#archlinux-ports | Logs for 2017-12-12

Back
[00:54:29] -!- because has quit [Ping timeout: 250 seconds]
[00:56:57] -!- because has joined #archlinux-ports
[01:49:46] -!- isacdaavid has joined #archlinux-ports
[01:53:35] -!- guys has quit [Ping timeout: 240 seconds]
[02:12:26] -!- because has quit [Ping timeout: 268 seconds]
[02:13:44] -!- because has joined #archlinux-ports
[02:14:48] -!- guys has joined #archlinux-ports
[02:33:36] -!- because has quit [Ping timeout: 260 seconds]
[02:36:50] -!- because has joined #archlinux-ports
[02:45:57] -!- because has quit [Ping timeout: 240 seconds]
[02:50:24] -!- because has joined #archlinux-ports
[05:20:54] -!- because has quit [Ping timeout: 246 seconds]
[05:23:37] -!- because has joined #archlinux-ports
[05:43:22] -!- hringriin_ has joined #archlinux-ports
[05:46:05] -!- hringriin has quit [Ping timeout: 240 seconds]
[05:46:05] hringriin_ is now known as hringriin
[06:12:05] -!- deep42thought has joined #archlinux-ports
[06:26:25] -!- deep42thought has quit [Quit: Leaving.]
[07:32:44] -!- oaken-source has joined #archlinux-ports
[07:33:34] -!- oaken-source has quit [Client Quit]
[08:09:35] -!- deep42thought has joined #archlinux-ports
[08:51:26] <mrkiko> deep42thought: good morning :)
[08:56:01] <deep42thought> <Hi
[08:57:19] <deep42thought> mrkiko: All good?
[08:57:35] <mrkiko> deep42thought: yeah! :) thank you! And at your side?
[09:06:49] <deep42thought> all fine, have some network trouble at work, so I don't really get to do my physics :-)
[09:07:45] <mrkiko> :D
[09:07:48] <mrkiko> ... :D
[09:43:09] -!- finsternis has quit [Ping timeout: 258 seconds]
[09:45:10] -!- finsternis has joined #archlinux-ports
[10:34:17] CalimeroTeknik_ is now known as CalimeroTeknik
[12:42:55] -!- oaken-source has joined #archlinux-ports
[14:19:57] -!- guys has quit [Ping timeout: 240 seconds]
[15:52:52] -!- guys has joined #archlinux-ports
[16:03:37] -!- deep42thought has quit [Quit: Leaving.]
[17:12:49] -!- because has quit [Ping timeout: 248 seconds]
[17:15:45] -!- because has joined #archlinux-ports
[17:22:52] -!- deep42thought has joined #archlinux-ports
[17:25:35] -!- because has quit [Ping timeout: 240 seconds]
[17:29:48] -!- because has joined #archlinux-ports
[18:00:24] -!- oaken-source has quit [Ping timeout: 246 seconds]
[18:45:38] -!- oaken-source has joined #archlinux-ports
[19:05:37] -!- abaumann has joined #archlinux-ports
[19:05:58] <abaumann> CalimeroTeknik: thanks for Arch586, just realized, I'm heading the same way. :-)
[19:16:05] -!- oaken-source has quit [Ping timeout: 240 seconds]
[19:20:01] <deep42thought> abaumann: what's that?
[19:21:19] <abaumann> an 586 ISO and packages.
[19:21:26] <abaumann> 2009, it looks like.
[19:21:27] <deep42thought> oh, nice
[19:21:33] <abaumann> yep. :-)
[19:21:48] <deep42thought> so this existed before archlinux32, then?
[19:21:52] <abaumann> but the bootstrapping process is not documented.
[19:22:11] <abaumann> it looks like a 686 standard archlinux from back then for i586.
[19:22:21] <deep42thought> ah, ok
[19:22:34] <deep42thought> does it receive current updates or is it just the iso?
[19:22:56] <deep42thought> I'm asking, because I see some potential to share common stuff :-)
[19:22:59] <abaumann> dunno. I'm trying to install it.. the installation procedure looked quite different back then..
[19:23:32] <abaumann> I was hoping, CalimeroTeknik has something to say about it. He is in the channel. :-)
[19:24:13] <abaumann> *argh* gcc. hard to cross-compile nowadays. plugins, funny libraries, all things which don't cross-compile.
[19:24:47] <deep42thought> maybe crosscompile a lighter compiler, then?
[19:24:48] <abaumann> I'll publish my way of porting to 486, so it can receive some critizism. :->
[19:25:03] <abaumann> oh. no. glibc/gcc is a more or less must package..
[19:25:25] <deep42thought> maybe, this can be compiled with the other compiler, too?
[19:25:34] <abaumann> clang and linux bootstrapping is not well-supported, I saw once a distro trying to do that.
