libera/#devuan-dev/ Wednesday, 2020-09-09

bgstack15Devuan meet tomorrow at 20:30 UTC. Pad is: https://pad.dyne.org/code/#/2/code/edit/7O9YAh6ECWv+Fr7ajJC+qzoE/16:01
masonbgstack15: thank you16:04
fsmithredmason, who is supposed to own the running /usr/sbin/unbound process?17:38
masonfsmithred: Launched by root, then it drops root privs and runs as unbound.17:45
fsmithredstop and restart don't seem to work in any version17:45
masonfsmithred: Make sure you've cleared away the old/bad PID. If it's owned by unbound:unbound that won't revert to root:root on its own, and process control will bail.17:46
masons/PID/pidfile/17:46
masonls -l /run/unbound.pid will probably show unbound:unbound - pkill unbound, rm that file, and the init script should start working henceforth.17:47
masonThe change in the packages I sent is to chop out the part that chowns the pidfile to unbound:unbound, but that's all it does, so you need to make sure /run/unbound.pid isn't bad when it starts. You can just rm it and it'll create a new one as root:root, and it'll stay that way.17:48
fsmithredok, it was the old pid file that gave me trouble. Your packages do restart17:48
fsmithredfwiw, the one in beowulf-backports is also broken17:49
masonI strongly suspect testing against sysvinit has just gone away, as systemd does it differently.17:52
masonThere's nothing in https://github.com/NLnetLabs/unbound/issues/303 yet17:53
masonfsmithred: I might have the wrong attitude, but as I quite like BIND and use it for all my non-forwarding installations, I've just worked out a forwarding config for it and I'm using that.19:41
masonMy unbound patch works, but I'm frankly happier just using BIND.19:42
fsmithredI've never used bind19:42
masonAh, I was thrown into it back when it was the only option.19:42
masonSame with Sendmail.19:43
masonfsmithred: Example: https://bpa.st/RQNQ19:44
fsmithredwhat does that do?19:45
masonIt's the equivalent of an unbound 'forward zone:\n\tname: "int.my.org"' block19:46
masoneach IP in "forwarders" is like an unbound "forward-addr" line, and unbound took the forward-first/forward-only concept from BIND, but called it forward-first and left forward-only as the unspoken complement to it.19:47
masonAnd with that: apt purge unbound19:48
masonI'll still help with whatever comes out of the unbound upstream. Just not going to run the stuff myself any more.19:49
golinuxThank you.  Happy to hear that.19:50
masonI could be wrong, but I don't think they'll be helpful. We'll see. Nothing stopping Devuan from shipping a corrected package, of course.19:51
fsmithredthey need the code block you removed?19:52
masonI don't think they reasonably do, but if they do they should note why in the bug.19:52
fsmithredis there an easy way to re-code it so it's optional, maybe with a setting in /etc/default?19:52
masonThat's what I was talking about with the runtime knob bit at the end of the bug. Wouldn't be wildly hard, but we'd never want the other behaviour ourselves, so they'd have to be willing to use it upstream or we might as well chop it out entirely.19:53
fsmithredyeah, I was just thinking of a way for us not to have to repackage it19:54
masonSadly, the only pidfile-related option now is "don't make one" which is even less helpful. We'd at least have to repackage to have more scaffolding, but if we're doing that, fixing it in the code isn't much more work.19:56
fsmithredI want to build refractainstaller for beowulf-proposed-updates but that's not a choice for a label. How do I add that?23:27

Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!