16 Oct
2005
16 Oct
'05
10:44 a.m.
Keith Burton wrote:
Given the nice simple definition of the container ( "hashed associative container that can be used without synchronization from multiple threads" ) , I believe find_and_visit should synchronise despite the breaking the "C don't pay for what is not used" mantra. Unless, of course, two functions are provided.
Two functions won't do, since the map provides several other ways to read the element that may be being modified, and all of these would need synchronization as well. The problem that I don't know how to solve yet is that a thread may be holding an iterator to the element, and there is no way to inject synchronization there. (I've never liked proxy references much :-) )