v1.4.0

Changelog

All notable changes to this project are documented in this file.

The format is inspired by Keep a Changelog and follows semantic versioning intent where possible.

[Unreleased]

Added

• Free-tier core additions (MIT):
  • lox_selfcheck() API + runtime structural checks for KV/TS/REL/WAL.
  • WCET package:
    • include/lox_wcet.h compile-time bound macros,
    • docs/WCET_ANALYSIS.md methodology and per-API formulas,
    • tests/test_wcet_bounds.c verification coverage.
  • TS logarithmic retention support:
    • new policy constant LOX_TS_POLICY_LOG_RETAIN,
    • extended registration API lox_ts_register_ex(...),
    • stream-level log-retain config (zones/zone_pct),
    • dedicated tests in tests/test_ts_log_retain.c.
  • New self-check coverage in tests/test_selfcheck.c.

Changed

• Build/test wiring:
  • registered test_selfcheck, test_wcet_bounds, and test_ts_log_retain in CMake.
  • added dedicated lox_ts_log_retain test library target so default TS policy behavior remains unchanged.

[1.3.6] - 2026-04-22

Added

• Offline verifier quality package:
  • deep decode passes for KV/TS/REL pages with warning surfacing.
  • WAL semantic summary counters (including orphaned TXN markers).
  • strict --check mode for CI gating.
• New coverage:
  • tests/test_offline_verifier.c extended with corruption/recovery/JSON/check-flag scenarios.
  • tests/test_capacity_estimator_model.c for capacity-model consistency checks.
  • tests/test_safety_invariants.c for safety-critical invariants (magic clear, WAL replay boundary, superblock switch, null-handle contract).
• Certification readiness artifacts:
  • docs/SAFETY_READINESS.md
  • scripts/run_static_analysis.sh

Changed

• tools/lox_capacity_estimator.html rewritten as single-file real-time planner:
  • preset-driven inputs (FOOTPRINT_MIN, CORE_MIN, CORE_WAL, CORE_PERF, CORE_HIMEM, Custom),
  • RAM/storage/wear outputs,
  • CMake define snippet generation,
  • formula comments tied to source files.
• docs/PROGRAMMER_MANUAL.md expanded with Capacity Planning section.
• CI workflow enriched with:
  • verifier integration smoke step (Linux build lane),
  • non-blocking static-analysis job with artifact upload.

Fixed

• Windows subprocess quoting and WAL entry construction in tests/test_offline_verifier.c:
  • stable verifier invocation on Windows (cmd /c quoting),
  • orphaned TXN WAL test now writes valid data_len in entry header.

[1.3.5] - 2026-04-22

Changed

• Release workflow/platform hardening:
  • publish job now explicitly depends on sanitizer gate and build jobs.
  • macOS release artifact label aligned to macos-arm64.
• Header-level KV iteration contract clarified as weakly-consistent under concurrent mutation.
• Programmer and thread-safety docs updated with explicit REL mutation semantics and lock/copy behavior notes.

Fixed

• REL consistency and error semantics:
  • lox_rel_find now returns LOX_ERR_MODIFIED (not LOX_ERR_INVALID) when table mutation is detected after callback re-lock.
  • lox_rel_iter now captures mutation_seq snapshot and returns LOX_ERR_MODIFIED when concurrent mutation is detected.
  • Added regression coverage in tests/test_rel.c for both rel_find and rel_iter concurrent mutation detection.
• REL schema guard:
  • lox_schema_seal now rejects rows larger than LOX_REL_ROW_SCRATCH_MAX with LOX_ERR_OVERFLOW.
  • Added regression test rel_schema_seal_rejects_oversized_row.
• TS test-suite stabilization for sanitizer/release profiles:
  • fixed test_ts collector overflow for capacities above 256 samples.
  • made test_ts_reject and test_ts_downsample capacity-aware and RAM-only in setup to avoid storage-path false failures.
  • updated downsample tests to avoid hardcoded query bounds and fixed-size output buffers.
• macOS release build compatibility:
  • fixed footprint baseline linker map flag on Apple toolchain (-Wl,-map,... instead of GNU -Map form).
  • tests/check_footprint_min_size.cmake now falls back to parsing size -m output when size -A is unavailable on macOS.

[1.3.0] - 2026-04-21

Added

• Real-data integration coverage on host (tests/test_realdata_integration.c) including KV/TS/REL flows, JSON wrapper, import/export, TXN, and reopen recovery assertions.
• ESP32-S3 run_real smoke command in bench/lox_esp32_s3_bench for on-device validation of realistic end-to-end data paths.
• C++ wrapper coverage extension in tests/test_cpp_wrapper.cpp for practical KV/TS/REL wrapper usage.

