fix(compartment-mapper): improve canonical name generator performance by boneskull · Pull Request #3157 · endojs/endo

boneskull · 2026-04-04T02:16:41Z

This improves performance of finding the canonical name for a Node in a Graph dramatically
by computing all paths up-front and retrieving data from a cache of TraversalContext objects.

changeset-bot · 2026-04-04T02:16:46Z

🦋 Changeset detected

Latest commit: cdb6eae

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 7 packages

Name	Type
@endo/compartment-mapper	Patch
@endo/bundle-source	Patch
@endo/check-bundle	Patch
@endo/daemon	Patch
@endo/import-bundle	Patch
@endo/test262-runner	Patch
@endo/cli	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

Copilot

Pull request overview

This PR improves canonical-name (shortest path) computation performance in @endo/compartment-mapper by caching Dijkstra traversal results per source node, avoiding repeated full-graph traversals during mapNodeModules() graph finalization.

Changes:

Refactors Dijkstra to compute single-source shortest paths to all reachable nodes (removing the per-target parameter).
Adds a per-source traversal-context cache in makeShortestPath() to reuse computed shortest-path data across repeated queries.
Adds a changeset for a patch release noting the performance improvement.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
packages/compartment-mapper/src/generic-graph.js	Updates shortest-path implementation to single-source traversal + per-source caching for performance.
.changeset/shaky-pigs-tan.md	Declares a patch changeset describing the performance improvement.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-04T02:21:59Z

+ * Dijkstra's algorithm is a _single-source_ shortest path algorithm: one run
+ * produces shortest paths to every reachable node. The returned function
+ * caches the traversal context by source, so the first call for a given source
+ * pays O(V²) and every subsequent call with the same source is O(path length).
 *
 * @template [T=string]
 * @param {GenericGraph<T>} graph Graph to use
 */
 export const makeShortestPath = graph => {
+  /** @type {Map<T, TraversalContext<T>>} */
+  const contextCache = new Map();
+
  /**
   * @param {NoInfer<T>} source Source node
   * @param {NoInfer<T>} target Target node
   * @returns {[T, T, ...T[]]} Nodes from `source` to `target` inclusive (minimum of two nodes)
   */
  const shortestPath = (source, target) => {
-    const context = dijkstra(graph, source, target);
+    let context = contextCache.get(source);
+    if (!context) {
+      context = dijkstra(graph, source);
+      contextCache.set(source, context);
+    }


The new contextCache has no invalidation; if the graph is mutated after makeShortestPath(graph) is created (e.g., nodes/edges added/removed), cached traversal contexts can become stale and return incorrect paths or incorrect membership assertions. If the intended contract is that the graph is immutable/finalized once makeShortestPath is created, please document that explicitly in this JSDoc; otherwise consider a strategy to clear/recompute cached contexts when the graph changes.

This happens during graph finalization and the result is a readonly FinalGraph object, so no, not a big deal.

gibson042

I'm very much in favor of this change to memoize graph traversal contexts by source node.

However, reviewing it led me back to #2872, which I don't think I'd previously looked at and seems to have introduced some bad code that I think should be addressed in a fast-follow PR: https://github.com/endojs/endo/pull/2872/changes#r3042170892

gibson042 · 2026-04-06T23:34:02Z

  const context = {
    distances: new Map(),
    predecessors: new Map(),
    queue: new Set(),
  };
  const { queue, distances } = context;

  for (const node of nodes) {
    distances.set(node, Infinity);
  }

  assert(
    distances.get(source) === Infinity,
    `Source ${q(source)} is not in the graph`,
  );
-  assert(
-    distances.get(target) === Infinity,
-    `Target ${q(target)} is not in the graph`,
-  );

  distances.set(source, 0);

  for (const node of nodes) {
    queue.add(node);
  }


If we're optimizing anyway, should this be refactored too?

Suggested change

const context = {

distances: new Map([...nodes].map(node => [node, Infinity])),

predecessors: new Map(),

queue: nodes,

};

const { queue, distances } = context;

assert(

distances.get(source) === Infinity,

`Source ${q(source)} is not in the graph`,

);

distances.set(source, 0);

I'll update this, but I can't apply it as-is, since the suggestion doesn't span all of the lines it intends to update.

This improves performance of finding the canonical name for a `Node` in a `Graph` dramatically by computing all paths up-front and retrieving data from a cache of `TraversalContext` objects.

