aszlig

**Spoiler:** We're both on Nix(OS), so the failures are not only reproducible, but I've also disabled the tests in the past (NixOS/nixpkgs@2acc258dff1a37974edd6475851e218bb09e281a) because I didn't have time to find out...

> +1, I'd love to host this myself as well. I might look into at least getting a docker container built with nix - it should be possible to version...

This is probably because we're using a custom `--version-script`: https://github.com/nixcloud/ip2unix/blob/1083369be4c2a09256f06507bacdf2accadd8d78/meson.build#L85-L92

~~Okay, just tested this and it's *not* because of the `--version-script`,~~ investigating...

Sooo, this essentially boils down to a `nullptr` dereference because the symbols *actually* resolve to `null`: With LTO: ```sh-session $ nm -D result/lib/libip2unix.so | grep socket 0000000000000000 A socket ```...

Okay, backtracking on what I wrote [earlier](https://github.com/nixcloud/ip2unix/issues/33#issuecomment-2212426927), it **is** the linker script indeed. I accidentally run a isolated test which ran into essentially a no-op without the linker script.

Maybe `power` is a better substitute for that, since it's more related to the old `sw` method. **edit:** actually, not really, the `power` and `power_long` methods are available only for...

I think the proper fix for this would be to use `sw` and fall back to `power` if `sw` is not available when using `soft`. Since we already query the...

Possibly related: #61

> Since the intention of upstream is to get rid of docbook for good, it probably makes sense to port our module documentation to markdown already when porting our manual...