| greenjeans | UsL: I'm pretty sure it was designed from the get-go to break stuff, that's what got LP hired at microsoft, they love that designed-to-fail stuff, keeps the billion dollar tech support industry humming right along, and trashes machines completely after a few years so you have to buy a new one | 00:06 |
|---|---|---|
| freem | <UsL> in my day and age I had flood prptection set to ten lines. But that was when all connected on dialup and connections were much more robust than today's fiber optical connections | 07:45 |
| freem | It is still required on IRC because IRC sends things line by line, this is not a technical issue but a protocol one, for which IRCv3 have WiP effort (https://ircv3.net/specs/extensions/multiline). Sorry for off-topic, but that explains why this particular chan have this limitation still. Also, regarding the topic, I see that debian's zfs-dkms package does not depends on anything systemd related on debian, so devuan probably just uses debian's package. | 07:45 |
| freem | So probably should be reported to upstream (devuan AFAIK only patches stuff to remove systemd dependency, nothing else). | 07:45 |
| freem | XFaCE: do you have some logs showing the error and it's context? With the informations you provided, guessing what may be the problem is going to require a lot of luck. Did you had the problem with stable ZFS package, for example? | 07:47 |
| freem | dmesg may be of some use for example, but zfs may also have it's own log somewhere, possibly sent to a syslog daemon | 07:48 |
| XFaCE | freem: good morning | 13:47 |
| XFaCE | The old ZFS package didn't have this issue, and I can be fairly certain it traces back to this change https://github.com/openzfs/zfs/commit/5e7a2f4665b5be32dab9c183e6fdb94e1f434b70 | 13:49 |
| XFaCE | Simply put, this change made it so zfs-mount.sh occurs in runlevel S, but also is supposed to follow zfs-load-key, which occurs at 2 at the earliest | 13:50 |
| XFaCE | When I removed zfs-load-key, the warning disappeared | 13:50 |
| freem | well, the commit is clearly about breaking non-systemd systems | 13:50 |
| freem | it's written directly in the title | 13:50 |
| freem | "Update zfs-mount to load before fstab, matches systemd service." | 13:51 |
| XFaCE | That part is sound... but the lack of diligence to not update zfs-load-key is... expected | 13:51 |
| freem | I never used ZFS myself, but, maybe you can simply stop using a daemon for that, and rely on fstab for mounting, since it's the whole job of fstab | 13:51 |
| XFaCE | Not from a pushing systemd standpoint, but rather from a "we don't check our work properly" standpoint common to many modern projects | 13:51 |
| freem | ah, this I know what you're talking about, sadly | 13:52 |
| XFaCE | In this case, the assumption of incompetence can take precedence over malice I think | 13:53 |
| XFaCE | freem: disabling zfs-load-key should be enough | 13:53 |
| XFaCE | I don't need it anyway because zfs encryption has been tenuous | 13:53 |
| freem | I try to assume incompetence all the time, but sometimes people make it very hard to do | 13:53 |
| XFaCE | To be fair, there's nothing wrong with zfs-mount changes per change... I did the same thing with Slackware, only manually patching rc.S in that case | 13:54 |
| XFaCE | *per say | 13:54 |
| XFaCE | The problem is as stated, not bothering to check the other scripts | 13:54 |
| freem | I never could manage to get into rc.d system, myself. I am actually thankful to systemd for it allowed me to learn about daemontools, and notably runit | 13:54 |
| UsL | freem: are you saying that irc protocol version 1998 did not have that limitation? We used to paste ascii art in the demo scene channels : ) | 13:55 |
| freem | I am not lol | 13:55 |
| XFaCE | rc.d on slackware is pretty pleasant due to its simplicity | 13:55 |
| XFaCE | The encouragement of manual patching and such | 13:55 |
| XFaCE | The lack of extra tools needed | 13:55 |
| freem | UsL: what I meant is that ascii art posting as I remember it from 20 years ago was annoying as f... | 13:55 |
| freem | and also we were more used to wait after things | 13:56 |
| * freem remembers images loading *very* slowly back in the 56k modem era | 13:56 | |
| XFaCE | freem: it's funny, I find myself missing rc.d when dealing with all the extra commands like insserv that the traditional sysvinit implementation uses | 13:56 |
| XFaCE | I'll probably end up switching to OpenRC because prior Gentoo experience | 13:57 |
