That makes sense.

I'm left wondering, then, why superlinter would enable two conflicting linters out of the box? That effectively makes superlinter broken/unusable out of the box?

I get why superlinter wouldn't want to "pick a winner" here. But it seems counterproductive to the benefits of superlinter, if superlinter is broken out of the box?

Shouldn't biome be disabled by default, if it's not possible to run with prettier? (Or at least automatically disabled while prettier is enabled, which it is by default)

Yeah, this is a discussion that came up a few times in the past. Ultimately, we decided not to introduce breaking changes.

I think that FAIL_ON_CONFLICTING_TOOLS_ENABLED might be helpful here. If you set it to true, it will make Super-linter exit with an error if you enable potentially conflicting tools.

I think that FAIL_ON_CONFLICTING_TOOLS_ENABLED might be helpful here. If you set it to true, it will make Super-linter exit with an error if you enable potentially conflicting tools.

Yeah, that would be helpful.

But barring that, introducing known-conflicting tools without a major-version bump would also be nice. Or maybe don't have them added on-by-default? Make new/known-incompatible tools opt-in?

Ultimately, we decided not to introduce breaking changes.

Not sure what this means. This biome tool is brand new to super-linter (heck, the tool itself isn't even that mature), it introduces breaking changes, and was in a minor version bump just a couple months ago. If this problem has been raised before, maybe biome shouldn't have been added in a minor version bump? or maybe it should have been opt-in so it doesn't break existing users?

Not sure what this means. This biome tool is brand new to super-linter (heck, the tool itself isn't even that mature), it introduces breaking changes, and was in a minor version bump just a couple months ago. If this problem has been raised before, maybe biome shouldn't have been added in a minor version bump? or maybe it should have been opt-in so it doesn't break existing users?

Adding more context: Super-linter enables all linters and formatters by default. Disabling some (such as Biome) by default would mean a breaking change.

To your point about updates: we might need to be more conservative, but linters and formatters add new checks all the time, so each time we bring a linter or formatter update, there's a risk of flagging new issues that weren't flagged before.

Having said that, I get your point of introducing new linters in major versions only. There might be room to improve things, although I'm not sure about how much effort it will take. I think that we might need a policy to decide which update goes in which release.

Disabling some (such as Biome) by default would mean a breaking change.

It's only a breaking change because biome is in now. It wouldn't have been a breaking change if it was added opt-in from the beginning. That's my point.

And I know new linters are added all the time. That's fine. That can cause it's own pain, sure. Because you never know what kind of codebase those new linters will suddenly be running in. But for any given new linter, there's only an X% chance of it breaking their build immediately, where X is probably somewhere below 50%.

Contrast that with biome in particular as an example. Given that biome lints a whole bunch of files (not just some obscure language or format), and that it's known to conflict with prettier which also lints a bunch of common files, then chance of breakage is more like 90%. That's why biome in particular is more problematic than any arbitrary linter.

A similar situation would have occurred with eslint and standardjs, back in the day. Those two were incompatible. But the blast radius for those was at least limited to javascript. Biome and prettier lint/format much more than just JS (markdown, yaml, json). So while prior conflicting linters did exist, the blast radius for biome being introduced is much much larger.

As a point of comparison: rubocop the ruby linter also releases new cops under feature (minor) bumps. However, new cops are introduced off by default. Then in major version releases, all the new cops that were introduced since the last major get enabled. So a possibility would be: given v3.x, new linters are added (off-by default) in 3.1, 3.2, 3.3, etc. Then those new linters are enabled by default with 4.x release. Just a possibility.