nsis 0.1.0

//! LZMA decompression for NSIS data blocks.
//!
//! NSIS LZMA streams begin with a properties byte (typically `0x5D` for
//! lc=3, lp=0, pb=2) followed by a 4-byte little-endian dictionary size.

use crate::error::Error;

/// Decompresses an NSIS LZMA stream.
///
/// NSIS LZMA streams use the raw LZMA format: a properties byte followed
/// by a 4-byte dictionary size, then the compressed data. The uncompressed
/// size is not stored in the NSIS LZMA header.
///
/// # Arguments
///
/// - `compressed`: the raw LZMA stream (properties + dict size + data)
/// - `max_output`: maximum decompressed size
/// - `expected_size`: if `Some`, the exact expected decompressed size is
///   written into the LZMA header so `lzma-rs` stops after that many bytes.
///   If `None`, the size is set to unknown and we rely on the EOS marker.
///
/// # Errors
///
/// Returns [`Error::DecompressionFailed`] if the LZMA stream is invalid.
pub fn decompress_lzma(
    compressed: &[u8],
    max_output: usize,
    expected_size: Option<usize>,
) -> Result<Vec<u8>, Error> {
    if compressed.len() < 5 {
        return Err(Error::DecompressionFailed {
            method: "lzma",
            detail: "LZMA stream too short (need at least 5 bytes for header)".into(),
        });
    }

    // Build a standard LZMA header for lzma-rs:
    // Bytes 0:   properties byte (from NSIS stream)
    // Bytes 1-4: dictionary size (from NSIS stream)
    // Bytes 5-12: uncompressed size (known or 0xFFFFFFFFFFFFFFFF for unknown)
    let uncompressed_size_bytes: [u8; 8] = match expected_size {
        Some(size) => (size as u64).to_le_bytes(),
        None => [0xFF; 8],
    };

    let mut lzma_header = Vec::with_capacity(13 + compressed.len() - 5);
    lzma_header.extend_from_slice(&compressed[..5]); // props + dict_size
    lzma_header.extend_from_slice(&uncompressed_size_bytes);
    lzma_header.extend_from_slice(&compressed[5..]);

    let mut output = Vec::with_capacity(max_output.min(compressed.len() * 4));

    // When expected_size is None (unknown decompressed size), lzma-rs will
    // decompress until it hits the EOS marker. If there are trailing bytes
    // after the EOS marker (CRC, padding, etc.), lzma-rs rejects them with
    // "Found end-of-stream marker but more bytes are available."
    //
    // We work around this by catching the specific error and returning the
    // output that was successfully decompressed before the error. The lzma-rs
    // decoder writes valid data to `output` before failing on the trailing bytes.
    let mut reader = std::io::BufReader::new(std::io::Cursor::new(&lzma_header));
    match lzma_rs::lzma_decompress(&mut reader, &mut output) {
        Ok(()) => {}
        Err(e) => {
            let msg = e.to_string();
            // If we got data and the error is about trailing bytes, that's OK —
            // the LZMA stream was fully decoded, just with leftover input.
            if !output.is_empty() && msg.contains("more bytes are available") {
                // Successfully decoded up to the EOS marker.
            } else {
                return Err(Error::DecompressionFailed {
                    method: "lzma",
                    detail: msg,
                });
            }
        }
    }

    if output.len() > max_output {
        output.truncate(max_output);
    }

    Ok(output)
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn too_short_input() {
        let result = decompress_lzma(&[0x5D, 0x00, 0x00], 1024, None);
        assert!(result.is_err());
    }
}