"Terminally deprecated method in sun.misc.Unsafe" : getByte

[christophe-riolo]

Hello,
When using Aircompressor, I saw the following warning, and followed the recommendation to report it !

WARNING: A terminally deprecated method in sun.misc.Unsafe has been called
WARNING: sun.misc.Unsafe::getByte has been called by io.airlift.compress.v3.lzo.LzoRawDecompressor (file:/home/amelia/.m2/repository/io/airlift/aircompressor-v3/3.4/aircompressor-v3-3.4.jar)
WARNING: Please consider reporting this to the maintainers of class io.airlift.compress.v3.lzo.LzoRawDecompressor
WARNING: sun.misc.Unsafe::getByte will be removed in a future release

[lizlooney]

Is there any plan to update Aircompressor to use VarHandle and/or MemorySegment::ofArray instead of sun.misc.Unsafe?

According to JEP 471, using sun.misc.Unsafe will throw exceptions by default, in JDK 26 or later and the memory-access methods will be removed from sun.misc.Unsafe in releases following JDK 26.

[dfa1]

hello @christophe-riolo @lizlooney

I just created #356 to propose a way out of Unsafe: I modified only LZ4, is not hard but performance will be worse for the non-native compressors/decompressors.

[dfa1]

Opened #356 as a starting point for discussion: an Unsafe-free LZ4 decompressor built on MemorySegment.

It's purely additive — Lz4JavaSafeDecompressor sits alongside the existing Unsafe decoder, which stays the default and serves as a byte-for-byte oracle in the tests. Both Decompressor overloads are implemented, including a zero-copy MemorySegment path for mmap'd input.

Tradeoff is a dataset-dependent throughput hit (~0.6–0.9× the Unsafe decoder on the project's NativeLz4DecompressBenchmark), so it's proposed as an opt-in / fallback rather than a replacement. Scope is LZ4 decompression only; the LZ4 compressor and the other codecs (snappy/zstd/lzo) are still on Unsafe.

Happy to follow whatever direction the maintainers prefer on impl selection and overall strategy.