Alpine Linux causes a segmentation fault

Reproduce the crash with cargo run --example networking on Alpine Linux and inspect src/lib.rs around load_xlib_with_error_hook at lines 169-171 and its initializer at line 196. Use the provided gdb backtrace to investigate the null error representation; done means the networking example no longer segfaults on Alpine.

Seen in winit-block-on:

$ cargo run --example networking           
    Finished dev [unoptimized + debuginfo] target(s) in 0.08s
     Running `target/debug/examples/networking`
[1]    17749 segmentation fault  cargo run --example networking

gdb gives this backtrace:

(gdb) bt
#0  std::io::error::{impl#8}::fmt () at library/std/src/io/error.rs:951
#1  0x00007ffff7b16a1c in core::fmt::rt::Argument::fmt () at library/core/src/fmt/rt.rs:138
#2  core::fmt::write () at library/core/src/fmt/mod.rs:1094
#3  0x00007ffff7b0c054 in core::fmt::Write::write_fmt<alloc::string::String> ()
    at library/core/src/fmt/mod.rs:192
#4  alloc::fmt::format::format_inner () at library/alloc/src/fmt.rs:610
#5  0x00007ffff76c6cb5 in alloc::fmt::format::{closure#0} ()
    at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/alloc/src/fmt.rs:614
#6  0x00007ffff76c66de in core::option::Option<&str>::map_or_else<&str, alloc::string::String, alloc::fmt::format::{closure_env#0}, fn(&str) -> alloc::string::String> (self=..., 
    default=..., f=0x0)
    at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/option.rs:1181
#7  0x00007ffff76c6c4f in alloc::fmt::format (args=...)
    at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/alloc/src/fmt.rs:614
#8  0x00007ffff76c73bf in tiny_xlib::XLIB___rust_ctor___ctor::initer::load_xlib_with_error_hook::error<std::io::error::Error> (e=...) at src/lib.rs:169
#9  0x00007ffff76c5fd3 in core::ops::function::FnOnce::call_once<fn(std::io::error::Error) -> std::io::error::Error, (std::io::error::Error)> ()
    at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/ops/function.rs:250
#10 0x00007ffff76c6a09 in core::result::Result<tiny_xlib::ffi::Xlib, std::io::error::Error>::map_err<tiny_xlib::ffi::Xlib, std::io::error::Error, std::io::error::Error, fn(std::io::error::Error) -> std::io::error::Error> (self=..., op=0x0)
    at /rustc/cc66ad468955717ab92600c770da8c1601a4ff33/library/core/src/result.rs:829
#11 0x00007ffff76c71e1 in tiny_xlib::XLIB___rust_ctor___ctor::initer::load_xlib_with_error_hook () at src/lib.rs:171
#12 0x00007ffff70d9fbe in tiny_xlib::XLIB___rust_ctor___ctor::initer () at src/lib.rs:196
#13 0x00007ffff7af9555 in libc_start_init () at ../src_musl/src/env/__libc_start_main.c:64
#14 0x00007ffff7af957a in libc_start_main_stage2 ()
    at ../src_musl/src/env/__libc_start_main.c:92
#15 0x00007ffff70da064 in _start ()

Setting a breakpoint inside of the error printer gives me this:

(gdb) print e
$1 = std::io::error::Error {repr: std::io::error::repr_bitpacked::Repr (core::ptr::non_null::NonNull<()> {pointer: 0x0}, core::marker::PhantomData<std::io::error::ErrorData<alloc::boxed::Box<std::io::error::Custom, alloc::alloc::Global>>>)}

So it's some kind of weird niche issue. My guess is that somehow the dynamically linked functions are null when Alpine links them together, thus causing Rust to assume that it's somehow gotten a null IO error.