Re: [Re: gcc-2.8.1 compiler problems]

["Martin v. Loewis" <martin at mira dot isdn dot cs dot tu-berlin dot de>]

> Yes, the library was there as you said, but it didn't work either
> and libgcc.a doesn't have those symbols. I also copied libgcc.a to
> /usr/local/lib and included it into the makefile, just in case. By
> doing a grep of one of the symbols (i.e. __rtti_user) it showed me
> that it is into libstdc++.a

grep is not appropriate for finding symbols, please use nm(1) instead.
What you've find is that libstdc++ *needs* this symbol, instead of
providing it. In nm output, this shows up as 'U' (rather than 'T', as
it should).

> Maybe I'm doing something wrong and I can't see what it could be. If
> you come up with another solution please do tell me, because I'm
> getting short of time and I need to solve this by this weekend if
> possible.

Unfortunately, it is hard to help here. I don't have access to HPUX,
and gcc 2.8.1 is not actively maintained anymore; the current version
of gcc is 2.95.1.

When you build gcc, please watch for the moment it builds
libgcc.a. Then, you need to see that it compiles tinfo.cc. This file
contains the implementation of __rtti_user.

In case you are looking for a quick hack: The implementation of
rtti_user is

extern "C" void
__rtti_user (void *addr, const char *name)
{ new (addr) __user_type_info (name); }

To compile this, you need to include tinfo.h from the gcc source
(gcc/cp/tinfo.h).

Hope this helps,
Martin