consider making it clear what subsystem error messages come from

[susanev]

overview

our error messages require users to understand how the subsystems are related and work together. for new users, this can be pretty frustrating. can we display the subsystem error message is coming from along with the error message? this will help users better understand how things work together and give them a better chance of being able to address the error.

[susanev]

Zaid

• worry about verbosity, would want a way to suppress if they wanted to
• does not feel that this would help beginners

[susanev]

notes from Pat:

• we can do this pretty easily
• pretty good support for attaching metadata about info we render to the user

[t0yv0]

This seems a good goal but I wonder how to make progress here. I wonder how can we scope this down or break down into reproducible use cases. I think we perhaps should scope out any presentation of errors that happen due to bugs that we're fixing, and just focus on the errors that will remain common after the bugs are fixed.

I also wonder if this is about errors displayed through the logging system (info=ERROR messages), or this is about errors shown directly through Pulumi CLI. When I mentioned it briefly to the team folks tended to assume this may be about errors being logged but now that I look at it fresh this is likely not the case.

Also what is the list of subsystems we are considering to classify the errors into that would be helpful?

• error in the user program
  • for example, code does not compile
• error from the cloud provider
  • for example, aws Bucket was given a name that is rejected by the provider as invalid, therefore failing deploy

Some overlap with https://github.com/pulumi/home/issues/2113 here potentially.