mirror of
https://github.com/HDFGroup/hdf5.git
synced 2026-09-25 04:09:44 +03:00
Scot Breitenfeld
19745508a3
Fixes all 3 Critical, 10 High, 12/13 Medium, and all 7 Low findings from the branch's comprehensive review (full detail in REVIEW-6153.md): - Critical: canonicalization-length stack over-read, false case-sensitivity doc claims (RFC amended to match), missing pipeline-v3 format-spec entry and golden regression file. - High: unbounded H5Pdecode() config length, H5Zregister() doc/validator mismatch, CHANGELOG errors, dead Fortran test code, JNI truncation bug, Java struct visibility, incomplete H5Pget_filter_params_by_idx docs. (H5P_ENCODE_VERS bump deliberately deferred as a separate, larger change.) - Medium: h5dump control-char injection, H5Z_find() type-confusion doc, missing C++ modifyFilterByIdx, doc wording fixes, TOML 0b/0o mis-lex, scaleoffset upper-bound check, missing error check, parallel stored-string coverage, C++ test gaps. (Java test porting deliberately skipped, no JDK available to verify.) - Low: dead H5Z_CLASS_T_VERS_MAX constant, duplicate nbit set_config, false JNI Javadoc, canonical-name cross-plugin collision check, stale doc anchors, compiler-ID visibility gate, CHANGELOG omissions. Also tightens newly-added comments across the branch (src, tools, tests, java, c++/fortran) to be concise and WHY-focused, and fixes two compiler warnings found during final verification: an intentional exact floating-point comparison in h5dump's hex-float annotation, and two oversized stack buffers in the same file's filter-listing loop (now heap-allocated). All changes verified via full rebuild (0 warnings/errors) plus tfilter2, testhdf5_fortran, parallel filter tests under mpirun, and the relevant h5repack/h5dump ctests -- all passing with 0 errors.