Ian Lilley

I started trying this out and the performance is really good. I didn't go too deep with code review because of the big question you brought up of what to...

Right now I'm thinking the best chance of getting this working is to convert the exaggerated node's box to a region and find out which cells from the full exaggerated...

>1. Perhaps the easiest, instead of ShaderProgram throwing a DeveloperError, maybe it could throw a new ShaderCompilationError that includes the vertex and fragment text. I like this. And if the...

I've wondered about this too and I think it could come down to heuristics since there are so many different factors involved. One idea is to take all of these...

I see. One could argue it's a bug that `wstring_to_utf8` returns a different utf8 string than expected for a valid wstring that has interior null characters. I guess I'll rephrase...

Oh wait, `wstring` is a multipointer, so the function would have to be called something else... I don't know how much confusion would be caused by keeping the name and...

I had this problem too. Here's a workaround for anyone else visiting this issue. ``` context.user_ptr = cast(rawptr)nil ``` ``` context.assertion_failure_proc = cast(runtime.Assertion_Failure_Proc)nil ``` etc.

It looks like this got fixed at some point Now it prints: ``` thing2 Foo{type = i64, print = proc() @ 7FF7A73D65E0, value = 36} ```

Interestingly the minimal repro in https://github.com/odin-lang/Odin/issues/1843#issuecomment-1185839425 doesn't crash on LLVM 15 with `-debug` But this does: ``` package main VeryLargeStruct :: struct { vals: [65536]u8, // change to 65535 to...

Oh wait did that commit fix the zeroing slowness / crash that this issue is about and I confused it with LLVM fixing the problem? Anyway, take a look at...