fix forceUpdate(): if several threads entered the method and there were files to update, all of the threads entered should leave the method synchronously after all the files scheduled for update are reindexed, no matter which thread will do actual reindexing job

platform/lang-impl/src/com/intellij/util/indexing/FileBasedIndex.java

@@ -1421,7 +1421,9 @@ public class FileBasedIndex implements ApplicationComponent {
     public VirtualFile[] queryNeededFiles() {
       r.lock();
       try {
-        if (myFilesToUpdate.isEmpty()) return VirtualFile.EMPTY_ARRAY;
+        if (myFilesToUpdate.isEmpty()) {
+          return VirtualFile.EMPTY_ARRAY;
+        }
         return myFilesToUpdate.toArray(new VirtualFile[myFilesToUpdate.size()]);
       }
       finally {
@@ -1446,7 +1448,17 @@ public class FileBasedIndex implements ApplicationComponent {
         }
         finally {
           myForceUpdateSemaphore.up();
-          myForceUpdateSemaphore.waitFor(); // possibly wait until another thread completes indexing
+        }
+      }
+
+      // If several threads entered the method at the same time and there were files to update,
+      // all the threads should leave the method synchronously after all the files scheduled for update are reindexed,
+      // no matter which thread will do reindexing job.
+      // Thus we ensure that all the threads that entered the method will get the most recent data
+
+      while (!myForceUpdateSemaphore.waitFor(500)) { // may need to wait until another thread is done with indexing
+        if (Thread.holdsLock(PsiLock.LOCK)) {
+          break; // hack. Most probably that other indexing threads is waiting for PsiLock, which we're are holding.
         }
       }
     }