Deflake LoopTest.WaitOneBlocking by increasing timeout. (#17857)
This test case has failed a few times on macos_x86, and possibly on
other platforms too. I suspect the OS isn't scheduling the test threads
soon enough - there's only a 150 millisecond window between the event
set on a worker thread and `iree_loop_wait_one`'s timeout on the main
thread:
```
main thread worker thread
----------- -------------
test start
spin up thread ------> wait 50ms
wait 200ms ...
... <------ set event
timeout if event not set
```
Sample logs:
*
https://github.com/iree-org/iree/actions/runs/9214985535/job/25352380335#step:10:1578
*
https://github.com/iree-org/iree/actions/runs/9882364677/job/27295096340?pr=17856#step:9:43
```
[ RUN ] LoopTest.WaitOneBlocking
iree/runtime/src/iree/base/loop_test.h:612: Failure
Value of: status
Expected: error code OK
Actual: 0x4, whose error code is DEADLINE_EXCEEDED: DEADLINE_EXCEEDED
[ FAILED ] LoopTest.WaitOneBlocking (200 ms)
```
We could also use an infinite timeout.
diff --git a/runtime/src/iree/base/loop_test.h b/runtime/src/iree/base/loop_test.h
index 480b9a7..1386b3d 100644
--- a/runtime/src/iree/base/loop_test.h
+++ b/runtime/src/iree/base/loop_test.h
@@ -606,7 +606,7 @@
bool did_wait_callback = false;
} user_data;
IREE_ASSERT_OK(iree_loop_wait_one(
- loop, wait_source, iree_make_timeout_ms(200),
+ loop, wait_source, iree_make_timeout_ms(2000),
+[](void* user_data_ptr, iree_loop_t loop, iree_status_t status) {
IREE_TRACE_SCOPE();
IREE_EXPECT_OK(status);
