|
Boost : |
From: Matthias Troyer (troyer_at_[hidden])
Date: 2002-10-13 08:23:03
Shouldn't we implement the workaround also for other platforms (e.g.
darwin or MacOS X) until the FSF has fixed the problem?
I will try to create a small test case.
Matthias
On Saturday, October 12, 2002, at 08:48 PM, David Abrahams wrote:
> Matthias Troyer <troyer_at_[hidden]> writes:
>
>> Hi,
>>
>> When trying to build the boost libraries under the current MacOS X
>> version I run into error messages of the type:
>>
>> template with C linkage
>>
>> I could pin this down to the use of the -isystem option
[cut]
>> Did anybody else have similar problems? Changing the -isystem option
>> to -I in the build file solved the problem for me, but I want to know
>> what the reason is, and if this is a bug that should be reported to
>> Apple or to the FSF.
>
> Absolutely it should be reported to FSF. It shows up on Sun as well,
> and we implemented a workaround in tools/build/gcc-tools.jam as you
> can see. I think (though I don't have a perfect memory of this) that
> the person who asked us to implement the workaround for Sun submitted
> a bug report, but with only a very large test case. I'm not sure how
> soon it will be addressed on that basis.
>
> Ah, no, it appears to be in a form of developer-disbelief-limbo see
> report #7327 at http://gcc.gnu.org/cgi-bin/gnatsweb.pl?database=gcc.
>
> The docs for -isystem say:
>
> -isystem dir
> Search dir for header files, after all directories specified by -I
> but before the standard system directories. Mark it as a system
> directory, so that it gets the same special treatment as is applied
> to the standard system directories.
>
>
> Someone seems to think that files with -isystem can be wrapped in
> extern "C", but then how do the standard C++ library headers get
> #included?
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk