Add error classification to Jekyll build retry logic

Implement smart error classification to only retry transient network errors,
preventing unnecessary retries on permanent build failures that waste 4-5 minutes.

Changes:
- Capture Jekyll build output to log files for error analysis
- Classify errors into retryable (network/DNS/SSL) vs permanent (Liquid syntax, missing files)
- Return different exit codes: 0=success, 1=permanent error, 2=transient error
- Update retry logic to immediately fail on permanent errors
- Only retry on actual network connectivity issues

Benefits:
- Eliminates 4-5 minute delays from retrying Liquid syntax errors
- Eliminates delays from retrying missing file reference errors
- Provides clear feedback explaining retry decisions
- Conservative approach: unclassified errors default to non-retryable

Addresses PR review feedback from #20409 to improve build experience
for writers encountering permanent errors.

Author: ebembi-crdb

src/current/netlify/build.sh

@@ -64,22 +64,46 @@ function log_attempt() {
 
 function build_with_monitoring {
     local config=$1
+    local build_log="build_${ATTEMPT_COUNT}.log"
+    
     echo "📝 Starting Jekyll build with config: $config"
     echo "⏰ Build start: $(date)"
+    echo "📄 Build log: $build_log"
 
-    # Run Jekyll build with network monitoring
-    bundle exec jekyll build --trace --config _config_base.yml,$config
-    local build_result=$?
-    
-    echo "⏰ Build end: $(date)"
-    echo "📊 Build result: $build_result"
-    
-    if [[ $build_result != 0 ]]; then
-        echo "❌ Jekyll build failed with exit code: $build_result"
-        return $build_result
-    else
+    # Capture Jekyll output for error analysis
+    if bundle exec jekyll build --trace --config _config_base.yml,$config 2>&1 | tee "$build_log"; then
+        echo "⏰ Build end: $(date)"
         echo "✅ Jekyll build completed successfully"
         return 0
+    else
+        local exit_code=$?
+        echo "⏰ Build end: $(date)"
+        echo "❌ Jekyll build failed with exit code: $exit_code"
+        
+        # Analyze build log for error classification
+        echo "🔍 Analyzing build errors for retry eligibility..."
+        
+        # Check for transient network errors that should be retried
+        if grep -qiE "(temporary failure in name resolution|connection refused|connection reset|SSL_connect|certificate verify failed|execution expired|timeout|network is unreachable|failed to open tcp connection|socketerror)" "$build_log"; then
+            echo "🌐 Transient network error detected - eligible for retry"
+            echo "📋 Network error details:"
+            grep -iE "(temporary failure in name resolution|connection refused|connection reset|SSL_connect|certificate verify failed|execution expired|timeout|network is unreachable|failed to open tcp connection|socketerror)" "$build_log" | head -3
+            return 2  # Retryable error
+        fi
+        
+        # Check for permanent errors that should NOT be retried
+        if grep -qiE "(liquid.*syntax error|liquid error|argumenterror|no such file or directory|undefined method|unknown tag|was not properly terminated|missing file)" "$build_log"; then
+            echo "🚫 Permanent build error detected - not retrying"
+            echo "📋 Error details:"
+            grep -iE "(liquid.*syntax error|liquid error|argumenterror|no such file or directory|undefined method|unknown tag|was not properly terminated|missing file)" "$build_log" | head -3
+            return 1  # Non-retryable error
+        fi
+        
+        # If we can't classify the error, treat it as non-retryable to be safe
+        echo "❓ Unclassified build error - treating as permanent (not retrying)"
+        echo "📋 Last few lines of build log:"
+        tail -5 "$build_log"
+        return 1  # Non-retryable error by default
     fi
 }
 
@@ -91,18 +115,30 @@ function build_with_retries {
         log_attempt $attempt $MAX_RETRIES
         ATTEMPT_COUNT=$attempt
 
-        if build_with_monitoring "$config"; then
+        build_with_monitoring "$config"
+        local result=$?
+        
+        if [[ $result == 0 ]]; then
             echo "✅ Build succeeded on attempt ${attempt}/${MAX_RETRIES}"
             success=true
             break
-        else
-            echo "❌ Build failed on attempt ${attempt}/${MAX_RETRIES}"
+        elif [[ $result == 1 ]]; then
+            echo "❌ Build failed on attempt ${attempt}/${MAX_RETRIES} with permanent error"
+            echo "🚫 Permanent error detected - failing immediately (no retry)"
+            break  # Don't retry permanent errors
+        elif [[ $result == 2 ]]; then
+            echo "❌ Build failed on attempt ${attempt}/${MAX_RETRIES} with transient error"
             if [[ $attempt -lt $MAX_RETRIES ]]; then
                 local next_delay=$((BASE_RETRY_DELAY * (1 << (attempt - 1))))
-                echo "🔄 Will retry in ${next_delay} seconds (exponential backoff)..."
+                echo "🔄 Transient error - will retry in ${next_delay} seconds (exponential backoff)..."
             else
-                echo "💀 All retry attempts exhausted"
+                echo "💀 All retry attempts exhausted for transient error"
             fi
+        else
+            # Fallback for unexpected return codes
+            echo "❌ Build failed on attempt ${attempt}/${MAX_RETRIES} with unexpected error code: $result"
+            echo "⚠️ Treating as permanent error - not retrying"
+            break
         fi
     done