chore: release 0.9.2

[stormslowly]

Release 0.9.2.

• Crate rspack_resolver: 0.9.1 → 0.9.2
• npm @rspack/resolver: 0.5.1 → 0.5.2

What's included since 0.9.1

• perf(cache): memoize package.json dep path per CachedPath (perf(cache): memoize package.json dep path per CachedPath #250)
• perf: optimize Specifier::parse (perf: optimize Specifier::parse #241)
• perf(cache): skip Box::pin on realpath cache hit (perf(cache): skip Box::pin on realpath cache hit #237)
• fix(cache): replay missing_dependency tracking in cached_node_modules warm path (fix(cache): replay missing_dependency tracking in cached_node_modules warm path #236)
• chore(deps): update dependency webpack to ^5.104.1 [security] (chore(deps): update dependency webpack to ^5.104.1 [security] #242)
• chore(deps): update napi (chore(deps): update napi #214)
• chore(deps): upgrade pnpm to 11.3.0 (chore(deps): upgrade pnpm to 11.3.0 #248)
• chore(bench): remove specifier parse microbenches (chore(bench): remove specifier parse microbenches #249)
• chore(bench): split specifier microbenches into a separate bench binary (chore(bench): split specifier microbenches into a separate bench binary #247)
• chore(bench): skip multi-threaded cases in CodSpeed memory mode (chore(bench): skip multi-threaded cases in CodSpeed memory mode #244)
• chore(ci): upload simulation-mode valgrind tmp files as artifact (chore(ci): upload simulation-mode valgrind tmp files as artifact #243)

[Copilot]

Pull request overview

Release PR updating the crate and npm package versions for rspack-resolver to ship v0.9.2 / @rspack/resolver v0.5.2.

Changes:

• Bump Rust crate version: rspack_resolver 0.9.1 → 0.9.2 (and reflect it in Cargo.lock).
• Bump npm package version: @rspack/resolver 0.5.1 → 0.5.2, including NAPI binding version checks.
• Update pnpm-lock.yaml (notably adding workspace importers for bindings/*).

Reviewed changes

Copilot reviewed 3 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
pnpm-lock.yaml	Updates pnpm lockfile; adds bindings/* importers entries.
npm/package.json	Bumps @rspack/resolver version to 0.5.2.
napi/index.js	Updates native binding version mismatch checks to expect 0.5.2.
Cargo.toml	Bumps crate version to 0.9.2.
Cargo.lock	Updates locked workspace crate version to 0.9.2.

Files not reviewed (1)

• pnpm-lock.yaml: Language not supported

---

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

[codspeed-hq]

Merging this PR will degrade performance by 3.95%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

❌ 1 regressed benchmark
✅ 9 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Memory	resolver[pnp resolve]	8.7 KB	9.1 KB	-3.95%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing release/0.9.2 (c509cb3) with main (888d0c3)}

[stormslowly]

@codspeedbot explain this regression

[codspeed-hq]

@codspeedbot explain this regression

The tool isn't available for this context since this is an explanation, not a code change. Let me provide my analysis directly.

Regression Analysis: resolver[pnp resolve] Memory (+3.95%)

This is not a real code regression — it is caused by a CI environment difference.

Evidence

I traced the pnp resolve Memory benchmark across all 4 runs on the release/0.9.2 branch and the baseline on main:

Commit	CPU	pnp resolve Memory
888d0c3 (main)	EPYC 7763	8.7 KB
1c23bf8 (release, initial)	EPYC 7763	8.7 KB
b6f205c (release, CI fix)	EPYC 7763	8.7 KB
48da9e8 (release, pnpm-lock)	EPYC 7763	8.7 KB
c509cb3 (release, latest)	EPYC 9V74	9.1 KB ⚠️

The regression only appears in the latest run (c509cb3), which is the only one that landed on an AMD EPYC 9V74 runner instead of the AMD EPYC 7763 used by all prior runs and the main baseline.

Why the environment matters for Memory mode

Memory mode uses Valgrind to count heap allocations. While this is typically deterministic for identical binaries, different runner environments can have different system library versions (e.g., glibc). These libraries may perform slightly different internal allocations during system calls (like path resolution), causing small measurement differences — exactly the ~400 bytes observed here.

Why this is safe to merge

1. No relevant code changes: The diff between main and release/0.9.2 contains only version bumps (0.9.1 → 0.9.2), CI workflow fixes, and a pnpm-lock update — nothing that touches Rust resolver logic or memory allocations.
2. Consistent baseline: The first 3 release runs reproduced the exact same 8.7 KB as main.
3. All other benchmarks unchanged: The remaining 9 benchmarks (including other Memory-mode benchmarks measuring MBs of allocations) are completely stable.
4. CodSpeed flagged the environment difference: The comparison report itself warns about the CPU change between the head and base runs.

Recommendation: This regression can be safely acknowledged on CodSpeed. A re-run that lands on the same runner type as the baseline would likely reproduce the original 8.7 KB result.

Summary

• Warming up the performance wizard...
• Investigating regression — Analysis complete — regression is environment-induced, not a real code change
• Handling comment — Comment handling finished successfully

_{Was this response helpful? React with 👍 or 👎 to help us improve CodSpeedBot.}