t/ds-leak: fix race

We need to ensure we run lsof on the sleep(1) process, and not
the fork of ourselves before execve(2).  This race applies when
we're using the default pure-Perl spawn() implementation.

1 files changed, 12 insertions, 0 deletions

diff --git a/t/ds-leak.t b/t/ds-leak.t
index dd8446dc..34ffc125 100644
--- a/t/ds-leak.t
+++ b/t/ds-leak.t
@@ -15,9 +15,21 @@ if ('close-on-exec for epoll and kqueue') {
 
         PublicInbox::DS->SetLoopTimeout(0);
         PublicInbox::DS->SetPostLoopCallback(sub { 0 });
+
+        # make sure execve closes if we're using fork()
+        my ($r, $w);
+        pipe($r, $w) or die "pipe: $!";
+
         PublicInbox::DS->AddTimer(0, sub { $pid = spawn([qw(sleep 10)]) });
         PublicInbox::DS->EventLoop;
         ok($pid, 'subprocess spawned');
+
+        # wait for execve, we need to ensure lsof sees sleep(1)
+        # and not the fork of this process:
+        close $w or die "close: $!";
+        my $l = <$r>;
+        is($l, undef, 'cloexec works and sleep(1) is running');
+
         my @of = grep(/$evfd_re/, `lsof -p $pid 2>/dev/null`);
         my $err = $?;
         SKIP: {