Changed

• ESP32 benchmark documentation expanded with:
  • WAL sync mode decision table (SYNC_ALWAYS vs SYNC_FLUSH_ONLY) with measured ESP32 values.
  • POSIX-vs-ESP32 interpretation guidance and follow-up notes for benchmark fidelity.
• Bench sources synchronized for ESP32 harness (loxdb, JSON, and import/export modules) to keep local bench build behavior aligned with core sources.
• Release hardening:
  • added macOS CI/release presets (ci-debug-macos, release-macos) and workflow matrix coverage.
  • release workflow now includes mandatory Linux ASan/UBSan sanitizer gate before packaging.
  • release page body now reads from CHANGELOG.md (body_path) instead of auto-generated commit-only notes.

Fixed

• KV JSON import/export TTL sentinel handling:
  • non-expiring keys exported via KV iter sentinel (UINT32_MAX) are now normalized to ttl=0 before JSON encoding.
  • prevents immediate expiry after import for persistent keys.
• KV admission performance:
  • lox_admit_kv_set now uses O(1) live_value_bytes accounting instead of O(n) bucket scan for compact-availability calculation.
  • verified on ESP32-S3 hardware with reduced admit_kv_set latency in run_real.
• REL/TS mutation/admission semantics hardening:
  • dedicated LOX_ERR_MODIFIED behavior for TS mutation detection in query flows.
  • deterministic budget signaling consistency in REL admission.

Repository Hygiene

• Removed tracked ESP32 build artifacts under bench/lox_esp32_s3_bench/build_sync_flush_only/ and added ignore rule to keep binary outputs out of git history.

[1.2.0] - 2026-04-20

Added

• Optional wrapper and backend adapter documentation in README.
• Automated GitHub Release workflow on tag push (v*).

Fixed

• UBSAN misalignment issues in TS/REL and WAL-related paths.
• ASAN leak paths in multiple recovery/reopen test flows.

[1.1.0] - 2026-04-12

Added

• Read-only diagnostics and admission APIs.
• Managed stress baseline tooling and threshold recommendation scripts.

Changed

• Preset-driven CI/release testing strategy (ci-debug-*, release-*).

[1.0.0] - 2026-04-01

Added

• Initial public release with KV, TS, and REL engines.
• WAL durability core and recovery flow.
• C and C++ wrapper surfaces.

v1.3.7

Changelog

All notable changes to this project are documented in this file.

The format is inspired by Keep a Changelog and follows semantic versioning intent where possible.

[Unreleased]

Added

• Free-tier core additions (MIT):
  • microdb_selfcheck() API + runtime structural checks for KV/TS/REL/WAL.
  • WCET package:
    • include/microdb_wcet.h compile-time bound macros,
    • docs/WCET_ANALYSIS.md methodology and per-API formulas,
    • tests/test_wcet_bounds.c verification coverage.
  • TS logarithmic retention support:
    • new policy constant MICRODB_TS_POLICY_LOG_RETAIN,
    • extended registration API microdb_ts_register_ex(...),
    • stream-level log-retain config (zones/zone_pct),
    • dedicated tests in tests/test_ts_log_retain.c.
  • New self-check coverage in tests/test_selfcheck.c.

Changed

• Build/test wiring:
  • registered test_selfcheck, test_wcet_bounds, and test_ts_log_retain in CMake.
  • added dedicated microdb_ts_log_retain test library target so default TS policy behavior remains unchanged.

[1.3.6] - 2026-04-22

Added

• Offline verifier quality package:
  • deep decode passes for KV/TS/REL pages with warning surfacing.
  • WAL semantic summary counters (including orphaned TXN markers).
  • strict --check mode for CI gating.
• New coverage:
  • tests/test_offline_verifier.c extended with corruption/recovery/JSON/check-flag scenarios.
  • tests/test_capacity_estimator_model.c for capacity-model consistency checks.
  • tests/test_safety_invariants.c for safety-critical invariants (magic clear, WAL replay boundary, superblock switch, null-handle contract).
• Certification readiness artifacts:
  • docs/SAFETY_READINESS.md
  • scripts/run_static_analysis.sh

Changed

• tools/microdb_capacity_estimator.html rewritten as single-file real-time planner:
  • preset-driven inputs (FOOTPRINT_MIN, CORE_MIN, CORE_WAL, CORE_PERF, CORE_HIMEM, Custom),
  • RAM/storage/wear outputs,
  • CMake define snippet generation,
  • formula comments tied to source files.
• docs/PROGRAMMER_MANUAL.md expanded with Capacity Planning section.
• CI workflow enriched with:
  • verifier integration smoke step (Linux build lane),
  • non-blocking static-analysis job with artifact upload.

Fixed

• Windows subprocess quoting and WAL entry construction in tests/test_offline_verifier.c:
  • stable verifier invocation on Windows (cmd /c quoting),
  • orphaned TXN WAL test now writes valid data_len in entry header.

