<mpark> Oh yeah, of course, since I do a lot of embedded work in my little home lab, I thought a musl build would be helpful farkuhar
<mpark> SiFuh_ queued me in on it, much appreciation
<ukky> farkuhar: that script is part of my runit shutdown script: http://cruxy.ca/crux/ports/locoverlay/runit/3
<farkuhar> ukky: thanks for identifying it.
<farkuhar> ukky: We got zorz to elaborate on what's wrong with using [ -x /sbin/lvm ] as the sole criterion for running the vgscan and vgchange commands. It looks like your runit shutdown script also tests how many such LVM mountpoints are actually in use, before proceeding to vgscan. Should /etc/rc and /etc/rc.shutdown have similar logic?
<ukky> farkuhar: I took that part of script (lvm+vgs) from Void 'as-is', adjusting style to my liking, but afaik, only crypt part is running on my systems (not using LVM)
<ukky> Somebody who uses LVM should step in and explain how shutdown for LVM should be handled.
<ukky> I only debugged shutdown process of full disk encryption (crypt volumes). If not done properly, fsck will run upon every reboot for all crypt volumes.
<farkuhar> ukky: Okay, we'll wait for someone who uses LVM to chime in. Meanwhile, I have a tentative patch that replaces beerman's colourization with something a little cleaner: https://n.uguu.se/QjsGdFoG.patch
<farkuhar> SiFuh_: ^
<ukky> First time I see famous /etc/rc. This line is priceless: printf "\n%s$(BOLD "$(RED)")%s\n%s\n%s\n%s\n%s\n%s\n%s\n%s\n%s\n"
<SiFuh_> farkuhar: Expired link
<SiFuh_> farkuhar: also the patch ukky provide at rough glance before coffee is way better. But I will review it later more closely
SiFuh has quit [Remote host closed the connection]
SiFuh has joined #crux-musl
<farkuhar> SiFuh_: Sorry for using a pastebin site with such a quick expiry policy. I should try to figure out why 0x0.st is blocking my IP address; their expiry policy is much more lenient.
<SiFuh_> No worries. I did look into ukky's version and it is suprisingly good. Very clean. He also seems to have removed the need for rc.function which I am for.
<SiFuh_> However, if you wish to modify the colours, you will need to edit multiple files. But that is fine.
<farkuhar> SiFuh_: I must have missed the patch that ukky wrote. Did he share it publicly, or only with you?
<SiFuh_> I think if we can get this lvm2 crap done, then mostly ukky's version will suffice.
<SiFuh_> Publicly
<farkuhar> Hmm, I'm not seeing it in my logs.
<SiFuh_> Oh wait. You pasted it
<SiFuh_> Hang on. Brain farts going here
<farkuhar> Heh, it's another expired link anyway. Good think you grabbed a copy before the URL stopped working.
<SiFuh_> Hah I think it is yours https://n.uguu.se/QjsGdFoG.patch
<SiFuh_> I downloaded it already. So I have it here anyway
<SiFuh_> I slept like a dead person after my dispute with beerman. I was quite happy with the outburst. So this morning, I was barely awake when skimming the logs :-P
<SiFuh_> farkuhar: Did you see my comment on #crux-social about how I odn't understand why beerman can introduce something new, he does a shit job of it. Expects use to clean it up for him?
<farkuhar> SiFuh_: Yes, I saw it. Any response from stenur, by the way?
<SiFuh_> Nope, been checking
<SiFuh_> farkuhar: I wonder if mpark will be a great asset in assisting us
<SiFuh_> Would be nice to have some fresh guys who are interested in advancing the CRUX project to a new level
<farkuhar> mpark: I admire your goal of provisioning both desktops and embedded systems using a common set of sources and ports. Just be aware that a few of the old-timers like darfo will be less supportive, disclaiming that "CRUX was never intended for embedded systems".
<SiFuh_> farkuhar: Do you remember there was a CRUX i586 built by morlenxus and a CRUX Sparc64 built by me?
<farkuhar> SiFuh_: That was before my time.
<SiFuh_> Crux was intended for i686. But there was always an option to expand by users who wanted to contribute. Now we have a CRUX-MUSL.
