Egregoros

@LinuxJedi I think what's most important is that a fix is pushed out ASAP before people start sticking WolfSSL in more places and now an unknown number of servers exist that are unable to handshake with RFC-compliant TLS 1.3 clients.

How many people are sticking WolfSSL in devices where they'll never update for years? That's what's incredibly concerning.

I'm also completely at a loss for why when even compiling with the middlebox compat mode it *still* couldn't complete the handshake. Something just seems extra broken about it.

@LinuxJedi I didn't want to mention this in the Github issue because it's just rage bait for passerbys, but I had the packet dump parsed by Claude which picked it apart for me with tshark and correctly analyzed the handshake frame by frame identifying the missing change_cipher_spec response from the server and how that was the root cause. Saved me so many hours of trying to figure it out myself.

(Claude did lead me down a wrong rabbithole first, but I caught that mistake pretty quick)

I was honestly expecting to be wrong and that the root cause was a broken client implementation on the Erlang side, but something in my gut said "it worked with OpenSSL, and we know OpenSSL is a trash fire, but Occam's Razor says the drop-in replacement of WolfSSL was the only variable that changed..." and I guess it was right.