[C++-sig] Re: STLPort problems

[David Abrahams]

"Mike Thompson" <mike.thompson at day8.com.au> writes:

> "David Abrahams" <dave at boost-consulting.com> wrote in message
> news:usmnkeirs.fsf at boost-consulting.com...
>> "Mike Thompson" <mike.thompson at day8.com.au> writes:
>>
>> > "Mike Thompson" wrote:
>> >> I'm getting compile time errors when building Boost.Python
>> >> using STLport.
>> >>
>> >
>> > Okay.  After far too long, this all seems to be down to two isses:
>> >
>> >      1.  A problem with Python2.3 and Boost 1.30.2 which now appears
>> >           fixed in cvs HEAD.  See my previous posts for details of the
>> >           missing #define for LONG_LONG.
>> >
>> >      2.  As suggested by David, there is an STLport bug.  Its been around
>> >           for a while but has not been fixed in the latest beta snapshots.
>> >           I've raised the following bug report as a further prompter:
>> >                       http://www.stlport.com/dcforum/DCForumID6/1364.html
>> >
>> >           NB: The bug is ONLY triggered when  you use STLport with the
> "I'll
>> > use
>> >           native iostream library" switch on STLport.
>>
>> Mike,
>>
>> STLPort "native iostreams" are a hack, and I'm inclined not to support
>> them.  Lots of subtle things don't work right because you can't make
>> the fact that they're in a different namespace truly invisible.  If
>> there's a small patch needed to get your stuff going, and you post it,
>> I'll probably be willing to apply it.  Can you use STLPort iostreams
>> instead?
>>
>
> I can't see any quick patch which might get STLport "native iostreams" working
> with Boost.Python.  A broken <complex> in STLport breaks a bit too much of
> boost, and fixing <complex> doesn't look easy.
>
> I'll now investigate using STLPort iostreams instead.
>
> Many thanks.
>
> P.S. If its not supported -- and I can now see why -- I wonder if Boost.Python
> should produce an #error if someone (like me) tries to go down the "native
> iostreams" path?

The build system does that already; you only miss the error if you
subvert Boost.Build.  OTOH if you want to submit a patch which will
work, I'll gladly fold it into the codebase.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com