> This is the accumulation of all build breaks test breaks, rather...
On Tue, Oct 14, 2014 at 9:32 AM, Todd Fiala <todd.fi...@gmail.com> wrote: > Hey all, > > I have good news and bad news. > > Good: our test runners now report failures correctly. > Bad news: our test runners have not been reporting failures since Aug 8 > 2014. > > As of svn r219689, dotest.py will once again report errors/failures via > non-zero return value, which will then propagate correctly through test run > results as reported by 'make test', 'ninja check-lldb', etc. > > You will almost certainly see test failures show up after r219689 that you > were not seeing before. This is the accumulation of all build breaks we > essentially made starting with r215256. We will need to fix the errors, > not revert r219689. (That would be akin to shooting the messenger). > > The lesson here is make sure we don't break our test runner, and ensure > return values from the test runner are preserved on exit. That is a hugely > critical facility. I will look into adding a test runner test so we don't > break that again. > -- > -Todd > -- -Todd
_______________________________________________ lldb-dev mailing list lldb-dev@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev
