Andrew Clayton

> In fact if we can configure the warnings to be build breaking even better in my opinion. That's currently the case as we compile with `-Werror` by default (you...

Regarding the (sizeof() -1) compile-time vs strlen(3) runtime evaluations. With optimisations enabled there are several cases where the strlen(3) will be turned into a runtime constant. E.g. ```c #define STR...

Rebased with master... ``` $ git range-diff 41d71d76...85f21b7c -: -------- > 1: de430eda Add flag for newline control in access log entries -: -------- > 2: 76cc071a Fix missing newlines...

Hey @alejandro-colomar I've been keeping an eye on these things. Good stuff!

I do have a [patch](https://git.sigsegv.uk/unit.git/commit/?h=wasm&id=8f6449f4e0152d9644b807cf2208e3734a0f8b4d) for this (though can't remember what actual state it's in) though it's an alternative for the original wasm language module that only supports WASM modules...

> The only configuration difference between Linux and FBSD is events use = "epoll and kqueue" Try using poll(2) on them both, does one perform markedly better than the other...

In [Unit](https://unit.nginx.org/) we [switched](https://github.com/ac000/unit/commit/2444d45ec969a539c6e1f4484783dd9e9ce21626) to using [sched_getaffinity(2)](https://man7.org/linux/man-pages/man2/sched_setaffinity.2.html) on Linux, keeping `sysconf(_SC_NPROCESSORS_ONLN)` as a fallback.

> When I configure a listener with *:8080 unitd only listens on 0.0.0.0:8080 and not on :::8080. Indeed. > The documentation states the following: A wildcard that matches any host...

_If_ we were to implement this, probably something simply like `:port` would be best so as not to change existing behaviour...

@pluknet Having reviewed the [PR](https://github.com/nginx/nginx/pull/517), the code is not unreasonable. But given your comment above which of these 3 options are you advocating for? 1) Stick with only the old...