The joint gcc 3.4.3 + 4.1.0 test runs appear to have turned up a bug (which you have since fixed I believe) in the LD_LIBRARY_PATH handling of BBv2.
When tests were run, I would get an LD_LIBRARY_PATH that included
both the 4.1.0 and 3.4.3 libdirs, and due to the issue in the next paragraph, this broke many of the 3.4.3 tests.
I've had to discontinue testing on gcc 4.1.0 for now due to it not supporting wide-characrter strings or streams on Solaris versions less than 10 (see
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6702
). This is why, when the 3.4.3 tests were picking up the 4.1.0 libstdc++, they were hitting missing symbol errors.
Here's my current test-config.jam:
--- cut ---
using gcc : 3.4.3_sunos : /home/ietdev/tools/sunos-sparc/gcc-
3.4.3/bin/g++ : <linker-type>sun ;
using gcc : 4.1.0_sunos : /home/ietdev/tools/sunos-sparc/gcc-4.1.0/bin/g++ : <linker-type>sun ;
using python : 2.4 : /home/ietdev/tools/sunos-sparc/python-2.4 ;
--- cut ----
Perhaps the value of LD_LIBRARY_PATH should be reported along with the rest of the test output?