Add a test case about "Conform lockingSourceCount is correct"

[QingShiLuoGu]

just add a failing test case for #6363 so so you can fix it.

[edith2727]

activar stripe

[AFZAL4911]

Hello dear
How is collect

[yschimke]

This should now be cleanly reproducible via Okio file system

See

okhttp/okhttp/src/test/java/okhttp3/internal/cache/DiskLruCacheTest.kt

Line 51 in ac63910

FakeFileSystem().apply { emulateWindows() } to true

Do you want to update the test and I'll take a look?

[QingShiLuoGu]

hello, do I need fork the latest version and then pull request again?
Or can I just edit this commit? @yschimke

[QingShiLuoGu]

I think it has been updated,please check it. @yschimke

[yschimke]

@swankjesse any opinion on this test?

I'm assuming that DiskLruCache.Snapshot, which is closeable, should have a checkNotClosed() method, and fail fast if you try to use it after closing it. But it breaks a bunch of other tests.

Using the snapshot sources will fail as they are closed, but the editor effectively reanimates them.

[swankjesse]

@yschimke yeah that reads to me like a small bug? The snapshot should not be used after it is closed, though I think there's a valid use case to using a snapshot after it's enclosing cache is closed.

This class isn't public API so we can improve it's behavior where that’s useful. But we don't really need to be super aggressive about enforcing invariants unless we're worried we might inadvertently violate them.

[yschimke]

Going to close this one out unless we can reproduce with a test case using public API. We can fix this, but it likely just incurs a tiny performance hit on each cache usage.