What happened
The scheduled Performance Tests: Nightly workflow aborts in global setup before any performance test executes.
perf-tests/globalSetup.ts still imports and calls canUseRipgrep, but current packages/core/src/tools/ripGrep.ts no longer exports that function. It was replaced by resolveRipgrepPath in 0750b01fe4e7b62d55d92918b1c4535a77acc334.
The first nightly after that change, run 25841561983, reports:
No test files found, exiting with code 1
TypeError: (0 , __vite_ssr_import_3__.canUseRipgrep) is not a function
Object.setup globalSetup.ts:27:66
The latest checked run, 29470500728, fails at the same call site. All 64 scheduled performance runs from May 14 through July 16 failed, with no successful run in that interval.
Expected behavior
Global setup should resolve the current ripgrep binary API and allow perf-tests/perf-usage.test.ts to execute, so the nightly workflow can detect CPU and latency regressions.
Client information
This is the repository-owned scheduled workflow rather than a local Gemini CLI session, so /about output is not applicable.
- Runner:
gemini-cli-ubuntu-16-core
- Node.js:
v20.20.2
- npm:
10.8.2
- Latest checked main SHA:
3ff5ba20fc1ad7d867218bbdb34756eb54d6eccb
Login information
Not applicable. The failure occurs before any performance test or authenticated CLI scenario runs.
Anything else we need to know?
Update performance global setup to use the current ripgrep resolver and add focused coverage at that setup boundary so a stale runtime import cannot silently disable the nightly performance gate again.
What happened
The scheduled Performance Tests: Nightly workflow aborts in global setup before any performance test executes.
perf-tests/globalSetup.tsstill imports and callscanUseRipgrep, but currentpackages/core/src/tools/ripGrep.tsno longer exports that function. It was replaced byresolveRipgrepPathin0750b01fe4e7b62d55d92918b1c4535a77acc334.The first nightly after that change, run 25841561983, reports:
The latest checked run, 29470500728, fails at the same call site. All 64 scheduled performance runs from May 14 through July 16 failed, with no successful run in that interval.
Expected behavior
Global setup should resolve the current ripgrep binary API and allow
perf-tests/perf-usage.test.tsto execute, so the nightly workflow can detect CPU and latency regressions.Client information
This is the repository-owned scheduled workflow rather than a local Gemini CLI session, so
/aboutoutput is not applicable.gemini-cli-ubuntu-16-corev20.20.210.8.23ff5ba20fc1ad7d867218bbdb34756eb54d6eccbLogin information
Not applicable. The failure occurs before any performance test or authenticated CLI scenario runs.
Anything else we need to know?
Update performance global setup to use the current ripgrep resolver and add focused coverage at that setup boundary so a stale runtime import cannot silently disable the nightly performance gate again.