From cf36c1614f7f05f146a314fba0727d0159866459 Mon Sep 17 00:00:00 2001
From: orbiter <mc@yacy.net>
Date: Sun, 12 May 2013 21:37:45 +0200
Subject: [PATCH] prevent that concurrent deletion process causes wrong
 double-check in crawl start

---
 source/net/yacy/search/Switchboard.java | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/source/net/yacy/search/Switchboard.java b/source/net/yacy/search/Switchboard.java
index 9b5c72607..f3563e896 100644
--- a/source/net/yacy/search/Switchboard.java
+++ b/source/net/yacy/search/Switchboard.java
@@ -2844,6 +2844,16 @@ public final class Switchboard extends serverSwitch {
         // remove url from the index to be prepared for a re-crawl
         final byte[] urlhash = url.hash();
         remove(urlhash);
+        // because the removal is done concurrenlty, it is possible that the crawl
+        // stacking may fail because of double occurrences of that url. Therefore
+        // we must wait here until the url has actually disappeared
+        int t = 100;
+        while (t-- > 0 && this.index.exists(ASCII.String(urlhash))) {
+            try {Thread.sleep(100);} catch (InterruptedException e) {}
+            Log.logFine("Switchboard", "STACKURL: waiting for deletion, t=" + t);
+            if (t == 20) this.index.fulltext().commit(true);
+            if (t == 50) this.index.fulltext().commit(false);
+        }
         
         // special handling of ftp protocol
         if (url.isFTP()) {