomDamage | I know I never did it because keeping the dependencies in my head was easier than writing a tool for it | 22:28 |
KatolaZ | I personally don't mind at all | 22:29 |
KatolaZ | the kernel is not something that anybody has the competence to delve into | 22:29 |
gnarface | muep: most the time when using a defconfig, i still had to diff the resulting config from my distro's kernel config to find stuff that was missing because it wasn't strictly required for the arch but was for normal debian/linux functionality | 22:29 |
KatolaZ | not anything can be reduced into a "cool" GTK app | 22:29 |
KatolaZ | the best way to make the configuration "easy" would be by lumping entire groups of options/modules together | 22:31 |
KatolaZ | which would result in pretty much overbloated kernels | 22:31 |
msiism | 'apt-get install vim' pulls libgpm2 as a dependency, but 'apt-get purge vim && apt-get autoremove --purge' does not remove it. you have to do that manually. why is that? | 22:31 |
muep | my complaint is not the amount of config options. I just think that things should have a well-working non-interactive UI in addition to the interactive ones | 22:32 |
KatolaZ | muep: go for it | 22:32 |
KatolaZ | I am serious | 22:32 |
RandomDamage | msiism: apt-get autoremove would then remove it, *if* nothing else depended | 22:32 |
gnu_srs | msiism: It depends on if the package is manually or automatically installed | 22:33 |
msiism | RandomDamage: hm... i'm pretty sure this is the case, since this is just a base system. | 22:33 |
muep | KatolaZ: I found my workarounds but I still think the kernel configuration system is way more difficult to use than it has to | 22:33 |
KatolaZ | muep: :) | 22:33 |
gnarface | heh | 22:34 |
muep | the menuconfig is pretty ok if you do not need to keep a history of your changes and the reasoning for them. and especially if you just want to browse around to see what options are available | 22:34 |
gnarface | wait until you try crossbuilding | 22:34 |
gnarface | tbh alot of these complaints you bring up have trivial workaround | 22:35 |
gnarface | *s | 22:35 |
muep | but in case I already know that I want FOO=y and BAR=y and BAZ=n, there is completely unnecessary manual hunting involved | 22:35 |
gnarface | for example, if you really need versioned changes to your config, nothing is stopping you from checking the file into a source repository | 22:35 |
gnarface | it's just $SRC_DIR/.config | 22:35 |
KatolaZ | muep: go for it | 22:36 |
KatolaZ | solve it | 22:36 |
KatolaZ | you'd discover that it is not as easy as it seems | 22:36 |
RandomDamage | The kernel community is also pretty small, as such things go. It takes someone who wants it badly enough to write it. | 22:36 |
KatolaZ | (even without considering the fact that you might want modules vs builtin) | 22:36 |
gnarface | i dunno i'm thinking a few hours reading the grep regexp syntax would save him weeks of time | 22:36 |
muep | I would like to but I do not have the resources. I am not even asking for it to be changed. I just noted that I prefer how most build configurator tools have an useful non-interactive mode | 22:37 |
KatolaZ | I see your point muep | 22:37 |
msiism | RandomDamage: if anything on that system would depend on libgpm2, it would already have been there, wouldn't it? but vim is the thing that pulls it in. and if i remove it again right after having it installed, libgpm2 will not be removed. but i think it should. | 22:37 |
muep | I considered just stuffing .config into version control as-is but then it is basically just a backup with no reasoning from me for most of the lines | 22:38 |
muep | based on my last version-controlled partial configuration, there are maybe about 100-200 options that I actually care about having enabled | 22:39 |
muep | in total there's about 250 settings in there but many of those are just the ones that I had to find and enable to make those relevants ones available | 22:40 |
RandomDamage | msiism: good point. I just looked and not a lot of things depend on libgpm2. So not an easy problem. | 22:42 |
msiism | RandomDamage: ok, i'll write it down for later. | 22:44 |
KatolaZ | msiism: what happens if you try to remove libgpm2? | 22:44 |
KatolaZ | muep: I just had a go at the KConfig files | 22:48 |
KatolaZ | it's not that easy at all | 22:48 |
msiism | KatolaZ: it just gets removed. | 22:49 |
KatolaZ | since not all the deps are stated | 22:49 |
KatolaZ | only those that are not directly implied by the position of the option/module | 22:49 |
