Author: toad
Date: 2009-03-18 18:21:02 +0000 (Wed, 18 Mar 2009)
New Revision: 26096
Modified:
branches/db4o/freenet/src/freenet/client/async/Db4oBugs.java
Log:
Update bug documentation now that I've found the real bug. Yes really!
Modified: branches/db4o/freenet/src/freenet/client/async/Db4oBugs.java
===================================================================
--- branches/db4o/freenet/src/freenet/client/async/Db4oBugs.java
2009-03-18 15:31:09 UTC (rev 26095)
+++ branches/db4o/freenet/src/freenet/client/async/Db4oBugs.java
2009-03-18 18:21:02 UTC (rev 26096)
@@ -17,20 +17,25 @@
query.constrain(HasKeyListener.class);
return query.execute();
}
+
+ /* http://tracker.db4o.com/browse/COR-1436
+ * ArrayList's etc must be stored with:
+ *
+ * container.ext().store(list, 2)
+ *
+ * Otherwise everything contained in the arraylist is updated to depth
3, which
+ * is not usually what we want, and can be catastrophic due to storing
deactivated
+ * objects and/or using up lots of memory. */
- /* This one needs to be worked around in the code:
- * Storing an object containing a HashMap without storing the HashMap
first results
- * in the HashMap being stored empty. After restart, loading it gives
an empty
- * HashMap. This causes all manner of problems! Look at SVN r26092:
manifestElements
- * (a HashMap, in this case a derivative of HashMap for debugging
purposes, which
- * is part of ClientPutDir, and can contain other HashMap's) becomes
empty on shutdown,
- * even though it was full in objectCanNew(), objectOnNew(), and
- * objectCanUpdate/objectOnUpdate (which don't get called since
changing the store
- * depth in FCPClient). To reproduce simply start the node, start an
insert for a
- * small directory, shutdown the node after a few seconds, start it
back up. Hook
- * into e.g. ClientPutDir.receive() (or wait for freeData(), which is
where it
- * matters). Even when it is activated, it is empty. */
-
+ /* http://tracker.db4o.com/browse/COR-1582
+ * Never activate a HashMap to depth 1. This will not only result in
its being
+ * empty, but activating it to depth 2 and thus loading its elements
will not be
+ * possible unless you deactivate it first. Combined with the previous
bug this
+ * can cause *really* annoying bugs: maps apparently spontaneously
clearing
+ * themselves, actual cause was it was accidentally activated to depth
1, and then
+ * was accidentally stored by e.g. being moved from one ArrayList to
another with
+ * the previous bug. */
+
/* We are using an oldish version of db4o 7.4 in ext #26 because the
newer versions,
* e.g. in ext-27pre2, have *really* horrible bugs - all sorts of wierd
things
* happen with them, the FEC object duplication bug happens with File's
as well,
_______________________________________________
cvs mailing list
[email protected]
http://emu.freenetproject.org/cgi-bin/mailman/listinfo/cvs
