RFR: 8299015: Ensure that HttpResponse.BodySubscribers.ofFile writes all bytes [v2]

[Dawid Weiss]

On Tue, 20 Dec 2022 10:37:34 GMT, Daniel Fuchs <dfuchs at openjdk.org> wrote:

>> src/java.net.http/share/classes/jdk/internal/net/http/ResponseSubscribers.java line 291:
>> 
>>> 289:                     long n = out.write(buffers);
>>> 290:                     if (n <= 0) {
>>> 291:                         throw new IOException("zero bytes written");
>> 
>> "Returns: The number of bytes written, possibly zero"
>> 
>> Sorry for being dumb, but why is the returned zero bytes treated as an error? If I were to write a fuzzifier for that class, it'd be my first thing on the to do list to randomly return zero bytes - if the contract allows it, it should be handled gracefully.
>
> Well this has the potential of creating a busy loop since bytesWritten will not increase if zero is returned.

I know that. But write should eventually make progress - even if it spins a few times. What's the point of having that JDK method explicitly state "may return 0" and then disregarding that as something that is an exceptional scenario? I don't see the consistency here... If it can return zero, fine - handle it by retrying. If this code falls into a spin loop and never makes progress - this would be a problem somewhere else (perhaps in the implementation of the file channel), not here.

-------------

PR: https://git.openjdk.org/jdk/pull/11722