"George Russell" <george@codeplay.com> wrote in message news:4427D1F3.4050509@codeplay.com...

Hi,

We're currently using boost::test in one of our projects. I'm seing a difference in behaviour between unittests compiled on two different machines. On one machine, the unittests run to completion, reporting failure(s) and fatal errors when accessing invalid memory. On the other, the unittests stop execution on the first fatal error. They also dump details of memory leaks. It's the same code, compiled in the same way.

Can someone explain why this is?

Hi, George Unfortunately you do not provide enough information to give you an exact explanation. An educated guess what could happened: On first machine system errors are caught on second are not. There could be several reasons for that. But I may need some answers first: Does the environment identical? Do you use the same build of Boost.Test? Which release? Does the test module produce fatal errors on both machines? Which environment you are running on? Same or different?. Are you sure that test are run the same way (meaning no --catch_system_error=no is defined in one of the cases)? It's possible that on one of these machines BOOST_TEST_CATCH_SYSTEM_ERROR environment variable is defined (and set to "no"). Gennadiy