| freem | well, when I think sysVinit+rc.d, I think about that horrible mess freedesktop.org had as a "standard", the thousands of lines of bourne shell split in thousands of files in random folders | 13:58 |
| UsL | yeah, it was rude to post ascii in ordinary channels. | 13:58 |
| XFaCE | freem: I see... yeah Slackware does a good job keeping everything in /etc/rc.d | 13:58 |
| XFaCE | One thing that's a bit of a curve ball with Devuan sysvinit is disabling services... in Slackware land all it takes is chmod -x | 13:59 |
| freem | XFaCE: I remember trying to understand some of those stuff debian have in /etc/init.d/* to replace them with runit stuff. I abandonned, too many indirections, I felt like I was going to be condemned to live in Italia of Code | 13:59 |
| freem | yes, disabling services is a nightmare, and the worst is, they are enabled by default when you install them | 14:00 |
| XFaCE | Ah so it's not just my instinctive dislike here | 14:00 |
| freem | I'm afraid it's not. Or we have quite similar instincts, dunno. | 14:01 |
| XFaCE | So in other words, switching to OpenRC is a good idea | 14:01 |
| freem | I myself have switched to runit 10 years ago, and live in paradise since then | 14:01 |
| XFaCE | The install I'm doing is a debootstrap + chroot, so it's about weighing what stuff to do before booting the actual install | 14:02 |
| freem | 95% of my services would only be one-liners, if it was not for me enjoying to have a clear indication that the service is being started in my logs, which are even handled on a per-daemon basis, so never rotated because another one is logging shit like crazy | 14:03 |
| freem | not to mention busybox-static includes a daemontools implementation :D | 14:03 |
| XFaCE | I think you may appreciate Slackware rc.d as well | 14:03 |
| XFaCE | It's been said it copies the BSD way of doing init/rc. I can't confirm having not used BSD very much | 14:04 |
| freem | possibly, but I like the fact systems can go up by themselves as well, with runit. This have proved to be invaluable in my last dev job, where I had ~300 street systems to keep in working state | 14:04 |
| XFaCE | Haha fair enough | 14:04 |
| freem | street system: you need to travel if you want to reboot them | 14:04 |
| XFaCE | I have nothing against runit personally, I just have used OpenRC more | 14:05 |
| freem | understandable | 14:05 |
| XFaCE | And depending on how OpenRC is maintained in the future I might be forced to move anyway | 14:05 |
| freem | I would likely have given a shot to openrc if was not python | 14:05 |
| freem | does it actually requires maintenance, thuogh? | 14:05 |
| XFaCE | I would argue no, but maintainers will want to anyway | 14:06 |
| XFaCE | What's that old saying... "If it isn't broken, break it" | 14:06 |
| freem | (as for python dep, this may be an old deprecated or totally wrong memory, I am very unsure about it) | 14:06 |
| XFaCE | I don't see a python dep in the ebuild | 14:07 |
| freem | me neither | 14:08 |
| freem | must be wrong. | 14:08 |
| freem | happens to me fairly often :) | 14:08 |
| XFaCE | Well... if I can figure out how to disable sysvinit services just for one reboot, I should be good | 14:09 |
| XFaCE | I'm just iffy whether I should swap init systems in the chroot minimal install state post-debootstrap | 14:09 |
| freem | those are not sysvinit stuff you'll touch, but debian's rc.d mess. IIRC there are readmes in i.e. /etc/rc2.d/ that describes the process | 14:10 |
| freem | in /etc/init.d/ as well | 14:11 |
| freem | (no, those readmes are not the sames) | 14:11 |
| freem | (would be too easy otherwise) | 14:11 |
| freem | To disable a service in this runlevel, rename its script in this directory so that the new name begins with a 'K' and a two-digit number, and run 'update-rc.d <script> defaults' to reorder the scripts | 14:12 |
| freem | according to dependencies. | 14:12 |
| XFaCE | yikes | 14:12 |
| freem | that's what the one in the rcX.d says | 14:12 |
| freem | yikes is the word yes | 14:12 |
| UsL | sysv-rc-conf | 14:13 |
| XFaCE | oh I see, it's all symlinks | 14:13 |
| XFaCE | UsL: Some ncurses thing? | 14:14 |
| UsL | works for me | 14:14 |
| XFaCE | So it some ncurses thing | 14:14 |
| XFaCE | *it is | 14:14 |
| UsL | editing files directly in rc.d is all fun and games, but when pressed for time there is sysv-rc-conf | 14:15 |
| XFaCE | I'll give it a shot | 14:15 |
| XFaCE | How does it relate to insserv though? | 14:15 |
| freem | heh, I wish I'd have known that command before! | 14:15 |
