i think it makes sense to have a minimum version in wasi-libc as an experimental target.
how do you think?

libc.so needs __wasm_call_ctors to be called. it doesn't seem like a standard.
i guess it makes more sense for libc.so exports _initiailize as specified for wasi reactor modules.
https://github.com/WebAssembly/WASI/blob/main/legacy/application-abi.md
probably by linking crt1-reactor, maybe reverting https://reviews.llvm.org/D153293 partly.

how do you think?

When reproducing “component-linking-demo“ project under a virtual machine running Ubuntu 20.04 on x86 architecture, I noticed that the resulting libc.so seems to import some functions and globals that i did't expected to see. I used WABT to translate libc.so and observed the following partial output in import section:

(import "env" "getentropy" (func $getentropy (type 4)))
(import "env" "isatty" (func $isatty (type 0)))
(import "env" "reallocarray" (func $reallocarray (type 6)))
(import "env" "crypt_r" (func $crypt_r (type 6)))
(import "GOT.mem" "errno" (global $errno (mut i32)))
(import "GOT.mem" "_CLOCK_REALTIME" (global $_CLOCK_REALTIME (mut i32)))
(import "GOT.mem" "__wasilibc_environ" (global $__wasilibc_environ (mut i32)))

In addition to these imports, there are numerous similar imports. In my understanding, it seems that libc.so should implement and export these functions rather than importing them. Is this because there is a problem with my reproduction of the project or these imports in libc.so are normal?At last,English is not my native language; please excuse typing errors.

why have you excluded __main_void.o from libc?
dicej/wasi-libc@4631fa6
i'm not sure how non-void main can work without it.
was it a temporary workaround?