RFR: 8325179: Race in BasicDirectoryModel.validateFileCache [v2]

[Tejesh R]

On Tue, 5 Mar 2024 11:07:13 GMT, Alexey Ivanov <aivanov at openjdk.org> wrote:

>> Ensure access to the `filesLoader` field of `BasicDirectoryModel` is synchronized.
>> 
>> Without synchronization, a thread checks if `filesLoader` is not null and creates a new `FilesLoader` thread. If the thread is pre-empted between these two operations, another thread or even several threads can see the `null` value and create new `FilesLoader` threads.
>> 
>> The same way, `BasicDirectoryModel.invalidateFileCache` needs to be synchornized to interrupt the current `filesLoader` and assign `null`.
>> 
>> This bug allows reproducing `ConcurrentModificationException` seen in [JDK-8323670](https://bugs.openjdk.org/browse/JDK-8323670) and [JDK-8307091](https://bugs.openjdk.org/browse/JDK-8307091) using the test in PR #18109.
>
> Alexey Ivanov has updated the pull request incrementally with four additional commits since the last revision:
> 
>  - Replace synchronized invalidateFileCache with synchronized block inside
>  - Declare DoChangeContents constructor private, wrap its parameters
>  - Space after synchronized in DoChangeContents.run
>  - Convert runnable to local variable

src/java.desktop/share/classes/javax/swing/plaf/basic/BasicDirectoryModel.java line 552:

> 550:             if (remSize > 0 && addSize == 0) {
> 551:                 fireIntervalRemoved(BasicDirectoryModel.this, remStart, remStart + remSize - 1);
> 552:             } else if (addSize > 0 && remSize == 0 && addStart + addSize <= fileCache.size()) {

Any reason for moving this portion out of `Synchronized` block? Because `fileCache.size()` might need to be inside the `Synchronized` block right?

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/18111#discussion_r1514242873