Mistakenly using `https` on outbound HTTP request results in hard-to-debug error

[itowlson]

I sent an outbound HTTP request to https://localhost:3001/ when I meant to write http://localhost:3001/. It failed with just RequestError. There was nothing in the error or in Spin logs. In the end I had to hack a println! into the Spin source to figure out what was going on. This is a suboptimal diagnostic experience.

[lann]

We really should have detail strings in these error results...

[itowlson]

We totally should, but that's a breaking change, so something to maybe batch up with removing params. As an interim measure I was going to try logging something from Spin - open to thoughts though.

[michelleN]

Next step on this will be to try this again and get the current experience.

[itowlson]

The current code looks like it's meant to log at warning level but I wasn't seeing it when I tested. Investigating.

[itowlson]

It does log but the incantation to see the log is non-obvious:

RUST_LOG=outbound_http=info spin up

But I am not sure how to make this less worse within the constraints of the current design. And given that I may be the only person to ever run into this (or at least, to be troubled enough by it to bring it up), I'm not sure it should be a high priority outside of a broader redesign of error handling and reporting.

[michelleN]

I wonder if in the near term we should simply document how to use the RUST_LOG env variable with spin up. In the long term, we could consider making these kinds of things easier to see with the --verbose flag as proposed in #608 ?