bitset ctor multiply defined

[Benjamin Kosnik]

> This was my hack/patch to the v3 implementation (not the SGI one).  It
> doesn't deserve to be called a patch, since all I did was comment out the
> ctor that's always present.  This is also the one that had endpoints for
> the string, with default arguments for both.  After commenting, doing
> "bitset<n>  b (a_string);" now calls the first member-template ctor inside
> the #ifdef's, the one which takes only a default starting point for the 
> string.

Hmm. This is fine in the SGI 3.20 sources, so the culprit is my
integration job. It's fixed now.

Sorry,
Benjamin