| freem | packages depending on it: 0. Not recommended, not suggested, nothing. Good luck finding this... | 14:16 |
| XFaCE | freem: http://www.slackware.com/config/init.php | 14:16 |
| freem | thanks. FYI on runit, you just symlink the service's folder in your runlevel one | 14:18 |
| XFaCE | This is the key point: "Since version 7.0, Slackware includes System V init compatibility. Many other Linux distributions make use of this style instead of the BSD style. Basically each runlevel is given a subdirectory for init scripts, whereas BSD style gives one init script to each runlevel. | 14:18 |
| XFaCE | " | 14:18 |
| freem | easy as well, but sadly it includes the supervise folder, which makes diskless systems a bit more annoying to do | 14:18 |
| XFaCE | runit does you mean? | 14:18 |
| freem | yes, sorry | 14:19 |
| XFaCE | how exactly does it impede diskless installs... can't a ramfs just be mounted in there? | 14:19 |
| freem | it does not impedes, just makes it more tedious since you need to copy the original folder in a ramdisk before exec'ing runsvdir | 14:20 |
| XFaCE | Ah I see | 14:20 |
| freem | I mean it could be easier, that's even something I'm planning to implement when I'll feel bored enough | 14:21 |
| XFaCE | probably could be done during initramfs | 14:21 |
| XFaCE | or sounds like a job for overlay | 14:22 |
| freem | I would just let runsvdir take a target dir as destination to store those, really | 14:22 |
| XFaCE | or that haha | 14:22 |
| freem | I already did diskless systems with runit anyway, so I know that can be done with the original program. But I clearly think runit could use some polish, someday. Not critical, but would ease my life. | 14:24 |
| XFaCE | One project I have in the backburner is converting a pihole to Devuan | 14:24 |
| fsmithred | the runit maintainer for debian interacts with us on the devuan forum | 14:25 |
| djph | I half did that with bind9 and whatever list pihole uses | 14:25 |
| XFaCE | Literal conversion from Debian | 14:25 |
| XFaCE | djph: bind9's featureset is sufficient to supplant pihole? | 14:26 |
| freem | I only have some toy commits uppon runit for my fun. There is a devuan forum, with a runit section or something? I could look at it and see if I can help with some things | 14:27 |
| fsmithred | no particular section for runit | 14:27 |
| fsmithred | wherever the subject fits | 14:27 |
| freem | this: https://dev1galaxy.org/ ? (damn, why do ALL websites have shitty contrast those years?!?) | 14:27 |
| fsmithred | yeah that's the place | 14:28 |
| freem | thx, will take a look then | 14:28 |
| fsmithred | if you log in you can change the colors | 14:28 |
| fsmithred | look for posts from Lorenzo | 14:28 |
| freem | thing is: I only log when I have something to do. Default stuff shuold be sane, not the other way. | 14:28 |
| freem | but I guess sane is not compatible with interweb *shrugs* | 14:28 |
| fsmithred | that color was the first devuan desktop theme | 14:29 |
| freem | there are tools to help with colour-picking and avoid bad contrasts... like this one https://webaim.org/resources/contrastchecker/ but I recently heard that android studio even gives warnings automatically | 14:30 |
| freem | well, that's one of the reasons I don't use the web much those years (also is off topic so I'll stop this ranting now) | 14:31 |
| XFaCE | freem: old forum tradition it seems. Gentoo Forums haven't changed at all really either. | 14:32 |
| XFaCE | Linux Questions | 14:32 |
| APic | 😌 | 14:32 |
| freem | not that old, I remember forums when I was discovering internet, they had sane contrast and fonts | 14:32 |
| amarsh04 | Interesting, I did an init 1, re-mounted root filesystem read-only, ran fsck -f on it, then remounted root as rw, crontrol-D to restart and the buffer cache was still there and previously running apps loaded fast | 14:32 |
| freem | amarsh04: that is to be expected, you never did anything to the kernel, and caches are handled by it | 14:33 |
| freem | if you want to drop kernel caches, there's some file you need to write in, around /sys or something. I don't remember the exact path+name, but shuold be easy to get from a search engine | 14:34 |
| amarsh04 | I also need to check the fsck options to do more caching while checking | 14:35 |
| ted-ious | Is it time to start downloading 6.0 beta and reporting installation bugs? | 18:10 |
| ted-ious | Or I guess any kind of bugs not just the installer. | 18:11 |
| fsmithred | ted-ious, latest installer isos are from five days ago, so yes you can do some testing. | 18:15 |
| ted-ious | fsmithred: It's not frozen yet before the release right? | 18:17 |
| cousin_luigi | Isn't Trixie itself projected to be released after Summer? | 18:17 |
| ted-ious | Oh I thought that was being released much sooner. | 18:19 |
| ted-ious | I must be confusing my rss feeds. :) | 18:20 |
| cousin_luigi | ted-ious: Looks like it is I who was wrong. | 18:23 |
| cousin_luigi | ted-ious: June/July 2025. | 18:23 |
| ted-ious | cousin_luigi: I think I was just about as wrong as you. :) | 18:25 |
| ted-ious | fsmithred: Do you think I should use those iso's to install or should I do a dist-upgrade to excalibur? | 18:29 |
| fsmithred | trixie just went into soft freeze | 18:40 |
| fsmithred | ted-ious, what's your intent? You just want to run excalibur because you need something newer? | 18:40 |
| fsmithred | or do you want to help troubleshoot the installer? | 18:41 |
| ted-ious | fsmithred: Basically I just want to see what's coming so I thought I would try out a beta and see what it's like. | 18:45 |
| ted-ious | Maybe I'll play around with the iso a few times and then do a 5.0 to 6.0 migration and keep using it to see if I can find anything that breaks. | 18:46 |
| ted-ious | I think I did that last time for some project but then I abandoned it so I didn't find anything reportable. | 18:47 |
| fsmithred | Please take notes about any issues you run into upgrading from 5 to 6. | 18:54 |
| ted-ious | I will try to do a much better job at completing my project this time. :) | 19:06 |
| XFaCE | If I knew how to use znc notes I'd note 'sysv-rc-conf' | 19:26 |
| XFaCE | landley: Are you Rob Landley? | 19:54 |
| landley | XFaCE: yes. | 20:00 |
| XFaCE | landley: I appreciate the thorough documentation around toybox, thanks much. A quick question, do you think toybox is mature enough to be embedded in an initramfs? | 20:02 |
| landley | Have you seen mkroot? | 20:02 |
| XFaCE | I've seen the documentation yes. | 20:02 |
| XFaCE | I'm assuming that answers my question indirectly. | 20:03 |
| landley | It uses toybox in the initramfs. | 20:03 |
| landley | That said, toysh is only about 2/3 finished. (It works, it's just not good enough to rebuild itself under itself and run its own test suite yet.) | 20:03 |
| XFaCE | I'm tempted to just include ksh93 for my shell needs in initramfs. | 20:04 |
| landley | And build Linux From Scratch under mkroot with the native toolchain squashfs files in https://landley.net/bin/toolchains/latest/ | 20:04 |
| landley | Toysh works pretty well, but I'm currently implementing "alias" support and finishing "trap" support and so on. | 20:04 |
| landley | (And I'm not doing command editing and history until it's reasonably complete because people assume if the gui is finished the rest of it must be too.) | 20:05 |
| XFaCE | landley: I also would like to voice appreciation for the extensive explanation as to which utilities you are including. I consider your choices the unofficial POSIX. | 20:05 |
| landley | Using a different shell for the moment is a reasonable choice. :) | 20:05 |
| landley | Thanks. | 20:05 |
| XFaCE | Well that and ksh93 provides the flexibility of Bash but isn't slow. And it has so many embed internal commands. | 20:06 |
| landley | I'm trying to make toysh be the best shell I know how to do, but life's kinda intervened the past few years... | 20:06 |
| XFaCE | I recall KSH93 beating even Dash in a speed comparison test | 20:06 |
| landley | But I'm not going for absolute speed, no. :) | 20:06 |
| XFaCE | No blame here... toybox alone is a project that would drive me nuts. | 20:06 |
| landley | I started out nuts. Saves time. | 20:07 |
| landley | What I really need to do is get a toybox+musl based debian root filesystem together. | 20:07 |
| landley | Erik Anderson did a busybox+uclibc one back when I first got into busybox, but did not document _how_. | 20:08 |
| landley | Taking apart rebootstrap is... elaborate. | 20:08 |
| XFaCE | Does the name Dermot Bradley ring any bells? | 20:08 |
| XFaCE | (not me) | 20:08 |
| landley | I don't even know how to set up my own debian server to add extra packages, which I need to do for the https://github.com/j-core/openlane-vhdl-build toolchains. | 20:09 |
| landley | I think I've seen the name before, but I'm bad with names... | 20:09 |
| XFaCE | He's been doing a lot with minimal scripted Alpine installs. | 20:09 |
| XFaCE | I'm sure you two have crossed paths before | 20:10 |
| landley | Most likely. | 20:10 |
