opt: Don't raise RLIMIT_NOFILE if it would prevent using posix_spawn()

If we raise RLIMIT_NOFILE, we have to lower it before calling exec() for
compatibility with select().  If posix_spawn() doesn't support that, we
fall back to fork(), which is quite a bit slower.

Therefore, if we're going to exec() on most files, it's better to keep
RLIMIT_NOFILE the same to avoid the fork() cost, even though it makes
bftw() somewhat slower.

1 files changed, 3 insertions, 0 deletions

diff --git a/src/eval.c b/src/eval.c
index 2ca4a1f..8863e34 100644
--- a/src/eval.c
+++ b/src/eval.c
@@ -1466,6 +1466,9 @@ done:
 static int raise_fdlimit(struct bfs_ctx *ctx) {
 	rlim_t cur = ctx->orig_nofile.rlim_cur;
 	rlim_t max = ctx->orig_nofile.rlim_max;
+	if (!ctx->raise_nofile) {
+		max = cur;
+	}
 
 	rlim_t target = 64 << 10;
 	if (rlim_cmp(target, max) > 0) {