IRC channel logs

2021-07-02.log

back to list of logs

<durango>I spent the evening trying to install Hurd 0.0 (with nethack included) on virtualbox . ‘97 Grub doesn’t recognize the partitions created by modern netbsd. And ‘97 netbsd doesn’t seem compatible with virtual box ide drive. Any ideas?
<durango>Do you guys remember the last version of netbsd before the transition from 4.2BSD disklabels to their own? I could try that.. another thing is to try bochs instead of virtualbox
<durango>Or finding a functioning 586..
<damo22>durango: it works in qemu with the image from the topic
<damo22>hurd does that is
<durango>damo22: I did that already. I want to install hurd-0.0 from 1997
<durango> https://www.gnu.org/software/hurd/history/hurd-flash13
<durango> http://www.ibiblio.org/pub/historic-linux/ftp-archives/prep.ai.mit.edu/Sep-29-1996/gnu-0.0/
<damo22>why?
<durango>because I find that more interesting
<durango>from an historical perspective, in '97 you had NetBSD 1.3, Linux 2.0 and the Hurd 0.0. And all of them could have become the GNU kernel
<damo22>well weve come full circle and now we are trying to use NetBSD 9 kernel drivers in userspace
<damo22>i mean who wants to reinvent drivers for hardware
<durango>damo22: I completely agree
<kilobug>durango: there is something I don't get with your problem, if you're using a virual machine (virtualbox or whatever) why do you need the partition table to be compatible between host and guest system ? can't you use a disk image with its own partition table for the guest ?
<durango>kilobug grub in '97 expected NetBSD partition/labels. It turns out it is extremely difficult to create such an image right now
<kilobug>durango: hum I'm pretty sure it worked with standard old-school DOS partition tables
<damo22>durango: what about fdisk and using a BSD label
<durango>kilobug read the announcement https://www.gnu.org/software/hurd/history/hurd-flash13
<durango>it seems like the development system for GNU was NetBSD
<damo22>durango: Command (m for help): b
<damo22>There is no *BSD partition on /dev/zero.
<damo22>The device (null) does not contain BSD disklabel.
<damo22>Do you want to create a BSD disklabel? [Y]es/[N]o: y
<damo22>or use parted, mklabel bsd
<durango>Hurd never worked: https://imgur.com/a/03DGYFm
<durango> https://imgur.com/a/DyMg9Qm
<durango>jrtc27: "it worked?"
<durango>on hurd version 0.0 I can run 'while true; yes & kill $!; done' without hanging the system. I'm satisfied
<durango>on hurd version 0.0 I can run 'while true; do yes & kill $!; done' without hanging the system. I'm satisfied
<durango>well.. not really
<durango>my macosx that runs my hw feels soulless in comparison
<psydroid>why couldn't much of the work done on Mach in Mac OS X be merged back into GNU Mach?
<psydroid>seeing that it got ported to several architectures
<youpi>because Mach kernel sources have diverged a lot
<youpi>so it's not an easy thing to do, far from it
<durango>I can imagine things like partition tables, device number and names have all been Linuxed
<dstolfa>to be fair, linux does have pretty reasonable device names
<dstolfa>way more reasonable than other systems
<jrtc27>linux is a weird halfway house
<jrtc27>freebsd has tons of different names for things based on which driver it is
<jrtc27>linux gives you sdX for most block devices (though I think NVMe is differentiated?) and ethN for most NICs
<jrtc27>but not always
<jrtc27>macos is the other extreme, everything is diskN
<jrtc27>also, why use letters rather than numbers for disks...
<durango>yep but I guess the greatest divergencies are substantial like multithreading and other features that are more common today than back then
<durango>porting glibc to xnu
<dstolfa>jrtc27: i haven't had linux give me `ethN` and `sdX` for at least 8 years now
<dstolfa>it was always `enpXsY`, `wlpX..........` and so on
<jrtc27>that's all done by udev in userspace
<dstolfa>i mean yeah, i should have said "linux in the eyes of an end user" instead of "linux the kernel implementation"
<jrtc27>and sdX are absolutely still around
<jrtc27>unless you have NVMe, when it's the saner nvmeN
<jrtc27>hm, nvmeNn1 apparently, not sure what the 1 is for
<dstolfa>jrtc27: i think the n1 is the controller namespace
<dstolfa>at least that's what i thought it was
<dstolfa>i'm not sure anymore
*dstolfa -> food
<youpi>it is, yes
<jrtc27>yep, found it
<paulusASol>durango: I guess the main advantage Linux had in 1996 over the HURD was already having attracted the developers.
<gnu_srs2>damo22: Seems like arbiter+ext2fs shows the problem already w/o rumpdisk:
<gnu_srs2>From https://paste.debian.net/plain/1201148:
<gnu_srs2>lrwxrwxrwx 0 4294967295 root 13 Jan 1 1970 /proc/2/exe -> /hurd/startup
<gnu_srs2>lrwxrwxrwx 0 4294967295 root 13 Jan 1 1970 /proc/5/exe -> /hurd/startup
<gnu_srs2>and no no fs process, as for normal boot: lrwxrwxrwx 0 4294967295 root 2 Jan 1 1970 5/exe -> fs
<durango>The amount of work that is still being poured in their Mach by Apple engineers still today is impressive: https://github.com/apple/darwin-xnu/tree/main/osfmk