Re: g++/map vs. clang

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Wed, Apr 18, 2012 at 10:45:09PM +0200, Thomas Klausner wrote:
> On Thu, Apr 12, 2012 at 12:35:00AM +0200, Thomas Klausner wrote:
> > On Tue, Apr 10, 2012 at 11:39:26AM -0400, Christos Zoulas wrote:
> > > Sure, but also add a comment saying that clang breaks with it, but the
> > > code might not be at fault.
> > 
> > Well, Dennis Ferguson <dennis.c.ferguson%gmail.com@localhost> posted the 
> > link
> >       http://clang.llvm.org/compatibility.html#deleted-special-func
> > which to me seems to point out just a standards issue.
> > 
> > I'll take a look if I can make copy constructors based on Jörg's link.
> 
> I took a look, but my C++ foo is not good enough to make whatever copy
> constructor clang wants here.
> 
> I claim that compilable code is preferable to probably faster
> non-compilable code, and that it can be made faster by introducing
> well-constructed copy&move constructors later on, so I'd like to go
> ahead with committing the attached diff (haven't changed it since my
> previous mail).

It's not "probably faster". It is a serious regression. I don't mind
walking through the changes, but I don't have time to extract a test
case from boost to verify correctness.

Joerg