Constant GRAPH_CACHE_VERSION 

pub const GRAPH_CACHE_VERSION: u32 = 48;

Persisted graph cache schema version.

Bump this whenever the serialized shape of the persisted graph (any of the graph types that derive serde for the cache, the manifest types, or the store envelope) changes, so a stale graph-cache.bin written by an older binary is rejected rather than deserialized into the wrong shape.

Bump it for import-resolution semantics changes too, not only for wire-shape changes. The manifest compares this constant, the cache mode, and per-file fingerprints, and carries no binary version, so a cache written before a classification change replays the old classification verbatim on an unmodified tree and silently hides the new behaviour. The same applies to plugin config extraction, which seeds entry points and path aliases.

Bumped to 17 for issue #2031: cached resolver output retains canonical test-root replacements and ESM/CommonJS mechanisms, while profiled graphs retain target-sparse reachability masks plus exact compact reference routes. Ordinary graphs omit unused provenance. Versions 7 through 16 were used by published development commits for this change, so the final version remains 17 rather than reusing a potentially stale intermediate cache version.

Bumped to 18 for issue #2083: reference provenance moved out of SymbolReference into the per-export reference_paths side table, changing the persisted layout of every export’s reference list.

Bumped to 19 for issue #2084 (PR #2096): profiled test-reachability masking now falls back to the legacy fail-open classification when a graph would need more than the mask-profile cap, so caches written by the unbounded profiling code must not replay their old profiled results on large monorepos where the cap now engages.

Bumped to 20: cached re-export edges now carry the enclosing statement span and the source string-literal span used to anchor unresolved-import findings on the specifier. Warm 19 caches lack both spans.

Bumped to 21: persisted module graphs now carry the canonical effective export index used for named and star re-export resolution. Warm 20 caches lack that index and would replay the previous propagation semantics.

Bumped to 22: effective export names are interned once per graph and type and value resolutions share one compact key. Warm 21 caches contain the allocation-heavy intermediate index layout.

Bumped to 23: effective bindings distinguish direct declarations, namespace objects, and implicit SFC default exports.

Bumped to 24: namespace-object bindings retain their source module so consumers can enumerate the namespace through the canonical index.

Bumped to 25: export references retain their semantic Type or Value namespace so one named re-export surface can route both without duplicate graph symbols.

Bumped to 26: the effective export index retains typed declaration merge groups for namespace-aware reference selection.

Bumped to 27: declaration merge groups are stored once and referenced by a compact per-slot group identifier instead of cloning every group per slot.

Bumped to 28: reference-site deduplication now includes the exact import span, so one consumer can retain multiple distinct imports of one binding.

Bumped to 29: resolved semantic facts retain directly required structural type members used to protect explicit interface implementations.

Bumped to 30: declaration-merge references now propagate through named and star barrel surfaces using the canonical effective binding group.

Bumped to 31 for issue #2213: speculative __mocks__ sibling candidates that resolve to package space are no longer emitted, so warm 30 caches would keep replaying the phantom @scope/__mocks__ package edges the resolver no longer produces.

Bumped to 32: the effective export index resolves the type namespace in two lanes (real declarations and value-derived fallbacks), stores a per-file name index, and retains opaque bindings for known external named and namespace re-export surfaces. Warm 31 payloads neither describe the same structure nor carry those external bindings, so a consumed barrel export could be falsely reported as unused.

Bumped to 33 for issue #2225: speculative root-level __mocks__/<specifier> candidates from factory-less bare-specifier mocks now resolve to project files. Warm 32 caches lack those edges, so root manual mocks would stay reported as unused files. Bumped to 34: the effective export index only seeds the value-derived type fallback lane along re-export paths that reach a type-only re-export, and a type query reads the value lane where that lane is absent. A warm 33 payload is still read correctly, but a 33 reader would take an absent fallback lane for an absent type meaning and report consumed exports as unused.

Bumped to 35 for issue #2348: JSX member-expression tags (<SC.UsedStyle />) now record member accesses, and namespace narrowing bakes the credited references into the persisted graph. Warm 34 caches carry the old empty accessed-members verdict, so exports rendered only through JSX would stay reported as unused on upgrade.

Bumped to 36 for issue #2356: export declarations inside a namespace declared without the export keyword no longer contribute file-level exports, and unused-export verdicts are read off the persisted export set. Warm 35 caches still carry those local namespace members as file exports and would keep reporting them as unused on upgrade.

