This is the mail archive of the gcc-help@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: problem linking with GCC 3.0 -- undefined reference to `std::codecvt<char, char, __mbstate_t>::id


I have tried many variants of building gcc 3.0 using Claudio's suggestion but 
each build has ended with a version of libstdc++.a with the wrong symbol for 
std::codecvt<char, char, __mbstate_t>::id.

System Configuration
---------------
Linux Mandrake 7.2 on an i686 machine.
gcc 2.95.2-12mdk
binutils 2.10.0.24-4mdk
glibc 2.1.3-18.3mdk

For the record, here is what I have tried.

1. Building gcc 3.0 with the installed tools and then building gcc 3.0
   with itself.

2. Building gcc 2.95.3 with the installed tools, building gcc 3.0 with
   gcc 2.95.3, and then building gcc 3.0 with itself.

3. Building gcc 2.96-85 (from the Red Hat 7.1 updates) with
   the installed tools without protoizing, building gcc 3.0 with gcc 2.96,
   and building gcc 3.0 with itself.

4. Building gcc 2.96-85 with the installed tools, protoizing gcc 2.96,
   building gcc 3.0 with gcc 2.96, and building gcc 3.0 with itself.

After each build, I used nm and grep to determine the values of codecvt::id 
in libstdc++.a, just as Claudio did.

I chose gcc 2.96-85 because Claudio said that he resolved the problem on a 
Red Hat 7.1 system.  Since I wanted to install gcc 2.96 under /usr/local, the 
rpm files were not relocatable, and my attempt to build the rpm files from 
the srpm file failed, I had to use the sources from the srpm file to build 
the package manually, following the steps from the spec file as closely as 
possible.  (I couldn't use the binary rpm files directly since my system 
lacks glibc 2.2.)

As you might imagine, my attempts to build gcc 3.0 have consumed an enormous 
amount of time.  I am very interested in obtaining a working version for C++ 
development since gcc 3.0 has so much better C++ standards compliance 
compared to gcc 2.95 but I am very close to giving up entirely.  If anyone 
sees something I'm missing or has more insight into this problem, I would 
very much appreciate additional suggestions.  I am also willing to gather 
whatever additional information from my system that might be necessary.

Thank you very much,

Mark


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]