*** rdale_ct has joined #baserock | 02:34 | |
*** rdale has quit IRC | 02:34 | |
*** leeming has joined #baserock | 06:56 | |
*** fay has joined #baserock | 06:59 | |
*** fay is now known as faybrocklebank | 06:59 | |
*** franred has joined #baserock | 07:33 | |
*** toscalix has joined #baserock | 07:43 | |
*** gtristan has joined #baserock | 07:55 | |
*** jude_ has joined #baserock | 08:04 | |
*** jonathanmaw has joined #baserock | 08:40 | |
*** jude_ has quit IRC | 09:09 | |
*** jude_ has joined #baserock | 09:26 | |
*** toscalix has quit IRC | 09:27 | |
*** toscalix has joined #baserock | 09:29 | |
*** toscalix has quit IRC | 09:32 | |
*** locallycompact has quit IRC | 13:27 | |
*** locallycompact has joined #baserock | 14:13 | |
*** gtristan has quit IRC | 14:55 | |
*** CTtpollard has quit IRC | 14:58 | |
*** CTtpollard has joined #baserock | 14:58 | |
jjardon | Hi, Have anyone here ever tried to bootstrap a baserock system not using busybox at all? | 15:09 |
---|---|---|
*** jmacs_ is now known as jmacs | 15:09 | |
rjek | I have only ever heard of people replacing elements of busybox with elements of coreutils etc, not the whole lot | 15:10 |
richard_maw | paulsherwood may have tried with toybox at some point | 15:10 |
jjardon | I meant using only gnu coreutils, not busybox/toybox | 15:11 |
rjek | I'd like to see that, but coreutils doesn't cover everything we use in busybox | 15:12 |
rjek | None of the mkfs or mounts for a start, or many other POSIX-essential things like awk, ed etc | 15:13 |
rjek | So a lot more integration needed for that to be an option | 15:13 |
* rjek ponders a "posix" stratum | 15:13 | |
* paulsherwood didn't succeed with toybox, iirc | 15:14 | |
jjardon | rjek: oh yeah, sorry; I meant coreutils, sed , diffutils ... my idea was to build without busybox/toybox | 15:14 |
paulsherwood | richard_maw: while you're on i think i've got a workaround for the tar order issue with http://paste.baserock.org/emotidokeb | 15:14 |
paulsherwood | (it appears to work, at least) | 15:15 |
rjek | jjardon: Even if you used everything from coreutils, you'd still need busybox to fill in the missing bits you'd need for a working system, though | 15:15 |
rjek | I would welcome movement in that direction though | 15:15 |
paulsherwood | richard_maw: thanks for explaining it so many times :) | 15:16 |
jjardon | rjek: what missing bits are you talking about exactly? | 15:16 |
rjek | 16:13:05 <rjek> None of the mkfs or mounts for a start, or many other POSIX-essential things like awk, ed etc | 15:17 |
rjek | Cross-reference https://github.com/goj/coreutils/tree/rm-d/src with all the links to busybox | 15:18 |
jjardon | rjek: mkfs is part of util-linux and we build gawk in build-essential | 15:18 |
*** jude_ has quit IRC | 15:18 | |
rjek | jjardon: Right, but I'm saying I bet there are functions of busybox that we have built that are not offered by coreutils | 15:19 |
rjek | They're not fungible | 15:19 |
jjardon | rjek: ok; do you know specifically of any? | 15:19 |
richard_maw | paulsherwood: how deep a directory tree have you tested this on? | 15:20 |
richard_maw | I'm not sure why you're using glob.glob, when os.listdir should work and also includes files beginning with . which aren't . or .. | 15:21 |
rjek | jjardon: All the mount helpers | 15:22 |
rjek | For a start | 15:22 |
paulsherwood | richard_maw: i've tested on devel-system | 15:22 |
paulsherwood | i can switch to listdir | 15:22 |
jjardon | rjek: mkfs.ext* is part of e2fsprogs, and mkfs.btrfs is part of btrfs-progs, which we build as well | 15:24 |
rjek | jjardon: mount.* | 15:24 |
jjardon | oops, sorry | 15:24 |
rjek | jjardon: http://paste.baserock.org/iyoxenuyem | 15:24 |
rjek | *lots* of this is not coreutils | 15:25 |
richard_maw | it looks like tar is a little more forgiving when it comes to subdirectories than files then, since your program will do files in /, / itself, files in the first subdir of /, the first subdir of /, files in the second subdir of /… | 15:25 |
rjek | jjardon: I'm not saying it's impossible, just more than building coreutils instead of busybox :) | 15:26 |