@kumavis

This PR was opened by the [Changesets release](https://github.com/changesets/action) GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to master, this PR will be updated. # Releases ## @endo/ocapn@1.0.0 ### Major Changes - [#3183](#3183) [`279c0c4`](279c0c4) Thanks [@kumavis](https://github.com/kumavis)! - Initial public release of `@endo/ocapn`. The package is no longer private and is now published to npm. Tested against the python test suite from 2026-01-06 <https://github.com/ocapn/ocapn-test-suite/commits/f0273f21c5ee05a28785b51c231535124f28bca9> ### Minor Changes - [#3172](#3172) [`6405b36`](6405b36) Thanks [@turadg](https://github.com/turadg)! - Parameterize CapTP slot types and improve TypeScript 6 conformance across the OCapN client surface. Compile-time type changes only; no runtime behavior changes. ### Patch Changes - Updated dependencies \[[`f65b000`](f65b000), [`d1d9625`](d1d9625), [`88bc2b9`](88bc2b9), [`e619205`](e619205), [`43165e5`](43165e5), [`6ada52b`](6ada52b)]: - @endo/eventual-send@1.5.0 - @endo/promise-kit@1.2.1 - @endo/pass-style@1.8.0 - @endo/marshal@1.9.1 - @endo/harden@1.1.0 - @endo/nat@5.2.0 ## ses@2.0.0 ### Major Changes - [#3153](#3153) [`e619205`](e619205) Thanks [@erights](https://github.com/erights)! - # Plug NaN Side-channel The JavaScript language can leak the bit encoding of a NaN via shared TypedArray views of an common ArrayBuffer. Although the JavaScript language has only one NaN value, the underlying IEEE 754 double-precision floating-point representation has many different bit patterns that represent NaN. This can be exploited as a side-channel to leak information. This actually happens on some platforms such as v8. @ChALkeR explains at <tc39/ecma262#758 (comment)> that the behavior of this side-channel on v8. At <https://junk.rray.org/poc/nani.html> he demonstrates it, and it indeed even worse than I expected. To plug this side-channel, we make two coordinated changes. - We stop listing the `Float*Array` constructors as universal globals. This prevents them from being implicitly endowed to created compartments, because they are not harmless. However, we still keep them on the start compartment (the original global), consider them intrinsics, and still repair and harden them on `lockdown()`. Thus, they can be explicitly endowed to child compartments at the price of enabling code in that compartment to read the side-channel. - On `lockdown()`, we repair the `DataView.prototype.setFloat*` methods so that they only write canonical NaNs into the underlying ArrayBuffer. The `@endo.marshal` package's `encodePassable` encodings need to obtain the bit representation of floating point values. It had used `Float64Array` for that. However, sometimes the `@endo/marshal` package is evaluated in a created compartment that would now lack that constructor. (This reevaluation typically occurs when bundling bundles in that package.) So instead, `encodePassable` now uses the `DataView` methods which are now safe. ### Minor Changes - [#3129](#3129) [`a675d8e`](a675d8e) Thanks [@erights](https://github.com/erights)! - `overrideTaming: 'moderate'` includes `overrideTaming: 'min'`. Previously `overrideTaming: 'min'` correctly enabled `Iterator.prototype.constructor` to be overridden by assignment, but due to an oversight, `overrideTaming: 'moderate'` did not. Now it does. To make such mistakes less likely, this PR also adopts a style where all records within larger enablements triple-dot the corresponding record from a smaller enablement, if present. ## @endo/bundle-source@4.3.0 ### Minor Changes - [#3180](#3180) [`7f7ae8e`](7f7ae8e) Thanks [@turadg](https://github.com/turadg)! - `BundleCache.load()` is now generic on the `format` option: - Omitted (default) → `Promise<BundleSourceResult<'endoZipBase64'>>` - Literal format → `Promise<BundleSourceResult<format>>` - Runtime-typed `ModuleFormat` → `Promise<BundleSourceResult<ModuleFormat>>` Previously `load()` returned `Promise<unknown>`, requiring callers to assert the bundle shape. ### Patch Changes - Updated dependencies \[[`154102b`](154102b), [`2b674ca`](2b674ca), [`d1d9625`](d1d9625), [`b4820dc`](b4820dc), [`acbacba`](acbacba), [`cdb6eae`](cdb6eae), [`6ada52b`](6ada52b), [`6ad084a`](6ad084a), [`1cd1246`](1cd1246)]: - @endo/compartment-mapper@2.1.0 - @endo/promise-kit@1.2.1 - @endo/harden@1.1.0 ## @endo/common@1.4.0 ### Minor Changes - [#3172](#3172) [`98c89b7`](98c89b7) Thanks [@turadg](https://github.com/turadg)! - Add `objectExtendEach` helper for merging a sequence of objects into an accumulator, with precise TypeScript inference of the resulting intersection type. ### Patch Changes - Updated dependencies \[[`f65b000`](f65b000), [`d1d9625`](d1d9625)]: - @endo/eventual-send@1.5.0 - @endo/promise-kit@1.2.1 - @endo/errors@1.3.1 - @endo/harden@1.1.0 ## @endo/compartment-mapper@2.1.0 ### Minor Changes - [#3132](#3132) [`b4820dc`](b4820dc) Thanks [@boneskull](https://github.com/boneskull)! - Expose `_redundantPreloadHook` option in `captureFromMap()`, which will be called for each item in the `_preload` array that was already indirectly loaded via the entry `Compartment`. Fixes a bug in the type of `_preload` option, which now allows for mixed arrays. Fixes a bug in the preloader, which was not exhaustively checking if a non-entry module was already loaded via the entry `Compartment`. - [#3048](#3048) [`6ad084a`](6ad084a) Thanks [@kriskowal](https://github.com/kriskowal)! - Add support for Node.js subpath pattern replacement in `package.json` `exports` and `imports` fields. Patterns like `"./features/*.js": "./src/features/*.js"` and `"#internal/*.js": "./lib/*.js"` are now resolved at link time using prefix/suffix string matching with specificity ordering. Null-target patterns exclude matching specifiers. Conditional pattern values are resolved through the standard condition-matching rules. Patterns are expanded to concrete module entries during archiving. ### Patch Changes - [#3111](#3111) [`154102b`](154102b) Thanks [@boneskull](https://github.com/boneskull)! - Fix type of `PackageDataHook.packageData` which now correctly allows `$root$` as a key. - [#3182](#3182) [`2b674ca`](2b674ca) Thanks [@kriskowal](https://github.com/kriskowal)! - Cull underscore-prefixed internal properties (like `__createdBy`) from serialized compartment maps in archives. The compartment map validator now also ignores underscore-prefixed properties when checking for extraneous fields. - [#3173](#3173) [`acbacba`](acbacba) Thanks [@boneskull](https://github.com/boneskull)! - Fixes potential issue wherein a canonical name may be computed incorrectly. Includes performance improvements. - [#3157](#3157) [`cdb6eae`](cdb6eae) Thanks [@boneskull](https://github.com/boneskull)! - Dramatically improve performance of canonical name (shortest path) computation in `mapNodeModules()`. - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - [#3115](#3115) [`1cd1246`](1cd1246) Thanks [@boneskull](https://github.com/boneskull)! - Remove unused "error" `ModuleSourceHookModuleSource` type. - Updated dependencies \[[`e619205`](e619205), [`6ada52b`](6ada52b), [`a675d8e`](a675d8e)]: - ses@2.0.0 - @endo/module-source@1.4.1 ## @endo/eventual-send@1.5.0 ### Minor Changes - [#3172](#3172) [`f65b000`](f65b000) Thanks [@turadg](https://github.com/turadg)! - Improve `E()` type inference and publicly export method-projection helpers. - `RemoteFunctions`, `PickCallable`, and `ECallableOrMethods` now short-circuit on `any`, preventing `E(anyValue)` from collapsing to an unusable type. - `EMethods`, `EGetters`, and related helpers are now part of the public type surface, so downstream packages can name the projected shapes `E()` produces. Compile-time type changes only; no runtime behavior changes. ### Patch Changes - Updated dependencies \[]: - @endo/harden@1.1.0 ## @endo/exo@1.7.0 ### Minor Changes - [#3172](#3172) [`88bc2b9`](88bc2b9) Thanks [@turadg](https://github.com/turadg)! - Improve TypeScript inference for patterns, exo, and pass-style. These are compile-time type changes only; no runtime behavior changes. - **pass-style**: `CopyArray<T>` is now `readonly T[]` so readonly tuples (e.g. `readonly ['ibc']`) satisfy `Passable`. Backward-compatible because `T[]` still extends `readonly T[]`. - **patterns**: `M.remotable()` defaults to `any` (matching `M.promise()`), so unparameterized remotables are assignable to concrete remotable typedefs. The parameterized form `M.remotable<typeof SomeInterfaceGuard>()` still yields precise inference. - **patterns**: `TFRemotable` returns `any` (not `Payload`) for non-`InterfaceGuard` arguments. - **patterns**: `TFOr` handles array-of-patterns and falls back through `TFAnd`; `M.undefined()` maps to `void`. - **patterns**: `TFOptionalTuple` emits truly optional elements; `M.promise()` maps to `PromiseLike`. - **patterns**: `TFSplitRecord` handles the empty-rest case correctly. - **patterns**: `TFRestArgs` unwraps array patterns. - **patterns**: `TypeFromArgGuard` discriminates by `toStringTag`, not structural shape. - **patterns**: `MatcherOf` payload is preserved through `InterfaceGuard`. - **patterns**: new `CastedPattern<T>` for unchecked type assertions in pattern position. - **exo**: `defineExoClass`, `defineExoClassKit`, and `makeExo` no longer intersect facet constraints with `& Methods`. The previous constraint collapsed specific facet keys into the `string | number | symbol` index signature, making `FilteredKeys` return `never` and erasing facet method inference (`Pick<X, never> = {}`). - **exo**: `Guarded<M, G>` is now structurally compatible across `G`, and the kit `F` constraint is widened. - **exo**: `defineExoClassKit` preserves facet inference when no guard is supplied. TypeScript consumers that were working around the previous inference gaps with casts may be able to remove those casts. Downstream code that depended on the narrower `CopyArray<T> = T[]` or the previous `M.remotable()` default may need minor adjustments. - [#3133](#3133) [`9111b4e`](9111b4e) Thanks [@turadg](https://github.com/turadg)! - feat: infer TypeScript types from pattern guards - `TypeFromPattern<P>` — infer static types from any pattern matcher - `TypeFromMethodGuard<G>` — infer function signatures from `M.call()` / `M.callWhen()` guards - `TypeFromInterfaceGuard<G>` — infer method records from interface guard definitions - `M.remotable<typeof Guard>()` — facet-isolated return types in exo kits - `M.infer<typeof pattern>` — namespace shorthand analogous to `z.infer` - `matches` and `mustMatch` now narrow the specimen type via type predicates - `makeExo`, `defineExoClass`, and `defineExoClassKit` enforce method signatures against guards at compile time These are compile-time type changes only; there are no runtime behavioral changes. Existing TypeScript consumers may see new type errors where method signatures diverge from their guards. ### Patch Changes - Updated dependencies \[[`8195a5a`](8195a5a), [`98c89b7`](98c89b7), [`f65b000`](f65b000), [`88bc2b9`](88bc2b9), [`9111b4e`](9111b4e), [`43165e5`](43165e5), [`df84eea`](df84eea), [`6ada52b`](6ada52b)]: - @endo/patterns@1.9.0 - @endo/common@1.4.0 - @endo/eventual-send@1.5.0 - @endo/pass-style@1.8.0 - @endo/errors@1.3.1 - @endo/far@1.1.14 - @endo/harden@1.1.0 ## @endo/pass-style@1.8.0 ### Minor Changes - [#3172](#3172) [`88bc2b9`](88bc2b9) Thanks [@turadg](https://github.com/turadg)! - Improve TypeScript inference for patterns, exo, and pass-style. These are compile-time type changes only; no runtime behavior changes. - **pass-style**: `CopyArray<T>` is now `readonly T[]` so readonly tuples (e.g. `readonly ['ibc']`) satisfy `Passable`. Backward-compatible because `T[]` still extends `readonly T[]`. - **patterns**: `M.remotable()` defaults to `any` (matching `M.promise()`), so unparameterized remotables are assignable to concrete remotable typedefs. The parameterized form `M.remotable<typeof SomeInterfaceGuard>()` still yields precise inference. - **patterns**: `TFRemotable` returns `any` (not `Payload`) for non-`InterfaceGuard` arguments. - **patterns**: `TFOr` handles array-of-patterns and falls back through `TFAnd`; `M.undefined()` maps to `void`. - **patterns**: `TFOptionalTuple` emits truly optional elements; `M.promise()` maps to `PromiseLike`. - **patterns**: `TFSplitRecord` handles the empty-rest case correctly. - **patterns**: `TFRestArgs` unwraps array patterns. - **patterns**: `TypeFromArgGuard` discriminates by `toStringTag`, not structural shape. - **patterns**: `MatcherOf` payload is preserved through `InterfaceGuard`. - **patterns**: new `CastedPattern<T>` for unchecked type assertions in pattern position. - **exo**: `defineExoClass`, `defineExoClassKit`, and `makeExo` no longer intersect facet constraints with `& Methods`. The previous constraint collapsed specific facet keys into the `string | number | symbol` index