Re: [boost] compiling with GCC 3.0.x (lexical_cast, regex_traits, threads)

From: larsbj_at_[hidden]
Date: 2001-11-15 19:15:40

Jens Maurer <Jens.Maurer_at_[hidden]> writes:

| larsbj_at_[hidden] wrote:
>> It seems that the checks for sstream of lexical_cast became a little
>> bit to strong. (GCC 3 has a good sstream.) This:
>
| Note the comment right above the preprocessor checks:
>
| // The GNU sstream implementation is broken for the purposes of lexical cast.

And I question if this is genrerally true.
(actually I think it is not)

| We're already checking with BOOST_NO_STRINGSTREAM whether <sstream>
| is available or not.
| The comment quoted above seems to indicate that tests with GNU's
| <sstream> implementation, although in principle available, produced
| problems.  You may want to review the CVS logs to find out who
| put the comment in and then contact that person directly, asking
| him/her to elaborate on the comment.

I belive I was one of those who reported problems with sstream in
gcc 2.96, and that is basically the same as the one in gcc 2.95.3

OTOH I have not experienced any problems with the sstream in
gcc 3.0.x. (which is a reimplementation)

| [...]
>>   But I have not checked if gcc < 3 handles this
>> ok, but I would guess so since previous gcc's came with the SGI STL.
>> Supposedly the "!defined(__STL_USE_NEW_IOSTREAMS)" should make
>> lexical_cast use sstream, but that does not happen.
>
| Note also that gcc <= 2.95.2 does not have <sstream> at all.
| 2.95.3 does have <sstream>.

I know.

>> Also in regex_traits.hpp I have to do this change:
>
| Please send an e-mail to John Maddock directly.

Sure I can do that... _but_ IMHO this is not the easy way to get users
to report problems. It is not as if the mail address jumps at you from
the source files.

-- 
	Lgb