ci: matrix builds with native ARM64 runners and PR test comments

[Copilot]

CI Workflow Upgrades

Plan

• Update release.yml to add merge commit extended description to release notes
  • Extract merge commit message body from the merge commit itself
  • Place it at the top of release notes before commit history
  • Check if PR was actually merged before running release steps
  • Use merge_commit_sha instead of HEAD for merge commit body
• Update build.yml to use matrix for multiple architectures
  • Create matrix job with x64, x86, and arm64 architectures
  • Use native windows-11-arm runner for ARM64 testing
  • Use x64 runner for x86/x64 (WoW64 compatibility)
  • Update test execution to run on appropriate runners
  • Fix ARM64 to use native compilation (not cross-compilation)
  • Fix test exit code capture
  • Remove unused test output variable
  • Use unique timestamp-based delimiter for multiline output
  • Create standard executable name (win-witr.exe) for tests
  • Capture test stdout/stderr in test results
  • Add job status field to test results JSON
  • Fix undefined variable in catch block
• Add PR comment with test results to build.yml
  • Collect test output during test runs
  • Format test results
  • Post comment to PR only when triggered by pull request
  • Show test pass/fail counts by architecture
  • Handle compile/setup failures and show appropriate messages

Changes Made

release.yml:

• Added extraction of merge commit extended description (commit body)
• Placed merge commit notes at the top of release notes as "Release Notes" section
• Followed by separator and then the auto-generated commit history
• Added guard to only run if PR was actually merged (github.event.pull_request.merged == true)
• Use github.event.pull_request.merge_commit_sha instead of HEAD for correct merge commit

build.yml:

• Converted to matrix-based build strategy with three architectures: x64, x86, arm64
• ARM64 now uses windows-11-arm runner for native compilation and testing (no cross-compilation needed!)
• x86 uses x64 runner with WoW64 compatibility
• Each architecture compiles and tests in parallel with fail-fast: false
• Creates architecture-specific binaries (win-witr-{arch}.exe) and standard name copy for tests
• Tests now capture stdout/stderr output and include in test results
• Added job status field to test results JSON
• Uses timestamp-based delimiter for multiline output to avoid conflicts
• Fixed undefined variable issue by initializing $output outside try block
• Test results are saved as artifacts and then posted as a formatted PR comment
• PR comments only appear on pull requests (not on push events)
• Comment shows test summary for each architecture with expandable details including test output
• Handles compile/setup failures by checking job status and showing appropriate error messages

Original prompt

---

This section details on the original issue you should resolve

<issue_title>ci upgrade</issue_title>
<issue_description>we don't have to cross compile arm anymore, github actions has windows-11-arm labels for runners.

the upgrades I want:

• in release.yml, I want the actual merge commit extended description to be added to the top of the release as a sorta personal "release notes" so i can write it myself, and then the commit history is after that.
• in build.yml, i want it to be a matrix that compiles all arch versions, then another matrix that tests them (it can be the same matrix job actually). for x86 there isn't a specific runner for that but we can just run it in x64 either way since they're compatible (WoW64). the only one that's really that different is arm64
• in build.yml, perhaps we can take the output of the batch tests and post them in a comment ONLY when the build runs in a pull request. that way, i can see a comment like:

Tests Completed

x/x tests passed

Test 1 (name of bat file)

``` output here ```

and so on</issue_description>

<agent_instructions>Do NOT make a documentation markdown file for your additions. Any of your "documentation" must be in the yml comments or the pull request description. Do NOT, I repeat, do NOT touch the c++ file(s)</agent_instructions>

Comments on the Issue (you are @copilot in this section)

• Fixes ci upgrade #61

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

Summary by CodeRabbit

Chores

• Build and test workflow expanded to support x64, x86, and arm64 architectures with independent execution and detailed per-architecture test result summaries posted to pull requests
• Release notes generation now includes merge commit information for improved traceability

_{✏️ Tip: You can customize this high-level summary in your review settings.}

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

• ✅ Full review completed - (🔄 Check again to review again)

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 3

🤖 Fix all issues with AI agents

In @.github/workflows/build.yml:
- Around line 143-208: The test result JSON needs a job status field and the
comment logic must consider it: update the step that writes test results to
include a stable field (e.g., "status" or "job.status") reflecting the job
outcome, then modify the GitHub Script that reads files (the files loop using
const files = ... and data = JSON.parse(...)) to mark allPassed = false when
either data.failed > 0 OR data.status !== 'success'; also adjust the
status/emoji logic (the const status = ... line) to use data.status alongside
failed so compile/setup failures show ❌ and the commentBody accurately reports
failures when tests never ran. Ensure you reference the same field name in both
the producer and consumer of the JSON.
- Around line 56-89: The test runner currently discards stdout/stderr with "|
Out-Null", losing output needed for PR comments; update the block that invokes
$_.FullName to capture the command output into a variable (e.g., $output = &
$_.FullName 2>&1), then use $LASTEXITCODE as before to determine pass/fail and
append the captured output to $testResults (include it alongside the
"PASSED"/"FAILED" message and exit code) so downstream post-test-results job can
post detailed logs; ensure the catch block also captures and appends $_ (the
exception) and any $output to $testResults for failed runs.

In @.github/workflows/release.yml:
- Around line 158-162: The workflow uses git log on HEAD to build
mergeCommitBody (variable mergeCommitBody) which is invalid when a PR is closed
without merging; add a guard that checks github.event.pull_request.merged ==
true before running release-note generation steps and replace the HEAD lookup
with the actual merge commit SHA (github.event.pull_request.merge_commit_sha)
when populating mergeCommitBody so the workflow always reads the correct merge
commit message.

[supervoidcoder]

@copilot In @.github/workflows/build.yml:

• Around line 143-208: The test result JSON needs a job status field and the
  comment logic must consider it: update the step that writes test results to
  include a stable field (e.g., "status" or "job.status") reflecting the job
  outcome, then modify the GitHub Script that reads files (the files loop using
  const files = ... and data = JSON.parse(...)) to mark allPassed = false when
  either data.failed > 0 OR data.status !== 'success'; also adjust the
  status/emoji logic (the const status = ... line) to use data.status alongside
  failed so compile/setup failures show ❌ and the commentBody accurately reports
  failures when tests never ran. Ensure you reference the same field name in both
  the producer and consumer of the JSON.
• Around line 56-89: The test runner currently discards stdout/stderr with "|
  Out-Null", losing output needed for PR comments; update the block that invokes
  $.FullName to capture the command output into a variable (e.g., $output = &
  $.FullName 2>&1), then use $LASTEXITCODE as before to determine pass/fail and
  append the captured output to $testResults (include it alongside the
  "PASSED"/"FAILED" message and exit code) so downstream post-test-results job can
  post detailed logs; ensure the catch block also captures and appends $_ (the
  exception) and any $output to $testResults for failed runs.

In @.github/workflows/release.yml:

• Around line 158-162: The workflow uses git log on HEAD to build
  mergeCommitBody (variable mergeCommitBody) which is invalid when a PR is closed
  without merging; add a guard that checks github.event.pull_request.merged ==
  true before running release-note generation steps and replace the HEAD lookup
  with the actual merge commit SHA (github.event.pull_request.merge_commit_sha)
  when populating mergeCommitBody so the workflow always reads the correct merge
  commit message.

[Copilot]

@copilot In @.github/workflows/build.yml:

• Around line 143-208: The test result JSON needs a job status field and the
  comment logic must consider it: update the step that writes test results to
  ...

Fixed in commit 8906b05:

build.yml:

• Test runner now captures stdout/stderr output instead of discarding with | Out-Null and includes it in test results
• Added status field to test results JSON based on steps.run_tests.outcome
• Updated comment logic to check both data.failed > 0 and data.status !== 'success'
• Comment now shows "Job failed" message when compile/setup failures prevent tests from running
• Exception handling also captures and includes output in test results

release.yml:

• Added if: github.event.pull_request.merged == true guard to only run when PR is actually merged
• Changed merge commit lookup from HEAD to github.event.pull_request.merge_commit_sha for the correct merge commit message

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In @.github/workflows/build.yml:
- Around line 197-205: The logic currently hides test counts when status !==
'success'; modify the branch around commentBody construction (variables: status,
total, passed, failed, arch, commentBody) so the "Job failed - Tests may not
have run..." message is only added when total === 0 (compile/setup failure). In
practice, change the condition to check both status and total (e.g., if (status
!== 'success' && total === 0) show the setup-failure message) and ensure the
else branch prints the `${passed}/${total} tests passed` line even when status
=== 'failure' but total > 0 so failed test runs still display counts.

🧹 Nitpick comments (1)

.github/workflows/release.yml (1)

160-166: Good use of merge commit SHA, but consider adding defensive validation.

The change correctly uses merge_commit_sha instead of HEAD to retrieve the actual merge commit body. With the merged guard at line 17, this SHA should always be valid.

However, adding a defensive check for an empty SHA would make the workflow more robust against edge cases:

🛡️ Optional defensive validation

          $mergeCommitSha = "${{ github.event.pull_request.merge_commit_sha }}"
          Write-Output "DEBUG: Merge commit SHA: $mergeCommitSha"
          
-          $mergeCommitBody = git log -1 --pretty=format:"%b" $mergeCommitSha
+          $mergeCommitBody = ""
+          if (-not [string]::IsNullOrWhiteSpace($mergeCommitSha)) {
+            $mergeCommitBody = git log -1 --pretty=format:"%b" $mergeCommitSha
+          }
          Write-Output "DEBUG: Merge commit body: $mergeCommitBody"