signature, making `FilteredKeys` return `never` and erasing facet method inference (`Pick<X, never> = {}`). - **exo**: `Guarded<M, G>` is now structurally compatible across `G`, and the kit `F` constraint is widened. - **exo**: `defineExoClassKit` preserves facet inference when no guard is supplied. TypeScript consumers that were working around the previous inference gaps with casts may be able to remove those casts. Downstream code that depended on the narrower `CopyArray<T> = T[]` or the previous `M.remotable()` default may need minor adjustments. - [#3184](#3184) [`43165e5`](43165e5) Thanks [@turadg](https://github.com/turadg)! - Unblock TypeScript declaration emit in downstream packages that structurally expose `PassStyled`/`Container` types. Compile-time type changes only; no runtime behavior changes. - `PASS_STYLE` is now typed as the string-literal `'Symbol(passStyle)'` rather than `unique symbol`. The runtime value is unchanged (still `Symbol.for('passStyle')`), and computed-key indexing like `obj[PASS_STYLE]` continues to work because JS computed keys accept any value. This removes TS4023 / TS9006 errors in consumers whose inferred types structurally contain `[PASS_STYLE]` (via `PassStyled`, `ExtractStyle`, object spread of a `PassStyled`, etc.). A `unique symbol` is only nameable via its original declaration module, which consumers have no reason to import; a string-literal type has no such nameability requirement. - `CopyArrayInterface`, `CopyRecordInterface`, and `CopyTaggedInterface` are now exported, so downstream `.d.ts` emit can name them when they appear through structural expansion of `Passable`/`Container`. - The `PassStyleOf` array overload is widened from `(p: any[]) => 'copyArray'` to `(p: readonly any[]) => 'copyArray'`, so `as const` tuples and `readonly T[]` values classify as `'copyArray'`. This aligns the classifier with `CopyArray<T>`, which is already `readonly T[]`. Backward-compatible because `T[]` still extends `readonly T[]`. Obviates the `@endo/pass-style` patch that agoric-sdk has been carrying in `.yarn/patches/`. TypeScript consumers that relied on `typeof PASS_STYLE` being `unique symbol` (e.g. annotating a value as `symbol` from `PASS_STYLE`) will need minor adjustments — widen the annotation to `symbol | string`, or cast via `unknown`. ### Patch Changes - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[[`98c89b7`](98c89b7), [`f65b000`](f65b000), [`d1d9625`](d1d9625)]: - @endo/common@1.4.0 - @endo/eventual-send@1.5.0 - @endo/promise-kit@1.2.1 - @endo/errors@1.3.1 - @endo/harden@1.1.0 ## @endo/patterns@1.9.0 ### Minor Changes - [#3067](#3067) [`8195a5a`](8195a5a) Thanks [@gibson042](https://github.com/gibson042)! - - Updates `containerHasSplit` to consider copyArray elements in forward order, better aligning with intuition. - [#3172](#3172) [`88bc2b9`](88bc2b9) Thanks [@turadg](https://github.com/turadg)! - Improve TypeScript inference for patterns, exo, and pass-style. These are compile-time type changes only; no runtime behavior changes. - **pass-style**: `CopyArray<T>` is now `readonly T[]` so readonly tuples (e.g. `readonly ['ibc']`) satisfy `Passable`. Backward-compatible because `T[]` still extends `readonly T[]`. - **patterns**: `M.remotable()` defaults to `any` (matching `M.promise()`), so unparameterized remotables are assignable to concrete remotable typedefs. The parameterized form `M.remotable<typeof SomeInterfaceGuard>()` still yields precise inference. - **patterns**: `TFRemotable` returns `any` (not `Payload`) for non-`InterfaceGuard` arguments. - **patterns**: `TFOr` handles array-of-patterns and falls back through `TFAnd`; `M.undefined()` maps to `void`. - **patterns**: `TFOptionalTuple` emits truly optional elements; `M.promise()` maps to `PromiseLike`. - **patterns**: `TFSplitRecord` handles the empty-rest case correctly. - **patterns**: `TFRestArgs` unwraps array patterns. - **patterns**: `TypeFromArgGuard` discriminates by `toStringTag`, not structural shape. - **patterns**: `MatcherOf` payload is preserved through `InterfaceGuard`. - **patterns**: new `CastedPattern<T>` for unchecked type assertions in pattern position. - **exo**: `defineExoClass`, `defineExoClassKit`, and `makeExo` no longer intersect facet constraints with `& Methods`. The previous