IRC channel logs

<damo22>Darelelve: how would it work if you need to specify a partition of a block device in the device string?

<damo22>for example part:4:device:wd0

<damo22>because wd0s4 does not resolve from rumpdisk

<Darelelve>demo22: Currently, the partfs translator (https://lists.gnu.org/archive/html/bug-hurd/2026-03/msg00319.html) works with a disk image and displays all partitions on that disk in the translator directory. I believe it would be possible to extend its functionality for the bootstrap system, if needed. However, perhaps you should pass partfs/wd0s4 to ext2fs instead of wd0s4.

<sneek>Welcome back Darelelve, you have 1 message!

<sneek>Darelelve, gnucode says: thanks for writing the partfs translator! I was pretty excited to read that email in the bug-hurd email list. I'm not a competent enough developer to review the code, but I look forward to getting it merged. And I look forward to you fixing licensing issues.

<Darelelve>damo22: And once again, I apologize...

<damo22>no apology needed

<damo22>does the block device for the partition appear there

<damo22>or just the name of the partition

<damo22>maybe a translator that creates a netfs with all the partition block devices under the top level would be useful, then you could attach the partfs translator to /dev/wd0 for example, and it would show /dev/wd0/s1

<damo22>but it would allow the whole disk device to be accessed on the root node

<damo22>hmm not sure how that would allow rump to drive the node

<damo22>since you can only have one translator on a node at a time?

<Darelelve>I'm not sure I understood the question correctly. I looked at the storeio code and tried to write something as similar as possible. Currently, you can write to and read from these partitions if you specify -w -T typed file:/home/usr/partfs/wd0s1 for ext2fs. Making the translator display all partitions from all disks doesn't seem like a difficult task, if I understood the meaning of /dev/wd0/s1 correctly.

<Darelelve>If I'm not mistaken, then yes, there is one translator on one node.

<damo22>the reason we need libparted is because the rumpdisk translator only knows about whole disk devices, eg wd0, so the part:X: device:wd0 allows storeio to select a partition of the opened block device

<damo22>but the whole disk is opened first

<damo22>there is no /dev/wd0s1 node involved

<damo22>you cant boot off a file:

<damo22>since there is no /

<Darelelve>I think there will be no problem booting from a block device; you can create a store from store_device_create by passing it a disk device.

<damo22>the point is, partitions of a block device need to also be presented as block devices

<damo22>that is what libparted helps with currently i think

<nexussfan>I like the new translator logos!

<Darelelve>I just checked by passing it to stat and file /partfs/device0p1 and they output that this is a block special file.

<damo22>Darelelve: the problem with your translator is that it cannot be used in bootstrap as it is, because it requires a node to attach to

<damo22>storeio + libparted accepts device:wd0 backed by rumpdisk and transparently offsets to the start of a partition

<damo22>with no root filesystem pre-existing

<Darelelve>damo22: This might be a stupid question, but how does the acpi translator work then, since it's also based on libnetfs? I thought I could try to do something similar.

<damo22>it uses libmachdev to expose a mach device in userspace

<damo22>there is a page on the wiki that explains the boot process

<Darelelve>Thanks for the explanation, I'll go study this process.

<damo22> https://darnassus.sceen.net/~hurd-web/hurd/bootstrap/

<dionysos> https://distfiles.gentoo.org/experimental/amd64/hurd/

<dionysos>you need a virtual e1000 ethernet card for rumpnet and qemu-system-x86_64 , otherwise all works the same as the i686 image

<dionysos>all packages natively rebuilt except rumpkernel (fails with strange error) and gcc (emerge hangs in python between compile and install)

<dionysos>happy holidays :)

<nexussfan>seems like this commit https://salsa.debian.org/gnome-team/pango/-/commit/11f32ade95ab71a3a09511fb6cc2e5677e42360d#58ef006ab62b83b4bec5d81fe5b32c3b4c2d1cc2_163_168 broke libpango1.0-dev on debian/hurd

<jab>nexussfan: thanks for the high praise! I really like the eth-multiplexor :)

<jab>dionysos: congrats on getting amd64 on gentoo working!

<dionysos>it was actually pretty easy, nearly everything just works

<dionysos>at least on the build level

<jab>Still is very cool!

<dionysos>next step is preparing for weekly autobuilds, but that may take a bit because our build infrastructure is quite linux-centered

<dionysos>so needs looking at it and figuring out replacements for some minor bits

<dionysos>:)

<jab>damo22: may I ask why rumpnet + pfinet on real hardware is a bit slow? I'm seeing ethernet speeds at 10 - 100 Kb/s. I'm glad that I can use the internet on modern devices. Just curious.

<jab>is anyone opposed to making the hurd's "rump" mascot a donkey? Because "rump" is slang for butt. And a donkey is also called an "ass."

<azeem>that's a bit far-fetched

<jab>so that's one vote against.

<Alicia>where did the name rump come from anyway?

<sam_> https://github.com/rumpkernel/wiki/wiki/Info%3A-FAQ#user-content-Why_the_name

<Alicia>ah! thank you sam_

<gnucode>it is a little annoying that the emacs manual is not included when you install emacs on Debian. And apparently you have to enable the non-free repo to install it...

IRC channel logs

2026-04-05.log