On 3 July 2011 02:17, Mathias Gaunard <mathias.gaunard@ens-lyon.org> wrote:
On 07/02/2011 06:08 PM, Daniel James wrote:
https://svn.boost.org/trac/boost/changeset/72799/
There's a constexpr method in that file which isn't just a return statement, which was causing random (and libraries which depend on it) to fail on gcc 4.6.0 in C++0x mode. So the change just disables use of constexpr. A proper fix is pretty simple so I'll do that for the next release if necessary.
Isn't it better to simply remove the temporary variable to make it a valid constexpr?
At this stage in the release process, I'd prefer to use the simplest, least disruptive change. Disabling the user of constexpr only affects gcc in C++0x mode (the preprocessed code is identical elsewhere), and causes it to use an implementation that's known to work. It's also a much easier change to review.