On Tue, Feb 28, 2017 at 4:09 PM Jason Molenda <jmole...@apple.com> wrote:
> > You mentioned that this kind of mechanical churn of the code is something > everyone else in the LLVM/LLDB community supports. I don't know about the > LLVM community, but I strongly argue that this is the wrong way to develop > a stable, reliable product. I know I'm on the conservative side here, but > these kinds of changes add no value to the product and add plenty of risk > as they're being done while everyone is half paying attention to them going > in. Maybe if there was a vastly more extensive testsuite we could make > large scale changes without fear of breaking the debugger, but we don't > have that, no one has had the time to add one yet, and it's definitely > something that we need to be putting energy into. We don't need to be > putting energy into switching to a different string representation across > the board simply because it shows benefits in a handful of methods. > The way to develop a stable, reliable product is with an extensive test suite. And the way to get an extensive test suite is to make a test infrastructure that encourages / enables low-footprint, low-overhead, fast, portable, and most importantly easy to write tests. I wish working on LLDB were still my primary responsibility, but sadly it's not. But rejecting obvious improvements to the codebase because there is insufficient test coverage for reasons outside the control of the person making the improvement is a quick way to send LLDB to the graveyard, not a way to make things better. I mean, Chris was 100% in agreement before he left Apple, as was the rest of the LLVM community, and I haven't heard anything about priorities changing amongst leadership (if anything I've heard the opposite), so it's a bit strange to see this disconnect here
_______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits