Update: RuleTester Improvements

[mysticatea]

Renderred RFC

I hope this RFC to be on the v7.0.0 line.

Summary

This RFC improves RuleTester to check more mistakes.

Related Issues

• Clarify non-standard properties of AST that custom parsers should create eslint#8956
• Fix: clarify AST and don't use node.start/node.end (fixes #8956) eslint#8984
• Fix: fails the test case if autofix made syntax error (fixes #11615) eslint#11798
• Nodes should include start and end for compatibility with espree typescript-eslint/typescript-eslint#405

[JamesHenry]

Would it be too much scope creep to suggest that preprocessor support is included here?

[mysticatea]

Oh. I had not considered processors. I will investigate it.

[mysticatea]

I made a separated RFC about preprocessor support #31 because this RFC focuses on improvements with breaking changes.

[platinumazure]

LGTM, thanks!

[mysticatea]

I updated this RFC:

• Remove about the deprecation of errors of a number.
• Add about to fail on syntax errors regardless of errors form.

[mysticatea]

I updated this RFC

• add about Breaking: Test with an unknown error property should fail in RuleTester eslint#12096.

[platinumazure]

Re: failing on unknown properties: It might be worthwhile to check with the typescript-eslint project, as well as anyone else who might be extending RuleTester, to see if that change breaks their cases and if we need to implement an escape hatch for them. (That said, for core ESLint rule testing, I absolutely support that change.)

[mysticatea]

@platinumazure I'm sorry for my late response. About extending RuleTester, I guess that it's a case that someone overrides run() method and implement additional checks. Therefore, I think the overriding method can remove the additional properties before passing it to super.run().

[btmills]

LGTM. These look like they'll do a good job catching some hidden mistakes. As mentioned previously, I don't know how checking unknown properties might affect third-party users of RuleTester. If it ends up causing problems, I would be fine with some sort of escape hatch or disabling that one check.