On Mon, Oct 21, 2013 at 9:00 PM, Brad Roberts <[email protected]> wrote:
> While I haven't re-examined this issue in years, the collector hasn't > changed and this property continues to hold... it's a natural artifact of > the stack scan process in the gc. Not every byte of the stack is > initialized from valgrind's point of view (or in reality), yet the stack > scan can't know that and takes every slot into account regardless. IE, > it's a standard part of conservative gc engines. The right 'solution' is > to suppress that stacktrace in valgrind. > Makes sense. Thanks for the explanation, Brad.
