Franziskus Kiefer

Yeah I also think that it's coming from copying the vector. It's still a little unexpected. The ref version is definitely better but not necessarily what I want to use...

Unfortunately I don't have to maintain this right now. But I'm happy to merge PRs to fix this.

oops, I guess I left the old versions in there... I'll push an update. v20 works in gnome 3.34 v21 works in gnome 3.36

Hm, this clearly shouldn't happen. I'll have to find some time to look into this. What version of GNOME do you use? I'm on Arch with GNOME 3.26.2 and didn't...

Thanks @franglais125 I'll try to port that when I get to it.

I agree that this would be a nice feature. But I don't know when I'll have time to get to it (PRs are as always welcome).

I'd second that. One of the recommendations in the audit was to use formally verified crypto. High-level crypto traits would allow rustls to do so, for example using evercrypt.

HKDF has a similar issue, trying to allocate memory of the size of `info` on the stack. https://github.com/project-everest/hacl-star/blob/43796f8827d3ef9232ef8cd75046431ff89b3327/dist/ccf/EverCrypt_HKDF.c#L41

DRBG is another instance of the same issue https://github.com/project-everest/hacl-star/blob/36eeec7cc2ff1f6f652a9cabb047c87da364378b/dist/c89-compatible/Hacl_HMAC_DRBG.c#L197-L200

> I think there's still the hashes that ought to be fixed, and perhaps a couple more algorithms. Yes, hashes are effected as well. I'll leave it to @msprotz whether...