[clang] 9a2a167 - [DirectoryWatcher] Increase timeout to make test less flaky

[Shoaib Meenai via cfe-commits]

Author: Shoaib Meenai
Date: 2021-03-05T17:49:14-08:00
New Revision: 9a2a167b6ca7c35b60846592d7c11332c1f424e3

URL: https://github.com/llvm/llvm-project/commit/9a2a167b6ca7c35b60846592d7c11332c1f424e3
DIFF: https://github.com/llvm/llvm-project/commit/9a2a167b6ca7c35b60846592d7c11332c1f424e3.diff

LOG: [DirectoryWatcher] Increase timeout to make test less flaky

We've observed this test being significantly flaky on our Mac CI
machines when we're running the full check-clang suite. It fails because
the wait_for condition isn't met within 3 seconds. We believe it's
because our CI machines are somewhat underpowered and pretty heavily
loaded when we're running the full check-clang suite.

I ran some experiments on increasing the timeout. I ran the full
check-clang suite 100 times with each timeout value and recorded how
many flaky failures we encountered in these tests. The results are:

3 second timeout (baseline): 20 failures
10 second timeout: 14 failures
20 second timeout: 4 failures
30 second timeout: 2 failures
40 second timeout: 1 failure
50 second timeout: 0 failures
60 second timeout: 0 failures

I ran another set of 100 tests for the 50 second timeout and observed
one flaky failure. By contrast, I ended up running check-clang 500 times
for the 60 second timeout and didn't observe a single flaky failure.
That's how the 60 second timeout value used in this patch was derived.

While a 60 second timeout might seem high, keep in mind that:
- This is a timeout, not a sleep; the test should require much less time
  the vast majority of instances, especially on more powerful machines.
- The long timeout is most likely to occur when other tests are also
  running at the same time, so the latency of the timeout will also be
  masked by the latency of the other tests.

See https://reviews.llvm.org/D58418?id=200123#inline-554211 for where
this timeout was originally introduced and the possibility of raising it
if it wasn't enough was discussed.

Reviewed By: plotfi

Differential Revision: https://reviews.llvm.org/D97878

Added: 
    

Modified: 
    clang/unittests/DirectoryWatcher/DirectoryWatcherTest.cpp

Removed: 
    


################################################################################
diff  --git a/clang/unittests/DirectoryWatcher/DirectoryWatcherTest.cpp b/clang/unittests/DirectoryWatcher/DirectoryWatcherTest.cpp
index 650c0fc49764..f0dc55a47bff 100644
--- a/clang/unittests/DirectoryWatcher/DirectoryWatcherTest.cpp
+++ b/clang/unittests/DirectoryWatcher/DirectoryWatcherTest.cpp
@@ -34,6 +34,17 @@ namespace {
 
 typedef DirectoryWatcher::Event::EventKind EventKind;
 
+// We've observed this test being significantly flaky when running on a heavily
+// loaded machine (e.g. when it's being run as part of the full check-clang
+// suite). Set a high timeout value to avoid this flakiness. The 60s timeout
+// value was determined empirically. It's a timeout value, not a sleep value,
+// and the test should require much less time in practice the vast majority of
+// instances. The cases where we do come close to (or still end up hitting) the
+// longer timeout are most likely to occur when other tests are also running at
+// the same time (e.g. as part of the full check-clang suite), in which case the
+// latency of the timeout will be masked by the latency of the other tests.
+constexpr std::chrono::seconds EventualResultTimeout(60);
+
 struct DirectoryWatcherTestFixture {
   std::string TestRootDir;
   std::string TestWatchedDir;
@@ -243,7 +254,7 @@ void checkEventualResultWithTimeout(VerifyingConsumer &TestConsumer) {
   std::thread worker(std::move(task));
   worker.detach();
 
-  EXPECT_TRUE(WaitForExpectedStateResult.wait_for(std::chrono::seconds(3)) ==
+  EXPECT_TRUE(WaitForExpectedStateResult.wait_for(EventualResultTimeout) ==
               std::future_status::ready)
       << "The expected result state wasn't reached before the time-out.";
   std::unique_lock<std::mutex> L(TestConsumer.Mtx);