[ 
https://issues.apache.org/jira/browse/GEODE-5591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16599036#comment-16599036
 ] 

ASF subversion and git services commented on GEODE-5591:
--------------------------------------------------------

Commit b6eb8faa855fdb44c3b30bad2ad47b3d3371413b in geode's branch 
refs/heads/develop from [~smgoller]
[ https://gitbox.apache.org/repos/asf?p=geode.git;h=b6eb8fa ]

[GEODE-5591] If there is a BindException we should always throw. (#2351)

* Replace start() logic with something more deterministic and fix test.
* Iterate over possible ports instead of picking random ones.
* Check for an exception of the appropriate type and replace old
  assertion logic.
* Refactor testing an exception was thrown
* Fix expected message to match exception.

Co-Authored-By: Patrick Rhomberg <prhomb...@pivotal.io>
Co-Authored-By: Sean Goller <sgol...@pivotal.io>


> GatewayReceiverImpl.start() retry logic is dependent on JDK's underlying C 
> runtime
> ----------------------------------------------------------------------------------
>
>                 Key: GEODE-5591
>                 URL: https://issues.apache.org/jira/browse/GEODE-5591
>             Project: Geode
>          Issue Type: Bug
>          Components: wan
>            Reporter: Sean Goller
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 1h 40m
>  Remaining Estimate: 0h
>
> {{org/apache/geode/internal/cache/wan/GatewayReceiverImpl.java:66}} denotes a 
> string comparison that relies on the specific C runtime that the JDK is run 
> on top of. This code will misbehave on Alpine Linux (A popular linux 
> distribution for containers) because it is based on the musl libc, not the 
> GNU libc. The intent of the code is to throw an exception if the receiver 
> cannot bind to the requested address. This will not occur if the JDK is 
> compiled against musl libc. As an aside, the entire retry logic seems 
> suspicious as it only retries in situations that are static and should not 
> change, so the reason for the code should be re-evaluated and made more 
> robust and not dependent on native os information.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to