fix(perf): audit and vitals agree on LCP via shared apply_lcp_fields (iter-125) - #165
Merged
Merged
Conversation
…er-125) `perf audit` rebuilt its .results.vitals LCP triple with a bare `lcp.map(|v| rate(v, 2500, 4000))` that lacked the iter-83 N7 unavailable-guard, so on a page with no measurable LCP — where the DOM approximation fabricates a `startTime: 0` entry (compute_lcp == Some(0.0), is_lcp_approximate == true) — audit reported `lcp_ms: 0.0, lcp_rating: "good"`: a false all-clear, while `perf vitals` on the same page correctly reported `lcp_ms: null, lcp_rating: "unavailable"`. Extract the guard + note/approximate annotation into a single `pub(crate) apply_lcp_fields` helper and call it from both `run_vitals` and `run_audit`, so the two commands share one LCP code path and cannot drift. render_audit_text renders `[unavailable]` (no `0 ms good`) since lcp_ms is now null. Adds 5 audit-side unit tests (twins of the vitals tests, incl. unit_perf_audit_lcp_unavailable_matches_vitals) and a new live test file live_perf_audit_lcp_unavailable.rs with live_perf_audit_lcp_unavailable and live_perf_audit_vitals_lcp_parity — both verified against real Firefox. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The `unit_perf_audit_lcp_unavailable_matches_vitals` AC slug starts with `unit_`, which ac-fidelity-check heuristic 1 (live_/test_/bench_) does not match; backtick-quote it (and `apply_lcp_fields`) so heuristic 2 resolves it against the perf.rs diff. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
iter-125 fixed an analogous vitals/audit drift by extracting the divergent logic into a shared helper plus an explicit parity assertion between the two call sites. Point iter-126 at the same pattern for the network JSON shape fix (shared builder + parity test between navigate and the standalone network command), and flag that firefox_refs line numbers should be re-verified against main before starting.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
perf auditfabricated a false all-clear on pages where LCP is unmeasurable: it rebuilt its.results.vitalsLCP triple with a barelcp.map(|v| rate(v, 2500, 4000))that lacked the iter-83 N7 unavailable-guard. When the DOM approximation fabricates astartTime: 0entry (compute_lcp == Some(0.0),is_lcp_approximate == true), audit reportedlcp_ms: 0.0, lcp_rating: "good"— whileperf vitalson the same page correctly reportedlcp_ms: null, lcp_rating: "unavailable"(dogfooding-session-61).lcp_note/lcp_approximateannotation into a singlepub(crate) apply_lcp_fieldshelper inperf.rs; bothrun_vitalsandrun_auditnow route the LCP triple through it, so the two commands share one code path and cannot drift again.render_audit_textnow prints[unavailable](no0.00ms good) sincelcp_msis null for the unmeasurable case; measurable-LCP pages (~587 ms regression-agent case) are unaffected — the helper reduces to the samerate(v, 2500, 4000)path.Test plan
unit_perf_audit_lcp_unavailable_when_missing— audit block withlcp = Noneyields"unavailable"/ null (not "good"/0.0).unit_perf_audit_lcp_unavailable_when_approximate_zero— the comparis.ch-class case:Some(0.0)+ approximate →"unavailable"/ null.unit_perf_audit_lcp_unavailable_matches_vitals— audit block equals vitals block field-for-field for both missing-LCP and approximate-zero inputs via the shared helper.unit_perf_audit_lcp_rated_when_approximate_nonzero/unit_perf_audit_lcp_rated_when_measurable— negative controls: real/measurable LCP is still rated, not nulled.live_perf_audit_lcp_unavailable— text-only fixture page, no LCP candidate:perf auditreports.results.vitals.lcp_rating == "unavailable",.lcp_ms == null(verified against real Firefox).live_perf_audit_vitals_lcp_parity— same page, same session:perf audit.results.vitals.{lcp_ms, lcp_rating}equalsperf vitals.results.{lcp_ms, lcp_rating}(verified against real Firefox).cargo fmt/cargo clippy --workspace --all-targets -- -D warnings/cargo test --workspace -qall clean.check-iteration-ready10/10 PASS (withFF_RDP_LIVE_TESTS=1).🤖 Generated with Claude Code## Claims vs code
<generated 2026-07-19T15:29:43Z by ralph-loop>
No claims extracted from commit messages.