IRC channel logs
2026-06-30.log
back to list of logs
<solid_black>pushed a part of getcwd implementation, namely io_identity, fortunately libfshelp does the hard part for us <solid_black>the other missing part is supporting '..' lookup at the root to go to the parent translator <solid_black>the hard part about that is understanding that we are, in fact, at the root :D <diegonc>hmm there's something wrong with diod. it goes out of the shared directory if enough .. are given :P <diegonc>demo@debian:~/dev/host-hurd$ cat ../../../etc/passwd <Alicia>are you sure that's from the remote filesystem and not the local filesystem? because /.. also leads to / for me (tested on linux) <diegonc>ok, that was not a good example. but yeah, if I'm at the root (the directory where diod export is "mounted") in Hurd and do "ls .." ,I get the list of files in the debian host where diod is running <diegonc>hm, it doesn't happen when mounted on linux with diodmount. maybe it's a manifestaion of what is missing for getcwd (9pfs knowing when it's at root) <damo22>hmm i wonder why smp fails with user32 <damo22>its stuck on timer calibration... <solid_black>sneek: later tell diegonc, I implemented tracking root & walking out of the root <solid_black>sneek: later tell diegonc, getcwd now works :) please test <solid_black>9P is really nicely designed here, it lets us both do server-side multi-part path lookups -- in a single roundtrip -- <solid_black>and returns the info we need to notice that we should have walked out of the root <solid_black>we're going to use the same trick to implement active translators <solid_black>walk server-side, but later notice if a node we traversed is translated, in which case re-do the walk from that point <jab>solid_black: you are up late! <jab>oh nevermind, it's afternoon there! <jab>I'm working on the latest qoth. <jab>Has anyone tried using Fil-C on the Hurd yet? We can write memory safe translators in C on the Hurd...