AMDG On 05/21/2012 04:48 PM, Ted Byers wrote:
Thank you one and all who replied. Your replies were quite helpful.
I have a couple final questions.
1) What does such a test program return to whatever invoked it. But that I mean, can I build the testsuite using make, have make execute the testsuite program, and fail with an error if one of the tests fail? If so, I can create another target in my makefiles that do just that, and still another one that checks code into an SVN repository after doing an update against the repository and rebuilding everything, and verifying the testsuite builds and runs successfully (and bales before the check-in if one of the tests fails).
If any test fails or crashes, the test program returns non-zero. What you want sounds like exactly what we do in our own regression tests.
2) Is there a tool included in Boost::Test that provides support for code coverage, that can be used to check to verify that all the code in the source code is tested by one or more test cases in the test suite? I have found several open source tools (gcov, trucov, and covtool), but have not yet had a chance to investigate them. Again, I want an automated way to ensure that all code in my project is subjected to one, or, preferably more, tests.
This isn't really something that can be handled from within C++. You need an external tool, or compiler support. In Christ, Steven Watanabe