perf(rust): evaluate gzip backend after installer optimization

[mohamedmansour]

Decision: Closed in favor of the miniz-only architectural improvements in #2676. We are not pursuing zlib-rs.

Proof: In the matched macOS benchmark, the architecture alone makes cold installation 27% faster, warm verification 46% faster, and repairs 23-24% faster. Warm retained physical memory falls from 174.35 to 1.76 MiB. Measured revisions: 9553d522 to 4ab7f873.

Safety: Keep the existing backend, bounded streaming, trusted file verification, and repair behavior. Native Windows/Linux validation remains tracked in #2676. Full measurements and reproduction.

[Copilot]

🟢 Approval recommended

The focused backend optimization preserves existing behavior and includes appropriate regression coverage.

Pull request overview

Optimizes Rust bundled gzip extraction without changing installer behavior or public APIs.

Changes:

• Enables flate2’s zlib-rs runtime backend.
• Uses bufread::GzDecoder for embedded byte slices.
• Adds gzip integrity and extraction regression tests.

File summaries

File	Description
rust/src/embeddedcli.rs	Updates decoding paths and adds contract tests.
rust/Cargo.toml	Selects the optional zlib-rs backend.
rust/Cargo.lock	Locks zlib-rs 0.6.8.

Review details

• Files reviewed: 2/3 changed files
• Comments generated: 0
• Review effort level: Balanced (auto)

Note

Copilot is running an experiment and ran this review at Balanced.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[stephentoub]

The underlying installer performance problem is real, but I recommend addressing the architectural cost first without adding a new decompression backend. install_runtime currently traverses the same embedded gzip archive three times for the normal runtime and four times with bundled-in-process; warm verification also reconstructs large entries. Consolidating installation into a single streaming traversal and using build-generated hashes or metadata to avoid decompression for valid warm installs should provide a larger end-to-end improvement while retaining the existing miniz_oxide dependency.

After that work lands, benchmark cold, warm, and repair paths again and compare release binary sizes. If raw inflate time is still material, zlib-rs can then be evaluated as a focused second-stage optimization with evidence that its remaining gain justifies the added dependency and unsafe implementation surface.

Generated by Copilot