[epompeii]

Problem

bencher run --github-actions creates a Bencher Report GitHub Check on the PR head commit. The check was actually created (GitHub returns 201 Created), but octocrab 0.51 failed to deserialize the response, surfacing as a misleading failure:

Failed to create GitHub Check: JSON Error in pull_requests[0]: missing field `node_id` at line 1 column 4954

Reported in #894 (#894 (comment)) and reproduced at NomicFoundation/slang#1873.

Root cause

octocrab's CheckRun model deserializes the response pull_requests array into the full models::pulls::PullRequest type. In octocrab 0.51, PullRequest.node_id was a required String, but GitHub's create-check-run response returns only a minimal pull request representation (no node_id), so deserialization failed. The pull_requests array is only populated when the check is attached to a PR, which is why a plain push never surfaced it.

Bug: XAMPPRocky/octocrab#902

Fix

octocrab 0.52 relaxed PullRequest fields to Option to accept GitHub's SimplePullRequest payload, so the minimal check-run pull_requests objects now deserialize. This bumps octocrab 0.51 -> 0.53.1; the existing create_check_run().send() path then just works.

• None of the octocrab APIs in use changed across 0.51 -> 0.53.1 (builder, create_check_run, issue comments, CommentId, CheckRunConclusion/CheckRunOutput). The one relevant breaking change (PR fields -> Option) does not touch Bencher, which never reads PullRequest fields directly. The only new transitive dependency is http-serde 2.1.1 (MIT/Apache-2.0).
• Added a regression test that deserializes GitHub's minimal check-run pull_requests payload, guarding against an octocrab downgrade reintroducing the bug.
• Log on successful check creation (previously the path only logged on failure).

Release with fix: https://github.com/XAMPPRocky/octocrab/releases/tag/v0.52.0

Verification

• cargo build -p bencher_cli (no breaking API changes)
• cargo test -p bencher_cli (17 pass, including the new test; doc tests pass)
• cargo clippy --no-deps -p bencher_cli --all-targets --all-features -- -Dwarnings (clean)
• cargo clippy on bencher_github_client (the other octocrab consumer; clean)
• cargo fmt (no changes)

cargo deny check was not run locally (cargo-deny not installed in this environment); worth confirming in CI for the new http-serde dependency.

[socket-security]

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	cargo/​octocrab@​0.51.0 ⏵ 0.53.1

View full report

[github-actions]

🤖 Claude Code Review

PR: #902
Base: devel
Head: claude/github-actions-workflow-docs-dzr53c
Commit: 6c693a8eca3b45f13d6ed7ebc1b87d9e354a01dc

---

This is a small, focused PR. Here's my review.

Review: Fix GitHub Check creation failing to parse response (#894)

Summary

The PR fixes a parsing failure when creating GitHub Checks by bumping octocrab 0.51 → 0.53. Newer octocrab models the minimal/simple PR representation in check-run responses with optional fields (node_id, etc.), which previously failed deserialization. It also adds a regression test and a user-facing log line on success.

Strengths

• Correct root-cause fix. The dependency bump addresses the actual deserialization issue rather than papering over it.
• TDD compliance. A regression test (octocrab_pull_request_accepts_minimal_check_run_payload) is added, matching CLAUDE.md's TDD requirement, and it documents the linkage to issue enable posting to GitHub Checks API #894.
• Version declared in workspace Cargo.toml, per the "all dependency versions go in the workspace Cargo.toml" rule. Feature flags are preserved across the bump.
• Consistent logging. The new cli_println_quietable!(log, "Created GitHub Check.") matches the existing quietable-logging idiom used throughout the file.

Observations / Minor

1. cargo deny check should be run. Dependencies changed: a new transitive crate http-serde 2.1.1 was pulled in, octocrab bumped, and heck 0.4 → 0.5 shifted. CLAUDE.md requires cargo deny check when dependencies are added/updated. Worth confirming license/advisory checks pass for http-serde.

2. Test is a serde-only guard, not an integration test. It validates Vec<PullRequest> deserialization in isolation rather than exercising create_check_run. That's a reasonable and cheap regression guard given the real path needs a live/mocked GitHub API, but it won't catch breakage elsewhere in the response shape (e.g., the CheckRun wrapper type). The intent is clearly documented, so this is acceptable; just be aware the guard is narrow.

3. Comment accuracy nit. The test comment says "octocrab >= 0.52 models these as optional," while the workspace now pins 0.53. Harmless, but pinning the comment to the actually-used version (or just "octocrab 0.53") avoids confusion.

4. Asymmetric success logging. create_check_run now logs "Created GitHub Check." on success, but the parallel create_pull_request_comment path (line 371-373) has no equivalent success message. Not a bug, but if the goal is observability for the check path, consider whether the comment path deserves the same for consistency. Optional.

Verdict

Low-risk, well-scoped fix with a sensible regression guard. No bugs, security, or performance concerns. Recommend running cargo deny check, cargo nextest run -p bencher_cli, and cargo fmt/clippy before merge to satisfy the project's dependency-update and code-quality gates.

---

_{Model: claude-opus-4-8}

[github-actions]

Bencher Report

Branch	claude/github-actions-workflow-docs-dzr53c
Testbed	intel-v1

Click to view all benchmark results

Benchmark	Latency	Benchmark Result microseconds (µs) (Result Δ%)	Upper Boundary microseconds (µs) (Limit %)
Adapter::Json	📈 view plot 🚷 view threshold	4.69 µs (+2.32%) Baseline: 4.59 µs	4.82 µs (97.28%)
Adapter::Magic (JSON)	📈 view plot 🚷 view threshold	4.56 µs (+2.17%) Baseline: 4.47 µs	4.68 µs (97.56%)
Adapter::Magic (Rust)	📈 view plot 🚷 view threshold	25.56 µs (+0.43%) Baseline: 25.45 µs	26.42 µs (96.72%)
Adapter::Rust	📈 view plot 🚷 view threshold	3.49 µs (+0.69%) Baseline: 3.47 µs	3.59 µs (97.38%)
Adapter::RustBench	📈 view plot 🚷 view threshold	3.49 µs (+0.67%) Baseline: 3.47 µs	3.58 µs (97.56%)

🐰 View full continuous benchmarking report in Bencher