On 12 Jun 2013, at 19:29, Chris Hegarty <chris.hega...@oracle.com> wrote:
> The source change looks fine to me, since our implementation treats the given > chunk length as the complete chunk size ( including header & footer ). > > I would prefer to not add a test, if the test will end up being problematic, > given this is a corner case. I mean, when running tests in the samevm concurrently. -Chris > > -Chris > > On 12 Jun 2013, at 18:48, Kurchi Hazra <kurchi.subhra.ha...@oracle.com> wrote: > >> >> >> >> Hi, >> >> In HttpUrlConnection, if the chunk length is set to Integer.MAX_VALUE, our >> code was trying to initialize a buffer of size >> greater than that, which was resulting in an integer overflow, and >> consequently a NegativeArraySizeException. This fix >> ensures that while initializing the internal buffer, we never exceed the >> maximum size of the chunk length fixed by the user, >> and hence avert the NegativeArraySizeException. >> >> I have added a test to cover the case. The test will however try to create a >> buffer of size Integer.MAX_VALUE. I could >> change the value of the JVM heap, but the required size (~2 GB) will be too >> large for most machines to support. >> However, I am open to advice as to what I should do in this case. >> >> Bug: http://bugs.sun.com/view_bug.do?bug_id=8015421 >> Webrev: http://cr.openjdk.java.net/~khazra/8015421/webrev.00/ >> >> >> >> Thanks, >> Kurchi
