denoland/deno · error · AnyError

malformed .msi: no _StringData stream

Error message

malformed .msi: no _StringData stream

What it means

Companion check to the _StringPool error: after locating the string-pool header stream, the repacker needs the _StringData stream (the concatenated UTF-8 string bytes the pool indexes into). Its absence with the same find_raw/de-mangle lookup means the MSI's string table is half-present — effectively a malformed or non-MSI input, since every valid MSI writes both streams together.

Solutions

1. Verify integrity of the template: compare its size/SHA-256 against the known-good source, re-download it.
2. Open the file with lessmsi/msidb to confirm both string-table streams are present before repacking.
3. If building the template in-house, ensure the MSI authoring tool completed successfully (check its exit status and logs).

Example fix

# before: template truncated by an interrupted download
ls -l template.msi          # 41 KB (expected ~3 MB)
# error: malformed .msi: no _StringData stream

# after: refetch and verify
sha256sum template.msi      # match against known-good digest, then rebuild

7z l template.msi | grep -q '_StringData' || { echo 'MSI string data stream missing — template truncated/corrupt'; exit 1; }
sha256sum template.msi   # compare against the digest recorded when the template was first obtained

Treat like 1795: catch, verify the template digest against its known-good value, and abort the build on mismatch; do not attempt partial recovery of a half-present string table.

When it happens

Trigger: find_raw("_StringData") misses after _StringPool was found: a truncated compound file where later streams were cut off, a hand-crafted/corrupted database, or stream-name mangling that only some names survive.

Common situations: Truncated template download (pool near the start survived, data stream at the end did not); disk-full while the template was written; a partial copy from an artefact store.

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

• malformed .msi: no _Columns stream
• malformed .msi: no _StringPool stream
• cabinet file missing
• Cannot build a .msi from an empty app directory
• deno.json `version` " " cannot be used as an MSI…

AI-assisted analysis of denoland/deno@f7822238ca (2026-08-20). Data as JSON: /api/errors/8a028ec075927898. Report an issue: GitHub.

Appendix: source

Thrown at cli/tools/desktop.rs:5303

      .cloned()
  };

  // The string-pool header's top bit selects 3-byte string references; tiny
  // databases use 2-byte refs, but honor the flag to stay correct at any size.
  let pool_raw = find_raw("_StringPool").ok_or_else(|| {
    deno_core::anyhow::anyhow!("malformed .msi: no _StringPool stream")
  })?;
  let pool = read_stream(&mut comp, &pool_raw)?;
  let str_w: usize = if pool.len() >= 4 && (pool[3] & 0x80) != 0 {
    3
  } else {
    2
  };

  // Map every string-pool id to its text so data-table streams (named by table
  // text) can be matched to their column schema (keyed by table string id).
  let data_raw = find_raw("_StringData").ok_or_else(|| {
    deno_core::anyhow::anyhow!("malformed .msi: no _StringData stream")
  })?;
  let str_data = read_stream(&mut comp, &data_raw)?;
  let mut strings = vec![String::new()]; // 1-based; index 0 is unused.
  {
    let mut off = 0usize;
    let mut i = 4;
    while i + 4 <= pool.len() {
      let len = u16::from_le_bytes([pool[i], pool[i + 1]]) as usize;
      let end = (off + len).min(str_data.len());
      strings.push(String::from_utf8_lossy(&str_data[off..end]).into_owned());
      off += len;
      i += 4;
    }
  }
  let id_of = |text: &str| -> Option<u64> {
    strings.iter().position(|s| s == text).map(|i| i as u64)
  };

View on GitHub (pinned to f7822238ca)