debug unordered_multi erase patch

[Paolo Carlini]

On 11/27/2010 02:15 PM, François Dumont wrote:
>     I know that as much as possible we should use pre-increment but in
> this case I really need post-increment. Do you mean that I should have
> introduce an intermediate copy ?
In general, you should yes, because being explicit about the copy makes
the life easier to the optimizers. Note that in the *entire* <algorithm>
postincrement is never used. Anyway, in this specific case I guess it's
fine to use it, because we are talking about debug-mode and in any case
we know the types we are working with.

Paolo.