apache/hadoop · error · AlreadyClosedException

decompress called on closed decompressor

Error message

decompress called on closed decompressor

What it means

Error "decompress called on closed decompressor" thrown in apache/hadoop.

Source

Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/io/compress/zlib/BuiltInGzipDecompressor.java:195

  /* From the caller's perspective, this is where the state machine lives.
   * The code is written such that we never return from decompress() with
   * data remaining in userBuf unless we're in FINISHED state and there was
   * data beyond the current gzip member (e.g., we're within a concatenated
   * gzip stream).  If this ever changes, {@link #needsInput()} will also
   * need to be modified (i.e., uncomment the userBufLen condition).
   *
   * The actual deflate-stream processing (decompression) is handled by
   * Java's Inflater class.  Unlike the gzip header/trailer code (execute*
   * methods below), the deflate stream is never copied; Inflater operates
   * directly on the user's buffer.
   */
  @Override
  public synchronized int decompress(byte[] b, int off, int len)
  throws IOException {
    int numAvailBytes = 0;

    if (state == GzipStateLabel.ENDED) {
      throw new AlreadyClosedException("decompress called on closed decompressor");
    }

    if (state != GzipStateLabel.DEFLATE_STREAM) {
      executeHeaderState();

      if (userBufLen <= 0) {
        return numAvailBytes;
      }
    }

    // "executeDeflateStreamState()"
    if (state == GzipStateLabel.DEFLATE_STREAM) {
      // hand off user data (or what's left of it) to Inflater--but note that
      // Inflater may not have consumed all of previous bufferload (e.g., if
      // data highly compressed or output buffer very small), in which case
      // userBufLen will be zero
      if (userBufLen > 0) {
        inflater.setInput(userBuf, userBufOff, userBufLen);

View on GitHub (pinned to 2add963021)

Solutions

1. Do not call decompress() after close(); allocate a new decompressor instead.
2. Ensure finish() and close() are invoked only after all input has been decompressed.

When it happens

Trigger: Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/io/compress/zlib/BuiltInGzipDecompressor.java:195 when the library encounters an invalid state.

Common situations: Occurs when decompress() is called after the decompressor has been closed. Guard calls with the closed flag and create a new decompressor rather than reusing a closed one.

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/83c96120fcc7b59b. Report an issue: GitHub.