Hi! Thanks for the PR! Do you have a repro for this issue? I suspect ${root_or_parent_with_any_config} might be a better setting if the issue is not finding the proper biome.jsonc file to run from.

Yes, it was noticed in this PR: freelensapp/freelens#874 You can get this commit after merge then remove settings from the .trunk.yaml and see the result: biome plugin silently ignores errors after you'll run trunk check --all -v.

If you say ${root_or_parent_with_any_config} is better then I'll switch to this option.

So it must be just ${parent} otherwise I get

          × Found a nested root configuration, but there's already a root configuration.
          
          i The other configuration was found in /private/var/folders/k5/wyp40c2n6kzfqm5_gzry576c0000gn/T/trunk-501/cMNcwD/7iasVC.
          
          i Use the migration command from the root of the project to update the configuration.
          
          $ biome migrate --write
          

        configuration ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

          × Biome exited because the configuration resulted in errors. Please fix them.

The problem is related to biomejs/biome#6662 which was not really fixed yet and decribed in biomejs/biome#6509

Biome sees the copy of working tree in sandbox as duplicated then refuses to work.

My workaround is to use explicit --config-path option. Unfortunately, we must use explicit config file name, then it is just hardcoded to biome.jsonc as I can be just a single file in this option.

I disabled cache as a first reason of the problem that it was for long time undetected. Biome is so fast that caching is not important.

FYI: I had a lot of troubles with using Biome and Trunk together. In the end my configuration now is: https://github.com/freelensapp/freelens/blob/a4d0d192398878caee3c73d9721f60071fdf9121/.trunk/trunk.yaml#L22

lint:
  definitions:
    # See https://github.com/trunk-io/plugins/pull/1063
    - name: biome
      files:
        - astro
        - css
        - graphql
        - html
        - javascript
        - json
        - typescript
      commands:
        - output: regex
          success_codes:
            - 0
            - 1
          batch: true
          cache_results: false
          name: lint
          parse_regex: ^::(?P<severity>.*) title=(?P<code>[^,]+),file=(?P<path>[^,]+),line=(?P<line>[^,]+),endLine=[^,]+,col=(?P<col>[^,]+),endColumn=[^:]+::(?P<message>.*)
          read_output_from: stdout
          run: biome check --config-path=${workspace}/biome.jsonc --reporter=github ${target}
          run_from: ${root_or_parent_with(biome.jsonc)}
        - output: regex
          success_codes:
            - 0
          batch: true
          cache_results: false
          formatter: true
          name: fmt
          parse_regex: ^::(?P<severity>.*) title=(?P<code>[^,]+),file=(?P<path>[^,]+),line=(?P<line>[^,]+),endLine=[^,]+,col=(?P<col>[^,]+),endColumn=[^:]+::(?P<message>.*)
          read_output_from: stdout
          run: biome format --config-path=${workspace}/biome.jsonc --reporter=github --write ${target}
          run_from: ${root_or_parent_with(biome.jsonc)}
      run_timeout: 1m

so it is now just ${root_or_parent_with(biome.jsonc)}.

And the main problem is that if biome uses vcs: {useIgnoreFile: true} option AND the .gitignore file contains tmp path then Biome does not work with Trunk Check!

Actually this is the same with yamlfmt tool: tmp cannot be in .gitignore.

Thanks for the follow-up. I'm reticent to accept this PR as-is, as it will break older versions of biome, and we try to preserve a reasonable level of backwards compatibility. However, I recognize that the multiple config file flow is a standard Biome idiom. Of the following alternatives, which do you think would work best for you:

1. Provide essentially 2 different biome linter configurations in Trunk, one for <2.0.0 and one for >=2.0.0. For the newer versions, we would only support the one config file name by default, and would use the run and run_from that you've set in your current config
2. The minimal change would be updating the README.md for biome (and Trunk docs) to include these overrides for users who have multiple config files
3. Some third middle ground? Although I don't believe there is a way we can support both nested and non-nested flows with flexibility given the biome issues you linked

I converted this PR back to a draft.

So: I think it is some limitation in Trunk that I miss some variable with the configuration file name. Now I must to use:

run: biome format --config-path=${workspace}/biome.jsonc --reporter=github --write ${target}
run_from: ${root_or_parent_with(biome.jsonc)}

and it will work only for 1 configuration file biome.jsonc. Biome can work also with biome.json and I would prefer to have some autodetection what file exists and I would like to use it in the command with some variable like --config-path=${config} or something else.

If it is not possible to detect more configuration files, then I prefer .jsonc of course, as it allows me to use comments.

My configuration, which I use almost every day, really works well in VS Code in real-time. I could prepare it as a proper PR if you allow me.