[1.3.5] - 2026-04-22

Changed

• Release workflow/platform hardening:
  • publish job now explicitly depends on sanitizer gate and build jobs.
  • macOS release artifact label aligned to macos-arm64.
• Header-level KV iteration contract clarified as weakly-consistent under concurrent mutation.
• Programmer and thread-safety docs updated with explicit REL mutation semantics and lock/copy behavior notes.

Fixed

• REL consistency and error semantics:
  • microdb_rel_find now returns MICRODB_ERR_MODIFIED (not MICRODB_ERR_INVALID) when table mutation is detected after callback re-lock.
  • microdb_rel_iter now captures mutation_seq snapshot and returns MICRODB_ERR_MODIFIED when concurrent mutation is detected.
  • Added regression coverage in tests/test_rel.c for both rel_find and rel_iter concurrent mutation detection.
• REL schema guard:
  • microdb_schema_seal now rejects rows larger than MICRODB_REL_ROW_SCRATCH_MAX with MICRODB_ERR_OVERFLOW.
  • Added regression test rel_schema_seal_rejects_oversized_row.
• TS test-suite stabilization for sanitizer/release profiles:
  • fixed test_ts collector overflow for capacities above 256 samples.
  • made test_ts_reject and test_ts_downsample capacity-aware and RAM-only in setup to avoid storage-path false failures.
  • updated downsample tests to avoid hardcoded query bounds and fixed-size output buffers.
• macOS release build compatibility:
  • fixed footprint baseline linker map flag on Apple toolchain (-Wl,-map,... instead of GNU -Map form).
  • tests/check_footprint_min_size.cmake now falls back to parsing size -m output when size -A is unavailable on macOS.

[1.3.0] - 2026-04-21

Added

• Real-data integration coverage on host (tests/test_realdata_integration.c) including KV/TS/REL flows, JSON wrapper, import/export, TXN, and reopen recovery assertions.
• ESP32-S3 run_real smoke command in bench/microdb_esp32_s3_bench for on-device validation of realistic end-to-end data paths.
• C++ wrapper coverage extension in tests/test_cpp_wrapper.cpp for practical KV/TS/REL wrapper usage.

Changed

• ESP32 benchmark documentation expanded with:
  • WAL sync mode decision table (SYNC_ALWAYS vs SYNC_FLUSH_ONLY) with measured ESP32 values.
  • POSIX-vs-ESP32 interpretation guidance and follow-up notes for benchmark fidelity.
• Bench sources synchronized for ESP32 harness (microdb, JSON, and import/export modules) to keep local bench build behavior aligned with core sources.
• Release hardening:
  • added macOS CI/release presets (ci-debug-macos, release-macos) and workflow matrix coverage.
  • release workflow now includes mandatory Linux ASan/UBSan sanitizer gate before packaging.
  • release page body now reads from CHANGELOG.md (body_path) instead of auto-generated commit-only notes.

Fixed

• KV JSON import/export TTL sentinel handling:
  • non-expiring keys exported via KV iter sentinel (UINT32_MAX) are now normalized to ttl=0 before JSON encoding.
  • prevents immediate expiry after import for persistent keys.
• KV admission performance:
  • microdb_admit_kv_set now uses O(1) live_value_bytes accounting instead of O(n) bucket scan for compact-availability calculation.
  • verified on ESP32-S3 hardware with reduced admit_kv_set latency in run_real.
• REL/TS mutation/admission semantics hardening:
  • dedicated MICRODB_ERR_MODIFIED behavior for TS mutation detection in query flows.
  • deterministic budget signaling consistency in REL admission.

Repository Hygiene

• Removed tracked ESP32 build artifacts under bench/microdb_esp32_s3_bench/build_sync_flush_only/ and added ignore rule to keep binary outputs out of git history.

[1.2.0] - 2026-04-20

Added

• Optional wrapper and backend adapter documentation in README.
• Automated GitHub Release workflow on tag push (v*).

Fixed

• UBSAN misalignment issues in TS/REL and WAL-related paths.
• ASAN leak paths in multiple recovery/reopen test flows.

[1.1.0] - 2026-04-12

Added

• Read-only diagnostics and admission APIs.
• Managed stress baseline tooling and threshold recommendation scripts.

Changed

• Preset-driven CI/release testing strategy (ci-debug-*, release-*).

[1.0.0] - 2026-04-01

Added

• Initial public release with KV, TS, and REL engines.
• WAL durability core and recovery flow.
• C and C++ wrapper surfaces.

v1.3.2

Full Changelog: v1.3.1...v1.3.2

v1.0.0

Full Changelog: https://github.com/Vanderhell/microdb/commits/v1.0.0