ci: Add remaining passing evergreen rdk unit tests by oxve · Pull Request #9216 · youtube/cobalt

oxve · 2026-02-24T07:40:52Z

Enable remaining passing unit tests and add test filters for:

cc:cc_unittests_loader
media/mojo:media_mojo_unittests_loader
third_party/blink/renderer/platform/wtf:wtf_unittests_loader

filters for targets that crash on exit

Also add an RDK-specific runfile generator. The new generator sets
environment variables and initializes the display system to work around
EGL init issues.

Bug: 484121488

github-actions · 2026-02-24T07:41:12Z

🤖 Gemini Suggested Commit Message

ci: Add remaining EG RDK unit tests

Enable the full suite of unit tests for the evergreen-arm-hardfp-rdk
platform in CI. This change adds new test targets and corresponding
filters for known failures on this specific platform.

The build system is updated to use an RDK-specific runfile generator.
This new generator sets essential environment variables and initializes
platform services (westeros-init, rdkDisplay) required for proper test
execution in the RDK environment.

Additionally, unit tests are configured to run in single-process mode,
and the CI test result processing is made more robust by explicitly
passing environment variables to ensure accurate failure detection.

Bug: 382508397

💡 Pro Tips for a Better Commit Message:

Influence the Result: Want to change the output? You can write custom prompts or instructions directly in the Pull Request description. The model uses that text to generate the message.
Re-run the Generator: Post a comment with: /generate-commit-message

gemini-code-assist

Code Review

This pull request adds more unit tests for the eg-rdk target, which is a great step towards improving test coverage. The changes to the CI workflow and test configurations are logical. However, I found a critical issue in the new script starboard/rdk/generate_runfile.py which is completely broken as written. I've provided a detailed comment with a full code suggestion to fix it.

starboard/rdk/generate_runfile.py

oxve · 2026-02-25T08:26:49Z

/gemini-review

oxve · 2026-02-25T08:37:23Z

/gemini review

gemini-code-assist

Code Review

The pull request enables more unit tests for the evergreen RDK target by adding new test executables and filter files for known failures. It also introduces a new RDK-specific runfile generator to handle platform-specific setup, and refactors a GitHub Action workflow for better clarity.

My review focuses on the new runfile generator script (starboard/rdk/generate_runfile.py), where I've identified a critical issue with resource cleanup, a high-severity issue with ASAN options that could mask memory errors, and a medium-severity issue regarding a fragile time.sleep. The other changes in CI configuration and test lists appear correct and improve the test infrastructure.

gemini-code-assist · 2026-02-25T08:39:09Z

starboard/rdk/generate_runfile.py

+try:
+    result = subprocess.run(command, check=False)
+    sys.exit(result.returncode)
+except subprocess.CalledProcessError:
+    # A subprocess failed, so don't log the python traceback.
+    raise SystemExit(1)
+except Exception as e:
+    print("An unexpected error occurred: " + str(e), file=sys.stderr)
+    sys.exit(1)
+
+subprocess.run(['rdkDisplay', 'remove'])


The subprocess.run(['rdkDisplay', 'remove']) call on line 72 is unreachable because sys.exit() is called within the try block on line 64, which terminates the script prematurely. This will lead to leaked display resources on the test device.

Additionally, the except subprocess.CalledProcessError: block (lines 65-67) is dead code, as this exception is only raised when check=True is used with subprocess.run().

To ensure cleanup always runs and to remove the dead code, the logic should be restructured using a try...finally block.

Suggested change

try:

result = subprocess.run(command, check=False)

sys.exit(result.returncode)

except subprocess.CalledProcessError:

# A subprocess failed, so don't log the python traceback.

raise SystemExit(1)

except Exception as e:

print("An unexpected error occurred: " + str(e), file=sys.stderr)

sys.exit(1)

subprocess.run(['rdkDisplay', 'remove'])

return_code = 1

try:

result = subprocess.run(command, check=False)

return_code = result.returncode

except Exception as e:

print("An unexpected error occurred: " + str(e), file=sys.stderr)

