Re: [boost] Interest in adding support for the Bloomberg Standard Library

Subject: Re: [boost] Interest in adding support for the Bloomberg Standard Library
From: Mathias Gaunard (mathias.gaunard_at_[hidden])
Date: 2015-08-28 06:12:52

Andrey Semashev <andrey.semashev <at> gmail.com> writes:

> Did you have a chance to test it with Boost locally? Or, if you have
> that knowledge, how compliant that implementation is? How many
> libraries other than Boost.Config and Boost.Container are going to be
> affected?

It is fairly standards-compliant, unless you consider extensions (as-is 
additional overloads to a standard function) to be non-conforming.
The problem is that it sits on top of another standard library, and 
inherits some of its bugs, but not all of them. Therefore anything 
detecting a particular standard library and working around bugs needs to 
be more careful.

So long as the boost library author is using boost config feature macros, 
there should be no problem, but if he's directly trying to detect which 
standard library is being used, for example directly assuming that this is 
libc++ because _LIBCPP_VERSION is defined, it might fail.

There is a small issue related to conformance: it requires strict aliasing 
to be disabled (-fno-strict-aliasing). I know that's a bit controversial, 
but it shouldn't affect anything at the source level.