Tai Chi Minh Ralph Eastwood
Tai Chi Minh Ralph Eastwood
I think I might have a solution! If we can override `__stack_chk_fail` with our own implementation, I think we might be on a winner. Then again that would introduce a...
There's a compromise flag: `-fstack-protector-explicit` which allows attributes to be passed to opt in for stack protection using `__attribute__((stack_protect))` but there's no equivalent opt out.
Fixed it. the unoptimisable vars and VLAs were actually doing more than I thought they were. TLDR; the VLAs are necessary to coerce the compiler to always generate a stack...
Yes, I have that in mind already :) and a comment saying it breaks if you build your application without -fno-stack-protector
There is a more stable alternative to this - libcoro and lwan do coroutines through having a fixed, single, argument for each coroutine and a stack pointer that is passed...
Some of the code I came up with in one of my attempts was mostly acceptable. With some extreme macro-wizardry it might be possible to wrap libcoro's style coroutines. I'd...
Hi, I'm implementing a simple http2 web server with libmill + nghttp2 c library. I need TLS with ALPN support for that. I've not fully looked into the protocol yet...
I've thought about that, but that does contaminate the namespace of libmill.h which should be avoided (I think?). I have made a concise extension to the API where you optionally...
Oops, I forgot to mention. Just a simple getter from `struct mill_sslconn` which stores the results of the callback from OpenSSL. This should be automatically set by the callback after...
1. The server name is for Server Name Indication extension. I don't think this is strictly necessary for ALPN but it does allow the client to specify the certificate it...