#archlinux32 | Logs for 2022-10-02

Back
[00:02:58] -!- GNUtoo has joined #archlinux32
[03:06:41] -!- drathir_tor has quit [Ping timeout: 258 seconds]
[03:18:31] -!- bill-auger has quit [Ping timeout: 260 seconds]
[03:33:39] -!- drathir_tor has joined #archlinux32
[03:36:18] -!- bill-auger has joined #archlinux32
[04:34:47] -!- drathir_tor has quit [Remote host closed the connection]
[04:45:46] -!- drathir_tor has joined #archlinux32
[10:31:29] <girls> bill-auger: may we borrow your kernel config for our iso?
[10:40:22] <bill-auger> of course - there are actually several variants
[10:41:17] <bill-auger> the last i looked your kernel was about 150MB IIRC - not a large as arch, but yea it could probably lose a little weight
[10:41:44] <girls> I propose a linux-slim, which we could put onto the i486 iso :D
[10:45:03] -!- abaumann has joined #archlinux32
[10:45:04] <buildmaster> Hi abaumann!
[10:45:04] <buildmaster> !rq abaumann
[10:45:04] <phrik> buildmaster: <abaumann> nono, it's self-cooking :-)
[10:45:10] -!- deep42thought has joined #archlinux32
[10:45:10] <buildmaster> Hi deep42thought!
[10:45:11] <buildmaster> !rq deep42thought
[10:45:11] <phrik> buildmaster: <deep42thought> otoh I have the feeling, archlinux32 dies anyway if I get overrun by a bus
[10:45:16] <deep42thought> Hi abaumann!
[10:45:22] <abaumann> morning deep42thought
[10:45:28] <abaumann> how are the rubies? ;-)
[10:45:35] <deep42thought> didn't have a look yet
[10:46:11] <abaumann> I was thinking, if I make a ruby-bootstrap providing ruby in build-support, it will always fallback to that version, if something fails to build from the normal repos, right?
[10:46:38] <deep42thought> yeah
[10:47:02] <deep42thought> but if the failure is unrelated to ruby, it will similarily fail with the ruby from build-support
[10:47:03] <abaumann> we could remove ruby-bootstrap from build-support after having build all packages from ruby-stdlib though.
[10:47:09] <bill-auger> this is the standard rolling kernel https://git.parabola.nu
[10:47:15] <deep42thought> abaumann: good idea
[10:47:46] <bill-auger> there are a few more which may be interesting linux-libre-hardened linux-libre-lts linux-libre-pae linux-libre-vanilla
[10:48:01] <abaumann> bill-auger: concernig the ISO size, I'm not aware on a on-the-fly compression flag in ISO9660, that would be handy..
[10:48:02] <bill-auger> not quite sure what 'linux-libre-vanilla' is thats new - but maybe it is smaller
[10:49:04] <abaumann> don't your kernels contain.. well.. not certain things? ;-)
[10:49:34] <bill-auger> all those changes are in the upstream though
[10:50:31] <bill-auger> most likely you could just change the source=() url to kernel.org and it would build - you really only need the kconfig though
[10:51:01] <bill-auger> build() only does the usual things `make all ; make htmldocs`
[10:53:12] <bill-auger> the "certain things" missing upstream are only blobs for special hardware - probably the config and build are exactly the same
[10:53:19] <abaumann> 164MB kernel modules, yeah, I see why optimizing the size of the kernel is the best we can do..
[10:53:28] <bill-auger> most non-free hardware support is in the linux-firmware package now
[10:53:46] <deep42thought> the issue could be, that the iso won't support all kinds of hardware
[10:53:52] <abaumann> exactly. That's why I omitted the linux-firmware package on the i486 ISO..
[10:53:57] <deep42thought> so one might be left without wifi on the install disk or so
[10:53:59] <abaumann> well, tough luck.
[10:54:15] <abaumann> old hardware is supposed to have an ethernet jack. :-)
[10:54:23] <deep42thought> but we could think of having two isos: one "normal" and one "slim"
[10:54:52] <abaumann> As the guy in the forum mentioned: he was quite happy to build his own ISO after spotting the build scripts for the ISO.
