This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
Re: gcc-4.0.0 for sparc-sun-solaris2.8: Almost all good!
- From: Benjamin Kosnik <bkoz at redhat dot com>
- To: Brad Spencer <spencer at infointeractive dot com>
- Cc: libstdc++ at gcc dot gnu dot org
- Date: Wed, 27 Apr 2005 11:24:11 -0500
- Subject: Re: gcc-4.0.0 for sparc-sun-solaris2.8: Almost all good!
- Organization: Red Hat / Chicago
- References: <20050426194621.GZ28225@eng.infointeractive.com>
> One library I have makes heavy use of Boost.Spirit, which is heavy on
> templates. Strangely enough, though, when I link this static library
> against code that uses it, ld 2.15 claims that some of
> std::vector<std::string>'s member functions are not available to
> link. It takes a very long time to conclude this, as if it is
> searching through every symbol (linearly?). I've attached some sample
> output and more details on the compilation.
>
> I fixed this by adding the following to the library:
>
> template class std::vector<std::string>;
>
> Clearly, I shouldn't have to do that. In my mind this could be
> libstdc++ or gcc. Ideas? Narrowing this down to a simple test case
> is very difficult; this is only happening in the complex case of using
> Boost.Spirit. :(
I haven't been following the Solaris build instructions for 4.0.0, but
it seems as if binutils or ld/as requirements are pretty strict. There
have been a lot of bugs about missing functions and link-time odditites.
Do you test results look something like:
http://gcc.gnu.org/ml/gcc-testresults/2005-04/msg01623.html
I would suggest you try and reproduce the Boost.Spirit issue on linux.
If you can, then it's probably not related to binutils lossage.
> Other than that, everything seems to be working fine, including wide
> character support after Benjamin's rewritten configury.
That part is only in 4.1, actually. However, it's still nice that 4.0.0 works.
;)
-benjamin