On Thu, 25 Apr 2024 16:37:39 GMT, Alexey Ivanov <aiva...@openjdk.org> wrote:
> This PR provides a regression test for > [JDK-8325179](https://bugs.openjdk.org/browse/JDK-8325179): _Race in > BasicDirectoryModel.validateFileCache_ reviewed in #18111. > > The test is inspired and based on `ConcurrentModification` that I wrote for > [JDK-8327137](https://bugs.openjdk.org/browse/JDK-8327137) / > [JDK-8323670](https://bugs.openjdk.org/browse/JDK-8323670). > > **How the test works** > > The test creates a temporary directory in the current directory and creates a > number of files in it. (The number of files is controlled by > `NUMBER_OF_THREADS` constant). Then the test creates `JFileChooser` in the > temporary directory. > > The test starts several scanner threads, the number is controlled by > `NUMBER_OF_THREADS`, which repeatedly call > `fileChooser.rescanCurrentDirectory()`. This results in calling > `BasicDirectoryModel.validateFileCache` which starts a background thread — > "Basic L&F File Loading Thread" — to enumerate the files. > > The test runner thread and scanner threads are synchronised with > `CyclicBarrier`, this ensures all the scanner threads start at the same time. > After a short delay, the runner thread takes a snapshot of live threads. > After a longer delay, the operation is repeated. See the `getThreadSnapshot` > method. (The number of snapshots is controlled by `SNAPSHOTS` constant.) > > The number of File Loading Threads in each snapshot is counted. There should > be no more than two threads. It is possible that thread two such threads > after JDK-8325179 is fixed: the existing thread is interrupted but hasn't > exited yet, and a new thread is already created. > > The test fails consistently without the fix for JDK-8325179. On Windows, the > output looks like this: > > > Number of snapshots: 20 > Number of snapshots where number of loader threads: > = 1: 0 > = 2: 0 > > 2: 20 > java.lang.RuntimeException: Detected 20 snapshots with several loading threads > at LoaderThreadCount.runTest(LoaderThreadCount.java:132) > at LoaderThreadCount.wrapper(LoaderThreadCount.java:72) > at java.base/java.lang.Thread.run(Thread.java:1583) > > > On Linux and macOS, there's more variation, for example I got the following > output on one of Linux systems: > > > Number of snapshots: 15 > Number of snapshots where number of loader threads: > = 1: 7 > = 2: 2 > > 2: 6 > > > The test passes on the builds where JDK-8325179 is present. Could you, @mrserb and @TejeshR13, take a look? You reviewed the test for [JDK-8323670](https://bugs.openjdk.org/browse/JDK-8323670) in #18109. Those who reviewed [JDK-8325179](https://bugs.openjdk.org/browse/JDK-8325179) in #18111 — @prrace and @turbanoff — could be interested too. ------------- PR Comment: https://git.openjdk.org/jdk/pull/18957#issuecomment-2077726248
