On 05/31/2010 03:29 PM, "C. Bergström" wrote:
Martin Sebor wrote:
On 05/30/2010 11:57 AM, "C. Bergström" wrote:
Hi
Is anyone interested or actively working on C++0x support? It's my
understanding it would be a huge effort, but that's coming from the
people working on libc++.
There is interest within the user community but I don't believe
anyone is actively working on C++ 0x at the moment. I have a few
patches in my local tree but nothing I could check in anytime
soon. As for the effort, back in 2008 we estimated full C++ 0x
support at 2,740 engineer hours. The big ones were:
Regular expressions (STDCXX-33): 1,080 hours
Containers (STDCXX-32): 400 hours
C compatibility (STDCXX-34): 340 hours
Numerical Facilities (STDCXX-31): 320 hours
General Utilities (STDCXX-28): 160 hours
Function Objects (STDCXX-29): 160 hours
Rvalue References: 160 hours
Also I had a build failure on FBSD-8.0 the other day, but didn't make a
note of it. I'll probably file a proper bug report later..
I'm interested to here if RogueWave's continued interest in the
project.. We potentially will be investing in it unless it's dead and
just purely maintenance mode..
Rogue Wave stopped putting resources into the project in the
summer of 2008. Which doesn't mean the project is dead -- just
on sabbatical :)
If someone is willing to put in development effort into stdcxx
I'm willing to help.
Thanks for the reply Martin...
Does that mean Rogue Wave stopped updating their internal copy as well
or just not pushing changes?
According to their documentation they stopped both shipping
and supporting stdcxx with the release of SourcePro 11 (in
September 2009). This is what they have to say about in the
Release Notes:
2.1 Apache Standard Library No Longer Distributed
With 11.0, SourcePro no longer distributes or supports
the Apache C++ Standard Library. Users should build against
the native standard library shipped with a compiler certified
by Rogue Wave.
Can anyone comment on why we'd want to
spend 2,740 man hours updating STDCXX vs contributing to libc++[1] ?
The biggest but possibly the only reason that occurs to me
is portability. The target platform of libc++ is gcc/clang
on Apple OS X. stdcxx on the other hand has been ported to
dozens of compilers and operating systems and versions,
and is easily portable to new ones (check out the nightly
test matrix: http://stdcxx.apache.org/builds/4.2.x/).
If you don't care about portability to this extent or if
your target platform is close to that of libc++ then going
with it is likely to be more economical than starting from
scratch with stdcxx.
That being said, I should mention that the time estimate
was based on the large number of platforms stdcxx targeted
and supported, and the goal of providing at least partial
C++ 0x support for compilers with an incomplete set of C++
0x language features.
Assuming our tests continue to go well I think we'll probably submit bug
fixes and any low hanging performance improvements.
That would be great.
Martin
Thanks!
./C
[1] http://libcxx.llvm.org/
ps. Martin please email me if you or anyone else is interested being
sponsored to work on the project.
