Hi Sergey, Overall looks good to me.
I’m a little confused by the main() test loop: 115 test(afw, afr, type, getStream(from, true)); Did you also intended to call (frameLength = NOT_SPECIFIED), to would be: test(afw, afr, type, getStream(from, true)); test(afw, afr, type, getStream(from, false)); Otherwise, I don’t see where getStream(…, false) is used. -Dan > On Nov 17, 2017, at 3:11 AM, Sergey Bylokhov <sergey.bylok...@oracle.com> > wrote: > > Hello, Audio Guru. > > Please review the fix for jdk10. > Bug: https://bugs.openjdk.java.net/browse/JDK-8191384 > Webrev can be found at: http://cr.openjdk.java.net/~serb/8191384/webrev > > > The bug occurs in the next code in WaveFloatFileReader.java: > final RIFFReader riffiterator = new RIFFReader(stream); > while (riffiterator.hasNextChunk()) { > RIFFReader chunk = riffiterator.nextChunk(); > if (chunk.getFormat().equals("data")) { > return new AudioInputStream(chunk, af, length); > } > } > In the code below we create a RIFFReader for the file stream, then iterates > over the chunks to find a data chunk and create an AudioInputStream. > > The problem is that the chunk has noop close() method, so if the user will > try to close AudioInputStream, the close() method of the file stream will not > be called. > > In the fix the close() method for the chunk was changed and now it always > close the stream on top of which it was created. > I also checked that this RIFFReader.close() method was never called in our > code(except the tests), its probably a reason why the NPE when it is called > twice was not found. > > -- > Best regards, Sergey.
