💅 noFloatingPromises is 3.89x slower in 2.5.12 on a Zod enum call

[jakeleventhal]

Environment information

Details

CLI:
  Version: 2.5.12

Platform:
  CPU Architecture: aarch64
  OS: macos 27.0
  CPU: Apple M1 Max

Environment:
  BIOME_DISTRIBUTION: npm
  JS_RUNTIME_VERSION: v24.12.0
  JS_RUNTIME_NAME: node

Biome Configuration:
  Status: Loaded successfully
  Path: biome.json
  VCS enabled: false
  Enabled rule: nursery/noFloatingPromises

Rule name

nursery/noFloatingPromises

Playground link

https://github.com/jakeleventhal/biome-promise-rules-perf-repro

The repository pins Biome 2.5.10 and 2.5.12 side by side. Run:

bun install --frozen-lockfile
bun run bench
bun run profile:floating:2.5.10
bun run profile:floating:2.5.12

The analyzed file contains one call expression:

import { enumSchema } from "./enum-schema";

enumSchema.safeParse("one");

enumSchema is a two-value Zod 4.3.6 enum.

Five-run averages on the machine above:

Version	End-to-end average	noFloatingPromises rule time
2.5.10	1.470s	1.481s
2.5.12	5.712s	5.343s

That is a 3.89x end-to-end slowdown and a 3.61x increase inside the single rule invocation.

Related: #11534. That report covers a broader noFloatingPromises regression present before 2.5.12. This reproduction isolates a further regression between 2.5.10 and 2.5.12 to one imported ZodEnum.safeParse() call. Please close this as a duplicate if both cases have the same cause.

Expected result

Biome 2.5.12 should process this single call in roughly the same time as 2.5.10. noFloatingPromises should not spend more than five seconds classifying one non-Promise Zod result.

Code of Conduct

• I agree to follow Biome's Code of Conduct