[19:25:54] <abaumann> I'm not a big fan of GNU and it's way of handling "special" extensions in glibc/gcc. :->
[19:26:23] <abaumann> to be fair.. it got much better.. used to be much worse..
[19:30:26] <abaumann> ceph, ends in a "/usr/bin/ld: final link failed: No space left on device
[19:30:35] <deep42thought> huh?
[19:30:40] <abaumann> before it failed somewhere in the unit tests..
[19:30:40] <deep42thought> didn't see that before
[19:30:45] <abaumann> build 4.
[19:31:22] <abaumann> don't harddisk have unlimited capacity nowadays? ;-)
[19:31:26] <deep42thought> /dev/sda2 99G 91G 3.0G 97% /
[19:31:32] <abaumann> oups. :-)
[19:32:01] <deep42thought> 13G build cache :-/
[19:32:17] <abaumann> well.. it's testing thoroughly.. ;-)
[19:32:23] <deep42thought> yeah
[19:35:51] <abaumann> something for upstream: /etc/passwd.pacnew with only root, systemd fills in the rest, the old /etc/passwd remains on the disk
[19:36:01] <abaumann> as soon uids/guids change you are in for a disaster.
[19:36:49] <abaumann> IMHO this is worth an explaination and a note in the news for how to do the proper transition..
[19:39:58] <deep42thought> I wondered about the empty files, too - why should uids/gids change?
[19:40:07] <deep42thought> and how does systemd fill in the rest?
[19:40:10] <deep42thought> I don't get it ...
[19:41:03] <abaumann> more /etc/passwd.pacnew
[19:41:05] <abaumann> root:x:0:0::/root:/bin/bash
[19:41:14] <abaumann> my old /etc/passwd contains the old entries.
[19:41:30] <abaumann> where exactly systemd gets the uids and guids is currently also a puzzle to me..
[19:41:59] <abaumann> ..but systemd it's on the never-ending quest of elliminating all plain text files from a linux system, so it gets even easier to configure it.. :->
[19:42:11] <abaumann> ..sorry, did I mention I don't like systemd? ;-)
[19:42:18] <deep42thought> :-D
[19:42:34] <abaumann> the problem is: people should really replace /etc/passwd with /etc/passwd.pacnew.
[19:42:41] <abaumann> dito for other files.
[19:43:22] <abaumann> one positive thing: you don't have to merge passwd and passwd.pacnew anymore (which is errorprone on it's own).
[19:43:55] <abaumann> fstab too? where are partitions configured then?
[19:45:39] <abaumann> ui. this is quite a change in 'filesystem'.
[19:46:42] <abaumann> system 235: dynamic user management..
[19:48:22] <abaumann> are we going now to randomized uids/guids for services?
[19:50:50] <abaumann> ok. moving passwd.pacnew and group.pacnew to passwd group gives me an unbootable system!
[19:51:25] <abaumann> let me restate my phrase from before: people should NOT replace passwd with passwd.pacnew!
[19:51:42] <abaumann> sorry.. this is a disaster!
[19:53:23] <deep42thought> ok, good to hear that my conservative approach (e.g. not installing *.pacnew) was right this time ...
[19:53:52] <abaumann> https://bugs.archlinux.org
[19:53:52] <phrik> Title: FS#56549 : [arch-install-scripts] arch-chroot and unpopulated /etc/passwd, /etc/group (at bugs.archlinux.org)
[19:54:17] <abaumann> I had the feeling I was completly not understood when I had trouble creating chroots showing the same issue..
[19:54:49] <abaumann> What's worse: I don't get where systemd has the default uids and gids stored.
[19:56:28] <deep42thought> the problem is, that the problem is (or is not) present when _creating_ the chroot (not when entering)
[19:56:43] <deep42thought> so it's mkarchroot's fault - if anyones
[19:57:07] <abaumann> yea, for the devtools.. for arch-chroot it's arch-chroot's job..
[19:57:19] <deep42thought> no
[19:57:20] <abaumann> otherwise people are in for a surprise when installing from the ISO :->
[19:57:22] <abaumann> no?
[19:57:25] <deep42thought> pacstrap?
[19:57:33] <deep42thought> I think
[19:57:35] <abaumann> pacstrap does it?
[19:57:39] <abaumann> lemme check.
[19:57:45] <deep42thought> you need to install the base system
[19:58:00] <deep42thought> arch-chroot is merely a chroot wrapper which also bind mounts system relevant dirs
[19:58:02] <deep42thought> IIRC
[19:58:12] <abaumann> ok.
[19:58:22] <abaumann> so it assumes the chroot has been installed correctly. you're right.
[19:59:56] <abaumann> So, which packages populates the missing entries in /etc/passwd?
[20:00:03] <deep42thought> dunno
[20:00:14] <abaumann> passwd in filesystem is empty.. maybe a install hook?
