swc-project/swc · error

invalid utf8

Error message

invalid utf8

What it means

The lazy SourceMap path duplicates serialize_range_mappings: it assembles the rangeMappings field as base64 characters plus ';' separators into a Vec<u8>, then asserts String::from_utf8 succeeds. All emitted bytes are ASCII by construction, so this expect firing means an internal swc_sourcemap bug or corrupted token state during re-serialization of a lazily decoded map, not bad user input.

Source

Thrown at crates/swc_sourcemap/src/lazy/mod.rs:648

                encode_rmi(&mut buf, &rmi_data);
                rmi_data.clear();
            }

            buf.push(b';');
            prev_line += 1;
            had_rmi = false;
            idx_of_first_in_line = idx;
        }
    }
    if empty {
        return None;
    }

    if had_rmi {
        encode_rmi(&mut buf, &rmi_data);
    }

    Some(String::from_utf8(buf).expect("invalid utf8"))
}

fn serialize_mappings(sm: &SourceMap) -> String {
    let mut rv = String::new();
    // dst == minified == generated
    let mut prev_dst_line = 0;
    let mut prev_dst_col = 0;
    let mut prev_src_line = 0;
    let mut prev_src_col = 0;
    let mut prev_name_id = 0;
    let mut prev_src_id = 0;

    for (idx, token) in sm.tokens.iter().enumerate() {
        if token.dst_line != prev_dst_line {
            prev_dst_col = 0;
            while token.dst_line != prev_dst_line {
                rv.push(';');
                prev_dst_line += 1;

View on GitHub (pinned to 5176682b65)

Solutions

1. Pin swc_core/swc to the last known-good version in your pipeline.
2. Drop range-mapping info before re-encoding (rebuild tokens without is_range) if the field is not required downstream.
3. Report upstream with a minimal reproducing map.

// Contain the internal assertion during re-encode of lazily decoded maps.
let out = std::panic::catch_unwind(std::panic::AssertUnwindSafe(|| sm.to_writer(&mut w)));
if out.is_err() {
    drop_range_tokens(&mut sm);
    sm.to_writer(&mut w)?;
}

When it happens

Trigger: Re-encoding (to_writer/to_data) a lazily decoded SourceMap that still carries range mapping tokens, e.g. flattening or rewriting a map produced by a tool that emits rangeMappings.

Common situations: Sourcemap processing pipelines that decode a range-mappings map and re-emit it (chained transforms, bundler concat of maps); swc_sourcemap regressions.

Understand the failure class

• Parsing and encoding errors: unexpected token, malformed input — why parsers reject input and how to find the real culprit.

Related errors

• invalid utf8
• Failed to convert RawValue to Data
• Invalid attempt to iterate non-iterable instance. In order t
• Duplicated methods (${element.key}) can't be decorated.
• Decorators can't be placed on different accessors with for t

AI-assisted analysis of swc-project/swc@5176682b65 (2026-08-17). Data as JSON: /api/errors/193a402dc053cf74. Report an issue: GitHub.