Verified on head of "mob" branch:

$ ./tcc test.c
tcc: error: library 'c' not found

This is due to the issue discussed here: https://news.ycombinator.com/item?id=23612772

Basically, the "*.dylib" files no longer exist on the filesystem, although when you call dlopen() on the paths it still works.  This breaks any code that tries to probe for the existence of a library itself rather than just calling dlopen() and seeing if it fails.

This seems to be happening inside tcc with tcc_add_library_err() -> tcc_add_library() -> tcc_add_library_internal() -> tcc_add_file_internal() -> _tcc_open()

Unfortunately after that point it gets ugly -- I don't even know if it is possible to do anything like what macho_load_dll() is currently doing.  On MacOS 11 the way that the system linker works is that it finds a yaml *.tbd file describing the library inside of the SDK (for example /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/usr/lib/libc.tbd ) and then does some magic to map that to a symbol to use.  It's very confusing to me and I'm not sure how easy it would be to emulate.