Bluetooth-Devices/bluetooth-data-tools

Bluetooth-Devices

bluetooth-data-tools

Blog Docs Changelog

Performance History

Latest Results

chore(pre-commit.ci): auto fixes

pre-commit-ci-update-config

2 days ago

chore(pre-commit.ci): auto fixes

pre-commit-ci-update-config

9 days ago

chore(pre-commit.ci): pre-commit autoupdate updates: - [github.com/commitizen-tools/commitizen: v4.18.0 → v4.18.1](https://github.com/commitizen-tools/commitizen/compare/v4.18.0...v4.18.1) - [github.com/astral-sh/ruff-pre-commit: v0.16.4 → v0.16.7](https://github.com/astral-sh/ruff-pre-commit/compare/v0.16.4...v0.16.7)

pre-commit-ci-update-config

9 days ago

chore(pre-commit.ci): auto fixes

pre-commit-ci-update-config

16 days ago

chore(pre-commit.ci): pre-commit autoupdate updates: - [github.com/astral-sh/ruff-pre-commit: v0.16.4 → v0.16.6](https://github.com/astral-sh/ruff-pre-commit/compare/v0.16.4...v0.16.6)

pre-commit-ci-update-config

16 days ago

fix(utils): reject ints outside the 48-bit Bluetooth address space int_to_bluetooth_address accepted any int and returned a plausible-looking address for out-of-range values -- and the two implementations disagreed on which one. Differential fuzzing of the built .so against the pure-Python fallback (200k random ints across the full 64-bit range plus negatives) found: 0x1000000000000 native 00:00:00:00:00:00 pure 10:00:00:00:00:00 0x1000000000001 native 00:00:00:00:00:01 pure 10:00:00:00:00:00 2**64 native OverflowError pure 10:00:00:00:00:00 -1 native OverflowError pure -0:00:00:00:00:01 The native path truncated to the low 48 bits via the uint64_t cast. The pure path formatted the value to more than 12 hex digits and then took the fixed [0:2]..[10:12] slices, i.e. its *high* half -- which also collapses distinct inputs onto one address and can emit a literal "-", a string mac_to_int then refuses. Every one of these is cached by the lru_cache wrapper. Both paths now raise ValueError outside 0..0xFFFFFFFFFFFF, matching mac_to_int's contract for input outside its domain. The guard is spelled identically in both so they cannot drift again. BEHAVIOR CHANGE: input outside the 48-bit range used to return a string (or raise OverflowError natively) and now raises ValueError. No in-range value changes; mac_to_int(int_to_bluetooth_address(x)) still round-trips for every x in 0..0xFFFFFFFFFFFF. The chained comparison was chosen over converting to a cdef uint64_t or wrapping the cast in try/except: those cost ~30ns per cache miss, this one measures ~5ns (98.7 -> 103.8ns, min of 25 runs x 4 interleaved rebuilds).

bluetoothbot:koan/int-to-address-range-check

16 days ago

test(bench): run one benchmark suite under both CI runners bench/ was a hand-maintained mirror of tests/benchmarks/ -- 15 files, 35 benchmarks, semantically identical (verified: same collected node IDs, same payloads, same benchmarked expressions; the only deltas were the CodSpeed BenchmarkFixture annotations and docstrings). Nothing enforced the mirror, so every new benchmark had to be written twice and either copy could drift unnoticed. pytest-benchmark and pytest-codspeed are both installed in the benchmark job, and whichever run is active owns the `benchmark` fixture -- so the CodSpeed suite already runs correctly under a plain pytest invocation. Point the walltime job at tests/benchmarks and delete the mirror. The now-dead bench.* mypy override goes with it.

bluetoothbot:koan/dedupe-benchmark-suites

16 days ago

dependabot/github_actions/github-actions-e4a0e5090a

22 days ago

Latest Branches

chore(pre-commit.ci): pre-commit autoupdate#322

2 days ago

ff34195

pre-commit-ci-update-config

+1%

fix(utils): reject ints outside the 48-bit Bluetooth address space#325

16 days ago

316dad8

bluetoothbot:koan/int-to-address-range-check

test(bench): run one benchmark suite under both CI runners#324

16 days ago

9d319ef

bluetoothbot:koan/dedupe-benchmark-suites

Home Terms Privacy Docs