On Thu, 29 Jan 2026 18:50:15 GMT, Chen Liang <[email protected]> wrote:
> I recommend using junit and fail fast for individual cases - fail fast stack > traces are much easier to locate compared to an error count. In addition, we > anticipate most of the test cases to pass, so an error count is not that > valuable. Yeah, I'll acknowledge the potential benefits of junit. During development, I find it helpful to have an error count since the count can better indicate how systematic the problem is. ------------- PR Review Comment: https://git.openjdk.org/valhalla/pull/1991#discussion_r2743329778
