Skip to content

QVAC-24089 test[skiplog]: skip redundant CosyVoice3 e2e tests on mobile - #4030

Merged
Victor-Rodzko merged 4 commits into
mainfrom
test/qvac-24089-skip-cosyvoice3-mobile
Aug 25, 2026
Merged

QVAC-24089 test[skiplog]: skip redundant CosyVoice3 e2e tests on mobile#4030
Victor-Rodzko merged 4 commits into
mainfrom
test/qvac-24089-skip-cosyvoice3-mobile

Conversation

@Victor-Rodzko

@Victor-Rodzko Victor-Rodzko commented Aug 24, 2026

Copy link
Copy Markdown
Contributor

🎯 What problem does this PR solve?

  • All 7 tts-cosyvoice3-* e2e tests run on both mobile platforms (Android + iOS), but several are very slow on Device Farm. In the last full mobile run, 4 of them timed out: emotion-conditioning (360s), native-streaming (180s), duplex-streaming (180s), default (120s).
  • Most of that coverage is redundant — the streaming/emotion variants overlap with TTS functionality already exercised by other e2e tests, so mobile pays a large wall-clock cost for little added signal.
  • Raised in the "QVAC full test-suite run" thread; redundancy confirmed by @gusttav-lang.

📝 How does it solve it?

  • Keep the two tests that carry distinct mobile signal: tts-cosyvoice3-default (happy path) and tts-cosyvoice3-invalid-emotion (error path).
  • Mark the other five as expected skips on mobile via the existing skipTests helper, alongside the other mobile-only exclusions in packages/sdk/e2e/tests/mobile/consumer.ts:
    • tts-cosyvoice3-emotion-conditioning
    • tts-cosyvoice3-streaming
    • tts-cosyvoice3-native-streaming
    • tts-cosyvoice3-sentence-streaming
    • tts-cosyvoice3-duplex-streaming
  • Unconditional (not Platform.OS-gated) since the reasoning applies equally to Android and iOS.

🧪 How was it tested?

  • Test-scope change only — no SDK source or public API touched. Desktop/Electron/Snap consumers are untouched, so those platforms keep full CosyVoice3 coverage as the source of truth.
  • The five test IDs were verified to match the definitions in packages/sdk/e2e/tests/tts-tests.ts exactly, so each resolves to a real test rather than silently matching nothing.
  • Expect the mobile skip count to go from 125 → 130 (Android) and 137 → 142 (iOS), with the corresponding drop in mobile run time.
  • CI run (android + ios full)

@Victor-Rodzko
Victor-Rodzko requested review from a team as code owners August 24, 2026 16:04
@github-actions

github-actions Bot commented Aug 24, 2026

Copy link
Copy Markdown
Contributor

Review Status

Current Status: ✅ APPROVED
Approvals so far: Team Lead: 1, Member: 1

@github-actions

Copy link
Copy Markdown
Contributor

License compliance — clean

No new dependency license findings in this PR.

Warn-only (shadow) mode — this check does not block merges yet.

Updated automatically by the canonical license compliance workflow.

NOTICE presence (advisory)

Missing NOTICE (advisory, does not block):

  • ./.github/actions/release-merge-guard
  • ./docs/website
  • ./packages/ggml-coload-smoke
  • ./packages/fabric/test/integration
  • ./packages/inference-addon-cpp/mobile
  • ./packages/sdk/e2e
  • ./packages/llm-llamacpp/benchmarks/performance
  • ./packages/llm-llamacpp/benchmarks/server
  • ./packages/vla-ggml/sim/server
  • ./packages/embed-llamacpp/benchmarks/performance
  • ./packages/embed-llamacpp/benchmarks/server
  • ./packages/asr-ggml/benchmarks/server

@Victor-Rodzko Victor-Rodzko added the test-e2e-full Triggers full e2e test suite [Currently SDK-only] label Aug 24, 2026
@Victor-Rodzko
Victor-Rodzko force-pushed the test/qvac-24089-skip-cosyvoice3-mobile branch from 86a4a70 to bdc342c Compare August 25, 2026 08:34
- all 7 tts-cosyvoice3-* e2e tests ran on both mobile platforms; several are
  slow on Device Farm (emotion-conditioning timed out at 360s,
  native-streaming and duplex-streaming at 180s in the last full mobile run)
  and overlap coverage already provided by other TTS tests
- keep tts-cosyvoice3-default and tts-cosyvoice3-invalid-emotion on mobile;
  mark the other five as expected skips via the existing skipTests helper
- desktop/Electron coverage for the skipped tests is unchanged

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@Victor-Rodzko
Victor-Rodzko force-pushed the test/qvac-24089-skip-cosyvoice3-mobile branch from bdc342c to 8c377ff Compare August 25, 2026 09:57
@Victor-Rodzko
Victor-Rodzko marked this pull request as ready for review August 25, 2026 12:40
- the harness derives its per-test timeout from estimatedDurationMs as
  max(estimate * 2, 120s), so a 60s estimate capped this test at 120s
- CosyVoice3 has no GPU path on Mali, so it synthesises on CPU on the Android
  Device Farm Pixel 9 Pro; the test measured 114015ms there (pass, run
  32718078677) and over 120s (timeout, runs 32827063693 and 32834986175) — it
  was sitting on its own cap and flipping run to run
- raise the estimate to 150s so the cap becomes 300s, ~2.5x the observed time
- iOS (Metal, 25s) and Adreno Android (~30s) are well inside either bound

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
@Victor-Rodzko
Victor-Rodzko force-pushed the test/qvac-24089-skip-cosyvoice3-mobile branch from ba10a18 to faadb0f Compare August 25, 2026 12:41
@Victor-Rodzko
Victor-Rodzko merged commit 2ce4ed8 into main Aug 25, 2026
35 checks passed
@Victor-Rodzko
Victor-Rodzko deleted the test/qvac-24089-skip-cosyvoice3-mobile branch August 25, 2026 15:02
@Victor-Rodzko Victor-Rodzko changed the title QVAC-24089 test: skip redundant CosyVoice3 e2e tests on mobile QVAC-24089 test[skiplog]: skip redundant CosyVoice3 e2e tests on mobile Aug 25, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

test-e2e-full Triggers full e2e test suite [Currently SDK-only]

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants