On 28/09/13 23:36, Antony Polukhin wrote:
2013/9/29 Rene Rivera
Note to the variant maintainers.. I commented out a test which was referring to a non-existent source file.
I forgot to commit the variant_swap_test.cpp file. I'll commit it at Monday, thanks for disabling the test.
But is it OK that all the regression testing could be stalled by a typo in one of the libraries?
It's typically the same for most projects. In some other projects it's even worse, you can't even build the libraries if someone refers to a test that doesn't exist. The right thing to do is to not commit changes that break the build system, and ensure that all changes are atomic. I personally still find it surprising that people can forget files when they're doing a commit, though this may be because of bad versioning tools or commit procedure. Surely you run "status" and "diff" commands before doing a commit?