RFR: 8237834: com/sun/jndi/ldap/LdapDnsProviderTest.java failing with LDAP response read timeout
Daniel Fuchs
daniel.fuchs at oracle.com
Thu May 7 14:12:04 UTC 2020
Hi Aleksei,
If that solves the issue, then LGTM!
best regards,
-- daniel
On 07/05/2020 14:57, Aleks Efimov wrote:
> Hi Daniel,
>
> As a follow-up to our off-list discussion, I've updated the
> generateUnseenPort method:
> a) Added comment to be more specific about the number of times it is called
> b) The port range has been extended to [1111, PortConfig.getUpper())
>
> Webrev with new version:
> http://cr.openjdk.java.net/~aefimov/8237834/01
>
> With Best Regards,
> Aleksei
>
> On 07/05/2020 11:10, Daniel Fuchs wrote:
>> Hi Aleksei,
>>
>> I agree with the general idea. However, the method:
>>
>> 215 private static int generateUnseenPort() {
>> 216 int port;
>> 217 do {
>> 218 port = 1111 + RND.nextInt(1000);
>> 219 } while (SEEN_PORTS.contains(port));
>> 220 SEEN_PORTS.add(port);
>> 221 return port;
>> 222 }
>>
>> might never return - if none of the port in the range
>> [1111, 2111] are free. Arguably, the test will have
>> failed in timeout before that. But maybe an assertion
>> error should be thrown if that happens.
>>
>> best regards,
>>
>> -- daniel
>>
>>
>> On 07/05/2020 00:52, Aleks Efimov wrote:
>>> Hi,
>>>
>>> LdapDnsProviderTest test expects to have no services running on 1111
>>> port on test machine. That could cause it to fail intermittently due
>>> to unexpected exception message thrown by LDAP client, i.e. timeout
>>> or disconnect instead of expected connection refusal.
>>>
>>> The proposed fix tries to make test more stable ('intermittent' key
>>> was still added) by running test cases with non-default port multiple
>>> times by trying to use different random port numbers ('randomness'
>>> key was added) for each run.
>>>
>>> Webrev: http://cr.openjdk.java.net/~aefimov/8237834/00/index.html
>>>
>>> JBS: https://bugs.openjdk.java.net/browse/JDK-8237834
>>>
>>> Test was not observed to fail for over 300 iterations.
>>>
>>> With Best Regards,
>>> Aleksei
>>>
>>
>
More information about the core-libs-dev
mailing list