KatolaZ | for instance | 22:49 |
KatolaZ | ethernet drivers do not depend on NET | 22:49 |
KatolaZ | they normally depend on the bus (ISA/PCI/etc.) on which they are installed | 22:49 |
KatolaZ | the "NET" dep is brought in "automaticallY: | 22:50 |
msiism | KatolaZ: and it doesn't try to take anything with it. | 22:50 |
KatolaZ | meaning that if you don't select NET you don't see ethernet drivers | 22:50 |
muep | I'm not saying that it is just missing some grepping | 22:50 |
KatolaZ | muep: ^^^ | 22:50 |
KatolaZ | tracking deps is not that easy | 22:50 |
KatolaZ | I remembered something like that, but it has been a couple of years since I built a really custom kernel | 22:51 |
KatolaZ | :) | 22:51 |
muep | I would say that it is a problem solved many times | 22:51 |
KatolaZ | muep: again, it's not that easy | 22:51 |
KatolaZ | it would require additional metadata | 22:51 |
KatolaZ | just have a look at a couple of KConfig files | 22:52 |
KatolaZ | the ethernet drivers are a good example | 22:52 |
muep | just having multiple ways to express a dependency (e.g. explicit or based on position in the hierarchy) does not add that much either. it should be a pretty simple to transform that into an uniform representation that is then processed by the dependency solver | 22:52 |
muep | dependency solving itself is a bit hard but e.g. multiple package managers have solved it for collections of a few tens of thousands of items | 22:53 |
muep | I'd say package managers even have more complex solvers because they need to deal with three levels of dependency (hard, recommend, suggest) and multiple versions of the same thing and package obsoletion and many other details that do not exist with kernel build options | 22:55 |
msiism | KatolaZ: just going through the expert install of the ascii RC. the missing translations problem (let me find that bug report...) is still there. | 22:57 |
muep | I can easily agree that it would take a non-trivial amount of work to have what I wish to have. but so far I do not see anything particularly hard about it | 22:58 |
KatolaZ | msiism: and it will remain there, if we don't have anybody that translates that :) | 22:58 |
KatolaZ | muep: please, have a look at a couple od Kconfig file | 22:58 |
KatolaZ | each option can be "select"ed by whatever other option | 22:58 |
KatolaZ | literally | 22:59 |
KatolaZ | CRC32 is select-ed by several ethernet drivers | 22:59 |
KatolaZ | and also by some FSs | 22:59 |
KatolaZ | and by who-knows-how-many other options | 22:59 |
muep | yes. so those would need to be transformed into explicit data that those several drivers depend on CRC32 | 22:59 |
msiism | KatolaZ: http://bugs.devuan.org/191. ok, i may be able to do that. | 23:00 |
KatolaZ | msiism: let me check if any translation is available there | 23:00 |
muep | the data is currently in an inconvenient format and it is a lot of work (as far as I can tell, only hard in the sense that there's lot of it) to write something that can construct the full set of deps | 23:01 |
KatolaZ | msiism: unfortunately not everything in d-i is or can be localised easily | 23:01 |
KatolaZ | muep: it's a lot, and it's changing | 23:01 |
KatolaZ | continuously | 23:01 |
KatolaZ | maybe not that quick for most parts, but still | 23:02 |
muep | I don't think the format is changing much | 23:02 |
msiism | KatolaZ: ok, i think the normal install is already fine now. (i've just done it an hour ago and would have noticed.) | 23:02 |
KatolaZ | the actual set of deps might | 23:02 |
KatolaZ | but that's not ereally a problem | 23:02 |
muep | yes, so of course the tool would collect the deps from the source that is being configured | 23:02 |
KatolaZ | still, looks like an insane amount of work to me :) | 23:03 |
KatolaZ | maybe I am wrong | 23:03 |
muep | it looks non-trivial to me too | 23:03 |
msiism | KatolaZ: but then again, if you're going for a new installer anyway (which i find very interesting), this is not too important. | 23:03 |
muep | comparable to writing a new package manager | 23:03 |
msiism | i just went through another install of the ascii rc, deselecting everything in tasksel. the base system i get after that already contains dbus. this could be connected to wpasupplicant, which will pull in dbus via libdbus-1-3 if not explicitly told not to. so, as it seems, it is not possible to get a dbus-free system using d-i on (any?) wifi-capable machine. (this may as well be the case in jessie.) | 23:59 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!