You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@sling.apache.org by Felix Meschberger <fm...@gmail.com> on 2010/09/09 14:53:57 UTC
integration tests
Hi all,
Maybe its just me, but currently the Sling (main) integration tests run
extremely unstable. Most of the time I only have one failure. But 1 out
of 4 runs, I have a random (but rather high) number of test failures.
I run "mvn -o clean install" in the launchpad/testing folder after a
full reactor build.
The single test that constantly fails is
ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
cannot find where it is created.
Do you observe the same ?
Thanks and Regards
Felix
Re: integration tests
Posted by Bertrand Delacretaz <bd...@apache.org>.
On Thu, Sep 9, 2010 at 12:53 PM, Felix Meschberger <fm...@gmail.com> wrote:
> ...Maybe its just me, but currently the Sling (main) integration tests run
> extremely unstable. Most of the time I only have one failure. But 1 out
> of 4 runs, I have a random (but rather high) number of test failures....
I haven't done much full builds recently, but that's usually a sign of
testing asynchronous things without caring about their async nature.
-Bertrand
Re: integration tests
Posted by Pontus Amberg <po...@milso.se>.
I was using Maven 3.0-beta-3 before and I had 3 different test failures
in a row today. After reading that you had better success with 2.2.1 I
downgraded to 2.2.1 and I have now built the current trunk 3 times with
no failures.
I have no idea if the 3 successful builds was just pure luck but I will
probably continue with 2.2.1 for a while.
/Pontus
On 09/10/2010 04:10 AM, Justin Edelson wrote:
> FWIW, which probably isn't much...
>
> On a clean Ubuntu 10.04 VM with all updates and OpenJDK 6b18 and Maven
> 2.2.1, I've had 5 clean integration test suite runs in a row.
>
> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>
>> Hi all,
>>
>> Maybe its just me, but currently the Sling (main) integration tests run
>> extremely unstable. Most of the time I only have one failure. But 1 out
>> of 4 runs, I have a random (but rather high) number of test failures.
>>
>> I run "mvn -o clean install" in the launchpad/testing folder after a
>> full reactor build.
>>
>> The single test that constantly fails is
>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>> cannot find where it is created.
>>
>> Do you observe the same ?
>>
>> Thanks and Regards
>> Felix
>>
>
>
Re: PermGen space problems Re: integration tests
Posted by Torgeir Veimo <to...@pobox.com>.
On 14 September 2010 18:56, Andreas Kuckartz <A....@ping.de> wrote:
> I am aware from another project that PermGen space problems can be
> really nasty...
>
> As a wild guess this might help:
>
> How to deal with “java.lang.OutOfMemoryError: PermGen space” error
> http://stackoverflow.com/questions/88235/how-to-deal-with-java-lang-outofmemoryerror-permgen-space-error
Jackrabbit leaves some timer threads running that prevents a webapp to
be fully GC'ed after shutdown. So even the fixes suggested in that
stackoverflow page aren't sufficient.
> Am 14.09.2010 10:14, schrieb Felix Meschberger:
>> It turns out, that my seemingly random failures might be related to Perm
>> Gen issues: After increasing the Perm Gen space (actually doubling it to
>> 512M) the full built ran through smoothly.
>>
>
--
-Tor
PermGen space problems Re: integration tests
Posted by Andreas Kuckartz <A....@ping.de>.
I am aware from another project that PermGen space problems can be
really nasty...
As a wild guess this might help:
How to deal with “java.lang.OutOfMemoryError: PermGen space” error
http://stackoverflow.com/questions/88235/how-to-deal-with-java-lang-outofmemoryerror-permgen-space-error
Cheers,
Andreas
---
Am 14.09.2010 10:14, schrieb Felix Meschberger:
> It turns out, that my seemingly random failures might be related to Perm
> Gen issues: After increasing the Perm Gen space (actually doubling it to
> 512M) the full built ran through smoothly.
>
Re: integration tests
Posted by Felix Meschberger <fm...@gmail.com>.
Hi,
It turns out, that my seemingly random failures might be related to Perm
Gen issues: After increasing the Perm Gen space (actually doubling it to
512M) the full built ran through smoothly.
Will try to find hints on why there are failures...
Regards
Felix
Am 10.09.2010 04:10, schrieb Justin Edelson:
> FWIW, which probably isn't much...
>
> On a clean Ubuntu 10.04 VM with all updates and OpenJDK 6b18 and Maven
> 2.2.1, I've had 5 clean integration test suite runs in a row.