[10:54:55] <bill-auger> yea it better - if the installer can not longer fit on a CD someday, network boot will be the only option for some machines
[10:55:04] <abaumann> I agree, that's not everybodys cup of coffee (speaking of which)..
[10:55:25] <abaumann> ..to build his own ISO.
[10:56:04] <abaumann> on the fly compression in the filesystem or in an overlay fuser filesystem?
[10:56:14] <abaumann> most likely comlicates the build and boot process..
[10:56:54] <abaumann> zisofs. mmh.
[10:57:09] <abaumann> a Rock Ridge extension..
[10:57:17] <deep42thought> I thought, the squashfs is already compressed or something
[10:57:33] <abaumann> yes, but that's exactly the idea of having an ISO with plain files.
[10:57:36] <bill-auger> i think the reason why our ISO still fit on a CD is also because ive not been following arch - the package list has changed only slightly since many years ago
[10:57:39] <abaumann> Assume, you have 16MB RAM or so
[10:57:49] <abaumann> not really space to decompress a ramdisk there
[10:57:55] <deep42thought> ok, I see
[10:58:24] <abaumann> so, reducing some unneeded packages first, I would say.
[10:58:40] <abaumann> The i486 ISO deviates heavily from upstream anyway (to be more precise, it's completely different)
[10:59:07] <bill-auger> yea i would try both - try the parabola kernel config to see if that shrinks, and try trimming down the installed packages
[10:59:22] <abaumann> yep.
[10:59:53] <bill-auger> im working on canonicalizing that package list today actually
[11:01:05] <abaumann> https://git.archlinux32.org
[11:01:06] <phrik> Title: mkarchiso486 « iso - archi486 - Archlinux32 i486 tools (at git.archlinux32.org)
[11:01:14] <bill-auger> right now there are multiple package lists to keep in sync - one for the ISOs, one for the CLI installer, and another for calamares
[11:01:17] <abaumann> mmh, base is just base. Otherwise I have to redefine base.. mmh.
[11:01:43] <abaumann> ah, having separate "base" sets for different tasks..
[11:02:01] <deep42thought> I think, base should not be touched
[11:02:14] <deep42thought> at least, we should not have different "base" packages for different purposes
[11:02:27] <abaumann> right, I meant, doing a pacstrap not with base but explicitely with the most important packages..
[11:02:33] <bill-auger> what i am doing today, is making a set of metapackages to consilidate them as parabola-base, parabola-desktop-lxde, parabola-desktop-mate, and so on
[11:03:28] <bill-auger> yes exactly, i dont want to fudge with base - base is not POSIX, and does not even include a kernel - its meant to be so
[11:03:32] <abaumann> the fact that I could use pacstrap from a 486 floppy busybox thing, with just minor things missing, means I could also but a busybox userland onto the 486 ISO..
[11:03:50] <abaumann> *put*
[11:04:12] <bill-auger> so parabola-base includes that stuff, that was in the old base group
[11:05:04] <abaumann> https://github.com
[11:05:16] <bill-auger> so the package list for building the ISO, is the same as the installers - one package 'parabola-base'
[11:05:21] <abaumann> urgh, that's quite minimal actually, it just draws in too many dependencies, I suppose.
[11:05:55] <abaumann> And don't want to start the debate now why omitting systemd could be interesting.. :->
[11:06:03] <deep42thought> :D
[11:09:47] <bill-auger> IIRC the smallest bootable image i can make is about 550 MB with no compression
[11:11:19] <bill-auger> that is the ARM system - the kernel is slightly smaller (90MB) - but i suppose there is still space to shrink on i686 too
[11:12:42] <bill-auger> thats not the ISO of course, just a raw disk image - i dont think the ISO stuff adds much over-head though - with compression it should be a net reduction
[11:12:43] <abaumann> mmh, the kernel and the modules are compressed with zstd, so a zisofs is not really effective..
[11:13:26] <abaumann> man pages are on the ISO, but no 'man' binary..
[11:13:32] <abaumann> 15 MB
[11:13:37] <abaumann> not that much.
[11:13:47] <deep42thought> hehe
[11:13:49] <abaumann> I would actually like to have man on the ISO.
