On 18/06/2021 18:30, Brian Burkhalter wrote:
On Jun 18, 2021, at 6:36 AM, Alan Bateman <al...@openjdk.java.net
<mailto:al...@openjdk.java.net>> wrote:
Adding an override of transferTo may require new tests. @bplb Do you
if we have good tests for all the scenarios where input stream
returned by Channels.newInputStream is the source?
There are not a lot of tests for the `InputStream` returned by
`Channels.newInputStream`:
`java/nio/channels/Channels`:
`Basic.java` - reading from a wrapped `FileChannel`
`ReadByte.java` - trivial
`Basic2.java` - reading at random offsets from a wrapped
`AsynchronousSocketChannel`
`ReadOffset.java` - trivial
Elsewhere:
`sun/nio/ch/TempBuffer.java`
`jdk/nio/zipfs/ZipFSTester.java`
I don’t see that `transferTo()` on such a stream is tested anywhere.
InputStream/TransferTo.java tests the default implementation. Maybe we
re-structured this to use a data provider so that the tests exercise
both the default implementation and the various overrides. Alternatively
a new test that exercises the various overrides in different scenarios.
In any case, I think the feedback for PR 4263 is that it is essentially
several implementations of transferTo and each of those need to have
test coverage.
-Alan
