labath added a comment.
I actually enjoy debugging things like this, so I tried playing around and came
up with this test case:
// Test that a signal which is not monitored by the MainLoop does not cause a
premature exit.
TEST_F(MainLoopTest, UnmonitoredSignal) {
MainLoop loop;
Status error;
struct sigaction sa;
sa.sa_sigaction = [](int, siginfo_t *, void *) { };
sa.sa_flags = SA_SIGINFO; // important: no SA_RESTART
sigemptyset(&sa.sa_mask);
ASSERT_EQ(0, sigaction(SIGUSR2, &sa, nullptr));
auto handle = loop.RegisterSignal(SIGUSR1, make_callback(), error);
ASSERT_TRUE(error.Success());
std::thread killer([]() {
sleep(1);
kill(getpid(), SIGUSR2);
sleep(1);
kill(getpid(), SIGUSR1);
});
ASSERT_TRUE(loop.Run().Success());
killer.join();
ASSERT_EQ(1u, callback_count);
}
It's not the nicest of tests: it uses sleep, which is necessary to make sure
the main thread gets a chance to block in kevent(2), and one needs to be very
careful when doing signal catching in a multithreaded environment (e.g.
creating the killer thread before registering the SIGUSR1 handler would make
the test flaky on linux, as the new thread would not inherit the signal mask
which has USR1 blocked). However, I think it should be sufficient for the
purposes of this patch.
In case anyone is interested, the fact that attaching a debugger produces an
observable side-effect (EINTR) in the debugged process is considered a bug by
the linux kernel folks and they have worked hard to fix it. Right now linux
kernel will automatically restart most syscalls after a debugger resume
regardless of the value of SA_RESTART flags or anything like that. There is
just a handful of syscalls that are really complex and they haven't figured out
how to resume them part-way. These still return EINTR.
================
Comment at: source/Host/common/MainLoop.cpp:108-112
+ do {
+ errno = 0;
+ num_events = kevent(loop.m_kqueue, in_events.data(), in_events.size(),
+ out_events, llvm::array_lengthof(out_events), nullptr);
+ } while (num_events == -1 && errno == EINTR);
----------------
There's an `llvm::sys::RetryAfterSignal` which implements the EINTR loop. I had
to pass `out_events` as `&out_events[0]` to make template deduction happy.
Repository:
rLLD LLVM Linker
https://reviews.llvm.org/D42206
_______________________________________________
lldb-commits mailing list
lldb-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
