Detect filtered chunk size overflow with small size of sizes

[mattjala]

For all chunk index types in version-5 chunk layout messages, and for the single chunk index in any version of the chunk layout message, the on-disk size of a filtered chunk is encoded in a fixed-width field equal to the file's "size of sizes" (set via H5Pset_sizes()). The encode check assumes this field is always 8 bytes, so when the size of sizes is set to 2 or 4 an oversized chunk has its encoded size silently truncated (e.g. 160000 encoded in 2 bytes becomes 160000 & 0xFFFF = 28928), silently corrupting the chunk.

The library now verifies that a filtered chunk's size fits in the file's size of sizes and reports an error at write time instead, via a check in the H5D_CHUNK_ENCODE_SIZE_CHECK macro (src/H5Dchunk.c), which runs from H5D__chunk_file_alloc() before any chunk storage is allocated.

The in-range case, when the filtered chunk is small enough to be encoded by a small size of sizes, is tested by test_chunk_size_of_sizes, and the case where a filtered chunk is too large to be encoded is tested by test_chunk_size_of_sizes_overflow.

Resolves #6023

[github-actions]

Review Checklist

This PR touches the following areas. Each needs a sign-off
from its listed owners before merging.

• src — @fortnern
• test — @fortnern
• release_docs ✅ — @hyoklee

[Copilot]

Pull request overview

This PR fixes a data-corruption risk when writing filtered chunked datasets to files whose configured “size of sizes” (via H5Pset_sizes()) is smaller than 8 bytes, by detecting and erroring out on filtered chunk sizes that do not fit in the on-disk encoded field width.

Changes:

• Add an overflow check in H5D_CHUNK_ENCODE_SIZE_CHECK to ensure filtered chunk sizes fit within the file’s configured “size of sizes” for v5 chunk layout messages and for the single-chunk index in any layout version.
• Add new dataset tests covering round-trip correctness for small “size of sizes” values and explicit overflow detection on write/flush.
• Document the fix in the release changelog.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
test/dsets.c	Adds new tests for filtered chunk size encoding with size of sizes set to 2/4/8, including an overflow-failure case.
src/H5Dchunk.c	Extends the filtered chunk size encoding check to honor the file’s configured “size of sizes” field width.
release_docs/CHANGELOG.md	Notes the bug fix and the prior silent truncation behavior in release documentation.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[hyoklee]

Copilot's suggestion for CHANGELOG.md:

### Fixed silent truncation of filtered chunk sizes when the file "size of sizes" is small

For all chunk index types in version‑5 chunk layout messages, and for the single chunk index in any version of the chunk layout message, the on‑disk size of a filtered chunk is encoded in a fixed‑width field equal to the file's "size of sizes" (set via `H5Pset_sizes()`). The encode check previously assumed this field was always 8 bytes. When the size of sizes was set to 2 or 4, an oversized chunk's encoded size could be silently truncated (for example, 160000 encoded in 2 bytes becomes 160000 & 0xFFFF = 28928), corrupting the chunk. The library now verifies that a filtered chunk's size fits within the file's size of sizes and reports an error at write time.

Fixes GitHub issue #6023

[github-actions]

This pull request has had no activity for 30 days and has been marked stale. Push a commit or comment to keep it open, or it will be flagged for maintainer review.