Franziskus Kiefer

> Do you have an exhaustive list of all the places where this specific UB happens? Unfortunately not. Any place that might get a null pointer as input and tries...

I'd suggest to use Travis only for platforms that aren't supported by GH actions such as FreeBSD to save credits. Everything in here we could do in actions as well.

Is FreeBSD the main issue you want to address @daachi ? You could do this with GH actions using for example this FreeBSD-vm action https://github.com/marketplace/actions/freebsd-vm

I don't know of any macro or similar things to handle this. There are a couple ways I can imagine to implement it. The existing krml macros could be changed...

> Does this mean if I have a struct that contains a vec256, I need to align it in the struct declaration _and_ align the outer allocation? This shouldn't be...

> I guess I would then have to align it unless I statically know that it's on a 16 byte boundary Yes. You have to know the sizes in your...

Sorry, I didn't expect that this will actually close this issue.

I'm not entirely sure... Looking at the evercrypt hash state it appears there have been issues before https://github.com/project-everest/hacl-star/commit/8f1d47e4b515c4390beb84c4d065dfc208556b5a. Note that the c89 code appears to be valid because the state...

> I believe this _used_ to be a an issue but with a recent change from Natalia, there is now a preprocessor guard that doesn't trigger __has_include unless we know...

I can confirm that the P256 performance on x64 regressed significantly. New code on main ``` ------------------------------------------------------------------- Benchmark Time CPU Iterations ------------------------------------------------------------------- P256_SHA256_ECDSA_Sign 170069 ns 170073 ns 4076 P256_SHA256_ECDSA_Verify 169703...