IRC channel logs

2026-03-29.log

back to list of logs

<damo22>BAM is block compressed and indexable by chromosome/position (if sorted) so you can easily locate the right blocks to uncompress and display
<damo22>if you cd into a file, it could use the file ext to decide how to parse it
<damo22>and auto mount a translator
<damo22>then when you cd .. the translator goes away
<damo22>but you need somewhere to add nested nodes
<damo22>in case you want to cd into a file inside your file
<azert>damo22: the people at samtools developed block-gzip because the gzip format is very bad for random access
<sneek>azert, you have 1 message!
<sneek>azert, youpi says: yes it's mach_host_self(). Apparently it was not meant to be virtualized indeed
<key69105>damo22: the hint about Mike's patches was really helpful. This is booting now! It seems hardcoding PCI0 like that, for the root, is not accurate. This machine uses PC00 (and a quick search indicates it's also a common string there).
<key69105>See a couple them at https://lists.gnu.org/archive/html/bug-hurd/2025-10/msg00149.html
<azert>key69105: we need people
<azert>fixing booting on all kind of hw
<key69105>(I mean... a couple of hardcoded usages in that patch)
<azert>key68105: can you send a patch?
<yelninei>Hi, what is supposed to happen for rename(NULL, "") ? Currently it just crashes
<azeem>it return EFAULT on Linux
<azeem>it crashes here in glibc/sysdeps/mach/hurd/renameat2.c:
<azeem>olddir = __file_name_split_at (oldfd, old, (char **) &oldname);
<azeem>guess there could be a guard added here about oldname not being NULL and returning EFAULT in that case
<youpi>C-library functions are not supposed to take care of returning EFAULT
<youpi>they are completely allowed to just crash
<youpi>and it's a safer side, rather than returning an error that the caller doesn't check and doesn't realize that it actually failed renaming the file
<yelninei>i encountered this as part of a test that asserts that this triggers an exception (same with remove(NULL)). I guess i could ignore these but wanted to ask first
<azeem>yelninei: what software project's tests are those?
<yelninei>azeem: the standard library for dlang, https://github.com/dlang/phobos/blob/master/std/file.d#L978-L982
<sam_>ignoring that doesn't sound right? if it's part of the D contract to allow it and throw an exception or whatever, you need to do that instead
<sam_>(by not passing down garbage on Hurd)
<sam_>(ofc it doesn't sound serious so you can ignore it for now)
<yelninei>sam_: I am running the unit tests to find issues with my druntime port. As that is not an issue in D directly it would not be my top priority right now
<sam_>ACTION nods
<yelninei>speaking of issues: I get an occasional crash in the rmdir test . Am i right to assume that this can only happen if the path string gets freed by the runtime while libc is still using it?
<jab>morning frineds
<Alicia>good morning jab
<nexussfan>i need to figure out how to set up the 2 pfinet translators at the same time
<nexussfan>right now setting 2 causes 26 to not work
<nexussfan>setting 26 causes 2 to not work
<youpi>you need to set them up as passive
<youpi>otherwise if you set them up as active, of course they conflict with each other
<nexussfan>oh
<youpi>it's not two translators
<youpi>it's just one that auto-attaches itself to the other place
<nexussfan>i thought you need /servers/socket/2 and /servers/socket/26?
<nexussfan>oooh
<nexussfan>it automatically does the other one
<youpi>not completely
<youpi>you still need to set the passive record on both
<youpi>so whenever you access one and thus trigger it, it attaches itself to the other
<nexussfan>so settrans -fgap /servers/socket/2 /hurd/pfinet -i /dev/wm0 -6 /servers/socket/26
<nexussfan>or wait
<nexussfan>not pfinet
<youpi>yes, but for the other one you wouldn't want the -a
<nexussfan>lwip
<nexussfan>youpi: oh okay
<youpi>otherwise you're killing your first instance
<youpi>you just want to set the passive record, in case at reboot you access 26 before 2
<nexussfan>alright
<nexussfan>now both 26 and 2 are set :D
<nexussfan>eh, still getting the Inappropriate ioctl for device
