Re: [boost] Clang Builds Boost

Subject: Re: [boost] Clang Builds Boost
From: Doug Gregor (doug.gregor_at_[hidden])
Date: 2010-05-21 11:05:28

On Fri, May 21, 2010 at 4:53 AM, Stewart, Robert <Robert.Stewart_at_[hidden]> wrote:
> Richard Webb wrote:
>> Beman Dawes <bdawes <at> acm.org> writes:
>>
>> > Is there a fighting chance of getting clang++ to build Boost's
>> > regression tests on Windows?
>>
>> Whats the status of Windows exception handling in LLVM? (i
>> guess that lack of
>> exception support would cause some problems with the tests.)
>
> That's an interesting question.  Doug mentioned no exception support at BoostCon, yet I see some exception support listed in the table at http://clang.llvm.org/cxx_status.html#specification.  There are no details, so it is unclear what is and isn't supported.

Remember, the C++ status table there [1] only covers the front end,
which has complete support for exceptions (modulo bugs!).

However, it does not capture platform support in the back end. LLVM
supports DWARF exceptions, which work on Linux, Darwin, and many other
*nix systems, including (I believe) Cygwin on Windows [2]. Visual C++
uses a different exception model that isn't implemented in LLVM.

The main pattern here is that the vast majority of LLVM (and Clang)
developers work on Linux, Mac OS X, or FreeBSD, so those platforms are
well-supported. Without more Windows developers involved in pushing
LLVM and Clang forward on their platform, support for Windows will lag
behind.

  - Doug

[1] The status table itself is a bit dated; I'll fix it when time
permits, but the general comment is: Everything in C++98/03 is
implemented except export; file bugs if Clang is wrong.
[2] As Eric mentioned, there are issues with Cygwin, too.