mflatt
2017-6-4 10:09:45
@leif I vaguely remembered there are issues with calling atexit()
directly, and I found the relevant code. See the PR comment. Although no one is likely to try to compile Racket on SunOS 4 where atexit()
is unavailable, the potential indirection through replacement_at_exit
is important for Windows due to libc sharing (or not) in DLLs.