On Thu, 2007-07-26 at 11:50 +0200, Stephan Bergmann wrote: > Caolan McNamara wrote: > > I suggest for new ports that use gcc to give configure-ing with > > --without-stlport4 > > a go. That way you avoid the need to build stlport and use the system > > gcc stl instead. That may get you further into the build, though of > > course a per-arch uno bridge will be required in bridges/source/cpp_uno > > that'll be the real hard bit. > > FYI, see > <http://porting.openoffice.org/servlets/ReadMsg?list=mac&msgNo=5826> for > some quick thoughts on --without-stlport4 in the context of the Mac OS X > port.
What bothers me is that *some* of the older gcc's have problems with visibility and system stl not working together. I don't know if the current default mac compiler is one of the broken ones. I know that our gcc-4.1.2 with backports is ok f.e. Ideally we would be able to detect this in configure, but I don't know how to test for bustage :-(. If we could detect potential bustage then I'd say we just setup configure to automatically default to --disable-stlport4 when using gcc for all platforms when it is known safe, except for a short table of established ports that are considered to need stlport so that 3rd party foo works with them, i.e. basically just linux i386. *Perhaps* also linux ppc but I wouldn't really consider that an established in a "people have actually created third party c++ addons for them" abi. FWIW: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19664 is the problem I refer to, so all gcc + libstdc++ 4.2.0 combos should be ok. We can test for that version no and be sure it's ok, but plenty of olders gcc's are also ok that a simple version test would exclude. C. Note that in stlport there is a little hack in place that bounces some of the sgi extensions which are in a __gnu_cxx namespace for system stl into the std namespace so that no other mods are required in our codebase, so on occassion the bouncer headers may need some other newly used methods added to them. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
