Hi. I use valgrind 3.2.3. When I run it on even simple Boost.Test unit tests, it complains about use of uninitialised data, the report looks like this: ==11809== Memcheck, a memory error detector. ==11809== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al. ==11809== Using LibVEX rev 1732, a library for dynamic binary translation. ==11809== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP. ==11809== Using valgrind-3.2.3, a dynamic binary instrumentation framework. ==11809== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al. ==11809== For more details, rerun with: -v ==11809== ==11809== Syscall param sigaltstack(ss) points to uninitialised byte(s) ==11809== at 0x8B1D3E7: sigaltstack (in /lib64/libc-2.6.1.so) ==11809== by 0x430CEA: boost::detail::signal_handler::~signal_handler() (execution_monitor.ipp:667) ==11809== by 0x431803: boost::execution_monitor::catch_signals(boost::unit_test::callback0<int> const&) (execution_monitor.ipp:731) ==11809== by 0x432675: boost::execution_monitor::execute(boost::unit_test::callback0<int> const&) (execution_monitor.ipp:1058) ==11809== by 0x40EF61: boost::unit_test::framework::init(boost::unit_test::test_suite* (*)(int, char**), int, char**) (framework.ipp:258) ==11809== by 0x41D740: boost::unit_test::unit_test_main(boost::unit_test::test_suite* (*)(int, char**), int, char**) (unit_test_main.ipp:169) ==11809== by 0x41DA70: main (unit_test_main.ipp:226) ==11809== Address 0x7FEFFEAB0 is on thread 1's stack Running 1 test case... *** No errors detected ==11809== ==11809== ERROR SUMMARY: 8 errors from 1 contexts (suppressed: 3 from 1) ==11809== malloc/free: in use at exit: 0 bytes in 0 blocks. ==11809== malloc/free: 200 allocs, 200 frees, 64,075 bytes allocated. ==11809== For counts of detected errors, rerun with: -v ==11809== All heap blocks were freed -- no leaks are possible. I think if you would add sigstk.ss_sp = 0; sigstk.ss_size = 0; in signal_handler::~signal_handler() It will not complain about this. I know, it should not be necessary, because it is disabling the signal stack, right? Jarl