Compare · Codebase Pulse vs. Trunk & BuildPulse
Point tools like Trunk Flaky Tests or BuildPulse ingest JUnit XML at the end of a pipeline and count pass/fail flips. They are blind to CPU saturation, disk stalls, and container crashes that happened on the runner while tests executed.
When a shared cloud runner hiccups or Redis dies mid-suite, XML-only tools flood dashboards with false flaky candidates. Codebase Pulse pairs TRX, nextest, and JUnit with 3-layer environment protection — and covers Build, Performance, and Codebase in the same workflow.
3-layer false-flaky protection
NoiseDetector (CPU, RAM, disk I/O, core count), IsMassFailure for suite-wide crashes, and selector drift detection keep outages off your flaky list.
Fix tests, do not only mute them
See shared-fixture co-failures and parallel execution overlap hints — not just auto-quarantine by flip rate.
One agent for the whole pipeline
One pulse-agent covers flaky tests, MSBuild/Cargo compile speed, benchmarks, load tests, and SARIF deltas — not a per-seat flaky-only SKU.
Feature comparison
Side-by-side capabilities for .NET and Rust teams evaluating a pre-production CI signal layer.
| Capability | Codebase Pulse | Trunk & BuildPulse |
|---|---|---|
| Test report formats | VSTest .trx, cargo nextest / libtest JSON, JUnit XML | Primarily JUnit XML |
| Runner hardware telemetry | Yes — excludes CPU saturation, RAM pressure, disk I/O stalls | No — VM starvation can look like flakiness |
| Mass infrastructure outage guard | Yes — skips flaky/baseline updates when shared deps crash the suite | No — mass outages pollute flaky statistics |
| Partial test filter detection | Detects selector drift when a subset of tests ran | Limited or manual |
| Root-cause diagnostics | Co-failure counts, parallel overlap, normalized signatures | Flip frequency and basic grouping |
| Automatic clean eviction | Evicts resolved flaky/slow tests after consecutive clean runs | Often manual status management |
| Broader CI observability | Build (.binlog/Cargo), Performance (BDN/k6/OTel), Codebase (SARIF/Git) | Flaky tests only (or separate products) |
On top of what you have
Keep Trunk or BuildPulse for quarantine workflows if they work for you. Pulse adds runner context and pipeline breadth flaky-only tools skip.
First-class VSTest TRX and cargo nextest — not JUnit XML alone for .NET and Rust suites.
Hardware telemetry and mass-failure guards so Redis/DB outages do not become hundreds of false flaky candidates.
Fixture co-failure and parallel overlap hints to fix races — not only flip-rate quarantine.
Same PR comment as MSBuild/Cargo regressions, benchmarks, load tests, and SARIF deltas from one pulse-agent install.
Common questions
Pulse focuses on classification, PR headlines, and optional --guard — plus eviction when tests stay clean. Quarantine workflows remain in your CI policy; Pulse gives evidence (noisy run vs same-commit flip vs new failure).
Quarantine hides symptoms. Pulse shows whether a flip was a noisy VM, a mass Redis outage, or a real same-commit race — and correlates it with compile and perf regressions on the same PR.