It seems that the problem is calling sizeof on a dependent type when the
containers have only been used by reference. So by putting in these
dummy structures with the containers as members, it helps the compiler
instantiate the class to the level where sizeof works. I hope.
[SVN r58130]
I'm trying to fix the codegear ICEs, but it's hard to tell the cause.
Since the error happens operator== I suspect it's either to do
with defining friend functions with different template names or
something to do with friend functions in general. This is the first stab
in the dark at fixing this.
[SVN r58062]
There's a bug in the emplace_hint implementation for unordered
containers with equivalent keys. Since my tests missed it, I'm just
going to use the normal emplace implementation until I write better
tests.
[SVN r57005]
- More template use, less preprocessor use.
- Removed some of the Visual C++ 6 workarounds.
- Reduced memory use of the main object.
- Split into smaller headers.
[SVN r55878]
anything which depends on the new move library.
------------------------------------------------------------------------
r52746 | danieljames | 2009-05-03 11:12:30 +0100 (Sun, 03 May 2009) | 1 line
Merge latest unordered container changes.
------------------------------------------------------------------------
r52747 | danieljames | 2009-05-03 11:15:35 +0100 (Sun, 03 May 2009) | 4 lines
Put the C++0x emplace implementations before the non-C++0x versions.
I'm going to change the non-C++0x to be macro heavy emulations of the
C++0x versions, so this will put the readable version first.
------------------------------------------------------------------------
r52748 | danieljames | 2009-05-03 11:15:44 +0100 (Sun, 03 May 2009) | 1 line
Refactor the unordered implementation a tad, to make implementing emplace less painful.
------------------------------------------------------------------------
[SVN r52884]
This feels like overkill but it seems to be the most reliable way to ensure
that the length is correct. I obviously can't be trusted to get it right, and
the template hack seems to prevent Boost.Range from working.
[SVN r52711]