| landley | I have the to-read list of doom. I haven't even gone through https://peertube.debian.social/w/45XroN9CnbYLNLKQH3GD9F yet. | 20:10 |
| XFaCE | I'm guessing Google's solution to the toysh work in progress is to stick with mksh | 20:11 |
| landley | (I did watch https://peertube.debian.social/w/chzkKrMvEczG7qQyjbMKPr but it was disappointing. Not "how to provide alternate packages for base dependencies", it was how to lobotimize the dependencies and then add packages on top of a basically empty chroot.) | 20:11 |
| landley | Google is mostly a consumer of toybox stuff. They did give me some money a couple times over the pandemic, but that ended when they started doing mass layoffs. | 20:11 |
| XFaCE | A shame | 20:11 |
| XFaCE | Nice though we've gotten to a point where I don't have to install a busybox on Android to make some applications work | 20:12 |
| landley | Elliott explained their position pretty well on a podcast once... http://androidbackstage.blogspot.com/2016/07/episode-53-adb-on-adb.html | 20:12 |
| XFaCE | Saved | 20:13 |
| landley | Get gets to toybox starting around 19 minutes. | 20:13 |
| landley | XFaCE: if you ever have to do that, let me know and I'll try to fix it. | 20:13 |
| landley | (Unless it's an selinux thing. :P ) | 20:13 |
| landley | I'm the guy who made busybox build under itself and built linux from scratch under the result in the first place. | 20:13 |
| landley | https://landley.net/aboriginal/about.html was like 15 years of my life. | 20:14 |
| XFaCE | Oh yes, I have read your Wiki page a few times. I know the rather... controversial history | 20:14 |
| landley | I have a wiki page? | 20:14 |
| XFaCE | yep | 20:14 |
| landley | ...where? | 20:14 |
| XFaCE | Oh wait, my bad. It's the busybox page | 20:15 |
| XFaCE | I could have sworn there was a standalone page though | 20:15 |
| XFaCE | Ah no, I'm thinking of the *toybox* page, my bad *2 | 20:16 |
| landley | Ah, thought so. Wikipedia's got a bunch of https://en.wikipedia.org/wiki/Linucon and https://en.wikipedia.org/wiki/Unix-like and https://en.wikipedia.org/wiki/Public-domain-equivalent_license pages talkinga bout stuff I did, but I myself am not "notable". | 20:17 |
| landley | (Because of course.) | 20:17 |
| XFaCE | I think anyone who rebuts Matthew Garrett should get a page haha | 20:17 |
| landley | Garrett's come around to my side! | 20:17 |
| landley | (I follow him on mastodon.) | 20:17 |
| XFaCE | Well that's good | 20:17 |
| landley | And Jeremy Allison publicly admitted that going GPLv3 did huge damage to samba. | 20:17 |
| XFaCE | What about Tridge? | 20:18 |
| XFaCE | *ll | 20:18 |
| landley | The allison talk was https://archive.org/details/copyleftconf2020-allison by the way. | 20:18 |
| landley | I fell out of touch with Tridge, I think he retired and went into cryptocurrency nonsense. | 20:19 |
| XFaCE | I'm guessing you keep thorough notes :) A habit I should try to cultivate | 20:19 |
| landley | I keep a blog. | 20:19 |
| XFaCE | That helps | 20:19 |
| landley | I search my old blog to find old links. Easiest way. | 20:20 |
| landley | That last link came from https://landley.net/notes-2020.html#13-05-2020 for example. | 20:20 |
| XFaCE | Well it's a pleasure to run into you. I'm literally nobody. | 20:21 |
| XFaCE | haha | 20:21 |
| landley | (Also, computer history's been a hobby of mine since 1999, and https://www.landley.net/history/mirror/ happened because I kept losing links where I knew I'd seen something but coulddn't find it again.) | 20:22 |
| XFaCE | Also sad to say that Gentoo and Slackware and this zfs debootstrap is the closest I've gotten to LFS | 20:22 |
| landley | Got in the habit of keeping track of WHERE I found something, so I could cite it. | 20:22 |
| XFaCE | Yeah, I have the bad habit of assuming I know how to search the same link again | 20:22 |
| landley | And then I went to work for Jeff Dionne, the founder of uclinux and the guy who ACTUALLY created uclibc and busybox (Erik Andersen was his employee at Lineo), and he knows where ALL the bodies are buried... | 20:22 |
| XFaCE | search *for | 20:22 |
| landley | It's scar tissue talking. I thought I knew and failed hard, way back when... | 20:23 |
| XFaCE | So when you say Debian MUSL root fs, you mean compiling that from scratch? | 20:23 |
| landley | Unless somebody has a better way? | 20:24 |
| landley | (I've looked, but maybe I just haven't found it.) | 20:24 |
| landley | The problem is "which libc" is basically part of the architecture, but most distros aren't set up to track it as such. | 20:24 |
