On Feb 18, 2011, at 10:55 AM, Jeff Squyres wrote:
> Let's do pthread_once() -- we don't have any of the infrastructure mojo for memory barriers.
Better yet, let's avoid all the memory barrier issues and pthread_once issues by just hard-coding the order of constants and putting a strongly-worded comment around it not to muck with the ordering.
I'll push that shortly up to bitbucket.
For corporate legal information go to: