IRC channel logs

<rsmarples>hello again :)

<rsmarples>does Hurd have any SO_PEERCRED support?

<youpi>nope

<youpi>there is some SCM_CREDS experimental support, but it's not finished (it's not actually checked...)

<rsmarples>SCM_CREDS is also racy :S

<youpi>racy to what extent ?

<rsmarples>iirc you have to set a socket option after creating the socket

<rsmarples>so there is a race with a client opening the socket and sending something

<youpi>no, it's not an option

<youpi>it's a message sent from one side and received by the other side

<youpi>so it's less convenient because the requester needs actions from the requestee

<youpi>but otherwise I don't see why there would be a race

<youpi>the requester will just wait for the requestee to provide the cred

<youpi>good thing being that the creds will be of the process that sends the message, and not of the process that created the socket (and possibly passed to another process)

<youpi>ah, maybe you are confusing with SCM_CREDENTIALS.

<rsmarples>quite likely

<youpi>which indeed needs an option

<jab>morning friends!

<yang>hello jab

<user_oreloznog>\o

<jab>what are ya'll up to ? I'm listening to the autobiography of St. Ignatius of Loyola: https://librivox.org/the-autobiography-of-st-ignatius-by-st-ignatius-loyola/

<yang>The Inquisitor?

<jab>as far as I know he was never an Inquisitor

<jab>he was the founder of the Jesuits.

<jab>apparently he "prayed for 7 hours a day on bended knee." That might have been when he first got started.

<yang> https://bigsiteofhistory.com/the-jesuits-and-the-inquisition-1540-1556-the-protestant-reformation/

<rsmarples>youpi: i think that behaviour must be FreeBSD specific? NetBSD and OpenBSD require socket options for SCM_CREDS

<jab`>good to know that the Society of Jesus was the main charge of the Inquisition. I didn't know that. The biography of St. Ignatius is still pretty cool. Miracles, lots of prayer, fasting, etc. Something that I want to try to emulate.

<jab`>hmmm, why did my irc nick change suddenly ?

<jab`>hmm

<jab>?

<jab>man irc is confusing.

<yang>jab`: the irc client uses secondary nickname, if the first one is taken

<youpi>rsmarples: afaik when the creds are explicitly pased, you don't need an option

<youpi>(cmsgcred explicit message)

<jab`>oh, I did just change my ethernet connection from a switch to a router...

<jab`>whatever.

<yang>yes, then your first nickname times out, and with new reconnect you get secondary nickname

<youpi>rsmarples: even in the implicit cred passing with the LOCAL_CREDS option, it doesn't seem to have a race, as in the cred will be attached on the next sent data

<youpi>so the receiver of the data will wait for the data to come anyway, and the creds will be along it

<youpi>at least that's what I understand from https://man.freebsd.org/cgi/man.cgi?unix(4)

<youpi>netbsd indeed doesn't seem to have an explicit cmsgcred in https://man.netbsd.org/unix.4

<youpi>what I don't know is whether a control message can be received without any data

<rsmarples>youpi: well I must be doing something wrong anyway or Gentoo/Hurd is broken as I can't get it to work

<youpi>the patch is not upstream

<youpi>since it's not finished

<rsmarples>.....

<youpi>(actually I probably shouldn't have applied it to debian at all, since apparently it dropped any incentive to actually finished it....)

<rsmarples>ok, so no point in writing that code.

<youpi>what code?

<rsmarples>ACTION thinks

<rsmarples>youpi: the SCM_CREDS code

<rsmarples>for dhcpcd

<youpi>well, in the end it should be working

<youpi>it's a matter of somebody taking the time to finish the support

<youpi>and I'll be happy to push it upsteram

<rsmarples>after dealing with 40+ AI detected "security" issues I have no time (at least short term, lol)

<youpi>that's unfortunately the case of a lot of people :/

<rsmarples>ok, I'll have to harden the socket at the filesystem level

<rsmarples>short term solution, just makes the reader group disabled and you have to be a writer

<gnuckb>good afternoon guys an' gals!

IRC channel logs

2026-08-05.log