[20:00:22] <abaumann> shadow would be a logical candidate.
[20:01:03] <abaumann> ok, not empty: root:x:0:0::/root:/bin/bash
[20:03:18] <deep42thought> I don't like/understand systemd either ...
[20:03:37] <abaumann> tachyons on my system :-)
[20:03:41] <abaumann> 29240 root 20 0 17.8m 6.0m 100.0 0.1 0:02.42 R tachyon
[20:03:51] <abaumann> building sagemath-doc
[20:03:52] <abaumann> :-)
[20:04:04] <deep42thought> especially since I tried to get rid of a circular dependency on my system
[20:04:18] <abaumann> cross-compiling is fun with systemd.
[20:04:32] <abaumann> glad some people thought about the --without-systemd configure flag.
[20:07:17] <abaumann> ok.. me read some more on systemd.. it's not like to go away soon. :-)
[20:07:27] <abaumann> s/like/likely
[20:10:44] <deep42thought> I wonder when systemd will start replacing the kernel
[20:11:31] <abaumann> :)
[20:11:50] <abaumann> I learned in school that super-daemons are bad as a design principle.
[20:12:16] <abaumann> I learned the strength of unix is that everything is a file and small pieces form a system.
[20:12:30] <abaumann> well linux is and was never a Unix most likely
[20:12:46] <deep42thought> :-)
[20:16:23] <abaumann> ah, this was tyzoid pointing me to https://lwn.net
[20:16:24] <phrik> Title: Building the kernel with clang [LWN.net] (at lwn.net)
[20:16:34] <abaumann> worth a try..
[20:16:52] <deep42thought> yeah, saw that, too
[20:18:12] <abaumann> ==> Leaving fakeroot environment.
[20:18:12] <abaumann> ==> Finished making: gcc 7.2.0-2 (Tue Dec 12 20:12:44 CET 2017)
[20:18:18] <abaumann> ah. cross-compilation finished. :-)
[20:21:24] <abaumann> /usr/lib/gcc/i486-unknown-linux-gnu/7.2.0/crtbeginS.o: ELF 64-bit LSB relocatable, x86-64, version 1 (SYSV), not stripped
[20:21:28] <abaumann> nice.
[20:21:35] <abaumann> back to the drawing board..
[20:21:38] <deep42thought> O.o
[20:27:50] <abaumann> checking for i486-unknown-linux-gnu-gcc... cc
[20:27:57] <abaumann> which is the host cc.
[20:28:02] <abaumann> this is used to build crtbeginS.o
[20:28:11] <abaumann> I specified --host and --build
[20:28:35] <abaumann> gcc seemed a little bit friendlier back in 4.7 days.. :-)
[20:28:52] <deep42thought> -march=
[20:28:54] <deep42thought> ?
[20:29:19] <abaumann> no, you have to use the right cross compiler. I have the crosstool-ng toolchain compiler there.
[20:29:32] <deep42thought> ah
[20:29:39] <abaumann> all the rest is compiled correcly, just the C startup code insists to get compiled with the host 'cc'
[20:29:48] <abaumann> checking build system type... x86_64-pc-linux-gnu
[20:29:51] <abaumann> checking host system type... i486-unknown-linux-gnu
[20:29:52] <abaumann> checking for i486-unknown-linux-gnu-gcc... cc
[20:29:58] <abaumann> really nice.
[20:30:12] <abaumann> I have the feeling, this is a wrong PATH or so.
[20:31:50] <guys> abaumann: pacnew for the [core]/shadow files should really not be merged; do not mess around with your user accounts database like that!
[20:32:31] <deep42thought> guys: so why are all the entries (also of system users) gone then?
[20:32:33] <guys> and quite frankly, having systemd-sysusers is no worse than the previous state of post_upgrade scripts running useradd 300 times
[20:32:49] <guys> They aren't, systemd-sysusers creates them via a pacman hook?
[20:32:52] <abaumann> guys: there I agree.
[20:33:08] <guys> I don't know how you can get them to be missing
[20:33:35] <abaumann> I'm good. :->
[20:33:42] <abaumann> and I never read documentation too..
[20:33:53] <abaumann> ls /usr/lib/sysusers.d/*.conf
[20:33:59] <abaumann> cat /usr/lib/sysusers.d/basic.conf
[20:34:09] <abaumann> no, there they are.
[20:34:24] <guys> Technically this is even more file-y than before :p
[20:34:31] <abaumann> the problem is: while booting before systemd is up, do all deamons have the right users?
[20:34:44] <abaumann> or in the ramdisk?
[20:34:48] <deep42thought> there are no daemons up before systemd, are there?
