On 09/Jul/2009 10:27, Charles Lee wrote: > On Thu, Jul 9, 2009 at 3:38 PM, Regis <xu.re...@gmail.com> wrote: > >> Charles Lee wrote: >> >>> It's a complicate situation. For example, user can override the read(), >>> which eats the exception thrown there. And this will make read(byte[], >>> int, >>> int) pass successful. Since PipedInputStream and PipedOutputStream are >>> >> He can override read(byte[], int, int) either. I'd like think it's a >> non-bug difference, and more, Harmony's implementation is better: reading >> batch data once is better than reading one by one byte. >> > I agree it's better to read data once. With the result of the test case, RI > seems to read one byte first to check the situation, and then read batch > data. > By the way, does harmony has a list about non-bug difference?
Yes, create a JIRA and assign it to the component "Non-bug differences from RI" Regards, Tim
