<Gamayun>Hmm... I'm trying to offload building to another machine. GUILE_LOAD_PATH seems to be set, and the guix modules seem to be in the load path, but whenever I run 'guix offload test', I get the error message about guile modules not found. Any ideas?
<roptat>you still need to know what you're looking for
<roptat>DHT indexes content by their infohash, which is derived from the content itself. Unless someone knows what that content is and tells you what the infohash is, you can't know that value in advance
<roptat>so you need a substitute server to build the substitute, compute the infohash and give you the info
<roptat>kristofer: you have to find a PID from the herd inside the container, then run something like "ip link add type veth peer ns <pid>" which adds a veth pair, one in the current namespace and one in the namespace of the container
<roptat>then you have to connect that veth to your actual network device somehow and assign ip adresses manually too
<kristofer>roptat: the container in question should have a static ip
<kristofer>roptat: I'm trying to decide which is better. I think a container would be more suitable on my somewhat antiquated hardware
<roptat>you can assign IPs from inside the container
<roptat>or outside the container: "ip netns exec <pid> ip a add veth0 10.0.0.2/8" or something
<roptat>the veth connects the container to the host, then it's up to you to decide how it's connected to the internet
<roptat>you can enable ip forwarding and use iptables to masquerade internal IPs, or you can use a bridge with brctl
<pkill9>so guix containers are safe to use as sandboxes?
<nckx>Is it just me or is issues.g.i more down than up? (504)
<g_bor>I have another version, so I will try that :)
<nckx>We're talking about a properly installed (i.e. not local git ./pre-inst-env'd) guix, right?
<g_bor>yes, absolutely, but a pre-inst-env one is also not working, same error.
<g_bor>This is somewhere so early, that issuing guix without any arguments fail.
<g_bor>I have ran gui gc earlier today, but not problem showed then.
<nckx>Then I really have no idea how the reference scanner allowed libgcrypt to be GC'd. It's either something so trivial that it didn't occur to ask, or a very grave bug. I'd guess the former for now :-)
<g_bor>Then did a pre-inst-env guix build --rounds=2 first, which failed because of a non-reproducible package.