17 Aug
2005
17 Aug
'05
12:47 p.m.
David Abrahams wrote:
Related your mentioning of incoming implicit conversions: where do you think these may take place? Do I miss something?
simple_future<int> f1 = bind(slow_fac, 4);
This won't work if the single-argument ctor for simple_future<int> used here is marked explicit, so I conclude that it is not.
The constructor was intended to be non-explicit, just to allow for such implicit conversions from any nullary functor. Is that too dangerous? What's currently missing in the code is a concept check allowing to spot the assignement of other types more easily (right now the compilation fails somewhere deep inside the library if another type - other than a nullary functor - is used to initialise a simple_future). Regards Hartmut