AP_Scripting: Allow generator to cope with windows line endings

Also fixes allowing an extra new line on the depends entries to be
emitted, as well as a possible stack overflow in the generator if
you inserted a malicious comment.

libraries/AP_Scripting/generator/src/main.c

@@ -202,7 +202,7 @@ char * next_token(void) {
   trace(TRACE_TOKENS, "Token %d:%d %s", state.line_num, state.token_num, state.token);
   if ((state.token!= NULL) && (strcmp(state.token, keyword_comment) == 0)) {
     trace(TRACE_TOKENS, "Detected comment %d", state.line_num);
-    while (next_token()) {} // burn all the symbols
+    state.token = NULL; // burn the line
   }
   return state.token;
 }
@@ -211,12 +211,27 @@ char * start_line(void) {
   while (fgets(state.line, sizeof(state.line)/sizeof(state.line[0]), description) != NULL) {//state.line = readline(NULL))) {
       state.line_num++;
     
+      const size_t length = strlen(state.line);
+      if (length > 1 && state.line[length - 2] == '\r') {
+        trace(TRACE_TOKENS, "Discarding carriage return");
+        if (length == 2) { // empty line of just carriage return, loop again
+          continue;
+        }
+        state.line[length - 2] = '\0';
+      } else if (length > 0 && state.line[length - 1] == '\n') {
+        trace(TRACE_TOKENS, "Discarding newline");
+        if (length == 1) { // empty line of just carriage return, loop again
+          continue;
+        }
+        state.line[length - 1] = '\0';
+      }
+
       state.token = strtok(state.line, token_delimiters);
       state.token_num = 1;
-      trace(TRACE_TOKENS, "Token %d:%d %s", state.line_num, state.token_num, state.token);
+      trace(TRACE_TOKENS, "Start of line token %d:%d %s", state.line_num, state.token_num, state.token);
 
       if (state.token != NULL) {
-          break;
+        break;
       }
   }