| XFaCE | I've wondered what it would take to compile Slackware using MUSL | 20:24 |
| landley | I learned long ago that compling uclibc binaries under glibc was cross compiling. | 20:24 |
| XFaCE | Thinking about that makes my brain hurt | 20:25 |
| landley | (This totally would have been called "why cross compiling sucks" if the people publishing it hadn't objected: https://landley.net/writing/docs/cross-compiling.html) | 20:25 |
| landley | If you can cross compile arm for x86 you can cross compile musl from glibc. It's exactly the same operation. Alas, the gnu/dammit people don't think so. (They are wrong.) | 20:26 |
| XFaCE | makes sense, but does it take a few compiles to do a stage 3 or whatever? | 20:26 |
| landley | (I blatered about that at length years and years ago... https://bootlin.com/pub/video/2008/ols/ols2008-rob-landley-linux-compiler.ogg is the link.) | 20:27 |
| landley | Eh, not exactly? | 20:27 |
| landley | blathered | 20:27 |
| landley | Eh, not entirely a word. Can't typo it I guess. | 20:27 |
| XFaCE | I have this Void Linux from source procedure written down that involves three compiles | 20:28 |
| landley | I need a whiteboard to properly go over that. :) | 20:29 |
| landley | I want to gesticulate at things! | 20:29 |
| landley | The bof link above probably covers most of it (albeit in a Pascal's Apology sort of way). | 20:30 |
| XFaCE | "gesticulate" wow | 20:30 |
| landley | (He apologized for writing a long letter because he didn't have time to write a short one.) | 20:30 |
| XFaCE | Your phraseology is unique | 20:30 |
| landley | That one _is_ a word. | 20:30 |
| XFaCE | Since it sounds like you've been acquainted with every OSS celebrity of the 90s (a bit exaggerating ;))... have you met Patrick Volkerding? | 20:32 |
| landley | Anyway, the multiple builds are mostly so dependencies don't leak from one context to another. Especially with dynamic builds. | 20:32 |
| XFaCE | Ah I see | 20:32 |
| XFaCE | That's an explanation for something that noone have given me an explanation for | 20:32 |
| landley | There's also making limited compilers (no threading support, etc) and then building a full compiler from that, which is also about constraining leakage between contexts. A bit like wringing out a towel. | 20:32 |
| landley | Linux From Scratch 3.0 back in the day had a section on this, I think. | 20:33 |
| XFaCE | I see you have the Carl Sagan gift of relating things with understandable metaphors | 20:33 |
| XFaCE | I appreciate that | 20:34 |
| landley | https://linuxfromscratch.org/museum/lfs-museum/3.0/LFS-BOOK-3.0-HTML/chapter05.html | 20:35 |
| landley | There's a lot of stuff where "the way to learn this is to be an old geezer who was around when it was being discovered", but what that often means is find the OLD references that didn't assume you already knew it. | 20:36 |
| XFaCE | LFS is the kind of thing everyone should try at least once. The issue for me is finishing my "real" install | 20:36 |
| landley | So many hobbyists are tinkering with old electronics because a great way to learn electronics is to find a book about it from 1965. :P | 20:36 |
| XFaCE | So kind of like trying to find archives of the Red Hat cgroup documentation pre-Systemd | 20:36 |
| landley | Yes, but A) Red Hat is Pointy Hair Linux that's been a corporate (not hobbyist) project since their IPO in 2000, B) there was a cgroups1 and cegroups2 (different filesystems with diferent APIs) | 20:37 |
| XFaCE | True, I should ask whoever what version that documentation was about | 20:38 |
| landley | (Sigh. There is SO MUCH BACKSTORY. But that's life...) | 20:38 |
| XFaCE | One of my favourite videos is some guy installing Slackware 1.1.2 on a 386SX | 20:38 |
| landley | git log and archive.org are, sadly, often your friends. As are lkml archives that go back along way, like lkml.iu.edu | 20:38 |
| XFaCE | I once found a mailing list conversation from 1997 that involved Theo De Raadt and Julian Assange. | 20:39 |
| landley | (Which has nothing on kclug, from which I harvested https://www.landley.net/history/mirror/linux/1991.html and https://www.landley.net/history/mirror/linux/1992.html) | 20:39 |
| landley | Did you know one of the founding arguments of Linux was Linus and the Minix maintainer basically disagreeing about everything? https://www.oreilly.com/openbook/opensources/book/appa.html | 20:40 |
| XFaCE | Oh yeah, I read Linus' autobio | 20:40 |
