IRC channel logs

2023-10-25.log

back to list of logs

<gnucode>hello friends!
<gnucode>is anyone going to be annoyed with me if I toot on fosstodon about our bcachefs talk ?
<gnucode>I realize that there are no Hurd developers actively porting bcachefs to the Hurd (and probably won't be).
<gnucode>but it was still a cool talk, and I would like others to be able to watch it.
<gnucode>and I want to encourage people to donate to Kent's patreon.
<psydroid>can you just link the talk here?
<gnucode>give me a second...
<gnucode>sure.
<gnucode>psydroid: also please don't post the link to phoronix. youpi didn't want me to do that. He didn't want people to accuse the Hurd project of being vaporware.
<psydroid>gnucode: I avoid phoronix myself, so that shouldn't be an issue
<gnucode>psydroid: thanks. I have to go and find the link again. Let me browse through bug-hurd archives.
<psydroid>gnucode, no problem, take your time
<gnucode> https://lists.gnu.org/archive/html/bug-hurd/2023-09/msg00073.html
<gnucode>also I am currently trying to transcribe solid_black and damo22 talk about a simplified Hurd bootstrap/boot up idea.
<gnucode>irc log 2023-07-18
<gnucode> https://logs.guix.gnu.org/hurd/2023-07-18.log
<gnucode>Any suggestions on where it should go on the wiki would be welcome.
<psydroid>I can't say anything about that. Let's hope someone who is officially affiliated with the project jumps into the channel later today.
<gnucode>at the very least I can transcribe it.
<psydroid>are you a kernel/driver developer yourself?
<psydroid>I only know Linux and I only wrote a driver a few years ago for my OS class
<gnucode>psydroid: I appreciate the compliment, but I am primarily a documentation writer for the Hurd.
<gnucode>I probably need to spend more time writing C code, to be a kernel/driver developer.
<gnucode>also weirdly
<gnucode>my Hurd on bare metal almost locked up just now.
<gnucode>I closed xfce4-terminal, and it began running smoothly again.
<gnucode>maybe I need to play with using a different terminal.
<gnucode>also git send-email appears to not work on the latest 32-bit hurd.
<gnucode>grrr.
<psydroid>being able to run it on bare metal is quite some progress
<gnucode>psydroid: I actually think that the Hurd is easiest to run on bare metal.
<gnucode>I have been trying to use the Hurd as an occasional daily driver for several years.
<gnucode>I bought this old T43 with 1.5 GB. And it works.
<gnucode>I have filesystem corruption that I have to deal with about once every two weeks.
<gnucode>there are some other minor annoyances, but mostly I'm satisified.
<gnucode>I need to go get lunch ready.
<psydroid>gnucode, I don't think I've ever tried to run it on bare metal
<danmorg>gnucode: you're are definitely an inspiration for the Hurd. I have been lurking here for weeks. I would try to more C and become a kernel / driver developer. You can even write about your experience and this will encourage others to get involved. If anything the Hurd needs, is the documentation you provide... Thanks! I installed the Hurd weeks ago on a VM. but didnt know what to do after
<psydroid>all-right
<danmorg>gnucode: I am sure you know about the https://wiki.osdev.org/Main_Page for those working on the internals of an OS. I found this tutorial on writing your own OS in assembly and C. Not so much to write my own, but it has helped me understand the low-level hardware details of protected mode 386 OS. Many OSes abstract the low-level details away.
<danmorg>hmmm. i should try the Hurd on bare metal. does it play well with other OSes on a MBR partioned drive? I have Debian Linux and Minx 3 on it and use GRUB as the boot manager. and i have an available primary partition i could use.
<kilobug>danmorg: GNU/Hurd requires Grub as bootloader so you should be fine, but its hardware support is limited, you'll have to check if it supports your hardware or not (sorry I didn't run it on bare hardware in a very long time so I don't know the current status)
<danmorg>I am using GRUB now.
<danmorg>I usually use a live DVD of Debian linux to repair the GRUB after installing an OS. or use a partioner such as gparted which i have to install in the live os. my first partition (MRB partioned) is used as root \ for kernels and grub. i assume the Hurd would put its kernel there too.
<gnucode>psydroid: thanks for the encouragement. That means a lot.
<gnucode>danmorg: I have heard about it before...I should give it a try.
<gnucode>danmorg: also I have been told that the Hurd runs on a Thinkpad 510. I actually bought one recently, but I have not tried to get it set up yet.
<gnucode>my blog gnucode.me will probably eventually have the details.
<gnucode>gotta run.
<luckyluke>are glibc's argz_next() and friends used only in hurd? I think I found a bug in argz_next() where a pointer is truncated to 32 bit...
<youpi>it's used in some code of glibc
<youpi>and other software, in debian
<luckyluke>uhm, let me add some context... I was trying to understand why the mtab translator crashes on my hurd64 vm, and I found that the strings returned fron fsys_get_children(), when unpacked with the argz_*() rountines , crash
<luckyluke>and apparently, if I take the same code as from argz_next() but change to 1ULL here https://sourceware.org/git/?p=glibc.git;a=blob;f=string/argz-next.c;h=dececc37120fe78de78d4551a3e43d2bae232422;hb=HEAD#l28 then it works
<luckyluke>otherwise, the value returned from argz_next is truncated to 32 bit
<luckyluke>it might be related to how I compile my code, or how glibc is compiled; if I copy the argz_next() code as-is, it still works, but if I use argz_next() from the library it fails
<luckyluke>I'm a bit confused :/
<youpi>changing 1 to 1ULL shouldn't change anything here: strchr returns a char *, so we get normal char * pointer arithmetic here
<luckyluke>yeah maybe it's the wrong lead, if I compile the code myself I don't need the ULL, but if I use the glibc version the result is truncated
<etno>This is mysterious, reminds me of that time when the compiler didn't know about a function prototype and decided that the return type is (int)
<youpi>luckyluke: you don't get any warning when building it ?
<luckyluke>no, I'm just using -Wall
<luckyluke>ah no wait, yes
<youpi>it's possibly missing some #include
<youpi>and then, yes, there's a truncation, but not due to +1, but to assuming the return type is int
<youpi>but string.h is supposed to bring the strchr prototype
<etno>That's what I would imagine, yes
<luckyluke>aaah yes that was it... missing argz.h :(
<etno>Yay!