[11:13:58] <abaumann> For a rescure medium or so, quite handy.
[11:14:00] <bill-auger> every 15 counts though - do that 20 times and yer there
[11:14:32] <deep42thought> don't you look at manpages via 'zless /usr/share/man/man1/bash.1.gz' ?
[11:15:05] <bill-auger> the thing is, anything which is not required to boot, could be put on a "extras" CD
[11:15:05] <abaumann> Only if I need my daily dose of troff/nroff ;-)
[11:15:17] <deep42thought> :)
[11:15:30] <abaumann> the 486 ISO is not read into memory, so you cannot easily swap out CDs
[11:15:34] <abaumann> and having a second drive..
[11:16:50] <bill-auger> that 486 prbably has a floppy drive - so could publish a set of "extras" floppy images
[11:17:10] <bill-auger> just thinking, thats how it was done , back in the day
[11:17:39] <abaumann> I made a tar/floppy boot loader in i486/floppy (the git above)
[11:18:11] <abaumann> you tar a bunch of things like bzIamge, ramdisk, etc. into a normal tar, then split it to say 1.44MB and then the bootloader asks you to swap floppies..
[11:18:14] <bill-auger> yea that was my next though - it probably need to boot from the floppy
[11:18:27] <abaumann> It's hard to get a Linux kernel below 1.44MB
[11:18:35] <abaumann> so this way it can be bigger than that.
[11:18:49] <abaumann> But, I don't think many people have working floppy drives around nowadays. :-)
[11:18:51] <bill-auger> fun with antiques
[11:18:58] <abaumann> always :-)
[11:19:07] <bill-auger> mine still work
[11:19:19] <abaumann> cool
[11:19:21] <bill-auger> i still have soe DOS floppies too
[11:21:00] <abaumann> 5 3 1/2 and 4 5 1/4 floppy drives, no all of them working though :-)
[11:21:05] <abaumann> *not*
[11:21:05] <bill-auger> and a GRUB floppy setup to boot a dynebolic CD - i have a pentium133 laptop with no hard disk - thats the only way to boot it
[11:21:31] <abaumann> I sometimes use a Plopp bootloeader on a floppy and a USB stick.
[11:22:27] <abaumann> mmh. a boot CD with ramdisk, loading all the modules and then a userland CD (as in old times a boot and a root floppy), that could also work for 486.
[11:23:00] <abaumann> 2 times 650MB, but a little bit more complicated to create..
[11:29:07] <abaumann> what are people doing? use xxd to embedd the formware blobs into network drivers?
[11:29:16] <abaumann> *firmware*
[11:31:21] <bill-auger> most of the firmware blobs are byte arrays in C files - _who_knows_ how they make them
[11:32:15] <bill-auger> thats basically the point of removing them - no one knows how to prepare them - yet many of those files have GPL license - its something of a loophole - they really should not be there
[11:32:46] <abaumann> agreed, they should be in /lib/firmware, so at least you actually _know_ you are loading a binary blob..
[11:38:44] <abaumann> oh, /usr/include, 27MB. I bet, everybody wants to program on an ISO, especially without a C compiler.. ;-)
[11:40:04] <deep42thought> :D
[11:41:28] <abaumann> 17MB *.a static libraries..
[11:42:49] <abaumann> 19 MB /usr/lib/udev, the hardware database mainly, well, somewhat needed..
