[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396766#comment-16396766
]
ASF subversion and git services commented on LUCENE-8201:
-
Commit 4645553a55d574c760fe637418c91229db64f2bb in lucene-solr's branch
refs/heads/branch_7x from [~dawid.weiss]
[ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=4645553 ]
LUCENE-8201: improve the test a bit to try to diagnose timeouts on jenkins.
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396767#comment-16396767
]
ASF subversion and git services commented on LUCENE-8201:
-
Commit cc3c4b61237619798a006217223fb6e78a688a4e in lucene-solr's branch
refs/heads/master from [~dawid.weiss]
[ https://git-wip-us.apache.org/repos/asf?p=lucene-solr.git;h=cc3c4b6 ]
LUCENE-8201: improve the test a bit to try to diagnose timeouts on jenkins.
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396747#comment-16396747
]
Dawid Weiss commented on LUCENE-8201:
-
Oh, that's better. I'll correct. ;)
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396740#comment-16396740
]
Uwe Schindler commented on LUCENE-8201:
---
bq. I know, but wanted to make sure that you won't incinerate me for not
wrapping those calls with access controller.
Hihi, I have a better nitpick:
{{Arrays.stream(t.getStackTrace()).map(Object::toString)}}
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396736#comment-16396736
]
Dawid Weiss commented on LUCENE-8201:
-
bq. Why did it fail at all?
We won't know because the subprocess doesn't print anything, it just got stuck
there for more than the timeout says. Looks longer than I/O delay, even on a
heavily loaded system... I improved the diagnostics a bit, let's see if we can
catch it.
bq. Security Manager should not be ne enabled, it's just a stupid "java -cp
..." subprocess.
I know, but wanted to make sure that you won't incinerate me for not wrapping
those calls with access controller. ;)
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396733#comment-16396733
]
Uwe Schindler commented on LUCENE-8201:
---
The question is indeed: Why did it fail at all? Because that should be tested
here :-) So we have to carefully check the codec bootup that SPI dependencies
do not block anywhere.
Dawid: Security Manager should not be ne enabled, it's just a stupid "java -cp
..." subprocess.
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396698#comment-16396698
]
Dawid Weiss commented on LUCENE-8201:
-
The patch should read 30 seconds in:
{code}
+ private static int MAX_TIME_SECONDS = 3;
{code}
I changed it locally for verifying that stack traces are dumped.
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396696#comment-16396696
]
Dawid Weiss commented on LUCENE-8201:
-
I've changed the test implementation slightly so that the timeout is ran within
the forked subprocess; we can then dump stack traces to see where they got
stuck.
I couldn't reproduce the failure though (beasting a number of times). Let's see
if it fails again and what it says. [~thetaphi] I don't think we need to bother
with security manager (privileged methods) since the subprocess was never
forked with a security manager anyway?
I will commit this in if there are no objections.
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Dawid Weiss
>Priority: Major
> Attachments: LUCENE-8201.patch
>
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
> Note:
> {code}
> git checkout -f d77f618a672d719b971684c55b9fa07d0ba59aaf
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16396646#comment-16396646
]
Dawid Weiss commented on LUCENE-8201:
-
Let me reopen this though, because it still did fail for some reason (and I
think we could improve the reporting here a bit).
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Assignee: Steve Rowe
>Priority: Major
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
[jira] [Commented] (LUCENE-8201) TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
[
https://issues.apache.org/jira/browse/LUCENE-8201?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16395835#comment-16395835
]
Dawid Weiss commented on LUCENE-8201:
-
This "reproduce with" line is printed with a test event listener, it's not
really part of randomized testing. This particular test has this in its header:
{code}
/* WARNING: This test does *not* extend LuceneTestCase to prevent static class
* initialization when spawned as subprocess (and please let default codecs
alive)! */
@RunWith(RandomizedRunner.class)
public class TestCodecLoadingDeadlock extends Assert {
{code}
Note it extends randomized runner, but does not extend the base test class from
Lucene (for a reason); that's why you can't see the "repro" line. The seed is
printed as part of the stack trace (and this is part of the randomizedtesting
framework), so if you use it, the failure should be reproducible.
{code}
ant -Dtests.seed=88B4FC32922379
{code}
> TestCodecLoadingDeadlock.testDeadlock failure has no "reproduce with" line
> --
>
> Key: LUCENE-8201
> URL: https://issues.apache.org/jira/browse/LUCENE-8201
> Project: Lucene - Core
> Issue Type: Test
>Reporter: Steve Rowe
>Priority: Major
>
> Is it expected that there are test situations where a "reproduce with" line
> is not printed? ({{reproduceJenkinsFailures.py}} assumes that all failures
> produce such a line.)
> Here's one from
> [https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.x/172/]:
> {noformat}
>[smoker][junit4] Suite:
> org.apache.lucene.codecs.TestCodecLoadingDeadlock
>[smoker][junit4] FAILURE 30.4s J0 |
> TestCodecLoadingDeadlock.testDeadlock <<<
>[smoker][junit4]> Throwable #1: java.lang.AssertionError: Process
> did not exit after 30 secs -> classloader deadlock?
>[smoker][junit4]> at
> __randomizedtesting.SeedInfo.seed([88B4FC32922379:DE355E834C88EAF]:0)
>[smoker][junit4]> at
> org.apache.lucene.codecs.TestCodecLoadingDeadlock.testDeadlock(TestCodecLoadingDeadlock.java:75)
>[smoker][junit4] Completed [132/466 (1!)] on J0 in 30.45s, 1 test, 1
> failure <<< FAILURES!
> {noformat}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
-
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
