On Fri, 2006-12-15 at 08:47 -0500, David Abrahams wrote:
"Aaron Windsor" <aaron.windsor@gmail.com> writes:
There was indeed a bug in the fibonacci heap code - the parent pointers weren't being updated in the make_child function. I just committed a fix for this to HEAD. I'd be grateful to hear from someone who has some heavy-duty code using the relaxed heap as to whether or not the fibonacci heap code works as a drop-in replacement now.
IMO critical bug fixes like this one ought to be mirrored to the RC_1_34_0 branch, unless they're very likely to break regression tests. But of course Thomas W. gets final say.
The Fibonacci heap is completely unused and has been broken for eons. It can't hurt the release at all to put this fix in. Cheers, Doug