<SiFuh_> There was also in the works a CRUX Alpha, but I never finished that.
<SiFuh_> And linux-hk was working on an embedded CRUX back in the day as well. This was before ARM became a thing.
<SiFuh_> In fact, I remember morlexus running CRUX on embedded chips with LCD screens back then.
<SiFuh_> But he shifted his focus to E17 and making it for portable devices.
<farkuhar> E17, as in the Enlightenment desktop environment?
<SiFuh_> Yes it switched from Desktop to portable devices and a complete system overlaying a Linux Kernel.
<SiFuh_> I think, (That is my thoughts) that it kind of lost steam when Android popped up on the scene
<farkuhar> mpark: yesterday in #crux you got no response to the comment about our startup and shutdown scripts being so reliant on bash idioms ("not exactly POSIX"). A few months ago I put together a rewritten set of rc-scripts targeting dash, and I had to eliminate the SERVICES array in favor of something more portable.
<farkuhar> Other than finding an alternative to SERVICES, there weren't too many changes needed, to get an rc script that works with #!/bin/sh as the shebang.
<farkuhar> The reliance on bash arrays has a long history; even pkgmk from the early days would source a Pkgfile that defines a bash array. As you noted, other distros that took inspiration from CRUX soon moved away from bash, allowing the build recipes in the repositories to be written in whichever language the packager found most convenient.
<SiFuh_> farkuhar: did you see ppetrov^s comment in CRUX?
<SiFuh_> Maybe the rc script for wpa_supplicant and wifi and net should have a conf file
<SiFuh_> farkuhar: Let's let ppetrov^ help him and we watch. Better to see how it goes then assist when we know the problem.
<SiFuh_> What is the saying "Too many chiefs and not enough Indians"?
<farkuhar> SiFuh_: Did zorz ever volunteer to share his lvm2 configuration, so you can test any changes to the rc scripts?
<SiFuh_> No dude
<SiFuh_> He's been absent minded for a while. Oh I mean absent. :-P
<farkuhar> "we hit a stump"? Is that how we're referring to zorz now?
<SiFuh_> Heh
<SiFuh_> I have a chainsaw.
<SiFuh_> [01 46 41] [jaeger> Anyone can write rc scripts.
<SiFuh_> Hehe apparently beerman can't
<SiFuh_> farkuhar: Interesting how jue is the silent one, jaeger is the understanding one and beerman is the cunt
<SiFuh_> I really think they just let beerman be because he does all the work.
<farkuhar> My first Linux distro was actually Red Hat, circa 1999. I vaguely remember manipulating the sequence of startup daemons by renaming symlinks under /etc/rcS.d/ or some other directory, so that the desired order is what you get from lexographic sorting. When I first tried CRUX (around version 3.4) it was a breath of fresh air, being able to accomplish the same thing by manipulating a bash array SERVICES instead.
<SiFuh_> Mine was Caldera
<SiFuh_> How fucking old am I?
<farkuhar> Little did I know that there's a price to be paid for relying on bash-specific features. You could argue that the tighter constraints of writing for a shell like dash, and using filesystem abstractions rather than non-POSIX shell code, is a more robust design.
<dlcusa> Mine was SLS.
<SiFuh_> dlcusa: I was an Alpha guy
<SiFuh_> SLS is fucking old school dude.
<dlcusa> Yup. I watched Debian but waited for 1.0^H1 to try it, and it became my primary distro.
<farkuhar> So it's interesting that Per decided to fully embrace the non-POSIX features of bash scripting, both for pkgmk and for the rc scripts. In some respects it achieves a cleaner design, but paradoxically also a more fragile one.
<SiFuh_> dlcusa: He is one to blow your mind I used on 2 floppy disks the original BSD before it beacme free open and net.
<SiFuh_> Hah dlcusa I remember that. Debian 1. My brother used it and tried to sell it to me
<dlcusa> SLS was many floppies to install.
