AMDG On 06/21/2011 08:55 AM, Jason Nye wrote:
Platform/toolset information: Windows, MSVC9.0, primarily 64-bit but we do have 32-bit, _SECURE_SCL=0
My company is using boost 1.45.0 and we build boost ourselves. We have a policy that for any off-the-shelf software we build from source, we run the unit tests for that package to verify a basic level of usability.
When running the tests for all libraries (by using a script to CD into their respective 'test' directories and running bjam), most pass but I do get a few failures. The bjam command line I'm using for each is:
You can run all tests by running bjam once in %BOOST_ROOT%/status
bjam.exe --build-dir=%BOOST_ROOT%\..\build_x64_tests --user-config=%BOOST_ROOT%\user-config.jam link=shared address-model=64 cxxflags=/D_SECURE_SCL=0
The user-config.jam simply forces the use of the msvc90 toolset.
-------------------------------------------------------------------------------------- The results:
Lexical Cast ============ ... lexical_cast_loopback_test.cpp(74): error in "test_round_conversion_double": check magic_msvc == lexical_cast<T>(magic_msvc_s) failed lexical_cast_loopback_test.cpp(74): error in "test_round_conversion_long_double": check magic_msvc == lexical_cast<T>(magic_msvc_s) failed
This is a known issue. It's relatively minor. It just means that streaming a double and reading it back doesn't necessarily return exactly the original value.
tr1.random ========== ... unknown location(0): fatal error in "test_main_caller( argc, argv )": This application has requested the Runtime to terminate it in an unusual way. Please contact the application's support team for more information.
run_random.cpp(48): last checkpoint C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC\INCLUDE\../include/random(850): invalid argument for subtract_with_carry[_01]::seed
*** 1 failure detected in test suite "Test Program"
This is from the TR1 implementation included in the standard library.
--------------------------------------------------------------------------------------
Does anyone know what might be the cause of these failures?
Maybe I'm running the tests incorrectly? Keep in mind that we absolutely need certain compiler flags - running bjam by itself will choose the wrong compiler, address model and _SECURE_SCL flags.
Any help would be greatly appreciated.
In Christ, Steven Watanabe