<nexussfan>looks like i can't use lwip for now :(
<youpi>what inappropriate ioctl ?
<nexussfan>or do I need to provide the -a -g and -m ?
<youpi>depends what you are trying to achieve
<youpi>remember the xy problem
<nexussfan>inetutils-ifconfig: SIOCSIFBRDADDR failed: Innappropriate ioctl for device
<nexussfan>trying to get lwip to work
<youpi>lwip does seem to have some implementation for that
<youpi> /* 19 SIOCSIFBRDADDR -- Set broadcast address of a network interface. */
<youpi>SIOCSIF (brdaddr, BRDADDR);
<youpi>but perhaps it's somehow not plugged correctly inside lwip
<nexussfan> https://paste.debian.net/hidden/68da8453
<nexussfan>it's certainly getting the IP it seems
<nexussfan>although i can ping 1.1.1.1?
<nexussfan>yeah for some reason connectivity works
<nexussfan>but it shows an error
<nexussfan>ifup: /etc/network/interfaces:18: unknown or no method and no inherits keyword specified
<nexussfan>ifupdown has ipv6 issues for me
<nexussfan>> Indeed, IPv6 now works properly, and the very machine hosting this wiki (darnassus.sceen.net) can be reached with that protocol.
<youpi>ah, no, we have ipv6 issues with the host of the vm
<gnucode>can someone try reproducing this netsurf bug for me? Install netsurf and then try to run netsurf. I cannot reliable start netsurf via "$ nsgtk3" (I've had to compile netsurf from source, because it's not packaged properly for Debian due to path_max issues).
<gnucode>the only way that I can reliably start netsurf is via "$ gdb nsgtk3"
<gnuhurdfan>i'll see if i can install it
<gnuhurdfan>weird, my libgtk-3-common is 3.24.51 while libgtk-3-dev is 3.24.52
<Alicia>it doesn't seem to be in the repository for x86_64, aside from netsurf-common
<nexussfan> https://buildd.debian.org/status/package.php?p=netsurf
<nexussfan>PATH_MAX failure
<nexussfan>might submit a patch to fix it upstream
<gnucode>Alicia: I had to compile netsurf from source. :)
<gnucode>nexussfan: it would be awesome if you would port it to the Hurd!
<gnucode>It's the main browser that I use on the hurd.
<Alicia>gnucode: ah. I think I misread, apologies
<gnucode>no worries.
<gnucode>Did you all see that doom has been ported to mostly css ?
<gnucode>a fair amount of the runtime and login is in js, but there's a dude that used some AI to code most of doom running on CSS.
<gnucode>nexussfan: let me know if you ever get lwip working semi or reliably. It would be nice to document how to actually use it.
<gnucode>Alicia: do you run the hurd? If so, what browser do you use?
<gnucode>damo22: may I add a link to your website on this web page? https://hurd.ion.nu/donate.html
<Alicia>gnucode: not consistently. I haven't had any browser working beyond maybe links
<Alicia>I tried to compile icecat and ran into some issues I could fix (mostly hacky fixes, but there's a patch for the multiprocessing python module I should submit), but then one I couldn't get past with rustc having memory allocation issues
<gnucode>if you run it again, I might recommend netsurf or dillo. Simple web sites work a-ok.
<gnucode>Alicia: I saw that you were working on that!
<gnucode>I thought I read a blog today that Firefox (and LadyBird) are allowing AI generated code into the code base...
<gnucode>this page is pretty out of date: https://hurd.ion.nu/microkernel/mach/gnumach/hardware_compatibility_list.html
<Alicia>oh!
<Alicia>I should submit the machines I've tried to there
<gnucode>I wonder if we should merge hardware_compatibility_list_ with https://hurd.ion.nu/faq/drivers.html ?
<gnucode>damo22: can you ask brauner if I can start linking to x15 from the Hurd ng wiki ? It would be nice to document his effort.
<Alicia>oh yeah, I was also trying to trim down icecat to isolate the rustc memory allocation issue, but I lack familiarity with rust and its cargo structures
<gnucode>I've heard packaging rust is not so easy for distrubitions to do.
<Alicia>it seems very stubborn about downloading things instead of compiling things
<azeem>gnucode: where did you read that he used AI for Doom CSS? I can't find a reference in the github repo itself
<azeem>also Rust is very stubborn in linking things statically
<ik5pvx>]
<gnucode>azeem: I think I saw it in the blog post.