Try compiling a debug version. That will give you line numbers to help you track it down easier. James On Nov 14, 2008, at 6:04 AM, Peter Klotz wrote:
Boost 1.37.0, Linux, x86_64, gcc 4.1:
Valgrind 3.3.1 complains when running this simple test:
#define BOOST_TEST_MAIN #include <boost/test/unit_test.hpp>
BOOST_AUTO_TEST_CASE( test1 ) { BOOST_CHECK( 1 == 1 ); }
The valgrind messages:
==1836== Memcheck, a memory error detector. ==1836== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al. ==1836== Using LibVEX rev 1854, a library for dynamic binary translation. ==1836== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP. ==1836== Using valgrind-3.3.1, a dynamic binary instrumentation framework. ==1836== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al. ==1836== For more details, rerun with: -v ==1836== Running 1 test case...
*** No errors detected ==1836== Invalid read of size 8 ==1836== at 0x41E41C: __tcf_1 (in RegressionTest) ==1836== by 0x3830632FA4: exit (in /lib64/libc-2.5.so) ==1836== by 0x383061D8BA: (below main) (in /lib64/libc-2.5.so) ==1836== Address 0x4c1ade0 is 40 bytes inside a block of size 48 free'd ==1836== at 0x4A0555C: operator delete(void*) (vg_replace_malloc.c:342) ==1836== by 0x42024C: std::_Rb_tree<unsigned long, std::pair<unsigned long const, boost::unit_test::test_unit*>, std::_Select1st<std::pair<unsigned long const, boost::unit_test::test_unit*> >, std::less<unsigned long>, std::allocator<std::pair<unsigned long const, boost::unit_test::test_unit*> > >::erase(unsigned long const&) (in RegressionTest) ==1836== by 0x417C11: boost::unit_test::test_unit::~test_unit() (in RegressionTest) ==1836== by 0x41E41B: __tcf_1 (in RegressionTest) ==1836== by 0x3830632FA4: exit (in /lib64/libc-2.5.so) ==1836== by 0x383061D8BA: (below main) (in /lib64/libc-2.5.so) ==1836== ==1836== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 5 from 1) ==1836== malloc/free: in use at exit: 0 bytes in 0 blocks. ==1836== malloc/free: 89 allocs, 89 frees, 38,935 bytes allocated. ==1836== For counts of detected errors, rerun with: -v ==1836== All heap blocks were freed -- no leaks are possible.
I was not able to figure out what valgrind is really complaining about.
The internal map of test units holds two entries. One corresponds to my BOOST_AUTO_TEST_CASE call (id=1), the other one seems to be used internally (id=65536).
Both are properly erased in deregister_test_unit() which is called by ~test_unit().
Maybe someone with a little bit more insight into Boost.Test is willing to look at this.
Regards, Peter. _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users