[boost] Re: RC_1_30_0 still broken -- More lexical_cast

From: Kevlin Henney (kevlin_at_[hidden])
Date: 2003-03-18 01:11:04

In article <4.3.2.7.2.20030317162109.02627fa0_at_[hidden]>, Beman
Dawes <bdawes_at_[hidden]> writes
>At 03:40 PM 3/17/2003, Terje Slettebø wrote:
>>
>>BOOST_NO_STRINGSTREAM
>>BOOST_NO_STD_WSTRING
>>BOOST_NO_INTRINSIC_WCHAR_T
>
>Are you sure disabling wide character support is really the solution, or 
>that it is really fully disabled?
>
>For the 2.95.x test at 
>http://boost.sourceforge.net/regression-logs/cs-linux-RC_1_30_0-links.html#confi
>g_info%20gcc2953, 
>BOOST_NO_STD_WSTRING is already defined, so presumably wide character 
>support is already disabled.
>
>Look at the error messages from date_time testperiod below, and the source 
>code lines they refer to. At least directly, they don't seem releated to 
>wide character support.

They are not, but the question is what is meant by
BOOST_NO_STRINGSTREAM? Does it mean that std::stringstream is not
supported and/or that std::basic_stringstream is not supported? The
lexical_cast code assumes that <sstream> is not standard if
BOOST_NO_STRINGSTREAM is defined, and that it is standard if it is not
defined. Clearly, the contents of <sstream> are not standard otherwise
the code would compile.

Either we need to clarify the intent of BOOST_NO_STRINGSTREAM or we need
to added another feature test macro, eg BOOST_NO_BASIC_STRINGSTREAM or
no BOOST_NO_STD_SSTREAM. My preference is that we generalise the meaning
of BOOST_NO_STRINGSTREAM, so that if that macro is defined the
programmer cannot assume standard string stream support, which --
judging by the error messages -- which is the case.

Kevlin
____________________________________________________________

  Kevlin Henney                   phone:  [phone]
  mailto:kevlin_at_[hidden]     mobile: [phone]
  http://www.curbralan.com        fax:    [phone]
  Curbralan: Consultancy + Training + Development + Review
____________________________________________________________