fix: ennhance windows binary testing with error handling and debugging and change linux runner

Author: Minh141120

.github/workflows/test-binaries.yml

@@ -27,22 +27,22 @@ jobs:
         include:
           - os: "linux"
             name: "noavx-x64"
-            runs-on: "ubuntu-20.04"
+            runs-on: "ubuntu-20-04"
             binary-name: "llama-server"
             artifact-name: "llama-linux-noavx-x64"
           - os: "linux"
             name: "avx-x64"
-            runs-on: "ubuntu-20.04"
+            runs-on: "ubuntu-20-04"
             binary-name: "llama-server"
             artifact-name: "llama-linux-avx-x64"
           - os: "linux"
             name: "avx512-x64"
-            runs-on: "ubuntu-20.04"
+            runs-on: "ubuntu-20-04"
             binary-name: "llama-server"
             artifact-name: "llama-linux-avx512-x64"
           - os: "linux"
             name: "vulkan-x64"
-            runs-on: "ubuntu-22.04"
+            runs-on: "ubuntu-22-04"
             binary-name: "llama-server"
             artifact-name: "llama-linux-vulkan-x64"
 
@@ -315,17 +315,53 @@ jobs:
         shell: pwsh
         run: |
           Write-Host "Testing binary basic functionality..."
+          
+          # Test version command
           try {
-            .\llama\build\bin\${{ matrix.binary-name }} --version
+            $versionOutput = & ".\llama\build\bin\${{ matrix.binary-name }}" --version 2>&1
+            Write-Host "Version output:"
+            Write-Host $versionOutput
           } catch {
-            Write-Host "Version check completed with exit code: $LASTEXITCODE"
+            Write-Host "Version check failed: $($_.Exception.Message)"
           }
+          
           Write-Host "Available arguments:"
           try {
-            .\llama\build\bin\${{ matrix.binary-name }} --help
+            # Use Start-Process to capture output properly and avoid exit code issues
+            $helpProcess = Start-Process -FilePath ".\llama\build\bin\${{ matrix.binary-name }}" `
+              -ArgumentList "--help" `
+              -RedirectStandardOutput "help_output.txt" `
+              -RedirectStandardError "help_error.txt" `
+              -Wait -PassThru -WindowStyle Hidden
+            
+            if (Test-Path "help_output.txt") {
+              $helpContent = Get-Content "help_output.txt" -Raw
+              if ($helpContent) {
+                Write-Host $helpContent
+              }
+            }
+            
+            if (Test-Path "help_error.txt") {
+              $errorContent = Get-Content "help_error.txt" -Raw
+              if ($errorContent) {
+                Write-Host "Help stderr:"
+                Write-Host $errorContent
+              }
+            }
+            
+            Write-Host "Help command exit code: $($helpProcess.ExitCode)"
+            
           } catch {
-            Write-Host "Help check completed with exit code: $LASTEXITCODE"
+            Write-Host "Help check failed: $($_.Exception.Message)"
+            Write-Host "This might be normal for some binary versions"
           }
+          
+          # Clean up temp files
+          Remove-Item -Path "help_output.txt" -ErrorAction SilentlyContinue
+          Remove-Item -Path "help_error.txt" -ErrorAction SilentlyContinue
+          
+          # Don't fail the step - verification is informational
+          Write-Host "Binary verification completed"
 
       - name: Test server startup (Linux/macOS)
         if: runner.os != 'Windows'
@@ -593,29 +629,93 @@ jobs:
         run: |
           Write-Host "Testing ${{ matrix.binary-name }} server startup..."
           
-          # Start server with CPU mode
+          # Start server with CPU mode and capture output
+          $logFile = "server_output.log"
           $process = Start-Process -FilePath ".\llama\build\bin\${{ matrix.binary-name }}" `
             -ArgumentList "--model", "models\Lucy-Q4_0.gguf", "--port", "8080", "--host", "127.0.0.1", "--n-gpu-layers", "0", "--ctx-size", "512" `
-            -WindowStyle Hidden -PassThru
+            -WindowStyle Hidden -PassThru -RedirectStandardOutput $logFile -RedirectStandardError "server_error.log"
           
           Write-Host "Server PID: $($process.Id)"
           
-          # Wait for server to start
-          for ($i = 1; $i -le 30; $i++) {
+          # Give server more time to start and check if process is alive
+          Start-Sleep -Seconds 10
+          
+          if ($process.HasExited) {
+            Write-Host "Server process exited immediately"
+            Write-Host "Exit code: $($process.ExitCode)"
+            Write-Host "Server output:"
+            if (Test-Path $logFile) { Get-Content $logFile }
+            Write-Host "Server errors:"
+            if (Test-Path "server_error.log") { Get-Content "server_error.log" }
+            exit 1
+          }
+          
+          # Wait for server to start responding with better error handling
+          $serverResponded = $false
+          for ($i = 1; $i -le 20; $i++) {
             try {
-              $response = Invoke-RestMethod -Uri "http://127.0.0.1:8080/health" -Method Get -TimeoutSec 2
-              Write-Host "[PASSED] Server started successfully and is responding"
-              Stop-Process -Id $process.Id -Force -ErrorAction SilentlyContinue
-              exit 0
+              # Try multiple endpoints
+              $healthResponse = $null
+              try {
+                $healthResponse = Invoke-RestMethod -Uri "http://127.0.0.1:8080/health" -Method Get -TimeoutSec 3
+                $serverResponded = $true
+                Write-Host "[PASSED] Server started successfully and is responding on /health"
+                break
+              } catch {
+                # Try root endpoint
+                try {
+                  $rootResponse = Invoke-RestMethod -Uri "http://127.0.0.1:8080/" -Method Get -TimeoutSec 3
+                  $serverResponded = $true
+                  Write-Host "[PASSED] Server started successfully and is responding on /"
+                  break
+                } catch {
+                  # Try basic connection test
+                  try {
+                    $tcpClient = New-Object System.Net.Sockets.TcpClient
+                    $tcpClient.Connect("127.0.0.1", 8080)
+                    $tcpClient.Close()
+                    Write-Host "[PASSED] Server started successfully (TCP connection established)"
+                    $serverResponded = $true
+                    break
+                  } catch {
+                    Write-Host "Attempt $i/20 - waiting for server... (HTTP and TCP failed)"
+                  }
+                }
+              }
             } catch {
-              Write-Host "Attempt $i/30 - waiting for server..."
-              Start-Sleep -Seconds 2
+              Write-Host "Attempt $i/20 - connection error: $($_.Exception.Message)"
             }
+            Start-Sleep -Seconds 3
+          }
+          
+          if (-not $serverResponded) {
+            Write-Host "[FAILED] Server failed to respond within timeout"
+            Write-Host "Server process status: Running = $(-not $process.HasExited)"
+            Write-Host "Server output (last 20 lines):"
+            if (Test-Path $logFile) { Get-Content $logFile | Select-Object -Last 20 }
+            Write-Host "Server errors:"
+            if (Test-Path "server_error.log") { Get-Content "server_error.log" }
+            
+            # Try to get more info about what the server is doing
+            Write-Host "Checking if server is listening on port 8080..."
+            try {
+              $netstat = netstat -an | Select-String ":8080"
+              if ($netstat) {
+                Write-Host "Port 8080 status:"
+                Write-Host $netstat
+              } else {
+                Write-Host "Port 8080 is not being listened on"
+              }
+            } catch {
+              Write-Host "Could not check port status"
+            }
+            
+            Stop-Process -Id $process.Id -Force -ErrorAction SilentlyContinue
+            exit 1
           }
           
-          Write-Host "[FAILED] Server failed to start or respond within timeout"
           Stop-Process -Id $process.Id -Force -ErrorAction SilentlyContinue
-          exit 1
+          Write-Host "Server test completed successfully"
 
       - name: Test inference (Windows)
         if: runner.os == 'Windows'