Re: Question regarding your post on gcc mail list

[Dan Kegel <dank at kegel dot com>]

Hi Wenhao,
you're right in that there has not been a definitive conclusion.
In the future, gcc and g++ may leave traces behind in the .o files
to let the linker know which support libraries to link in.
For now, the cleanest solution may be to link with g++ if you
know there are c++ object files involved.
Does that work ok for you?
- Dan

Wenhao Wu wrote:
> Hi Dan,
>
> I am writing this email to you regarding your following post on gcc list:
>
> http://gcc.gnu.org/ml/gcc/2004-04/msg01032.html
>
> I just met the exact same problem: My project is a C library that is coded
> in C++ internally with a C wrapper. The user API is pure C based. With gcc
> 2.95.x, I need to use "gcc -lstdc++" for linking. For recent gcc (3.3.3),
> I was having this undefined reference to "_Unwind_*" when I use "gcc
> -lstdc++"  to link. So I googled around and fortunely found your post at
> the gcc mail list. And it fixed my problem :)
>
> So, first of all, very good post. Thanks for the summary / explanation of
> the problem and putting it in a central place for people to looking it up.
>
> Second, after reading all the pointers and discussions in the same
> thread, I had more questions and hoped that you may be able to provide
> more insights:
>
> 1. I also have gcc 3.3.1 installed on a machine with same configuration
> and it seems to work ok by only linking with "-lstdc++". I tried "gcc -v"
> and "g++ -v" on both machines, and found that both gcc 3.3.1 and gcc 3.3.3
> were linking with the same libraries:
>
> gcc: -lgcc -lgcc_eh -lc -lgcc -lgcc_eh
>
> g++: -lstdc++ -lm -lgcc_s -lgcc -lc -lgcc_s -lgcc
>
> So, is this problem only pertinent to gcc 3.3.3 or higher? Maybe these
> symbols were still in -lgcc or -lgcc_eh in gcc 3.3.1 but got removed to
> -lgcc_s in gcc 3.3.3? Do you know of any place that documents these?
>
> 2. Do you think that linking "-lgcc_s" this way is a good solution? From
> you following post here (http://gcc.gnu.org/ml/gcc/2004-04/msg01079.html),
> you seemd to still hesitate on that. I work on the embedded space, so
> things have been tricky already, as you might already know :)
> (http://www.kegel.com/crosstool/).
>
> 3. Do you think that I should just use "gcc" for compiler and "g++" for
> linker? That might make the things a bit clearer.
>
> I tried to follow on all the articles that you mentioned, but could not
> reach any clear conclusions (possibly due to my lack of background in this
> area). So I figured that I'd ask you directly. Sorry if I had been lazy
> this way :)
>
> I will certainly appreciate more pointers to follow if there is no
> confirmative answer on these issues. I also tried to look up on
> gcc.gnu.org and didn't find much useful information on this topic. Maybe
> we are not supposed to mess around with these symbols, and I would not
> want to anyway should the gcc works perfectly :)
>
> Thanks for your time for reading this email. I know it's probably a
> bit long considering that it's the first communication between us :)
>
> I look forward to your reply. Thanks again.
>
> Regards,
> Wenhao


--
My technical stuff: http://kegel.com
My politics: see http://www.misleader.org for examples of why I'm for regime change

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com