summaryrefslogtreecommitdiff
path: root/tools/aapt2/java/JavaClassGenerator_test.cpp
diff options
context:
space:
mode:
authorKevin Han <kevhan@google.com>2020-03-17 13:55:49 -0700
committerKevin Han <kevhan@google.com>2020-03-19 11:27:57 -0700
commita6d5bb47502e90e983c1e44a82190ea8f6dfd9c2 (patch)
treee542049f8788c82068564996e210793bdc5b05e3 /tools/aapt2/java/JavaClassGenerator_test.cpp
parent035a0710827ee23e098ad6bf3b2cade5c690478f (diff)
Deflake NotificationEntryManagerInflationTest
Deflake NotificationEntryManagerInflationTest by moving back to the countdown latch approach and providing our own executor to AsyncInflationTask. It was infeasible to depend on the exact number of main thread messages as other main thread messages from other tests could interfere. Thus. we move back to the countdown latch for when NotificationEntryManager finishes inflation. This makes us dependent on the listener API for determining when inflation is finished, but it's unlikely we can do better unless we can inject an executor into RowInflaterTask. In addition, using AsyncTask's SERIAL_EXECUTOR on hwasan builds seemed to be much slower than normal. This lead to other tests' AysncTask work delaying the work in this test and leading to the timeout happening before inflation finished. By providing a test executor instead and synchronously controlling its execution, we avoid this issue. Bug: 150618180 Test: atest --iterations 100 NotificationEntryManagerInflationTest Test: atest SystemUITests Test: A forrest run coming soon! Change-Id: I87c6e2d216c8f26aaf340d311f618c9dccaba8af
Diffstat (limited to 'tools/aapt2/java/JavaClassGenerator_test.cpp')
0 files changed, 0 insertions, 0 deletions