Skip to content

Conversation

@dkruces
Copy link
Contributor

@dkruces dkruces commented Oct 18, 2025

No description provided.

This allows to leverage linux-ci and kdevops-ci on any kdevops's
linux fork git ref [1]. Use mirror repo and name kdevops-linus that is
avaialble on the host where github selfhosted runers execute the
jobs/workflows.

Link: https://github.com/linux-kdevops/linux.git [1]

Signed-off-by: Daniel Gomez <[email protected]>
The linux-ci test mode was timing out after 2 hours when executing full
fstests baselines. This mode runs make fstests-baseline which executes
the complete test suite, taking 6, 11 or even 24 hours depending on the
profile complexity. The xfs_reflink_4k profile, for example, typically
requires 4-6 hours to complete all tests.

The workflow now uses conditional timeouts based on the test mode.
Quick validation via kdevops-ci continues using a 2-hour timeout as it
only runs a single test. Full test suite execution via linux-ci and
scheduled runs now gets 6 hours, which provides adequate time for
baseline completion with a reasonable safety margin.

GitHub Actions enforces a hard limit of 5 days (7,200 minutes) for jobs
on self-hosted runners. This limit is documented in CI-TIMEOUT-NOTES.md
along with implications for future soak duration testing support. The
pathological soak duration setting (48 hours per test across 46 tests)
would require approximately 92 days of serial execution, which exceeds
GitHub Actions limits and will require dedicated infrastructure or
parallel execution strategies.

Generated-by: Claude AI
Signed-off-by: Daniel Gomez <[email protected]>
The blktests result parsing was failing with a broken pipe error when
processing test results in linux-ci scheduled runs. The issue occurred
at line 153 of the test action where a massive list of test files was
being echoed through a pipe to head -1. With set -o pipefail enabled,
when head closed the pipe after reading the first line, the subsequent
SIGPIPE caused the shell to exit with an error.

The fix eliminates the intermediate variable and unnecessary echo by
piping find output directly to head. This avoids the broken pipe issue
while achieving the same result of selecting the first sample test file
for the summary report.

Generated-by: Claude AI
Signed-off-by: Daniel Gomez <[email protected]>
The blktests result collection was using intermediate variables and
echo pipes for counting tests, which is inefficient and could
potentially hit broken pipe issues with set -o pipefail. The test
name extraction piped through echo to count total tests, and failed
test detection stored results in a variable before piping through
echo to wc.

Optimize by piping find output directly to the counting pipeline,
eliminating the intermediate test_names and bad_files variables. This
makes the code more efficient with less memory usage and avoids any
potential SIGPIPE issues. The failed_tests count is now always set
directly from the find pipeline, eliminating the conditional logic
that was previously needed to handle the empty case.

Generated-by: Claude AI
Signed-off-by: Daniel Gomez <[email protected]>
@dkruces dkruces merged commit e1ab612 into main Oct 18, 2025
22 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants