Hi,
I just tracked down a subtle and annoying netsync bug that has been
pestering us for a while. It involves a not-very-hard-to-trigger case
wherein the nesting of queries and subqueries gets broken. The symptom
is just netsync going dead awaiting i/o, with no obvious logic errors or
data corruption.
If netsync has been hanging for you recently (post-0.26), I'd be curious
to hear feedback on whether this change fixes your problems. The head
containing a merge of my change is
502422b06d27c0bd826d4ace41ad433cec2c1256
I also added a randomizing unit tester for the netsync refiner (which is
what isolated the bug, actually) so this sort of thing should be
slightly more readily caught in the future.
-graydon
_______________________________________________
Monotone-devel mailing list
Monotone-devel@nongnu.org
http://lists.nongnu.org/mailman/listinfo/monotone-devel
