Skip to content

fix: md/raw switch alignment, and effect 4.0.4 span changes breaking CI - #1450

Merged
Makisuo merged 2 commits into
mainfrom
fix/view-switch-alignment
Oct 14, 2026
Merged

Makisuo merged 2 commits into
mainfrom
fix/view-switch-alignment

fix(effect-sdk): keep the 5xx response marker on effect 4.0.4 spans

4d3b5b2
Select commit
Loading
Failed to load commit list.
Maple Review Bot / Maple / review succeeded Oct 14, 2026 in 3m 23s

Confidence 8/10 · No issues found

🟢 Confidence 8/10 · likely safe to merge
The 5xx rule is narrow and pinned, but the Effect 4.0.2 HttpServerError/ResponseError shape is only mirrored by a hand-built fixture, never a real router round trip.
quality 100/100 · no findings · tests covered · risk medium

Two unrelated threads: ViewSwitch is pinned to the top of the system-prompt bodies (self-start), and the SDK tracer now treats Effect's own HttpServerError/ResponseError for a 5xx as the rendered 5xx. Both are contained; the tracer branch is narrowly gated and tested.

  • ViewSwitch takes self-start at the two body-level call sites
  • makeOtlpSpan accepts an HttpServerError/ResponseError failure for a 5xx as renderedServerError
  • Worker-bridge and warehouse span tests now select server spans by kind
  • New test pins HttpServerErrorResponse on a failed 5xx span
What was checked
  • New branch only fires on a single Fail reason tagged HttpServerError/ResponseError, so Die causes keep their own exception name (flushable-tracer.ts:263)
  • renderedServerError still requires kind server and status ≥ 500, so client spans and 4xx keep their old status (flushable-tracer.ts:254)
  • Anticipated identifiers are all 4xx (apps/ai/src/mcp/expected-failures.ts:19), so the new branch cannot override an anticipated-Ok span

4d3b5b2 · Updated on every push. Reply "won't fix" to dismiss a finding, or mention @maple-review-bot to ask about one.