On Sun, 2024-09-22 at 19:20 +0000, Gary D. Gregory wrote: > Hi All, > > After updating an app from HC Core 5.2.5 to 5.3, my custom version of > > https://github.com/apache/httpcomponents-core/blob/master/httpcore5-h2/src/test/java/org/apache/hc/core5/http2/examples/H2FileServerExample.java > > started failing tests because it is replying with a HTTP status 421 > instead of 200. > > Narrowing down the Core versions, 5.3-alpha2 was OK and the failure > started in 5.3-beta1. > > I test this app with 2 different clients, Jetty 12 (12.0.14-SNAPSHOT) > and HttpClient 5.3.1. > > The app flow is: > - HC HTTP Client <-> Jetty 12 > org.eclipse.jetty.ee9.proxy.AsyncMiddleManServlet (a custom subclass) > <-> Custom H2FileServerExample > - Jetty HTTP Client <-> Jetty 12 > org.eclipse.jetty.ee9.proxy.AsyncMiddleManServlet (a custom subclass) > <-> Custom H2FileServerExample > > Both Jetty and HC clients see the 421 come back from > AsyncMiddleManServlet. > > This is a hint that something changed in Core since this is what > H2FileServerExample uses. > > It turns out that the problem is that the client sends request to > "localhost" and the > H2FileServerExample’s AsyncServerBootstrap ends up using > InetAddress.getLocalHost().getCanonicalHostName() which replies > "host.docker.internal" on my machine and not “localhost”, so the Core > issues a 421 on behalf of H2FileServerExample. > > If I change the test clients to send requests to > "host.docker.internal" instead of "localhost", it still fails with a > 421. > > In the file server, these are the headers we get from the > AsyncMiddleManServlet: > [Accept-Encoding: gzip, x-gzip, deflate, User-Agent: Apache- > HttpClient/5.3.1 (Java/17.0.12), Via: 1.1 MY_COMPANY_COMPUTER_NAME, > X-Forwarded-For: 192.168.1.2, X-Forwarded-Proto: http, X-Forwarded- > Host: host.docker.internal:60975, X-Forwarded-Server: 192.168.1.2, x- > header-a: a1, Host: localhost:60974] > > So I updated the configuration of my AsyncMiddleManServlet subclass > to use getCanonicalHostName(). > > Now, it all works. > > My question is: Did 5.3 become stricter and that is undocumented in > the release notes, or, is this a regression? >
This is per RFC 9110, section 15.5.20 https://www.rfc-editor.org/rfc/rfc9110.html#name-421-misdirected-request Oleg --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@hc.apache.org For additional commands, e-mail: dev-h...@hc.apache.org
