IRC channel logs

<damo22>raid is not a good idea if its in hardware

<damo22>if the controller dies, you lose all data

<damo22>would be cool if hurd supported mdadm

<damo22>block based software raid should be trivial in hurdish land

<nikolar>trivial is a strong word :)

<nikolar>but yeah, it should be relatively easy

<damo22>i mean, mirrored block raid would just be a translator that opens two or more block devices and performs writes to both?

<nikolar>how do you recover in that case

<nikolar>from corruption

<nikolar>or from one disk resetting while you are writing to it

<damo22>something like what linux does i guess

<damo22>syncs all the blocks?

<nikolar>i meant, if you read two different blocks, how do you know which one is correct

<nikolar>*two different blocks from the two drivers

<damo22>i dont know actually

<nikolar>it's non-trivial

<Gooberpatrol66>usually you need checksums to prevent bitrot

<Gooberpatrol66>like btrfs has a merkle tree of all the data

<Gooberpatrol66>then the checksum demonstrates which raid copy is correct

<sneek>almuhs, you have 1 message!

<sneek>almuhs, youpi says: file_name_lookup is the client part of looking up a file, it's inside glibc, hurd/hurdlookup.c

<almuhs>youpi: i don't file the path hurd/hurdlookup.c . I neither find the glibc sources

<almuhs>**i don't find

<damo22>almuhs: the glibc salsa does not have glibc sources

<damo22>you have to download the tarball from somewhere and overlay the debian git

<almuhs>in savannah neither there are glibc sources. The repository seems empty, and the secundary branchs are deprecated

<almuhs>and i find in Debian GNU/Hurd filesystem, and i don't found neither

<almuhs>although i don't understand this ternary operation in return at all

<almuhs> return err ? (__hurd_fail (err), MACH_PORT_NULL) : result;

<azeem>which file system is this on Debian?

<azeem>root 103 0.0 0.0 166M 1.11M - S<o 7:38AM 0:00.06 /hurd/tmpfs --nosuid --size=10% --mode=755 tmpfs

<azeem>and/or is there a way to figure out where it is mounted by asking the process?

<azeem>(cause I had a crash on it recently)

<youpi>df -h should be showing the potential places

<youpi>(a translator can attach to several places actually, e.g. pfinet)

<azeem>ok right, so it is/was /run

<youpi>(using stat on /run will tell the pid in the "Device" part)

<azeem>the crash was the following, not sure whether that is useful: https://paste.debian.net/1392506/

<gnucode>morning friends!

2025-08-20.log