Bumped to 37 for issue #2357 (36 was taken by issue #2356 while this change was in review): a star re-export inside a declare module '<specifier>' body no longer becomes a ReExportEdge on the declaring module; the target’s full ES star surface is credited through a whole-module namespace edge instead, in both the type and the value namespace and following the target’s own export * and export * as ns chains. Warm 36 caches persist the old edge, the laundered re-export references, the runtime package usage for a bare-specifier ambient star, and type-lane-only credits that leave the value half of a same-name type and value pair unreferenced, and a graph-cache hit would replay all of them.

Bumped to 38 for issue #2355 (37 was taken by issue #2357 while this change was in review): Astro markup and MDX bodies now record member accesses for member-expression component tags (<SC.Card />), and namespace narrowing bakes the credited references into the persisted graph. Warm 37 caches carry the old empty accessed-members verdict for those consumers, so exports rendered only in Astro or MDX markup would stay reported as unused on upgrade.

Bumped to 39 for issues #2372 and #2373: a consumer that observes a whole namespace object (a whole-object namespace use of an import * as or of an export * as binding imported by name, a dynamic-import pattern match) and an export * as ns chain an entry point exposes now credit the names the target only exposes through its own export * and export * as chains, and those references are baked into the persisted graph. Warm 38 caches carry only the direct-export credit, so the star-forwarded and nested-namespace exports would stay reported as unused on upgrade. The same version covers the one direction that moves the other way: a plain export * hop no longer carries a downstream export * as default onward, so a chain behind such a namespace object can report one finding more than a warm 38 cache holds.

Bumped to 40 for issue #2374: the per-module export-name index now treats default as one importable name however each side spells it, so import { default as x } from './impl', an ambient declare module '<specifier>' { export { default } from './impl' }, and a plain import x from './impl' against an export { x as default } credit the target’s default export, and those references are baked into the persisted graph. Warm 39 caches carry the uncredited verdict, so the default export would stay reported as unused on upgrade.

Bumped to 41 for issue #2376 (40 was taken by issue #2374 while this change was in review): an MDX prose line opening with the word “import” (and any other line the statement parser rejects) no longer drops every import of the file, so those files now resolve their imports and credit their bodies. Warm 40 caches persist the import-less module and its missing edges, so the imported modules would stay reported as unused files on upgrade.

Bumped to 42 for issue #2377: a namespace import handed over whole (a call argument, a JSX attribute value, an alias, an array or object literal element, an initializer, an assignment right-hand side, a return value) no longer narrows to its dotted accesses, and those mark-all verdicts are baked into the persisted graph. Warm 41 caches carry the narrowed verdict, so the siblings the consumer can still reach would stay reported as unused on upgrade.

Bumped to 43 for issue #2365: import X = require('./x') now resolves to a CommonJS namespace import edge, and the references it credits are baked into the persisted graph. Warm 42 caches hold the module without that edge, so the target would stay reported as an unused file on upgrade.

Bumped to 44 for issue #2375: export type * inside a declare module body now carries a type-only-star symbol edge instead of a file-level star re-export, and the type-lane-only credit it hands the target is baked into the persisted graph. Warm 43 caches hold the re-export edge, the laundered entry surface, and the value-lane credits, and a graph-cache hit skips the build entirely.

Bumped to 45 for issues #2391, #2395, and #2397: namespace handover now covers require, dynamic-import, Vue, and CSS Module bindings; ambient plain stars retain star-surface exposure; equivalent default spellings share the same duplicate-export rules; and proven CommonJS object maps use member narrowing while resolved CommonJS and CSS Module default imports preserve whole-object handoffs. Those reference outcomes are baked into the persisted graph, so a warm 44 cache would skip the corrected build logic.

Bumped to 46 for PR #2436: a tsconfig reached through references without include or files now applies only to files under its own directory instead of every file, so paths from a referenced package no longer resolve imports in sibling packages. Resolver output is persisted with the graph and the cache key does not cover tsconfig scope, so a warm 45 cache would keep replaying the leaked cross-package resolutions.

Bumped to 47 for issue #2444 (PR #2435): Yarn Plug’n’Play projects now resolve bare specifiers through the inlined .pnp.cjs manifest, anchored to the manifest directory, instead of missing on the empty node_modules and falling back. The resolver output persisted in a warm 46 cache holds those misses and would replay them as unresolved imports.

Bumped to 48: a directory a framework serves at a URL mount now resolves root-absolute references from any HTML document, not only from Storybook’s preview fragments, and SvelteKit declares its static/ directory as such a mount. A warm 47 cache holds the resolver’s earlier miss and would replay it as an unresolved import plus an unused asset file.