RFR: 8257736: InputStream from BodyPublishers.ofInputStream() leaks when IOE happens
Daniel Fuchs
dfuchs at openjdk.java.net
Fri Dec 4 10:02:52 UTC 2020
On Fri, 4 Dec 2020 00:47:51 GMT, Yasumasa Suenaga <ysuenaga at openjdk.org> wrote:
> `InputStream` from `BodyPublishers.ofInputStream()` is usually closed when the stream reaches EOF. However IOE handler returns without closing.
>
> I confirmed this problem in `BodyPublishers.ofInputStream()`, but I think `BodyPublishers.ofFile()`has same problem because it also use `StreamIterator`as well as `ofInputStream()`.
>
>
> # How to reproduce:
>
> Following code (Test.java) attempts to post body from `TestInputStream` which throws IOE in `read()`. "close called" is shown on the console if `close()` is called.
>
> import java.io.*;
> import java.net.*;
> import java.net.http.*;
>
> public class Test{
>
> private static class TestInputStream extends InputStream{
>
> public TestInputStream(){
> super();
> System.out.println("test c'tor");
> }
>
> @Override
> public int read() throws IOException{
> System.out.println("read called");
> throw new IOException("test");
> }
>
> @Override
> public void close() throws IOException{
> System.out.println("close called");
> super.close();
> }
>
> }
>
> public static void main(String[] args) throws Exception{
> var http = HttpClient.newHttpClient();
> var request = HttpRequest.newBuilder()
> .uri(URI.create("http://httpbin.org/post"))
> .POST(HttpRequest.BodyPublishers.ofInputStream(() -> new TestInputStream()))
> .build();
> http.send(request, HttpResponse.BodyHandlers.discarding());
> System.out.println("Press any key to exit...");
> System.in.read();
> }
> }
Hi Yasumasa,
Thanks for filing the issue and providing a fix.
Before integrating, can you please provide a non regression test that verifies the fix? You can place it in `test/jdk/java/net/httpclient/`;
Please also make sure that there is no regression in existing tests.
best regards
-- daniel
src/java.net.http/share/classes/jdk/internal/net/http/RequestPublishers.java line 422:
> 420: } catch (IOException ex) {
> 421: need2Read = false;
> 422: haveNext = false;
This shouldn't be required - `need2Read`/`haveNext` will be set by `hasNext()`; I'd prefer if we kept the logic there.
-------------
Changes requested by dfuchs (Reviewer).
PR: https://git.openjdk.java.net/jdk/pull/1614
More information about the net-dev
mailing list