[jdk8u-dev] RFR: 8258855: Two tests sun/security/krb5/auto/ReplayCacheTestProc.java and ReplayCacheTestProcWithMD5.java failed on OL8.3

[Andrew John Hughes]

On Tue, 16 Apr 2024 18:33:12 GMT, Andrew John Hughes <andrew at openjdk.org> wrote:

> In testing `sun/security/krb5/auto/ReplayCacheTestProc.java` following [JDK-8168518](https://bugs.openjdk.org/browse/JDK-8168518), I found both it and `sun/security/krb5/auto/rcache_usemd5.sh` failed on RHEL 8.  This did not seem to be a regression caused by the 8u412 change as it failed on the current release, 8u402.  The failure was further confirmed by @martinuy who found it was due to the newer Kerberos version on RHEL 8, and it passed on Fedora 31.
> 
>  @jerboaa alerted me to [JDK-8258855](https://bugs.openjdk.org/browse/JDK-8258855) and backporting this change to 8u makes the tests pass.
> 
> The code changes apply as is. The headers have to be adjusted slightly to take into account the different DNS options used in 8u.  The change to `test/jdk/sun/security/krb5/auto/ReplayCacheTestProcWithMD5.java` also moves to `jdk/test/sun/security/krb5/auto/rcache_usemd5.sh` on 8u and has to be manually applied. I looked briefly at backporting [JDK-8180569](https://bugs.openjdk.org/browse/JDK-8180569) which refactors these Kerberos shell tests into Java tests, and I think it has merit, but is likely quite involved and requires other test library backports. For now, we can just backport this change as the shell script is different to the one removed by 8180569 with or without this patch applied first.

* Test failures are the Certigna certificate and a sporadic HotSpot GC test failure, `gc/logging/TestGCId.java`
* Build failures seem to be down to Buster inaccessibility for the cross-compile. With it going EOL at the end of June, we may need to backport [JDK-8314658](https://bugs.openjdk.org/browse/JDK-8314658) to upgrade to bullseye.

None of them relate to this change.

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

PR Comment: https://git.openjdk.org/jdk8u-dev/pull/481#issuecomment-2060222244