[Boost-users] strange "fatal errors" with Boost.Test

Subject: [Boost-users] strange "fatal errors" with Boost.Test
From: Christoph Groth (cwg_at_[hidden])
Date: 2010-03-12 13:54:31

Dear boost experts,

while adding regression tests to my project I experience very strange
errors.  I am using boost 1.40 and g++ 4.4 as packaged in Debian
testing.

The problem is that a very simple test suite, one which does almost
nothing, produces the following runtime error:
################################################################
cwg_at_argo:/tmp/lg-bug$ ./test_all 
Running 1 test case...
unknown location(0): fatal error in "sq_dist": signal: illegal operand; address of failing instruction: 0x00410106

*** 1 failure detected in test suite "lg"
################################################################

A quite minimal program which allows me to reproduce the bug is available from
http://ilorentz.org/~cwg/lg-bug.tar.gz

This is how I can reproduce it on any of my two Debian testing systems.
The problem does not appear on an Ubuntu system to which I have access
(with boost 1.38, but I do not think that this problem is specific to
some boost version).
################################################################
cwg_at_argo:/tmp$ wget -q http://ilorentz.org/~cwg/lg-bug.tar.gz
cwg_at_argo:/tmp$ tar xfz lg-bug.tar.gz 
cwg_at_argo:/tmp$ cd lg-bug/
cwg_at_argo:/tmp/lg-bug$ g++ -g -o test_all test_glass.cc glass.cc
cwg_at_argo:/tmp/lg-bug$ gdb test_all
GNU gdb (GDB) 7.0.1-debian
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /tmp/lg-bug/test_all...done.
(gdb) run
Starting program: /tmp/lg-bug/test_all 
Running 1 test case...

Program received signal SIGILL, Illegal instruction.
0x0000000000410106 in boost::itest::expectations_logger::data_flow (this=Cannot access memory at address 0x3fdffffffffffe38
) at /usr/include/boost/test/impl/logged_expectations.ipp:188
188             BOOST_CHECK_EQUAL( *tit, DATA_SIG ); ++tit;
(gdb) bt
#0  0x0000000000410106 in boost::itest::expectations_logger::data_flow (this=Cannot access memory at address 0x3fdffffffffffe38
) at /usr/include/boost/test/impl/logged_expectations.ipp:188
Cannot access memory at address 0x3fe0000000000008
################################################################

The test case (file test_glass.cc) is just:
################################################################
#define BOOST_TEST_MODULE lg
#include <boost/test/included/unit_test.hpp>

#include "defs.hh"
#include "glass.hh"

BOOST_AUTO_TEST_CASE(sq_dist)
{
    Levy::Vecd a(2, 1);
    Levy::Glass sys(Levy::Vecd(1, 1), Levy::Vecb(true, true));
}
################################################################

The two constructors in the body of the testcase do not really do
anything except initializing a few numbers and std::vectors.  However,
the bug disappears when I comment out one of these lines.

The backtrace (see log) is not really helpful to me...  This is the
backtrace of the problem as it appears in my full program:

################################################################
Program received signal SIGSEGV, Segmentation fault.
0x0000000000432728 in ~lazy_ostream_impl (this=0xffffccc0, __in_chrg=<value optimized out>) at /usr/include/boost/test/utils/lazy_ostream.hpp:64
64      class lazy_ostream_impl : public lazy_ostream {
(gdb) bt
#0  0x0000000000432728 in ~lazy_ostream_impl (this=0xffffccc0, __in_chrg=<value optimized out>) at /usr/include/boost/test/utils/lazy_ostream.hpp:64
#1  0x000000000041010f in expectations_logger (this=0x6f3730, log_file_name=..., test_or_log=false) at /usr/include/boost/test/impl/logged_expectations.ipp:87
#2  0x0000000000475c66 in boost::unit_test::ut_detail::invoker<boost::unit_test::ut_detail::unused>::invoke<void (*)()> (this=0x7fffffffceaf, f=@0x6f2a08)
    at /usr/include/boost/test/utils/callback.hpp:56
#3  0x00000000004718e5 in boost::unit_test::ut_detail::callback0_impl_t<boost::unit_test::ut_detail::unused, void (*)()>::invoke (this=0x6f2a00) at /usr/include/boost/test/utils/callback.hpp:89
#4  0x000000000043b773 in boost::unit_test::callback0<boost::unit_test::ut_detail::unused>::operator() (this=0x6f2ad0) at /usr/include/boost/test/utils/callback.hpp:118
#5  0x000000000042ec4f in operator() (this=0x6f7378) at /usr/include/boost/test/impl/unit_test_monitor.ipp:41
#6  0x000000000042ec32 in invoke<boost::unit_test::<unnamed>::zero_return_wrapper_t<boost::unit_test::callback0<boost::unit_test::ut_detail::unused> > > (this=0x7fffffffcf3f, f=...)
    at /usr/include/boost/test/utils/callback.hpp:42
#7  0x000000000042ec13 in invoke (this=0x6f7370) at /usr/include/boost/test/utils/callback.hpp:89
#8  0x0000000000443257 in boost::unit_test::callback0<int>::operator() (this=0x7fffffffdb60) at /usr/include/boost/test/utils/callback.hpp:118
#9  0x000000000043b872 in boost::detail::do_invoke<boost::scoped_ptr<boost::detail::translate_exception_base>, boost::unit_test::callback0<int> > (tr=..., F=...)
    at /usr/include/boost/test/impl/execution_monitor.ipp:244
#10 0x000000000040f156 in boost::execution_monitor::catch_signals (this=0x6f1d00, F=...) at /usr/include/boost/test/impl/execution_monitor.ipp:956
#11 0x000000000040f21a in boost::execution_monitor::execute (this=0x6f1d00, F=...) at /usr/include/boost/test/impl/execution_monitor.ipp:1282
#12 0x0000000000418eee in boost::unit_test::unit_test_monitor_t::execute_and_translate (this=0x6f1d00, tc=...) at /usr/include/boost/test/impl/unit_test_monitor.ipp:69
#13 0x00000000004311ee in boost::unit_test::framework_impl::visit (this=0x6f1340, tc=...) at /usr/include/boost/test/impl/framework.ipp:150
#14 0x000000000041d587 in boost::unit_test::traverse_test_tree (tc=..., V=...) at /usr/include/boost/test/impl/unit_test_suite.ipp:193
#15 0x000000000041daff in boost::unit_test::traverse_test_tree (id=65536, V=...) at /usr/include/boost/test/impl/unit_test_suite.ipp:232
#16 0x000000000041d748 in boost::unit_test::traverse_test_tree (suite=..., V=...) at /usr/include/boost/test/impl/unit_test_suite.ipp:207
#17 0x000000000041db1c in boost::unit_test::traverse_test_tree (id=1, V=...) at /usr/include/boost/test/impl/unit_test_suite.ipp:234
#18 0x000000000040ab87 in boost::unit_test::framework::run (id=1, continue_test=true) at /usr/include/boost/test/impl/framework.ipp:436
#19 0x0000000000418c5c in boost::unit_test::unit_test_main (init_func=0x41eed8 <init_unit_test_suite(int, char**)>, argc=1, argv=0x7fffffffe378) at /usr/include/boost/test/impl/unit_test_main.ipp:185
#20 0x0000000000418e56 in main (argc=1, argv=0x7fffffffe378) at /usr/include/boost/test/impl/unit_test_main.ipp:237
################################################################

When I add change some irrelevant code the program which just shifts
everything a bit the backtrace becomes different or the error
disappears.  So it looks like some memory corruption problem.  But I'm
clueless...

Thank you in advance for any hints

Christoph