Chunked reads beat an h5py process pool; unlimited writer B-trees; Blosc2; 599/697 conformance #16

Merged
osobh merged 48 commits from feat/p2b-scale into main 2026-09-26 17:42:16 +00:00
Showing only changes of commit 55e0e7e9cf - Show all commits
+6 -3
View File
@@ -3,10 +3,13 @@
## Unreleased
### Remaining conformance errors (2026-09-26)
Conformance on tank, `conformance/run.sh --no-fetch`: 597 of 697 files
ok (575 before). Of the 6 our-errors left, 4 are corrupt data HDF5 2.0
Conformance on tank, `conformance/run.sh --no-fetch`: 598 of 697 files
ok (575 before). Of the 5 our-errors left, 3 are corrupt data HDF5 2.0
reads only through a bug (listed in `CONFORMANCE.md`), 2 are the Blosc2 and
ZFP filters; the 2 mismatches are the known h5py big-endian VL bug.
ZFP filters; the 2 mismatches are the known h5py big-endian VL bug. The
five files whose cache image libhdf5 cannot load (`cve-2025-6269-*`,
`cve-2025-6516`) count as ok because the library, like libhdf5, opens them
and fails their objects (see below).
- **Metadata cache images are read.** A file written with a metadata cache
image keeps its metadata cache entries in an image block the superblock
extension points at, and libhdf5 reads them in place of the file's own