Community
Participate
Working Groups
+++ This bug was initially created as a clone of Bug #235008 +++ Cloning this bug for the 3.4.2 stream
While doing a large reorg in the UI code for 3.5, I realized a simple workaround for this that can be backported to 3.4.2. Will need to spend some time creating the exact steps to reproduce. This typically requires a clean installation to see it.
I tried to reproduce on a clean 3.4, clean 3.4.1, and clean M20080910-1030. Cleared out repo caches and tried different combinations of sites and expansions/refreshes, in particular using the Ganymede repo since it has so many references. Was never able to get the duplicate categories/sites/IUs. As discussed in other bugs, the timing window in which this problem could occur was made very small for 3.4, but we knew it could still happen. This fix is surgical and addresses the specific timing hole. It uses a local flag in a model element to prevent that element from adding its contents to the deferred element collector more than once. This same patch has been in the I-build stream for nearly 3 weeks, and although not seeing the problem doesn't prove it's fixed, we haven't seen any new problems introduced by this local flag. Committed to R3_4_maintenance >20081029. I'll pick up the next M-build and use that for all my maintenance stream work.
Clarification for IP purposes: the changes occurred in RemoteQueriedElement, you can compare revision 1.7 with revision 1.7.2.1
This "seemingly harmless" surgical fix has indeed caused a regression in 3.4.2 testing. Bug 262044. I think we will need to pull this out, given the regression is worse than the problem it was trying to fix.
Marking WONTFIX as we have backed out this change in bug 262044.
*** Bug 284961 has been marked as a duplicate of this bug. ***