[12:41:11] <bill-auger> you could also use this beauty for a down-and-dirty overview
[12:41:17] <bill-auger> pkguse ()
[12:41:17] <bill-auger> {
[12:41:17] <bill-auger> LANG=C pacman -Qi | egrep '^(Name|Installed)' | paste -d" " - - | awk '{ print $7$8" "$3 }' | sed -E 's|\.[0-9]+([GKM]?)i?B |\1 |' | sort -h
[12:41:17] <bill-auger> }
[12:41:54] <bill-auger> eg output like:
[12:41:54] <bill-auger> 169M clang
[12:41:54] <bill-auger> 169M gcc
[12:41:54] <bill-auger> 171M boost
[12:41:54] <bill-auger> 192M mariadb
[12:41:56] <bill-auger> 230M iceweasel
[12:41:58] <bill-auger> 237M icecat
[12:42:00] <bill-auger> 597M qemu-user-static
[13:07:10] <deep42thought> hey, thanks - I didn't knew "paste" command - now I do
[13:07:53] <bill-auger> i think egrep just became deprecated recently
[13:07:59] <deep42thought> it did
[13:08:25] <deep42thought> I always used grep (w/o any options) and just got used to the quoting
[13:08:47] <bill-auger> i had to fix several scripts last week because of grep's new warnings
[13:08:48] <deep42thought> I never understood, why one would prefer -e
[13:09:08] <deep42thought> err, I mean -E
[13:09:36] <bill-auger> you need E to use real regex
[13:09:50] <bill-auger> same as with sed
[13:09:53] <deep42thought> what's unreal about grep's regexes?
[13:10:06] <bill-auger> some features are missing like the ? and +
[13:10:16] <deep42thought> no, you just have to escape them
[13:10:24] <deep42thought> grep '\.[0-9]\+'
[13:10:51] <deep42thought> that's exactly my point: the escaping differs, but besides that, I didn't find any problems (so far)
[13:11:01] <deep42thought> afk, lunch
[13:11:59] <bill-auger> nio kidding - i never knew that
[13:12:29] <bill-auger> sed 's|a\+bc|by golly, it works|' <<<'abc'
[13:13:07] -!- GNUtoo has quit [Ping timeout: 258 seconds]
[13:35:00] -!- GNUtoo has joined #archlinux32
[13:49:18] <deep42thought> I'm not 100% sure with look-aheads and look-behinds, though
[13:53:18] -!- GNUtoo has quit [Remote host closed the connection]
[13:53:39] -!- GNUtoo has joined #archlinux32
[13:58:59] -!- deep42thought has quit [Quit: Leaving.]
[14:24:43] <bill-auger> or the {N} notation
[14:25:18] <bill-auger> that works too
[14:25:51] <bill-auger> sed 's|a\{2\}bc|it works|' <<<'aabc'
[14:30:44] <abaumann> -rw-r--r-- 1 root root 599M Oct 2 13:21 archlinux32-2022.10.02-i486.iso
[14:30:49] <abaumann> this looks better :-)
[14:32:25] <abaumann> argh, /var/tmp is on the iso and not writeable.. well.. things are still a little bit rough..
[16:51:26] -!- abaumann has quit [Quit: leaving]
[17:14:28] -!- abaumann has joined #archlinux32
[17:14:28] <buildmaster> Hi abaumann!
[17:14:28] <buildmaster> !rq abaumann
[17:14:29] <phrik> buildmaster: <abaumann> let me stop my slaves, they need a new-year cleaning anyway. :-)
[17:25:28] -!- abaumann has quit [Quit: leaving]
[21:14:14] -!- abaumann has joined #archlinux32
[21:14:14] <buildmaster> Hi abaumann!
[21:14:14] <buildmaster> !rq abaumann
[21:14:15] <phrik> buildmaster: <abaumann> "works on my laptop and I buy a laptop every year"
[21:14:40] <abaumann> So, rebuilding ruby didn't help and my ruby-bootstrap doesn't work at all. This is just a huge mess of cycles and dependencies..
[21:14:52] <abaumann> Packages (8) libyaml-0.2.5-1.1 ruby-3.0.2-1.0 ruby-bundledgems-bootstrap-3.0.4-9.2 ruby-irb-1.4.1-1.0 ruby-reline-0.3.1-1.0 ruby-stdlib-bootstrap-3.0.4-9.2 rubygems-3.3.15-1.1 ruby-bootstrap-3.0.4-9.2
[21:15:04] <abaumann> for instance when trying to build ruby-abbrev..
[21:15:16] -!- abaumann has quit [Client Quit]
[21:35:19] -!- drathir_tor has quit [Remote host closed the connection]
[22:00:27] -!- drathir_tor has joined #archlinux32
[22:43:05] -!- epony has quit [Quit: QUIT]
[23:38:09] -!- epony has joined #archlinux32