#unmarshal error

1 messages · Page 1 of 1 (latest)

versed marsh
#

Could you share the module's source code? And/or execution logs (ideally a dagger cloud run url)

prisma elbow
#

Unrelated to the error, I've just updated this to 0.9.10 and I noticed two things immediately - the --help output now lists all of the SDK functionality (which doesn't feel necessary?), and the module returns Container evaluated. Use "dagger call echo --help" to see available sub-commands. without adding stdout to the call method

fallen nexus
#

@prisma elbow can you verify if locally is also working correctly with your custom engine image

#

?

#

you can do this by setting the _EXPERIMENTAL_DAGGER_RUNNER_HOST variable to docker-image://$image

prisma elbow
#

I tried that earlier, it doesn't run - can't create directory '/sys/fs/cgroup/init': Read-only file system

fallen nexus
#

it'll be faster than troubleshooting here

versed marsh
#

@prisma elbow I was able to load your module cleanly:

$ dagger -m github.com/Mjb141/daggerverse/echo call echo --word 'bonjour monde!' stdout
Here's a word... bonjour monde!
fallen nexus
#

@versed marsh we managed to find the issue around the module loading problem. Different CLI and engine versions. Now @prisma elbow is having a different issue which seems to be related to connecting to the engine via the TCP transport unstead of unix sockets

#

I'll try to make a repro now

#

yep, I was able to repro. cc @prisma elbow

#

@prisma elbow as @granite moon mentioned here: #maintainers message

the solution is to add another --addr flag to your command definition with unix:///run/buildkit/buildkitd.sock

prisma elbow
#

Will give that a go now

fallen nexus
#

so basically: command: ["--addr", "tcp:....", "--addr","unix:///run/buildkit/buildkitd.sock"]

prisma elbow
#

Putting runner back up, should have a test in a minute or two

fallen nexus
prisma elbow
#

That's worked - job succeeded 🎆

#

Thanks for your help with this!

fallen nexus