| landley | (That whole book is pretty good if you're in to computer history.) | 20:40 |
| landley | "Just for fun?" | 20:40 |
| XFaCE | Yep | 20:40 |
| XFaCE | Rented it from a public library more than 10 years ago | 20:41 |
| landley | That's good. Robert Young's "Under the radar" was good too. | 20:41 |
| XFaCE | How Linus had a Sinclair, and then bought a PC to run Minix | 20:41 |
| landley | And Peter Salus' "A quarter century of unix". (I haven't read his newer one about the demon and the penguin.) | 20:41 |
| XFaCE | Any of your books you'd recommend? | 20:41 |
| landley | sinclair ql with an 80009 processor. And before that his grandfather's vic 20. | 20:41 |
| landley | I haven't written any books. Kept meaning to, but too busy with other things. | 20:42 |
| landley | Although back before Eric Raymond went crazy he almost called me a co-author of The Art of Unix Programming. | 20:42 |
| XFaCE | And he used that Sinclair as a terminal to connect to the University system | 20:42 |
| landley | (At least that's what he wrote in the introduction. http://www.catb.org/esr/writings/taoup/html/pr01s06.html ) | 20:43 |
| XFaCE | Speaking of toybox and POSIX, perhaps you can help me with this mystery | 20:43 |
| XFaCE | so it's known that <<'EOF' means a here-document that ignores shell characters like $ | 20:43 |
| landley | XFaCE an then sold it to buy his 386, and then minix couldn't keep up with his modem because microkernel dropping characters task switching between each IRQ, so he wrote a terminal that booted from floppy, and linux evolved from that. :) | 20:43 |
| XFaCE | Yep :) | 20:44 |
| landley | Moral: dogfooding is vital to real progress. | 20:44 |
| XFaCE | landley: but it appears that <<\EOF does the same thing as <<'EOF' but is much easier to type | 20:44 |
| landley | So is <<\EOF | 20:44 |
| XFaCE | Yeah | 20:44 |
| landley | Turns out it's any escaping in the delimiter. (I had to know this to make toysh do it right.) | 20:44 |
| XFaCE | But if you look through opengroup docs <<\EOF isn't mentioned anywhere | 20:44 |
| XFaCE | Even though it works on every shell I've tried | 20:44 |
| landley | Posix is a frame of reference to diverge from. | 20:45 |
| landley | It's a common subset, but it doesn't even mention "init" or "mount". | 20:45 |
| XFaCE | So is <<\EOF an accident? | 20:45 |
| landley | A system that doesn't go beyond posix can't BOOT. | 20:45 |
| landley | It's not an accident, it's basically "if unparsed token != parsed token then call it quoted". | 20:46 |
| landley | <<EOF'' | 20:46 |
| XFaCE | I love the official recommendation to deal with hashbangs | 20:46 |
| landley | ? | 20:46 |
| XFaCE | I'd have to find it again, but it was in relation to people saying that #!/bin/sh isn't recommended by POSIX whatsoever | 20:46 |
| XFaCE | the recommendation is to write an "install" script that replaces the hashbang in the script being installed to match the system's environment | 20:47 |
| landley | Sigh. | 20:47 |
| landley | The bsd people all use "#!/usr/bin/env bash" and so on, which raises the question of why you trust env to be at an absolute path but not bash. | 20:47 |
| XFaCE | Indeed | 20:48 |
| XFaCE | BTW, I hope toysh doesn't allow one to set a path to be a function name, or if it does the path on the drive takes precidence | 20:48 |
| landley | (Which turns out to be a bsd thing: they refuse to install anything that isn't in the BSD repo into /bin or /usr/bin. On android there's no /usr, the $PATH points to /system something.) | 20:48 |
| landley | My reponse was always to root cause grandma's roast: https://lists.busybox.net/pipermail/busybox/2010-December/074202.html | 20:48 |
| landley | (P.S. that was off the top of my head and the numbers are slightly wrong. The system disk was .5 megs, each rk05 is 2.5 megs. It's all on dennis ritchie's web page...) | 20:49 |
| XFaCE | Oh wow I said hashbang, I meant shebang | 20:49 |
| landley | Neither is real. The words are made up. :) | 20:49 |
| XFaCE | But yeah, on Bash one can do /usr/bin/sudo() { ; } as a function name | 20:50 |
| XFaCE | Unless using POSIX mode | 20:50 |
| landley | In $PATH? | 20:50 |
| landley | That's insane. When did that start. (Wasn't in 2.05b.) | 20:50 |
| XFaCE | Irrespective of $PATH | 20:50 |
| XFaCE | it overrides anything on the drive | 20:50 |
| landley | Oh, you mean it #! ? | 20:50 |
| XFaCE | I'm not sure what you mean | 20:51 |
| landley | Oh, you mean you can put / in a function name. | 20:51 |
| XFaCE | yeah | 20:51 |
| XFaCE | that's technically a POSIX extension | 20:51 |
