experimental-code-coverage-swarming-output-validator

Code Coverage Swarming Output Validator

Use this skill exclusively to audit remote Swarming bot executions for code coverage on Chromium tryjobs. It validates whether code coverage profile artifacts (.profraw) were generated and returned to Content Addressable Storage (CAS), verifies isolate packaging dependencies, and reproduces shard executions inside pinned LUCI containers.

---

When to Use This Skill

• Diagnosing why code coverage numbers dropped or returned empty profiles on CQ.
• Verifying whether a remote Swarming shard successfully produced .profraw outputs before attempting slow local compilations.
• Auditing CAS isolate inputs (.isolate, test_env.py) for missing data deps.
• Replaying a Swarming shard locally inside LUCI's hermetic CIPD/Python runtime.

---

Inputs

This skill requires the following inputs:

• swarming_server: The Swarming host (default: "chromium-swarm.appspot.com").
• cas_instance: The CAS instance name (default: "projects/chromium-swarm/instances/default_instance").
• test_suites: The target test suites list (e.g., ["net_unittests", "browser_tests"]).
• build_url: The Buildbucket build URL/link or build ID (e.g., "https://ci.chromium.org/b/8679...").

---

Workflow

1. Extract Task Digests & Swarming Task IDs

First, resolve the numeric Buildbucket build ID (build_id) from build_url. Note that LUCI URLs may use direct Buildbucket IDs (/b/8679...) or sequential builder paths (.../builders/try/linux-rel/2772707/...). When given a builder path URL, pass the canonical builder path (chromium/try/linux-rel/2772707) directly to bb get or use the buildbucket skill to resolve the numeric build ID.

Then, for each test suite in test_suites, invoke the standalone extraction helper script tools/code_coverage/get_task_digest.py passing build_id:

vpython3 tools/code_coverage/get_task_digest.py \
  --build {{build_id}} --step <test_suite>

This helper prints JSON containing the swarming_task_id and cas_input_digest for each target test suite.

2. Fast Remote Output Audit

For each extracted swarming_task_id, download and audit only the CAS output artifacts returned by the bot shard. This verifies remote execution integrity with zero local compilation.

mkdir -p scratch/swarm_out
vpython3 third_party/depot_tools/swarming.py collect \
  -S chromium-swarm.appspot.com -output-dir=scratch/swarm_out \
  {{swarming_task_id}}

• Profile Audit: Check if .profraw files exist in scratch/swarm_out/.
• Merge Check: Run llvm-profdata show on downloaded profile data:

  third_party/llvm-build/Release+Asserts/bin/llvm-profdata show \
    scratch/swarm_out/profraw/*.profraw
  

• Diagnosis:
  • Empty / ~2 Functions: If function count is abnormally low (e.g., only 2 Rust wrapper functions instead of thousands of C++ functions), immediately diagnose a Remote Siso/Recipe Collection Bug. The test executed remotely, but Siso/RBE failed to return intermediate profile artifacts.

3. Pre-Run CAS Input Audit

If remote outputs are missing entirely, audit the uploaded CAS input root to confirm whether GN data_deps bundled required runtime files.

• Download the isolate bundle using cas_input_digest:

  mkdir -p scratch/cas_in
  cas download \
    -cas-instance projects/chromium-swarm/instances/default_instance \
    -digest {{cas_input_digest}} -dir scratch/cas_in
  

• Inspect .isolate mappings and test_env.py to ensure profile environment variables (LLVM_PROFILE_FILE) are properly initialized.

4. Hermetic Container Replay

If remote profiles are corrupt or crash on the bot, replay the exact shard run locally on your workstation inside LUCI's pinned CIPD/Python environment using swarming_task_id.

mkdir -p scratch/container_repro && cd scratch/container_repro
vpython3 ../../third_party/depot_tools/swarming.py reproduce \
  -S chromium-swarm.appspot.com {{swarming_task_id}}

• Note: This abstracts away local OS differences and runs the test binary inside the exact virtualenv used by the LUCI compilator bot.

---