Errors and GNU compatibility

Nobody has claimed this yet.

TL;DR: I think we need a policy for error compatibility with GNU. We can pick either:

Errors should be identical to GNU.
Errors can differ from GNU and we should aim to make them better than GNU.

We advertise a goal of (almost) 100% GNU compatibility and I think we need to figure out whether that includes error messages. We currently have some parts of the code which try really hard to give the same errors and other parts which blatantly ignore the GNU messages. As a community, we need to decide what we want do to here.

My personal opinion is that we should not make error messages identical to GNU, because I think that this is an area where we could provide a significant improvement over GNU, by making prettier and more informative messages. The code also usually becomes simpler because we need to jump through fewer hoops to match GNU. However, this means that we need to patch the GNU test suite a lot where they test for specific messages.

Note: I'm not advocating that we change when messages pop up, only what the messages look like.

Questions to discuss:

Is an improvement in error messages worthwhile?
Are there scripts which rely on specific messages that we would break by introducing different messages?
Would you expect a drop-in replacement for coreutils to have identical error messages?
What is your personal preference on this?

Once we settle on something, I propose we add it to the documentation and the contributors guide.

Random selection of some relevant issues/PRs:

Start by reading this issue and reviewing the referenced issues and pull requests, especially the GNU compatibility discussions and test-suite changes. Inspect the current documentation and contributors guide to see where a policy belongs. Done means the community has decided how error-message compatibility should work and the decision is documented.

Written by the indexing model from the issue text.

Description

Contributor guide

First steps

Research direction

Assessment