eval/cargo · pac · Tangled

ryan.freumh.org / pac

This repository has no description

pac eval cargo

12 files

Ryan Gibb · SNAPSHOTS records a held-out crates.io-index of 2026-09-27, cargo's scale.sh runs against it through CARGO_INDEX and still refuses an index no row records, and SEED=20260927 draws the held-out sample · 1d ago

Cargo's check removes its cargo-home once cargo has answered, so a run's out/ keeps about 0.2 MB a query rather than 12

cargo: a crate's records of one name share a slot exactly when they ask the same requirement and features, as cargo's resolver then gives them one version; the cargo harness has one semver and one index path

cargo's check reads an index row off cargo's row schema as unreadable, and leaves unchecked a target or feature name whose Unicode class Python cannot settle

comments: say only what the code, the proofs and the tools' sources bear out; check cited Rocq names in runtest

eval: cut what-comments; drop the host-specific runbook

eval/cargo: an edge lost out of a crate cargo keeps is INVALID, whatever else holds the version; one reader of pac's answer

eval: a regression line is a whole query in the tool's syntax, in queries.txt; opam adds its two --with-test queries

run_query.py imports sys, which check_toolchain exits through

cargo: a crate's records of one name share a slot exactly when they ask the same requirement and features, as cargo's resolver then gives them one version; the cargo harness has one semver and one index path

SNAPSHOTS records a held-out crates.io-index of 2026-09-27, cargo's scale.sh runs against it through CARGO_INDEX and still refuses an index no row records, and SEED=20260927 draws the held-out sample

sparse_proxy.py

eval: one serve for shim.py and sparse_proxy.py, which asks the server what it serves; npm's command line and module loader shared

npm's and cargo's checks judge an answer consistent, npm's by the lock holding every edge and cargo's by an independent checker, and report the tool's lock apart