jjardon | rjek: cheers. Yeah, I know, but that doesn't mean is not possible to build a system without busybox at all | 15:26 |
jjardon | rjek: ok, seems we agree :) | 15:26 |
rjek | jjardon: OOI, why do you want to build without busybox, even if it's only to fill in the gaps that coreutils does not? | 15:27 |
jjardon | rjek: I do not want busybox in the final system at all. and I do not want to use busybox utils to build other chunks; I had compilation problems before | 15:28 |
rjek | Ah, because people's Makefiles assume GNU or Linux versions of tools? | 15:28 |
jjardon | thas why Ive created strata/coreutils-common.morph , but I was wondering if a more radical aproach can be done | 15:29 |
rjek | There aren't many holes left to fill after using all of coreutils, mind | 15:29 |
* paulsherwood wonders why s/files = glob.glob(os.path.join(d, '*'))/files = os.listdir(d)/ fails completely | 15:29 | |
jjardon | rjek: yes, or the busybox tools doesnt support some parameters the "real" tools have | 15:29 |
richard_maw | paulsherwood: ah, it'll be because listdir returns just the file name, rather than the whole path including the `d` prefix | 15:29 |
paulsherwood | ah | 15:29 |
paulsherwood | good spot | 15:31 |
*** jjardon_matrix has quit IRC | 15:32 | |
richard_maw | files = [os.path.join(d, f) for f in os.listdir(d)] | 15:32 |
paulsherwood | ooh :) | 15:32 |
richard_maw | ah, I've worked it out, the reason why this isn't complaining about making the directories top-down is that you're adding the directories multiple times | 15:33 |
paulsherwood | ack | 15:34 |
richard_maw | my approach would have been: http://paste.baserock.org/gonigumeqe | 15:41 |
richard_maw | do the traversal top-down with reverse-sorted directories, then reverse that traversal | 15:41 |
paulsherwood | richard_maw: thanks! i had already submitted though - https://gitlab.com/baserock/ybd/commit/766b244b87e86882d06ef88f55479773095960a7 | 15:57 |
paulsherwood | jjardon: as of that, it should be possible to do deploys | 15:57 |
* richard_maw normally takes "reviewed the code" as "reviewed it and had no problems with it" | 15:59 | |
*** jjardon_matrix has joined #baserock | 15:59 | |
* paulsherwood believes #241 is now actually, really, properly closed | 15:59 | |
jjardon | paulsherwood: AWESOME | 15:59 |
* jjardon tests | 16:00 | |
paulsherwood | richard_maw: that's quite an assumption :) but ok i'll bear that in mind for future | 16:01 |
* richard_maw would make a note of the fact that this implementation adds directory entries twice, in case you get a later bug from tar about having the directories in twice | 16:01 | |
paulsherwood | richard_maw: i could apply your patch if you prefer :) | 16:01 |
richard_maw | paulsherwood: is it? I wouldn't normally bother mentioning a review if it didn't say "no problems" | 16:02 |
richard_maw | paulsherwood: I'd be interested in knowing how my version behaves in YBD, but if you're fine with duplicate directory entries in the tarball then I wouldn't worry about it. | 16:03 |
*** faybrocklebank has quit IRC | 16:14 | |
*** jonathanmaw has quit IRC | 16:49 | |
*** locallycompact has quit IRC | 16:53 | |
jjardon | paulsherwood: seems it worked \o/, now we are limited by the btrfs dependency for the deployment: https://gitlab.com/baserock/definitions/builds/3799426#down-build-trace | 17:42 |
*** persia_ has quit IRC | 18:01 | |
*** persia has quit IRC | 18:01 | |
*** thrace has quit IRC | 18:01 | |
*** toscalix has joined #baserock | 18:25 | |
*** gtristan has joined #baserock | 18:26 | |
*** toscalix has quit IRC | 18:33 | |
*** toscalix has joined #baserock | 18:33 | |
*** persia_ has joined #baserock | 19:27 | |
*** persia has joined #baserock | 19:27 | |
*** toscalix has quit IRC | 20:45 | |
*** paulwaters_ has joined #baserock | 20:58 | |
*** paulwaters_ has quit IRC | 21:08 | |
*** toscalix has joined #baserock | 21:41 | |
*** radiofree has quit IRC | 22:01 | |
*** toscalix has quit IRC | 22:58 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!