[
https://issues.apache.org/jira/browse/KUDU-2593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16634463#comment-16634463
]
Adar Dembo commented on KUDU-2593:
----------------------------------
Todd and I discussed a related issue in [this
gerrit|https://gerrit.cloudera.org/c/10224/2/build-support/parse_test_failure.py#99].
> Make LeakSanitizer errors in build-and-test.sh more incriminating
> -----------------------------------------------------------------
>
> Key: KUDU-2593
> URL: https://issues.apache.org/jira/browse/KUDU-2593
> Project: Kudu
> Issue Type: Bug
> Components: build, test
> Reporter: Andrew Wong
> Priority: Major
>
> Currently, if a run of build-and-test.sh yields a test that hits a memory
> leak without failing any tests, the result is that all the tests show up
> green on the test reporting server (if reporting is enabled), but the test
> XML gets a <failure> field added to it.
> To make matters worse, the XML that gets added uses the generic test case
> name "LeakSanitizer" and generic classname "LSAN" to describe the failure,
> making it non-obvious upon a quick glance at just the XML failures what has
> broken.
> A quick solution would be to include the test name with the injected XML
> failure, making it easier to determine the test culprit. A more complete
> solution might entail parsing the leak before reporting the test to the test
> reporting server so leaks get reported as failed runs (which echoes what
> Jenkins does today).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)