I just backported the multiset fix to 7.6 since we are getting it by it on a regular basis. The test suite runs fine even with TEST_MULTISET, but someone more familiar with the code should probably have a look to make sure I didn't mess something up.
Check. And check. Does it fix your problem?
Maybe at some point you'd like to investigate why you're getting destructed objects in your multisets. Could be a bug in itself, perhaps.
We'll know in a while if it fixes anything. There are many thousands of CPU hours between each crash of this kind.
Indeed the origin of this is interesting, but the amount of CPU that has to be burned to hit it makes it comlicated to track down. Top men are working on it.
It will take a couple of days before I can phase in a new Pike on most of the nodes.
pike-devel@lists.lysator.liu.se
-
Martin Nilsson (Opera Mini - AFK!) ＠ Pike (-) developers forum -
Martin Stjernholm, Roxen IS ＠ Pike developers forum
-
Peter Bortas ＠ Pike developers forum