* Gary V. Vaughan wrote on Sun, Aug 21, 2005 at 05:44:05PM CEST: > On 21 Aug 2005, at 16:16, Ralf Wildenhues wrote: > > > >Can you get a demangled name of the symbol __Z2a2c? > > I dunno. This is the kind of horror that makes me avoid C++ as if it > were perl ;-)
Bob answered this already, but if you have GNU binutils nm, you can compare the difference of `nm' and `nm --demangle' of the objects created in this test. I assume g++ uses the same mangling on darwin, so that would be: a2(char) This function is used in libb, and also in a template in the b.hh header which is instantiated in main(). I wrote the test with different link lines to find out what is portable and what is not. If you comment out the AT_CHECKs below the comments # lib convenience and # both installed (both of these do not link libb.la against liba.la), does the test succeed then? If so, could you commit a patch doing this for host_os darwin? (I would still like to test this for other systems, even if our link lines are not portable, it provides nice information. Maybe add a comment to this extent.) > >I think this is a genuine failure with template libraries on darwin. > >IOW: I don't think this ever worked. Our test is doing the right > >thing and exposes the failure. > > Quite possibly. It does mean that I can't run distcheck on HEAD > though :-( Oh, if above does not work, I would not mind an AT_XFAIL_IF(test "X$host_os" = ...) in tests/template.at. Cheers, Ralf