# return_code is already 1

finally:

subprocess.run(['rdkDisplay', 'remove'], check=False)

sys.exit(return_code)

gemini-code-assist · 2026-02-25T08:39:09Z

starboard/rdk/generate_runfile.py

+    # TODO: Move?
+    "ASAN_OPTIONS": "exitcode=0",


Setting ASAN_OPTIONS to exitcode=0 will cause the process to exit with code 0 even if AddressSanitizer detects a memory error. This can mask critical bugs and cause failing tests to be reported as passing. This option should be removed to ensure that memory errors are properly reported as test failures. The associated TODO comment suggests uncertainty about this option, reinforcing that it should likely be removed.

lol , might be good to have a proper bug assovciated with this ( more formal todo)

gemini-code-assist · 2026-02-25T08:39:09Z

starboard/rdk/generate_runfile.py

+if subprocess.run(['pgrep', 'westeros-init'], capture_output=True).returncode != 0:
+    subprocess.Popen(['/usr/bin/westeros-init'])
+import time
+time.sleep(2)


Using time.sleep(2) to wait for westeros-init to start can lead to flaky tests. The 2-second delay may not be sufficient on a heavily loaded system, causing subsequent commands to fail. A more robust approach would be to implement a polling mechanism that actively checks if the westeros-init service is ready before proceeding, for example by checking for a PID file or a listening socket.

niranjanyardi · 2026-02-25T20:41:56Z

starboard/rdk/generate_runfile.py

+if subprocess.run(['pgrep', 'westeros-init'], capture_output=True).returncode != 0:
+    subprocess.Popen(['/usr/bin/westeros-init'])
+import time
+time.sleep(2)


can you add this as a variable and write a comment about why "2) was selcted ?

niranjanyardi · 2026-02-25T20:43:31Z

starboard/rdk/generate_runfile.py

+    "ASAN_OPTIONS": "exitcode=0",
+}})
+
+# TODO: Remove this when westeros-init is started by default.


this should be tracked in a partner bug

niranjanyardi · 2026-02-25T20:45:06Z

starboard/rdk/generate_runfile.py

+import subprocess
+import sys
+
+# TODO: Remove when lab devices are updated to RDK_V6_20251218+.


good to track in actual buganizer issue

niranjanyardi · 2026-02-25T20:48:02Z

cobalt/build/modular_executable.gni


  action("${original_target_name}_runfile_gen") {
-    script = "//starboard/build/util/generate_runfile.py"
+    # TODO: Investigate why starboard_target_platform is not defined for evergreen-arm.


can you add it to rdk's args.gn ?

bit surprised this didnt trigger: https://source.corp.google.com/h/github/youtube/cobalt/+/main:starboard/build/platform_path.gni;l=70?q=starboard_target_platform%20%3D&ss=h%2Fgithub%2Fyoutube%2Fcobalt%2F%2B%2Frefs%2Fheads%2Fmain

this seems wrong , this file is suppsoed to be platform agnostic. Is there a better way to do this inside the RDK code ?

oxve added 2 commits February 18, 2026 08:07

Single process tests

2269815

ci: Add remaining eg-rdk unit tests

6c420ce

oxve marked this pull request as ready for review February 24, 2026 07:40

gemini-code-assist bot reviewed Feb 24, 2026

View reviewed changes

starboard/rdk/generate_runfile.py Show resolved Hide resolved

oxve added 4 commits February 24, 2026 19:00

Fix target paths

5e63745

Fix rdk runfile generator

95225c2

Move imports

fd36784

zlib tests pass but return non-zero status

e19206d

oxve changed the title ~~ci: Add remaining eg-rdk unit tests~~ ci: Add remaining passing evergreen rdk unit tests Feb 25, 2026

oxve added 4 commits February 25, 2026 05:54

Fix end of file newlines

ce7fb4e

Also print tests passed, skipped and total

1204139

Split disabled tests to separate json fields

cb215d4

Merge branch 'main' of github.com:youtube/cobalt into rdk-tests-part-2

7fa3af6

gemini-code-assist bot reviewed Feb 25, 2026

View reviewed changes

niranjanyardi reviewed Feb 25, 2026

View reviewed changes

Conversation

oxve commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Feb 24, 2026

🤖 Gemini Suggested Commit Message

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

oxve commented Feb 25, 2026

Uh oh!

oxve commented Feb 25, 2026

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

niranjanyardi Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

niranjanyardi Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

niranjanyardi Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

niranjanyardi Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

niranjanyardi Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

niranjanyardi Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

oxve commented Feb 24, 2026 •

edited

Loading