jzmaddock

Seems like a perfectly reasonable request, I guess we would just have to macro-ise the namespace open and closing?

I would tend to weakly prefer the latter.

I had assumed (perhaps incorrectly) that we would maybe use something like: ``` #define BOOST_MATH_NAMESPACE_START namespace boost{ namespace math{ #define BOOST_MATH_NAMESPACE_END }} ```

> But (thinking out loud) would a prepended namespace give the client the desired isolation? I assume that anything that changes the name mangling would work equally well. > Do...

Personally, I would make BOOST_MATH_NAMESPACE equivalent to `boost::math`. I have no great objection to the changes required, but it does increase the number of configurations we need to test quite...

Works for me here, but may possibly be fixed by https://github.com/boostorg/math/commit/fa8f763df9e6635cf255d60766af65950eec58e8. Can you: 1) check to see if the above commit fixes your issue, otherwise, 2) Can you please provide...

Well.... I've just spent 2 rather fruitless days getting Android Studio up and running (how does anyone actually get any work done with this thing?) only to resolutely fail to...

It might have been more useful to see the output from the first iteration, but whatever those derivative values (f1 and f2) are completely wrong :( Quick sanity check -...

>What I wrote was the entire output - should there have been more? Yes that was only the final iteration of the root finder's main loop. >With -DBOOST_MATH_PROMOTE_DOUBLE_POLICY=false the call...

Unfortunately, I don't think this is solvable in all cases. In this situation, the initial NR step goes out of bounds to the left, we therefore assume (erroneously in this...