On Sun, 18 Oct 2026 05:46:45 GMT, Alan Bateman <[email protected]> wrote:
> > It sounds like it was the right call to deduplicate JDK-8394737 into then > > JDK-8387022, since my issue is one part of the original. > > Would it be possible to zoom out for a moment and explain how you got here? > Did you encounter a class that extends FIS and overrides read with a broken > implementation or it was this a drive-by observation? It was a drive-by observation; I have been reading `ChannelInputStream` and `FileInputStream` implementations closely in the course of investigations that led me to file https://bugs.openjdk.org/browse/JDK-8394231. I have seen the `(byte)read()` cast bug in a number of `InputStream` implementations over the years, so it stuck out. > > JDK-8387022 is about the problematic readNBytes implementation in FIS and > sun.nio.ch.CIS.readNBytes. It's the same code where it doesn't handle read > returning 0 correctly. Right, and my impulse was to delete these blocks, since they can only happen for non-conforming subclasses or implementations. ------------- PR Comment: https://git.openjdk.org/jdk/pull/33401#issuecomment-6246279662
