What causes "list_filtered failed: {e}" in BoundaryML/baml?

Check the --include/--exclude filter arguments for syntax errors and correct them

list_filtered failed: {e}

Error message

list_filtered failed: {e}

What it means

Before running, the CLI lists the testset names matching the invocation's filters via a list_filtered call into the registry; any failure is wrapped as `list_filtered failed`. This is a discovery/enumeration step, so failures here mean the filter set could not be evaluated.

Source

Thrown at baml_language/crates/baml_cli/src/test_command.rs:882

) -> Result<Vec<String>> {
    let call_ctx = FunctionCallContextBuilder::new(CallId::next())
        .with_cancel_token(ctx.cancel.clone())
        .build();
    let value = ctx
        .rt
        .block_on(ctx.engine.call_function(
            "testing.TestRegistry.list_filtered",
            vec![
                registry.clone(),
                string_array(&invocation.profile_include),
                string_array(&invocation.profile_exclude),
                string_array(&invocation.cli_include),
                string_array(&invocation.cli_exclude),
            ],
            call_ctx,
            true,
        ))
        .map_err(|e| anyhow!("list_filtered failed: {e}"))?;
    Ok(string_array_values(&value))
}

/// Print a flat report and fold its counts into the running totals.
///
/// Print the aggregate testset result and fold its counts into the running
/// totals. Both filtered and unfiltered runs go through the same aggregate path
/// (filtered selections still run under their testset runners), so a failing
/// leaf whose runner still passes the suite is reported as a *tolerated* failure
/// — kept out of the hard `failed` count and not failing the command. Every
/// leaf line uses the same canonical ID emitted by `baml test --list`; aggregate
/// runner verdicts are labeled as aggregates rather than presented as test IDs.
fn consume_flat_report(
    flat: &FlatReport,
    passed: &mut usize,
    failed: &mut usize,
    tolerated: &mut usize,
    total: &mut usize,

View on GitHub (pinned to bd85ce9dee)

Solutions

Check the --include/--exclude filter arguments for syntax errors and correct them
Ensure the project compiles/tests are discoverable (run a plain `baml test` without filters)
Inspect the chained inner error for the registry-level root cause

Example fix

# before
baml test --include "[[bad-filter"

# after
baml test --include "my_testset.*"

Defensive patterns

Strategy: try-catch

Try / catch

match list_selected_testset_names(&ctx, &registry, &invocation) {
    Ok(names) if !names.is_empty() => run(names),
    Ok(_) => eprintln!("no testsets matched the given filters"),
    Err(e) => eprintln!("list_filtered failed: {e:#}"),
}

Prevention

Validate include/exclude filter syntax before invoking
Ensure the project compiles before filtering testsets
Start with unfiltered runs to confirm testsets exist
Log the filter arguments on failure

When it happens

Trigger: Calling list_selected_testset_names (from run) where list_filtered errors — bad include/exclude filter values, registry lookup failure, or an invalid call context while converting the returned value with string_array_values.

Common situations: Malformed --include/--exclude CLI filter expressions; testsets not yet compiled/registered; internal registry state unavailable.

Related errors

AI-assisted analysis of BoundaryML/baml@bd85ce9dee (2026-09-12). Data as JSON: /api/errors/efb5bbf139e5f480. Report an issue: GitHub.