winsvc: Fix bug where service accidentally stops after starting (#4954)

**Which issue(s) this PR fixes**:
This fixes a bug of #4909.

**What this PR does / why we need it**:
After 53bcd3c (#4909), the service
accidentally stops after starting, without stopping the supervisor and
workers.

**Docs Changes**:
Not needed.

**Release Note**:
The same as the title.

Signed-off-by: Daijiro Fukuda <[email protected]>
Co-authored-by: Takuro Ashie <[email protected]>
Co-authored-by: Kentaro Hayashi <[email protected]>
Signed-off-by: Daijiro Fukuda <[email protected]>

Author: 3 people

.github/workflows/test.yml

@@ -27,7 +27,6 @@ jobs:
       matrix:
         os: ['ubuntu-latest', 'macos-latest', 'windows-latest']
         ruby-version: ['3.3', '3.2', '3.1', '3.0', '2.7']
-
     name: Ruby ${{ matrix.ruby-version }} on ${{ matrix.os }}
     steps:
       - uses: actions/checkout@v3
@@ -43,3 +42,23 @@ jobs:
         run: bundle install
       - name: Run tests
         run: bundle exec rake test TESTOPTS="-v --no-show-detail-immediately"
+
+  test-windows-service:
+    runs-on: windows-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        ruby-version: ['3.3', '3.2', '3.1', '3.0', '2.7']
+    name: Windows service (Ruby ${{ matrix.ruby-version }})
+    steps:
+      - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+      - name: Set up Ruby
+        uses: ruby/setup-ruby@e34163cd15f4bb403dcd72d98e295997e6a55798 # v1.238.0
+        with:
+          ruby-version: ${{ matrix.ruby-version }}
+      - name: Install dependencies
+        run: |
+          bundle install
+          rake install
+      - name: Run tests
+        run: test\scripts\windows_service_test.ps1

lib/fluent/winsvc.rb

@@ -63,7 +63,7 @@ def service_main
         loop do
           sleep 5
           break unless running?
-          raise Errno::ECHILD unless Process.waitpid2(@pid, Process::WNOHANG)
+          raise Errno::ECHILD if Process.waitpid(@pid, Process::WNOHANG)
         end
       rescue Errno::ECHILD
         @pid = 0

test/scripts/windows_service_test.ps1

@@ -0,0 +1,73 @@
+$ErrorActionPreference = "Stop"
+Set-PSDebug -Trace 1
+
+$default_conf_path = (Resolve-Path fluent.conf).Path
+$current_path = (Get-Location).Path
+$log_path = "$current_path/fluentd.log"
+
+ruby bin/fluentd --reg-winsvc i --reg-winsvc-fluentdopt "-c '$default_conf_path' -o '$log_path'"
+
+# Test: must not start automatically
+if ((Get-Service fluentdwinsvc).Status -ne "Stopped") {
+    Write-Error "The service should not start automatically."
+}
+
+Start-Service fluentdwinsvc
+Start-Sleep 30
+
+# Test: the service should be running after started
+if ((Get-Service fluentdwinsvc).Status -ne "Running") {
+    Write-Error "The service should be running after started."
+}
+
+# Test: no warn/error/fatal logs
+Get-ChildItem "*.log" | %{
+    Get-Content $_
+    if (Select-String -Path $_ -Pattern "[warn]", "[error]", "[fatal]" -SimpleMatch -Quiet) {
+        Select-String -Path $_ -Pattern "[warn]", "[error]", "[fatal]" -SimpleMatch
+        Write-Error "There are abnormal level logs in ${_}:"
+    }
+}
+
+Stop-Service fluentdwinsvc
+Start-Sleep 10 # Somehow it is possible that some processes stay alive for a while. (This could be not good behavior...)
+
+# Test: status after stopped
+if ((Get-Service fluentdwinsvc).Status -ne "Stopped") {
+    Write-Error "The service should be in 'Stopped' status after stopped."
+}
+# Test: all Ruby processes should stop
+$ruby_processes = Get-Process -name ruby -ErrorAction SilentlyContinue
+if ($ruby_processes.Count -ne 0) {
+    Write-Output $ruby_processes
+    Write-Error "All Ruby processes should stop."
+}
+
+# Test: service should stop when the supervisor fails to launch
+# https://github.com/fluent/fluentd/pull/4909
+$test_setting = @'
+<source>
+  @type sample
+  @id DUPLICATED_ID
+  tag test
+</source>
+<match test>
+  @type stdout
+  @id DUPLICATED_ID
+</match>
+'@
+Add-Content -Path "duplicated_id.conf" -Encoding UTF8 -Value $test_setting
+ruby bin/fluentd --reg-winsvc-fluentdopt "-c '$current_path/duplicated_id.conf' -o '$log_path'"
+Start-Service fluentdwinsvc
+Start-Sleep 30
+if ((Get-Service fluentdwinsvc).Status -ne "Stopped") {
+    Write-Error "The service should be in 'Stopped' status when the supervisor fails to launch."
+}
+$ruby_processes = Get-Process -name ruby -ErrorAction SilentlyContinue
+if ($ruby_processes.Count -ne 0) {
+    Write-Output $ruby_processes
+    Write-Error "All Ruby processes should stop."
+}
+
+ruby bin/fluentd --reg-winsvc u
+Remove-Item $log_path