> FYI: There was a thread in Nov 2023 about rand() which might be related.
>
> rand is not ISO C compliant in Cygwin
> https://cygwin.com/pipermail/cygwin/2023-November/254735.html
>
> conclusion of that thread:
> https://cygwin.com/pipermail/cygwin/2023-November/254764.html
I don't think that has much relevance here, as this is regarding
arc4random's seed after a fork, not rand's behavior in threads.
In any case I think I have found the cause of the bug. In
arc4random.c, _rs_stir_if_needed() attempts to detect a fork by
calling _rs_forkdetect(), which is located in arc4random.h. However,
that function is nothing more than a stub:
static inline void
_rs_forkdetect(void)
{
#ifdef _ARC4RANDOM_FORKDETECT
_ARC4RANDOM_FORKDETECT();
#endif
}
If _ARC4RANDOM_FORKDETECT is defined, _rs_forkdetect will attempt to
use it for detecting if a fork occured (the LibreSSL implementation
then zeros out `rs' with 'memset(rs, 0, sizeof(*rs));'.
Yet Cygwin has no such function or macro, so _rs_forkdetect does
nothing and _rs_stir subsequently cannot reseed after a fork() occurs.
- Johnothan King
--
Problem reports: https://cygwin.com/problems.html
FAQ: https://cygwin.com/faq/
Documentation: https://cygwin.com/docs.html
Unsubscribe info: https://cygwin.com/ml/#unsubscribe-simple
