Artem Pelenitsyn

3.8 looks beyond salvation. Should we close this?

## Meta I'm a little worried about how this fits into a bigger picture of improving solver's output. See e.g. #8475. Especially I'm always nervous when we add more information...

> I'd really rather not tackle unicode output on this pull request but would be happy to investigate it on a separate issue. Fair enough. Let's wait up to one...

> especially in the case where there's only one cabal.project with no further imports. Indeed. I'd potentially look into special-casing one project file and see if the output can be...

> Do the tests' .out files show enough detail of before and after? Well, diffs are not always pleasant to read. Also, it'd be interesting to see deeper nesting than...

> I hope all the project files are actually used and not dead code themselves... I'm afraid many of them are dead wood. Most used are cabal.project and cabal.project.validate —...

Good! Then this is the third one we know is in use. The rest is still under a big question mark.

Thank you for the analysis. I expected them to show up in the Makefile. The issue is: I don’t think the Makefile itself is widely used. Just look at all...

Oh, right, the `style` (and related) target is the recommended way to do it. So, I patch the rule above: > More radically, I propose to cut all the project...

CI wouldn't help if tests depend on something in the root project file because the file is present in CI as well? I think it's the right thing to do...