[20:34:50] <guys> It's idempotent configuration files now, rather than some obscure CSV database
[20:35:06] <guys> abaumann: it should be, systemd starts first
[20:35:11] <abaumann> oh.. not XML or binary.. I'm disappointed. ;-)
[20:35:12] <guys> Also the install hooks
[20:35:39] <abaumann> So, when /etc/passwd and /etc/group are set to the pacnew versions, they only have root in there.
[20:35:54] <abaumann> and then I had several deamons complaining, and the system didn't come up.
[20:36:03] <guys> they've always been an exception to "merge the pacnew"
[20:36:10] <abaumann> :)
[20:36:34] <abaumann> I admit: I did it so see what happens, when you actually do.
[20:37:38] <guys> Well, that isn't fair unless you also do what pacman -Sy base would do, which is run systemd-sysusers on each installed sysusers.d file :p
[20:38:23] <abaumann> I like to do impossible stuff to get error behaviour.. so in case somebody does the same stupid thing I know exactly the symptoms.
[20:41:40] <guys> Well, try using pacstrap on an empty directory (or mkarchroot) and see what the etc/passwd file looks like afterwards -- everything's fine!
[20:42:07] <abaumann> yep.
[20:42:18] <abaumann> that's settled. I had a slightly older version of mkarchroot
[20:42:25] <abaumann> me bad.
[20:43:26] <guys> mkarchroot just uses the current version of pacman --root /path/to/archroot -Sy "${packagelist[@]}" though...
[20:43:42] <guys> it's all handled in the updated repo packages
[20:46:10] <abaumann> mmh.. puzzle..
[20:47:01] <abaumann> this brings me to bad problems when cross-compiling.. can't run a systemd there just to fill in /etc/passwd, so I have to go with an old populated passwd..
[20:48:11] <abaumann> ah. util-linux insists of having a stty during installation: chgrp: chgrp: invalid group: 'tty'invalid group: 'tty'
[20:48:35] <abaumann> but I think usually packages never rely on certain entries in groups or passwd.
[20:55:32] <guys> hmm, so confused as util-linux does not have an install script
[20:55:45] <abaumann> aeh. no: in 'make install'
[20:57:19] <abaumann> term-utils/Makemodule.am: chgrp tty $(DESTDIR)$(usrbin_execdir)/wall
[20:57:22] <abaumann> term-utils/Makemodule.am: chgrp tty $(DESTDIR)$(usrbin_execdir)/write
[20:57:24] <guys> oh, well as long as the OS is properly bootstrapped you'll have all that
[20:57:45] <abaumann> nothing you can do here but have a tty group in place when building the package (preferable with the same uid)
[20:57:47] <guys> But really, other init systems should probably implement sysusers.d -- I know openrc did
[20:58:25] <abaumann> From a simplicity point of view it's much better to separate services users (the _xxx ones in OpenBSD) and the real users
[20:58:52] <abaumann> only the distribution nows how to set 'tty' or 'nbd'.
[20:59:00] <abaumann> s/nows/knows
[20:59:16] <abaumann> you can not have users merge the passwd, shadow, gshadow, group files :-)
[20:59:57] <abaumann> nis/yp had a nice approach with /etc/passwd and a continuation "file" which got appended by nis(+)/yp, if I recall correctly.
[21:00:23] <abaumann> so you could distribute groups and users in an organization..
[21:01:27] <abaumann> creating a jail, have a daemon with temporary uids/guids is also a nice thing to have, I can see, why the systemd people went that way.
[21:01:45] <abaumann> for old-timers like me changes happen sometimes a little bit too fast to catch up. :-)
[21:35:12] -!- because has quit [Ping timeout: 248 seconds]
[21:35:12] -!- titus_livius has quit [Ping timeout: 248 seconds]
[21:36:26] -!- titus_livius has joined #archlinux-ports
[21:40:52] -!- because has joined #archlinux-ports
[21:42:24] <abaumann> beginner mistake: it's --target=i486-unknown-linux-gnu --host=i486-unknown-linux-gnu --build=x86_64-pc-linux-gnu
[21:45:15] -!- phrik has quit [*.net *.split]
[21:45:15] -!- City-busz has quit [*.net *.split]
[21:45:15] -!- anyone has quit [*.net *.split]
[21:46:21] -!- City-busz has joined #archlinux-ports
[21:46:21] -!- City-busz has quit [Changing host]
[21:46:21] -!- City-busz has joined #archlinux-ports
[21:46:23] -!- anyone has joined #archlinux-ports
[21:48:26] -!- phrik has joined #archlinux-ports
[21:53:59] -!- abaumann has quit [Quit: leaving]
[22:04:43] -!- deep42thought has quit [Quit: Leaving.]
[22:55:58] -!- oaken-source has joined #archlinux-ports
[23:26:39] -!- oaken-source has quit [Ping timeout: 248 seconds]