jeffatrad

Hi, Thomas - that method seems to just be using Apple-clang? I'm trying to build without MacOS at all...

Innnnteresting... I'll look into that!

Good catch Nik - that is REALLY hard to trigger, but yeah, that's a bug. Let me double check the other si128 casts, and I'll submit to Sean...

Yeah, this is slightly more complicated. Are you calling stbir_build_samplers_with_splits() back out, prior to the resizing? What should happen here is that if we need to do a reinit of...

Hmmm, it actually already does fail if you need to rebuild coeffs. I would expect a warning that we are setting the values to the same value when calling from...

> It's writing when stbir__update_info_from_resize is called and it's reading a bit later from the guts of the resize Correct, but every thread does this - all writes will happen...

Can you tell me the command line you use to build this test app?

Wait, nevermind, got it to happen in clang.

Hmmm, the point sample filter isn't used very often (it's just for pixel art, really) - I must have a bug in there, it definitely shouldn't be using 2 coeffs....