- * approaches than that. However, as we're currently only
- * handling variable-length radiotap headers, and as all
- * filtering expressions other than raw link[M:N] tests
- * require the length of that header, doing more for that
- * header length isn't really worth the effort.
+ * approaches than that.
+ *
+ * However, those strategies would be more complicated, and
+ * as we don't generate code to compute a length if the
+ * program has no tests that use the length, and as most
+ * tests will probably use those lengths, we would just
+ * postpone computing the lengths so that it's not done
+ * for tests that fail early, and it's not clear that's
+ * worth the effort.
+ */
+ insert_compute_vloffsets(p->head);
+
+ /*
+ * For DLT_PPI captures, generate a check of the per-packet
+ * DLT value to make sure it's DLT_IEEE802_11.