[jira] [Assigned] (KUDU-1624) Possible data races in util/trace

["Todd Lipcon (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/KUDU-1624?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Todd Lipcon reassigned KUDU-1624:
---------------------------------

    Assignee: Todd Lipcon

> Possible data races in util/trace
> ---------------------------------
>
>                 Key: KUDU-1624
>                 URL: https://issues.apache.org/jira/browse/KUDU-1624
>             Project: Kudu
>          Issue Type: Bug
>          Components: util
>    Affects Versions: 1.0.0
>            Reporter: Adar Dembo
>            Assignee: Todd Lipcon
>         Attachments: rpc_stub-test.txt, rpc_stub-test.txt.gz
>
>
> I'm running clang 3.9 against a Kudu source tree using libcpp, so it's possible this is partly due to differences in how libstdc++ implements std::vector. Or maybe it's real.
> At first glance, it looks like the problem is that Trace::MetricsToJSON(JsonWriter*) doesn't hold lock_ while accessing child_traces_. It would need to some refactoring to avoid taking lock_ recursively, though, since it's a recursive function.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)