Hey all 👋 Is it possible to pass host env vars to pants commands executed through pantsd? I'm using an executable as a pants external tool and the binary is downloaded from S3. If I run my pants command with

--no-pantsd

the file is downloaded from S3 and it works fine, however, without the

--no-pantsd

flag, I get an error

IntrinsicError: Client error (403) downloading file

. Digging on the S3 download handler I've noticed that the

AWS_PROFILE

env var is not present when pants is running through pantsd, and therefore the AWS credentials are not correctly loaded to download the file, so I'm wondering if it would be possible to have

AWS_PROFILE

when pants is running through pantsd?

Hmmm, maybe @bitter-ability-32190 knows how to get this working?

Nothing comes to mind other than ensuring the daemon was started with the env var set. I forget which config setting allows you to add those 🤔

I tried to find an

extra_env_vars

option or something similar like we have on some subsystems but didn't find anything. Also setting the env var inline with the command doesn't seem to work, e.g.

AWS_PROFILE=foo pants lint ::

.

There's a chance it doesn't work, and I only ever loaded AWS stuff from the config on disk. I no longer work at that job, so memory is fuzzy

Potentially these values should be passed through from the client? Rather than inherited from pantsd' environment? (thanks for filing https://github.com/pantsbuild/pants/issues/21168 btw)

I think that would also work. In this case we could have an option somewhere to define which env vars should be passed through from the client? Possibly a global option? I've seem we have other

pantsd_*

global options, right?

Yeah I think thats the best solution. That way we don't invalidate the daemon too often unless we have to