added bug report note to Cache.java; investigate later

darcs-hash:20040213061807-5007d-a99cfb527736dfcccf78f20ea944c2ea800cbd11.gz

diff --git a/src/org/ibex/util/Cache.java b/src/org/ibex/util/Cache.java
index 47f4faf..af88c88 100644 (file)
--- a/src/org/ibex/util/Cache.java
+++ b/src/org/ibex/util/Cache.java
@@ -5,6 +5,35 @@
 // the portion of clause 6a after the semicolon (aka the "obnoxious
 // relink clause")
 
+/*
+
+Bug report from a user:
+
+I looked at your Cache.java and tried to make good use of it, but I was
+out of luck - it wouldn't run here. Digging deeper into the code, I came
+across something that might be considered a bug. But maybe it's just a
+feature :-)
+
+
+Starting with an empty cache, Cache.put() immediately followed by
+Cache.get() on same keys / same object will set Node lru back to null in
+Node.remove() which is called in get().
+
+Assuming this put()-get() sequence is fixed, it will fill the cache, but
+lru will remain null.
+
+When cache is filled 100%, we have, at the end of the get(), where
+size>maxSize is checked, a state that mru == lru == n (from
+if(lru==null) thus deleteting the newly inserted object. Oops.
+
+
+Hope I made this clear enough. Maybe it's not a problem in xwt due to a
+different usage scheme of the cache.
+
+
+
+*/
+
 package org.ibex.util;
 
 // FIXME needs to be a weak hash