Hi Bruno, Bruno Haible <[email protected]> writes:
> Collin Funk wrote: >> ASAN was not top of mind when I wrote that test. I >> should probably get in the habit of using it no matter what > > You don't need to. That's what we have the CI for. So that you can > concentrate on the algorithm and a maintainable code. True. I still like to avoid making it fail. But if it never failed, I guess it would have no purpose. :) By the way, I suggest adding "-fsanitize=bounds" to the CI, so that any violations of the counted_by attribute are caught. With both GCC and Clang, it can be enabled with ASAN and/or UBSAN. So you could use the same job, unless you prefer using a separate one for some reason. Collin
