ADDITIONAL SYSTEM INFORMATION :
Reproduced with 11.0.14, 17.0.3 and 19-ea+22 on Ubuntu 18.04.

A DESCRIPTION OF THE PROBLEM :
In some circumstances (see the reproducer) InputStreamReader::read(char[], int, int) can return zero while actually consuming a decoded character and writing it into the destination buffer. It happens because StreamDecoder.lockedRead() appears to assume that it's correct to always return "n + implRead(cbuf, off, off + len)" at [1], but it's not so if n == 1 and implRead() returns -1.

Note that the reproducer below is just one of the ways to trigger the bug. Another potential case is when implReady() returns true at [2], but the stream is actually at EOF (which typically doesn't happen, but is not forbidden by the contract of ready()).

[1] https://github.com/openjdk/jdk/blob/69ff86a32088d9664e5e0dae12edddc0643e3fd3/src/java.base/share/classes/sun/nio/cs/StreamDecoder.java#L218
[2] https://github.com/openjdk/jdk/blob/69ff86a32088d9664e5e0dae12edddc0643e3fd3/src/java.base/share/classes/sun/nio/cs/StreamDecoder.java#L204

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Compile and run the following code.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
a
1
-1

ACTUAL -
a
0
-1


---------- BEGIN SOURCE ----------
import java.io.*;
import java.nio.charset.CodingErrorAction;
import java.nio.charset.StandardCharsets;

class Test {
    public static void main(String[] args) throws Exception {
        var r = new InputStreamReader(
                new ByteArrayInputStream(new byte[] {'a', 'b', (byte) 0xC2}),
                StandardCharsets.UTF_8.newDecoder().onMalformedInput(CodingErrorAction.IGNORE));
        System.out.println((char)r.read());
        System.out.println(r.read(new char[3], 0, 3));
        System.out.println(r.read());
    }
}
---------- END SOURCE ----------