Am 02.02.26 um 23:40 schrieb Mark Thomas:
On 23/01/2026 13:19, Rainer Jung wrote:
I did some debugging: for Java 8 (failing) and Java 11+ (OK), the test
case testIPv6LinkLocal use the same link local address in
TestStartupIPv6Connectors. For example:
fe80:0:0:0:2e0:4cff:fee2:a74a%enx00e04ce2a74a
But in org.apache.coyote.AbstractProcessor.parseHost(), different
values appear. For Java 8:
[fe80:0:0:0:2e0:4cff:fee2:a74a%enx00e04ce2a74a]:PORTNUM
For Java 11+:
[fe80:0:0:0:2e0:4cff:fee2:a74a]:PORTNUM
The thread name in both cases is http-nio-
fe80:0:0:0:2e0:4cff:fee2:a74a%enx00e04ce2a74a-auto-3-exec-1.
The % character in the Java 8 case leads to a failure in parseHost
(german error strings):
I've tracked this down. This issue is this bug:
https://bugs.openjdk.org/browse/JDK-8027308
This has never been applied to Java 8:
https://github.com/openjdk/jdk11u/
commit/66135672e1fb2290834277d9c359478fff4ddd12
I've configured this test to be skipped on Java < 11.
Thanks for finding the root cause. Interesting, that they never fixed it
for Java 8, only 9+.
Best regards,
Rainer
Mark
[junit] 23-Jan-2026 14:09:43.062 INFORMATION [http-nio-
fe80:0:0:0:2e0:4cff:fee2:a74a%enx00e04ce2a74a-auto-3-exec-1]
org.apache.coyote.AbstractProcessor.parseHost Der Host
[[fe80:0:0:0:2e0:4cff:fee2:a74a%enx00e04ce2a74a]:38323] ist nicht gültig.
[junit] Info: Weitere Vorkommen von Fehlern beim Parsen der
Anfragen werden mit DEBUG Level ausgegeben
[junit] java.lang.IllegalArgumentException: Das Zeichen [%]
ist kein valides Zeichen in einer IPv6 Addresse.
[junit] at
org.apache.tomcat.util.http.parser.HttpParser.readHostIPv6(HttpParser.java:901)
Some there is some conversion involved for the more modern Java
versions, that prevent the failure.
Regards,
Rainer
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
