From: Huw Davies Subject: Re: [PATCH v2 6/6] gdi32/tests: Add StretchBlt HALFTONE mode tests. Message-Id: <20210614071214.GE29266@merlot.physics.ox.ac.uk> Date: Mon, 14 Jun 2021 08:12:14 +0100 In-Reply-To: References: On Mon, May 31, 2021 at 04:22:47PM +0800, Zhiyi Zhang wrote: > Signed-off-by: Zhiyi Zhang > --- > I'm not sure the tests should be in tests/dib.c. The framework in dib.c hashes results and compares them. > I don't think that's going to help us figuring out the exact HALFTONE algorithm details with the results > being hashes. Also, the tests that succeed have all input pixels has the same value. So if I only add the > tests that succeed, it provides little value. The tests here are mostly for helping figuring out the > algorithm in the future. Since the actual Wine implementation in the previous patch uses a different approach. > It doesn't reproduce identical results. So if you feel that such tests are not necessary, you can drop this. Yes, let's drop this. If we end up switching to an implemention that matches native, we can add tests to the hashing framework. I'm concerned about adding corner cases now, which will be a pain to get right later. Of course if there's an app that depends on a specific result that's a different matter. Huw.