Performance History
Latest Results
fix(module-graph): assert the primary type store stays frozen after finalization
register_type grows the primary store, and every lazy id is offset past
that store's length, so growing it after finalization would shift the
lazy/primary id partition and misroute already-minted ids. The invariant
held structurally — the resolver is shared as `&self` behind an `Arc` once
constructed — but was only documented; assert it so any future caller that
grows the store post-finalization fails loudly.
minseong0324:fix/no-misleading-return-type-generic-alias-unions
6 hours ago
test(lint): cover a cross-module imported generic-alias union
Exercises the lazy store's cross-module import path end to end: a
`Maybe<string>` imported from another module resolves to `string | null`,
so a function returning only a string is flagged by noMisleadingReturnType.
minseong0324:fix/no-misleading-return-type-generic-alias-unions
7 hours ago
fix(module-graph): materialize generic aliases only after inference finalizes
Lazy generic-alias materialization mints ids offset past the primary type
store's length. When it ran during `run_inference` (while the primary
store was still growing) the lazy and primary id ranges could overlap,
misrouting lookups. Defer materialization until the primary store is
frozen (a `finalized` flag set after inference), and add debug assertions
at the id-routing and insertion sites that fail loudly if the lazy base
ever shifts. Also bound the reference-following loop and document the
lazy store, its id-offset invariant, and the on-demand materialization.
minseong0324:fix/no-misleading-return-type-generic-alias-unions
8 hours ago
refactor(module-graph): document and guard the lazy instantiation invariant
Document the lazy instantiation side store (the boxcar-backed append-only
store, the LazyInstantiationResolver, and the on-demand materialization),
and add a debug assertion that the primary-store base used for lazy ids
never shifts. The lazy/primary id partition relies on the primary store
being frozen before any lazy type is minted; the assertion turns a future
violation of that invariant into a loud test failure instead of silent
id misrouting.
minseong0324:fix/no-misleading-return-type-generic-alias-unions
8 hours ago
refactor(xtask): drop --ts-tag/--ts-sha/--offline codegen overrides
minseong0324:chore/global-types-codegen-pr1
8 hours ago
perf(type-info): short-circuit the conditional type-parameter check
Determining whether a conditional type depends on a type parameter
walked every ancestor up to the module root, scanning each ancestor's
children for a type-parameter list — including the module node, whose
children are all top-level declarations. For a declaration-heavy .d.ts
with many conditional types this dominated module-graph collection.
Collect the check/extends references first, bail when there are none,
and stop the ancestor walk at the first enclosing parameter that
matches. The result is unchanged.
minseong0324:fix/no-misleading-return-type-generic-alias-unions
9 hours ago
Latest Branches
0%
6 hours ago
d67df72
minseong0324:fix/no-misleading-return-type-generic-alias-unions
0%
9 hours ago
8fa41bd
minseong0324:chore/global-types-codegen-pr1
+6%
10 hours ago
d390d94
renovate/serde_json-1.x