Closed morinmorin closed 6 years ago
Testing on Travis CI fails, because
Oh, cyclic dependencies of PR's...
The change to Boost.Iterator is reapplied now.
The change to Boost.Iterator is reapplied now.
Thanks! I just reran the Travis testing of this PR (by force-pushing).
I had already re-run it. It passed except for the last job, which fails because libc++ wants random engines to have constexpr min/max, and the one in random_shuffle.cpp appears not to.
OK, I'm going to merge this and see how it goes.
Would be nice if we could fix the (unrelated) Travis failure though.
Thanks for merging.
Would be nice if we could fix the (unrelated) Travis failure though.
Definitely! (though I don't have any insight into how to fix that...)
Move
distance
function template into the ADL barrier namespace and pull the name by a using-directive. This fix should be OK, since Boost.Range documentation explicitly forbids unqualified call todistance
.std::distance(rng)
function template. This fix prevents ambiguity errors betweenboost::distance(rng)
andstd::distance(rng)
caused by accidental ADL.boost::distance(it1, it2)
, but Boost.Iterator documentation advertises onlyboost::iterators::distance(it1, it2)
. So let's use it.