feat: apply ESLint bulk suppressions below ESLint 10

[alexander-akait]

Summary

Closes #291. Supersedes #292, whose conflicts this carries forward instead.

Bulk suppressions now work on every ESLint this plugin supports:

• ESLint 10 takes applySuppressions and suppressionsLocation as constructor options, and the plugin already hands unrecognised ESLint options to the ESLint class as they are. Nothing needed.
• ESLint 9.24 and later ships the same SuppressionsService but wires it into its CLI alone, and the constructor rejects the options outright with Invalid Options: - Unknown options: applySuppressions. The plugin now drives the service itself after linting.
• Below 9.24 the option raises an error naming the version it needs, rather than silently reporting the very violations the user recorded as suppressed.

The ESLint 9 path resolves the file and the paths recorded inside it against ESLint's cwd, and reads the default name eslint-suppressions.json when no location is given — the same rules ESLint 10 applies, so one config behaves identically on both and the tests run unmodified against each.

Two details behind that:

• ESLint's exports hides the service, so it is loaded by absolute path rather than by specifier. The package directory comes from resolving eslint/package.json, which respects eslintPath and works on Windows. feat: support bulk suppressions #292 stripped /lib/api.js off the resolved entry with a POSIX-only regexp, so its fallback silently did nothing there.
• DEFAULT_SUPPRESSIONS_FILENAME is an ESLint 10 addition. ESLint 9 hardcodes the name in its CLI, so reading the constant gave undefined and a paths[1] must be of type string crash — caught only by running the suite against 9.

Credit for finding the bug and for the shape of these cases goes to @viddo.

What kind of change does this PR introduce?

feat.

Did you add tests for your changes?

Yes. test/suppressions.test.js covers eight cases: no file, the default file, a named suppressionsLocation, an absolute one, a nested file recorded relative to cwd, a suppressions file outside cwd reached with ../, partial suppression that still reports the rest, and a location naming no file. They are gated at ESLint 9.24 and run on both majors — 132 passing on ESLint 10, 134 on ESLint 9 (the two eslintrc suites also run there). I checked they hold something: deleting applySuppressions from getESLintOptions turns three of them red on ESLint 10, and the ESLint 9 path was written against a failing suite.

Does this PR introduce a breaking change?

No.

If relevant, what needs to be documented once your changes are merged or what have you already documented?

The README's Suppressions section under ## ESLint covers the options, the cwd caveat and the version requirement. The changeset is a minor.

Use of AI

Written with Claude Code, driven interactively. I asked it to fold #292 into this PR; it established that ESLint 10 makes #292's private-API module unnecessary, wrote the ESLint 9 fallback against the same service, and found the two portability defects above by running the suite on both majors. I reviewed the result.

Note

The first commit of this PR reached main already, through #311's squash while this was stacked under it. What is left here is the feature.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GzZci4NQeiqwdrVfd7dGXy