stage2: ask for more file descriptors

[xackus]

Follow up to #6906.

[andrewrk]

I believe you also get to delete this code now too :)

zig/src/stage1/os.cpp

Lines 981 to 1003 in 445d808

	#if defined(ZIG_OS_POSIX)
	// Raise the open file descriptor limit.
	// Code lifted from node.js
	struct rlimit lim;
	if (getrlimit(RLIMIT_NOFILE, &lim) == 0 && lim.rlim_cur != lim.rlim_max) {
	// Do a binary search for the limit.
	rlim_t min = lim.rlim_cur;
	rlim_t max = 1 << 20;
	// But if there's a defined upper bound, don't search, just set it.
	if (lim.rlim_max != RLIM_INFINITY) {
	min = lim.rlim_max;
	max = lim.rlim_max;
	}
	do {
	lim.rlim_cur = min + (max - min) / 2;
	if (setrlimit(RLIMIT_NOFILE, &lim)) {
	max = lim.rlim_cur;
	} else {
	min = lim.rlim_cur;
	}
	} while (min + 1 < max);
	}
	#endif

[xackus]

I don't have a mac, so I would appreciate help with the EINVAL failure from someone who does. cc @kubkon
I made my linux machine go through the binary search path and it worked fine.

[kubkon]

@xackus 8dda64f should fix it on Darwin :-) The issue on Darwin is that you can't trust getrlimit() for NOFILE. According to the man pages:

setrlimit() now returns with errno set to EINVAL in places that historically succeeded. It no longer accepts "rlim_cur = RLIM_INFINITY" for RLIM_NOFILE. Use "rlim_cur = min(OPEN_MAX, rlim_max)".

[xackus]

@kubkon Thank you, I missed the "COMPATIBILITY" section of the man page.

[xackus]

This looks like another argument against EINVAL => unreachable. See #6389, #6925.

[kubkon]

This looks like another argument against EINVAL => unreachable. See #6389, #6925.

I guess, in this case, we could already map EINVAL to some proper error value. However, we just need to ensure we make it impossible to specify an incorrect resource value. For instance, on Darwin EINVAL can be returned in two instances. From the man pages:

 [EINVAL]           resource is invalid.
 [EINVAL]           The specified limit is invalid (e.g., RLIM_INFINITY or lower than rlim_cur).

Since we constrain the input resource to resource_limit enum, we should be OK mapping EINVAL exclusively to invalid limit error condition.

EDIT: proposing that for another PR, not necessarily this one.

[kubkon]

The CI failure is unrelated to this PR, so proceeding with the merge.