Breakwater

Backtest / #1259

perf(specifiers): push PEP 440 logic into the range engine

pypa/packaging#1259 · merged 2026-06-15 · base 155760e37a → head 664f049980 · ran on Docker

PASS

No behaviour divergence found by any probe.

† Merged after Nemotron 3 Ultra was released (2026-06-04): the model cannot have seen this PR or its fixes.

Performance · SpecifierSet#contains

No realistic input family measured slower.

Measured in backtest packaging-v2 (same commits, run bt-1259-8d2b3c, image bw-pair:ec9525174f993117). The semantic results above are from a later run; perf doesn't use the semantic planner, so it was not re-measured.

head slowerhead faster1×1001,00010,000input size (characters)F2F2: 1.00× at ~28 charsF2: 1.00× at ~44 charsF2: 1.00× at ~147 charsF2: 0.97× at ~579 charsF2: 1.00× at ~2,464 charsF2: 1.00× at ~10,169 charsF2: 1.00× at ~43,961 chars
  • F2
Head/base cachegrind instructions per call, by input size (log-log). Each line is one input family Nemotron proposed; points above 1× are where the "faster" code does more work.
FamilyInputsRatio range Slope base → headVerdict
F2SpecifierSet containing many simple comparators (>=, <=, ==, !=, <, >) with plain release versions and no local segment.claimed speedup · plan0.97× – 1.00×0.88 → 0.88neutral

Receipt

TargetBackendImageRun AuditAdjudication
packaging.specifiers:SpecifierSet#containslocal Dockerbw-pair:b2b8cb16aee35160bt-1259-562491$0.082$0.000

Nemotron cost on Nebius Token Factory. Backend: where the untrusted code ran. local Docker (an image cited by its content-addressed tag). Every image is content-addressed (repo at each commit + harness), so every card re-runs from the image and seed 0. Raw data: audit.