On Thu, 27 Jun 2024 05:06:30 GMT, Jaikiran Pai <[email protected]> wrote:
>> Currently HttpUrlConnection accepts truncated responses: if the server sends
>> a `Content-Length` header, and then closes the connection before
>> transferring all promised bytes, the input stream reports a clean EOF.
>>
>> In this PR I modify the MeteredStream class to throw an IOException when it
>> detects EOF before receiving all promised response bytes. MeteredStream (or
>> its subclass KeepAliveStream) is used when the response contains a
>> content-length header.
>>
>> The included test fails without the change, passes with it.
>> The same exception message and type is reported when a chunked response
>> (`Transfer-Encoding: chunked`) is truncated.
>> Unknown length responses that are terminated by EOF continue to work.
>>
>> 2 tests depended on the old behavior and had to be adjusted. The remaining
>> tier 1-3 tests continue to pass.
>
> src/java.base/share/classes/sun/net/www/MeteredStream.java line 58:
>
>> 56: if (n == -1) {
>> 57: if (expected > count) {
>> 58: throw new IOException("Premature EOF");
>
> Hello Daniel, should we instead throw `java.io.EOFException` which is much
> more precise than a generic `IOException`? `EOFException` states:
>
>
> * Signals that an end of file or end of stream has been reached
> * unexpectedly during input.
>
> which I think accurately depicts what's happening here.
Hi Jaikiran, I used IOException because that's already used by the
ChunkedInputStream:
https://github.com/openjdk/jdk/blob/409a3fe601c61451e36554d57f29e661fe16058a/src/java.base/share/classes/sun/net/www/http/ChunkedInputStream.java#L257
No code in sun.net uses EOFException today. I suppose I could change that, but
it might require changes in more places.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/19909#discussion_r1656424221