| landley | Huh, hadn't thought of that. | 20:51 |
| landley | I'm not sure Chet thought of that either. :) | 20:51 |
| landley | He might just not have _excluded_ it. | 20:51 |
| XFaCE | However, David Korn made it so KSH priortizes drive paths over functions | 20:52 |
| landley | (But every time I ask him about a bash corner case, he has like a 50% chance of "fixing" it, which means I'm chasing a MOVING TARGET...) | 20:52 |
| XFaCE | I confirmed this with mksh, openbsd ksh, and ksh93 | 20:52 |
| landley | I'd just make it so anything with a / in it doesn't check the function list. I'm already doing that in toybox's multiplexer. | 20:52 |
| XFaCE | dash and busybox sh don't support special chars in function names at all | 20:52 |
| landley | Sorry, this is spamming the devuan list with off-topic discussion... | 20:52 |
| XFaCE | True, I can move to offtopic | 20:53 |
| landley | Talking loudly in a public library sort of thing, just realized. | 20:53 |
| XFaCE | Or DM, whatever you wish | 20:53 |
| landley | There's a #toybox | 20:53 |
| UsL | /topic | 20:55 |
| XFaCE | UsL: Now back to ontopic deboostrapping for zfs purposes | 22:03 |
| XFaCE | which I have now attempted... 5 times? | 22:07 |
| fsmithred | XFaCE, check the forum for posts about installing zfs. There are a few. I don't know which one is best. | 22:14 |
| XFaCE | I shall, but I think I have a good handle the available documentation, the failures aren't really ZFS related as much as figuring out the best way to install stuff after debootstrap | 22:16 |
| XFaCE | *with the | 22:16 |
| XFaCE | Practice with apt and whatnot | 22:17 |
| ted-ious | XFaCE: I found out the easiest process for me is to just do a regular install into the swap partition and then move everything to the root pool later. | 22:52 |
| ted-ious | Just don't try to recompile firefox until you finish moving. :) | 22:52 |
| XFaCE | True, that is always easier than bootstrap | 22:52 |
| XFaCE | on the other I can zfs destroy and recreate datasets indefinitely ;) | 22:52 |
| ted-ious | Right. :) | 22:53 |
| ted-ious | After I did it once and lost track of how many hours it takes I decided to just cheat. | 22:53 |
| XFaCE | There's probably a way to trick the installer into a dataset somehow | 22:53 |
| XFaCE | Slackware's installer is simple enough to do that | 22:53 |
| XFaCE | But the swap partition is a fine trick... the only thing I don't like is having to resize partitions | 22:54 |
| XFaCE | ted-ious: does your approach avoid that somehow? | 22:54 |
| ted-ious | Yes. | 22:54 |
| ted-ious | I just use large swap partitions. :) | 22:55 |
| XFaCE | Yeah but do you delete that partition after and resize the pool? | 22:55 |
| ted-ious | No. | 22:55 |
| rustyaxe | its 2025, ram is cheap. stop using swap :O | 22:55 |
| ted-ious | rustyaxe: Tell that to my laptop. :) | 22:55 |
| rustyaxe | all swap does in these modern times is give a broken process an opportunity to make the machine unresponsive for a while, while destroying the flash | 22:56 |
| rustyaxe | ted-ious: idk my crappiest laptop has 48gb of ram | 22:56 |
| XFaCE | True, 128 GB is enough to fit a base install | 22:56 |
| XFaCE | Just don't lose power | 22:56 |
| ted-ious | XFaCE: Or you can do the server install and it's much smaller. | 22:56 |
| XFaCE | True but at that point what's the difference to a bootstrap? | 22:57 |
| XFaCE | so long as I can get to tasksel it should be smooth sailing | 22:57 |
| ted-ious | It's much faster to get it done. | 22:57 |
| rustyaxe | debootstrap<333 | 22:57 |
| rustyaxe | ive used debootstrap to stuff debian and devuan into some strange places ;) | 22:58 |
| ted-ious | And I don't have to mess around with getting grub and everything installed the way devuan expects it. | 22:58 |
| XFaCE | true, I won't argue the amount of time I've sunk | 22:58 |
| ted-ious | I think in a few years when bcachefs replaces zfs it will make all these hacks obsolete and I can't wait. | 22:59 |
| UsL | XFaCE: I actually tried to display topic and had some spaces in thr way : ) | 23:03 |
| XFaCE | well the message was understood regardless | 23:11 |
| XFaCE | ted-ious: yeah I thought so too until the bcachefs dev did what the bcachefs did | 23:12 |
| ted-ious | XFaCE: I think the point is that there was no message intended. :) | 23:12 |
| XFaCE | Oh | 23:12 |
| XFaCE | I thought he was emphasizing the "Take offtopic chat to" part | 23:12 |
| XFaCE | Which was fair | 23:13 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!