<SiFuh_> farkuhar: Point taken. I remember bash was a thing then
<SiFuh_> dlcusa: Actually, true story. I was mad at Linux and said I like BSD but I like the X desktop. Think it was R6 then. My brother said but BSD has the same. I switched instanlty. FreeBSD
<SiFuh_> I didn't even know. But ended up a FreeeBSD to OpenBSD guy. Still using OpenBSD since. Even now, messaging you.
<dlcusa> I'm still planning to take a hard look at OpenBSD when I can free up the time.
<SiFuh_> dlcusa: It is really good. Pure and clean. I will update 7.6 to 7.7 soon. Doesn't take long. But I have a lot of shit running so...
<SiFuh_> dlcusa: There are no OS's out there that are perfect. But for me, OpenBSD is the closest
<dlcusa> I like that it's pretty stable after all these years.
<SiFuh_> dlcusa: Well firefox does crash it. When too many videos are opened
dlcusa has quit [Ping timeout: 252 seconds]
<farkuhar> So jaeger went ahead and tested the startup and shutdown scripts, on a system with lvm2 installed but no logical volumes defined. No errors were reported, so maybe zorz was making a fuss over nothing. On the other hand, ukky's runit unmounting script http://cruxy.ca/crux/ports/locoverlay/runit/3 does have special treatment for lvm2 that isn't in our current rc.shutdown.
<SiFuh_> I am not sure farkuhar, I realy need zorz to actually explain it.
<SiFuh_> Or at-the-least make me build his system and test it myself
<farkuhar> zorz never replied to my question about /sbin/bootlogd either. He's too absent to be relied upon for this work.
<SiFuh_> Or mine
<SiFuh_> If he can't explain it in a timely manner, I think we push the beerman fix. And then later look at his issues.
<farkuhar> ukky's runit unmounting script also has special treatment for dm-crypt volumes, but I don't see anybody complaining about how the current /etc/rc.shutdown is working for those.
<SiFuh_> It actually worked for me. Since 3.5
<SiFuh_> And after jaeger introduced my modifications and ports for 3.6 there was still no issues.
<farkuhar> "01:55 <ukky> If not done properly, fsck will run upon every reboot for all crypt volumes." <- this in reference to the stock runit files, not to the BSD-style scripts that come with CRUX.
<SiFuh_> Yes I saw that
<SiFuh_> But never noticed a problem.
<SiFuh_> But ukky is quite trustworthy when it comes to stuff like this
<farkuhar> And maybe there's no problem with lvm2 either. zorz could be raising a false alarm, just to fan the flames of the fight between SiFuh_ and beerman.
<SiFuh_> I am thinking that farkuhar but ukky and zorz is probably our best source
<SiFuh_> We need to get to the bottom of it.
<SiFuh_> Even if it is a dead end
<farkuhar> ukky said he copied the lvm code straight from Void, only making minor style changes. I have no idea whether the current logic in /etc/rc.shutdown is inadequate, but I'll wait for zorz to chime in.
<SiFuh_> Yes
dlcusa has joined #crux-musl
<farkuhar> After jaeger showed SiFuh_ how to navigate the history of git commits, SiFuh_ looked at https://git.crux.nu/tools/rc/commit/e7cbffdeae4e51a0aa26af1ab65965131b5a28ad and didn't recognize it as his patch. So was the "18-year-old bug" in fact fixed even earlier, in a different commit?
<farkuhar> https://git.crux.nu/tools/rc/commit/ede456ef4488ee85e3bff9f2d00fafbc77082f25 from 8 years ago is interesting, in light of zorz's criticism. Apparently the earlier invocation of vgchange was throwing spurious warnings, and changing one flag silenced them.
<farkuhar> If we want to go back to the commit that first introduced the lvm stuff, that was 16 years ago: https://git.crux.nu/tools/rc/commit/3e2c1d35146c6837f3677a9ea8dbb19a6dd1f2f2
<farkuhar> So it looks like nobody touched the lvm commands except for those two commits by jue. And if zorz doesn't come back to elaborate on his claim, then I'm inclined to think it's a false alarm (like the report of broken `prt-get dependent` on a system without gcc).
<ukky> farkuhar: you need full disk encryption (i.e. root, swap) to check whether /etc/rc.shutdown properly unmounts LUKS volumes. The only non-encrypted MP is /boot on my system.