一篇对各种 C Unit Test 工具进行评测/选择的文章。

Evaluation Criteria Ease of use Features Portability Support/community Frameworks CppUnit

Evaluation Criteria Ease of use Features Portability Support/community Frameworks CppUnit

Perhaps the best solution is indeed to use a SignalHandler for all platforms to prevent being dependant on certain compiler switches.

Thanks for the great writeups, always enjoy the read!

You are using a catch(…) to capture system exceptions such as GPFs and such. However, in VS2005, there was a change in the way exceptions are dispatched. As a result, normally a system exception is not catchable by a catch(…) statement. This means the SystemExceptionShouldFailTest unit test crashes instead of reporting a failure, because the exception inside is not handled.

I found two ways to fix this, and I’m not sure which is better (both work well):

- use /EHa instead of /EHsc in both CppUnitLite2 project configs. This makes catch(…) work with system exceptions, like previous versions of VS.

- in Test::Run(), use the same trick as the Linux version (ie SignalHandler) and declare a local object that temporarily installs an UnhandledExceptionFilter, and that handler just throws a C++ exception.

For reference, I found a web page that explains this issue very well: http://members.cox.net/doug_web/eh.htm

Thanks for sharing the framework, I’m quickly becoming test-infected

I’d heard that he released it under those terms. Kinda crazy in my mind, but there you go.

Nevertheless, is it fair or ethical for you to benefit from the ‘cppunit’ name and Feather’s reputation while restricting his original liberal terms?

Your choice of the MIT license in this case is ideal. In keeping with Feather’s intent, I’d ask you to consider dropping any exceptions. Or change the name of the library as I mentioned above.

(Note that I’m not associated with the military in any way, in case you were wondering.)

Michael Feathers released the code with a comment along the lines of “do whatever you want with it”, so this seems well within those parameters.

Also, there’s probably not a single line left from the original code, so I can’t really see how there can be any objection.

Michael Feathers released the code with a comment along the lines of “do whatever you want with it”, so this seems well within those parameters.

Also, there’s probably not a single line left from the original code, so I can’t really see how there can be any objection.

While I commend your effort to build upon and improve the work of another (Michael Feathers in this case) it seems highly presumptive to impose a new restriction that you have no idea whether or not he endorses.

Seek his permission if you can. If you cannot obtain it, you should consider renaming the tool, lifting the restriction or removing the distribution from the website.

In any case, I’m looking forward to giving your changes a try.