http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53028
--- Comment #9 from Mike Stump <mikestump at comcast dot net> 2012-04-24 00:31:35 UTC --- Since little proof was added to support the assertion that the additional testing is useful, I can remain skeptical about it, though, the CFE people certainly are free to require it, what they say goes. I like the idea of testing new code, just I think it could be over done. I'll give you an example, I think we all can agree on. Let's say we add a warning. The code goes in, and a single testcase: main() { i = code to produce warning; // warning } but, we don't also test: -pedantic-errors main() { i = code to produce warning; // warning } even though, it is reasonable. We could, but don't. Why don't we, because it isn't worth the testing time and the maintenance time. Testing it once, and assuming that no one will accidentally change the warning is reasonable. We also don't test that the warning goes away with -w. We don't test the warning turns into an error with -Werror. How many times has one of these tests caught someone modifying a pedantic into a warning? How many times has one of these tests caught someone modifying a pedantic into an error? How many times did someone modify a pedantic in one of the two ways because a testcase wasn't present? Feel free to use your best recollection for the above answers. I can't help but think the numbers are so low, as to not be worth the cost of the additional testcases.