Utilize GitHub Actions for building and running via QEMU

[G36maid]

Implement GitHub Actions CI pipeline for testing Linmo across GNU and
LLVM RISC-V toolchains with automated reporting.

Objective: Ensure the Linmo kernel builds cleanly with both supported toolchains

Features:

• Builds Linmo kernel with both supported toolchains
• Matrix testing for GNU and LLVM toolchains
• Automated app discovery and crash detection
• Functional test framework with criterion-based validation
• TOML-based result aggregation across toolchains
• Automated PR comments with detailed test results
• Proper failure handling and artifact collection

Components:

• .ci/ci-tools.sh: Unified tool for data collection, aggregation,
  and PR comment formatting
• .ci/run-app-tests.sh: Automated app testing with QEMU
• .ci/run-functional-tests.sh: Criterion-based functional tests
• .ci/setup-toolchain.sh: Automated toolchain installation
• .github/workflows/ci.yml: Complete CI workflow with summary job
• arch/riscv/build.mk: Builds Linmo kernel with both supported toolchains

The workflow ensures all tests run to completion, captures full output
for debugging, and fails CI based on the aggregated overall status

Slove #2

• step1 : Build Linmo Kernel
• step2 : Application Crash Sanity Check
• step3 : Run Dedicated Functional Tests (partial)

[jserv]

Rebase the latest main branch.

[jserv]

There should be only a commit, discarding unrelated ones.

[G36maid]

Rebase the latest main branch.

Already.

There should be only a commit, discarding unrelated ones.

Sure.

[jserv]

Carefully read https://cbea.ms/git-commit/ and strictly apply its rules when writing full sentences in commit messages.

[jserv]

Always rebase properly.

[G36maid]

Always rebase properly.

I'll make sure to rebase properly before merging. For now, I prefer to keep the commits.
Once everything is finalized, I'll squash or clean them up as needed.

[G36maid]

Do these results match what we wanted?
@jserv @Damien-Chen
Please check this run:
G36maid#6 (comment)

[cubic-dev-ai]

6 issues found across 7 files

Prompt for AI agents (all 6 issues)

Understand the root cause of the following 6 issues and fix them.


<file name=".ci/run-app-tests.sh">

<violation number="1" location=".ci/run-app-tests.sh:32">
Treating exit code 124 as success causes hung QEMU runs to be marked as passing; exit status 124 from `timeout` indicates the command timed out, so this branch should fail the test instead.</violation>
</file>

<file name="arch/riscv/build.mk">

<violation number="1" location="arch/riscv/build.mk:22">
The clang-toolchain detection probes $(CROSS_COMPILE)gcc, so on LLVM jobs (where only riscv32-unknown-elf-clang is installed) CC_IS_CLANG remains empty and CC falls back to the nonexistent riscv32-unknown-elf-gcc, breaking the LLVM build.</violation>
</file>

<file name=".ci/run-functional-tests.sh">

<violation number="1" location=".ci/run-functional-tests.sh:79">
A timeout from the QEMU run (exit code 124) is currently treated the same as a clean exit, so a test that hangs after emitting the PASS lines will still be marked as passing. Please fail the test when timeout terminates QEMU to avoid masking hangs.</violation>
</file>

<file name=".ci/ci-tools.sh">

<violation number="1" location=".ci/ci-tools.sh:187">
The aggregated GNU app entries are written as `app=passed`, which is invalid TOML because the string value is unquoted. This makes the generated summary file unparsable; please quote both the key and value here (and update the matching LLVM line as well).</violation>

<violation number="2" location=".ci/ci-tools.sh:197">
Functional test results are emitted as `test=passed` without quotes, yielding invalid TOML in the `[gnu.functional_tests]` section. Please emit quoted keys/values here (and mirror the change for the LLVM line).</violation>

<violation number="3" location=".ci/ci-tools.sh:207">
Functional criteria entries are written as `fft:max_error=passed`, leaving both the key (contains a colon) and the string value unquoted, which violates TOML syntax. Quote the key and value here (and apply the same adjustment to the LLVM line).</violation>
</file>

---

Since this is your first cubic review, here's how it works:

• cubic automatically reviews your code and comments on bugs and improvements
• Teach cubic by replying to its comments. cubic learns from your replies and gets better over time
• Ask questions if you need clarification on any suggestion

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

[cubic-dev-ai]

4 issues found across 7 files

Prompt for AI agents (all 4 issues)

Understand the root cause of the following 4 issues and fix them.


<file name=".ci/run-app-tests.sh">

<violation number="1" location=".ci/run-app-tests.sh:32">
Treating exit code 124 (timeout) as a pass causes hung QEMU runs to be reported as successful. A timeout should fail the test instead of being marked as passed.</violation>
</file>

<file name=".ci/ci-tools.sh">

<violation number="1" location=".ci/ci-tools.sh:187">
The GNU apps entries are emitted as `key=value` without spaces or quotes, so the generated TOML is invalid and format_comment cannot parse the section.</violation>

<violation number="2" location=".ci/ci-tools.sh:197">
GNU functional test entries are emitted without TOML-compliant formatting, preventing format_comment from parsing and making the TOML invalid.</violation>

<violation number="3" location=".ci/ci-tools.sh:207">
GNU functional criteria entries are written as `key=value`, which is not valid TOML and breaks parsing of the criteria section.</violation>
</file>

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

[cubic-dev-ai]

3 issues found across 6 files

Prompt for AI agents (all 3 issues)

Understand the root cause of the following 3 issues and fix them.


<file name=".ci/run-app-tests.sh">

<violation number="1" location=".ci/run-app-tests.sh:32">
A timeout (exit code 124 from `timeout`) indicates the QEMU run hung, but this branch treats it as a pass. That masks failures and lets broken apps appear healthy; only exit code 0 should be considered passing.</violation>
</file>

<file name="arch/riscv/build.mk">

<violation number="1" location="arch/riscv/build.mk:22">
The clang detection runs $(CROSS_COMPILE)gcc --version, but the LLVM toolchain only ships $(CROSS_COMPILE)clang, so CC_IS_CLANG never becomes 1 and the build selects the missing GNU binaries, breaking LLVM builds.</violation>
</file>

<file name=".ci/ci-tools.sh">

<violation number="1" location=".ci/ci-tools.sh:187">
The aggregated gnu apps section writes values like `app1=passed` without quotes, producing invalid TOML that downstream tools cannot parse.</violation>
</file>

_{React with 👍 or 👎 to teach cubic. Mention @cubic-dev-ai to give feedback, ask questions, or re-run the review.}

[cubic-dev-ai]

No issues found across 6 files

[visitorckw]

Again, why does this PR include adding LLVM support, while the description makes it look like it's only adding CI tests?

[G36maid]

Sorry, just updated the description.
It’s based on a single issue that covers both CI and LLVM support.
Would you prefer that I split it into separate PRs?

[visitorckw]

Yes, or IMHO, at least they should be in separate commits.
Ref: https://www.kernel.org/doc/html/v6.17/process/submitting-patches.html#separate-your-changes