constraint collapsed specific facet keys into the `string | number | symbol` index signature, making `FilteredKeys` return `never` and erasing facet method inference (`Pick<X, never> = {}`). - **exo**: `Guarded<M, G>` is now structurally compatible across `G`, and the kit `F` constraint is widened. - **exo**: `defineExoClassKit` preserves facet inference when no guard is supplied. TypeScript consumers that were working around the previous inference gaps with casts may be able to remove those casts. Downstream code that depended on the narrower `CopyArray<T> = T[]` or the previous `M.remotable()` default may need minor adjustments. - [#3133](#3133) [`9111b4e`](9111b4e) Thanks [@turadg](https://github.com/turadg)! - feat: infer TypeScript types from pattern guards - `TypeFromPattern<P>` — infer static types from any pattern matcher - `TypeFromMethodGuard<G>` — infer function signatures from `M.call()` / `M.callWhen()` guards - `TypeFromInterfaceGuard<G>` — infer method records from interface guard definitions - `M.remotable<typeof Guard>()` — facet-isolated return types in exo kits - `M.infer<typeof pattern>` — namespace shorthand analogous to `z.infer` - `matches` and `mustMatch` now narrow the specimen type via type predicates - `makeExo`, `defineExoClass`, and `defineExoClassKit` enforce method signatures against guards at compile time These are compile-time type changes only; there are no runtime behavioral changes. Existing TypeScript consumers may see new type errors where method signatures diverge from their guards. - [#3133](#3133) [`df84eea`](df84eea) Thanks [@turadg](https://github.com/turadg)! - Add optional `label` parameter to `M.promise()`, aligning its signature with `M.remotable(label?)`. When a label is provided, runtime error messages include it for diagnostics (e.g., "Must be a promise Foo, not remotable"). ### Patch Changes - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[[`98c89b7`](98c89b7), [`f65b000`](f65b000), [`88bc2b9`](88bc2b9), [`e619205`](e619205), [`43165e5`](43165e5), [`6ada52b`](6ada52b)]: - @endo/common@1.4.0 - @endo/eventual-send@1.5.0 - @endo/pass-style@1.8.0 - @endo/marshal@1.9.1 - @endo/errors@1.3.1 - @endo/harden@1.1.0 ## @endo/check-bundle@1.1.1 ### Patch Changes - [#3182](#3182) [`2b674ca`](2b674ca) Thanks [@kriskowal](https://github.com/kriskowal)! - Cull underscore-prefixed internal properties (like `__createdBy`) from serialized compartment maps in archives. The compartment map validator now also ignores underscore-prefixed properties when checking for extraneous fields. - Updated dependencies \[[`154102b`](154102b), [`2b674ca`](2b674ca), [`b4820dc`](b4820dc), [`acbacba`](acbacba), [`cdb6eae`](cdb6eae), [`6ada52b`](6ada52b), [`6ad084a`](6ad084a), [`1cd1246`](1cd1246)]: - @endo/compartment-mapper@2.1.0 - @endo/errors@1.3.1 - @endo/harden@1.1.0 ## @endo/errors@1.3.1 ### Patch Changes - Updated dependencies \[[`e619205`](e619205), [`a675d8e`](a675d8e)]: - ses@2.0.0 - @endo/harden@1.1.0 ## @endo/import-bundle@1.6.1 ### Patch Changes - Updated dependencies \[[`154102b`](154102b), [`2b674ca`](2b674ca), [`b4820dc`](b4820dc), [`acbacba`](acbacba), [`e619205`](e619205), [`cdb6eae`](cdb6eae), [`6ada52b`](6ada52b), [`6ad084a`](6ad084a), [`1cd1246`](1cd1246), [`a675d8e`](a675d8e)]: - @endo/compartment-mapper@2.1.0 - ses@2.0.0 - @endo/errors@1.3.1 - @endo/harden@1.1.0 ## @endo/lockdown@1.0.19 ### Patch Changes - Updated dependencies \[[`e619205`](e619205), [`a675d8e`](a675d8e)]: - ses@2.0.0 ## @endo/lp32@1.2.1 ### Patch Changes - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[]: - @endo/errors@1.3.1 - @endo/harden@1.1.0 - @endo/stream@1.3.1 ## @endo/marshal@1.9.1 ### Patch Changes - [#3153](#3153) [`e619205`](e619205) Thanks [@erights](https://github.com/erights)! - # Plug NaN Side-channel The JavaScript language can leak the bit encoding of a NaN via shared TypedArray views of an common ArrayBuffer. Although the JavaScript language has only one NaN value, the underlying IEEE 754 double-precision floating-point representation has many different bit patterns that represent NaN. This can be exploited as a side-channel to leak information. This actually happens on some platforms such as v8. @ChALkeR explains at <tc39/ecma262#758 (comment)> that the behavior of this side-channel on v8. At <https://junk.rray.org/poc/nani.html> he demonstrates it, and it indeed even worse than I expected. To plug this side-channel, we make two coordinated changes. - We stop listing the `Float*Array` constructors as universal globals. This prevents them from being implicitly endowed to created compartments, because they are not harmless. However, we still keep them on the start compartment (the original global), consider them intrinsics, and still repair and harden them on `lockdown()`. Thus, they can be explicitly endowed to child compartments at the price of enabling code in that compartment to read the side-channel. - On `lockdown()`, we repair the `DataView.prototype.setFloat*` methods so that they only write canonical NaNs into the underlying ArrayBuffer. The `@endo.marshal` package's `encodePassable` encodings need to obtain the bit representation of floating point values. It had used `Float64Array` for that. However, sometimes the `@endo/marshal` package is evaluated in a created compartment that would now lack that constructor. (This reevaluation typically occurs when bundling bundles in that package.) So instead, `encodePassable` now uses the `DataView` methods which are now safe. - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[[`98c89b7`](98c89b7), [`f65b000`](f65b000), [`88bc2b9`](88bc2b9), [`43165e5`](43165e5), [`6ada52b`](6ada52b)]: - @endo/common@1.4.0 - @endo/eventual-send@1.5.0 - @endo/pass-style@1.8.0 - @endo/errors@1.3.1 - @endo/harden@1.1.0 - @endo/nat@5.2.0 ## @endo/memoize@1.2.1 ### Patch Changes - [#3107](#3107) [`05cdb5f`](05cdb5f) Thanks [@erights](https://github.com/erights)! - `@endo/memoize` no longer depends on `ses`, just `@endo/harden` - Updated dependencies \[]: - @endo/harden@1.1.0 ## @endo/module-source@1.4.1 ### Patch Changes - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[[`e619205`](e619205), [`a675d8e`](a675d8e)]: - ses@2.0.0 ## @endo/netstring@1.1.1 ### Patch Changes - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[[`d1d9625`](d1d9625)]: - @endo/promise-kit@1.2.1 - @endo/harden@1.1.0 - @endo/stream@1.3.1 ## @endo/promise-kit@1.2.1 ### Patch Changes - [#3108](#3108) [`d1d9625`](d1d9625) Thanks [@erights](https://github.com/erights)! - `@endo/promise-kit` no longer depends on `ses`, just `@endo/harden` - Updated dependencies \[]: - @endo/harden@1.1.0 ## @endo/ses-ava@1.4.1 ### Patch Changes - Updated dependencies \[[`e619205`](e619205), [`a675d8e`](a675d8e)]: - ses@2.0.0 - @endo/harden@1.1.0 ## @endo/stream@1.3.1 ### Patch Changes - Updated dependencies \[[`f65b000`](f65b000), [`d1d9625`](d1d9625), [`e619205`](e619205), [`a675d8e`](a675d8e)]: - @endo/eventual-send@1.5.0 - @endo/promise-kit@1.2.1 - ses@2.0.0 - @endo/harden@1.1.0 ## @endo/stream-node@1.2.1 ### Patch Changes - [#3127](#3127) [`6ada52b`](6ada52b) Thanks [@turadg](https://github.com/turadg)! - Remove stale runtime dependencies from package manifests. - Updated dependencies \[]: - @endo/errors@1.3.1 - @endo/harden@1.1.0 - @endo/stream@1.3.1 ## @endo/daemon@2.5.3 ### Patch Changes - Updated dependencies \[[`8195a5a`](8195a5a), [`154102b`](154102b), [`2b674ca`](2b674ca), [`f65b000`](f65b000), [`d1d9625`](d1d9625), [`b4820dc`](b4820dc), [`88bc2b9`](88bc2b9), [`9111b4e`](9111b4e), [`acbacba`](acbacba), [`e619205`](e619205), [`df84eea`](df84eea), [`cdb6eae`](cdb6eae), [`6ada52b`](6ada52b), [`6ad084a`](6ad084a), [`1cd1246`](1cd1246), [`a675d8e`](a675d8e)]: - @endo/patterns@1.9.0 - @endo/compartment-mapper@2.1.0 - @endo/eventual-send@1.5.0 - @endo/promise-kit@1.2.1 - @endo/exo@1.7.0 - ses@2.0.0 - @endo/marshal@1.9.1 - @endo/netstring@1.1.1 - @endo/stream-node@1.2.1 - @endo/captp@4.5.0 - @endo/errors@1.3.1 - @endo/far@1.1.14 - @endo/harden@1.1.0 - @endo/import-bundle@1.6.1 - @endo/stream@1.3.1 ## @endo/stream-types-test@1.0.19 ### Patch Changes - Updated dependencies \[[`e619205`](e619205), [`a675d8e`](a675d8e)]: - ses@2.0.0 - @endo/nat@5.2.0 - @endo/stream@1.3.1 ## @endo/test262-runner@0.1.50 ### Patch Changes - Updated dependencies \[[`154102b`](154102b), [`2b674ca`](2b674ca), [`b4820dc`](b4820dc), [`acbacba`](acbacba), [`e619205`](e619205), [`cdb6eae`](cdb6eae), [`6ada52b`](6ada52b), [`6ad084a`](6ad084a), [`1cd1246`](1cd1246), [`a675d8e`](a675d8e)]: - @endo/compartment-mapper@2.1.0 - ses@2.0.0

boneskull self-assigned this Apr 4, 2026

boneskull added the performance label Apr 4, 2026

boneskull force-pushed the boneskull/shortest-path-fix branch from 902c10e to 2ce1968 Compare April 4, 2026 02:18

boneskull requested review from Copilot, kriskowal and naugtur April 4, 2026 02:19

boneskull added the lavamoat label Apr 4, 2026

Copilot started reviewing on behalf of boneskull April 4, 2026 02:20 View session

Copilot AI reviewed Apr 4, 2026

View reviewed changes

boneskull force-pushed the boneskull/shortest-path-fix branch from 2ce1968 to 87a6aeb Compare April 4, 2026 02:29

gibson042 approved these changes Apr 6, 2026

View reviewed changes

naugtur approved these changes Apr 8, 2026

View reviewed changes

boneskull force-pushed the boneskull/shortest-path-fix branch from 87a6aeb to 64947b9 Compare April 8, 2026 21:53

fix(compartment-mapper): improve canonical name generator performance

cdb6eae

This improves performance of finding the canonical name for a `Node` in a `Graph` dramatically by computing all paths up-front and retrieving data from a cache of `TraversalContext` objects.

boneskull force-pushed the boneskull/shortest-path-fix branch from 64947b9 to cdb6eae Compare April 8, 2026 21:54

boneskull enabled auto-merge April 8, 2026 22:00

boneskull merged commit 26a96aa into master Apr 8, 2026
26 checks passed

boneskull deleted the boneskull/shortest-path-fix branch April 8, 2026 22:00

kriskowal mentioned this pull request Apr 7, 2026

Version Packages #3104

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(compartment-mapper): improve canonical name generator performance#3157

fix(compartment-mapper): improve canonical name generator performance#3157
boneskull merged 1 commit into
masterfrom
boneskull/shortest-path-fix

boneskull commented Apr 4, 2026

Uh oh!

changeset-bot Bot commented Apr 4, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Copilot AI Apr 4, 2026

Uh oh!

boneskull Apr 4, 2026

Uh oh!

gibson042 left a comment

Uh oh!

gibson042 Apr 6, 2026

Uh oh!

boneskull Apr 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

boneskull commented Apr 4, 2026

Uh oh!

changeset-bot Bot commented Apr 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Copilot AI Apr 4, 2026

Choose a reason for hiding this comment

Uh oh!

boneskull Apr 4, 2026

Choose a reason for hiding this comment

Uh oh!

gibson042 left a comment

Choose a reason for hiding this comment

Uh oh!

gibson042 Apr 6, 2026

Choose a reason for hiding this comment

Uh oh!

boneskull Apr 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

changeset-bot Bot commented Apr 4, 2026 •

edited

Loading