(arrow-rs) branch main updated: fix(arrow-ipc): write 8-byte i64 length prefix for compressed IPC buffers on wasm32 (#10989)

[jeffreyvo]

This is an automated email from the ASF dual-hosted git repository.

Jefffrey pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/arrow-rs.git


The following commit(s) were added to refs/heads/main by this push:
     new b1aa94b05c fix(arrow-ipc): write 8-byte i64 length prefix for 
compressed IPC buffers on wasm32 (#10989)
b1aa94b05c is described below

commit b1aa94b05c88ef4cc6d0293d784bbc254537e942
Author: Narendran K T <naren.cse...@gmail.com>
AuthorDate: Sat Sep 5 07:06:35 2026 +0530

    fix(arrow-ipc): write 8-byte i64 length prefix for compressed IPC buffers 
on wasm32 (#10989)
    
    # Which issue does this PR close?
    
    * Closes https://github.com/apache/arrow-rs/issues/10986.
    
    # Rationale for this change
    
    On `wasm32`, `usize` is 32-bit, causing the compressed IPC buffer length
    prefix to be written as 4 bytes instead of the required 8-byte `i64`
    value.
    
    # What changes are included in this PR?
    
    Serialize the uncompressed buffer length explicitly as `i64` to ensure
    the IPC compression prefix is always 8 bytes across platforms.
    
    # Are these changes tested?
    
    Yes. Verified the generated compressed IPC buffer and confirmed the
    8-byte length prefix is correctly written on `wasm32`.
    
    # Are there any user-facing changes?
    
    No. This is a bug fix for IPC compression compatibility on `wasm32`;
    there are no API or breaking changes.
    
    Co-authored-by: Narendran <narendra...@lumel.com>
---
 arrow-ipc/src/compression.rs | 18 +++++++++++++++++-
 1 file changed, 17 insertions(+), 1 deletion(-)

diff --git a/arrow-ipc/src/compression.rs b/arrow-ipc/src/compression.rs
index c40ee56055..59e10f7698 100644
--- a/arrow-ipc/src/compression.rs
+++ b/arrow-ipc/src/compression.rs
@@ -201,7 +201,7 @@ impl CompressionCodec {
             // empty input, nothing to do
         } else {
             // write compressed data directly into the output buffer
-            output.extend_from_slice(&uncompressed_data_len.to_le_bytes());
+            output.extend_from_slice(&(uncompressed_data_len as 
i64).to_le_bytes());
             self.compress(input, output, context)?;
 
             let compression_len = output.len() - original_output_len;
@@ -443,4 +443,20 @@ mod tests {
             "unexpected error: {err}"
         );
     }
+
+    #[test]
+    #[cfg(feature = "lz4")]
+    fn test_compress_to_vec_writes_8_byte_length_prefix() {
+        // The length prefix must always be 8 bytes (i64),
+        // even on platforms where `usize` is narrower (e.g. wasm32).
+        let input_bytes = vec![42u8; 132];
+        let codec = super::CompressionCodec::Lz4Frame;
+        let mut output_bytes: Vec<u8> = Vec::new();
+        codec
+            .compress_to_vec(&input_bytes, &mut output_bytes, &mut 
Default::default())
+            .unwrap();
+
+        let prefix: [u8; 8] = output_bytes[..8].try_into().unwrap();
+        assert_eq!(i64::from_le_bytes(prefix), input_bytes.len() as i64);
+    }
 }