>
> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>> Hi all,
>>
>> Maybe its just me, but currently the Sling (main) integration tests run
>> extremely unstable. Most of the time I only have one failure. But 1 out
>> of 4 runs, I have a random (but rather high) number of test failures.
>>
>> I run "mvn -o clean install" in the launchpad/testing folder after a
>> full reactor build.
>>
>> The single test that constantly fails is
>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>> cannot find where it is created.
>>
>> Do you observe the same ?
>>
>> Thanks and Regards
>> Felix
>
>
Re: integration tests
Posted by Justin Edelson <ju...@gmail.com>.
FWIW, which probably isn't much...
On a clean Ubuntu 10.04 VM with all updates and OpenJDK 6b18 and Maven
2.2.1, I've had 5 clean integration test suite runs in a row.
On 9/9/10 8:53 AM, Felix Meschberger wrote:
> Hi all,
>
> Maybe its just me, but currently the Sling (main) integration tests run
> extremely unstable. Most of the time I only have one failure. But 1 out
> of 4 runs, I have a random (but rather high) number of test failures.
>
> I run "mvn -o clean install" in the launchpad/testing folder after a
> full reactor build.
>
> The single test that constantly fails is
> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
> cannot find where it is created.
>
> Do you observe the same ?
>
> Thanks and Regards
> Felix
Re: integration tests
Posted by Justin Edelson <ju...@gmail.com>.
On 9/9/10 2:18 PM, Bertrand Delacretaz wrote:
> On Thu, Sep 9, 2010 at 6:06 PM, Justin Edelson <ju...@gmail.com> wrote:
>> On 9/9/10 2:04 PM, Bertrand Delacretaz wrote:
>>> ... IIRC mvn clean -> delete sling folder is configured (didn't check
>>> right now) but maybe doesn't work for full reactor build?
>>>
>> mvn clean will delete [root]/launchpad/testing/sling. It will NOT delete
>> [root]/sling. When you run a full reactor build, sling.home is set to
>> [root]/sling....
>
> Ah, that's the difference then, thanks for clarifying. I agree that
> $root/sling should be deleted by mvn clean.
>
> -Bertrand
It should, but that's probably not enough because you can do something like:
mvn clean install -rf launchpad/testing
and sling.home will be set to [root]/sling, yet the root clean won't be run.
Argh....
Re: integration tests
Posted by Bertrand Delacretaz <bd...@apache.org>.
On Thu, Sep 9, 2010 at 6:06 PM, Justin Edelson <ju...@gmail.com> wrote:
> On 9/9/10 2:04 PM, Bertrand Delacretaz wrote:
>>... IIRC mvn clean -> delete sling folder is configured (didn't check
>> right now) but maybe doesn't work for full reactor build?
>>
> mvn clean will delete [root]/launchpad/testing/sling. It will NOT delete
> [root]/sling. When you run a full reactor build, sling.home is set to
> [root]/sling....
Ah, that's the difference then, thanks for clarifying. I agree that
$root/sling should be deleted by mvn clean.
-Bertrand
Re: integration tests
Posted by Justin Edelson <ju...@gmail.com>.
On 9/9/10 2:04 PM, Bertrand Delacretaz wrote:
> On Thu, Sep 9, 2010 at 5:56 PM, Justin Edelson <ju...@gmail.com> wrote:
>> ...$ mvn clean install
>> $ mvn clean install
>>
>> consistently fails. This should be reasonably simple to resolve by
>> adding an execution of the clean plugin to the root pom...
>
> IIRC mvn clean -> delete sling folder is configured (didn't check
> right now) but maybe doesn't work for full reactor build?
>
> -Bertrand
mvn clean will delete [root]/launchpad/testing/sling. It will NOT delete
[root]/sling. When you run a full reactor build, sling.home is set to
[root]/sling.
Justin
Re: integration tests
Posted by Bertrand Delacretaz <bd...@apache.org>.
On Thu, Sep 9, 2010 at 5:56 PM, Justin Edelson <ju...@gmail.com> wrote:
> ...$ mvn clean install
> $ mvn clean install
>
> consistently fails. This should be reasonably simple to resolve by
> adding an execution of the clean plugin to the root pom...
IIRC mvn clean -> delete sling folder is configured (didn't check
right now) but maybe doesn't work for full reactor build?
-Bertrand
Re: integration tests
Posted by Justin Edelson <ju...@gmail.com>.
On 9/9/10 2:21 PM, Felix Meschberger wrote:
> Hi,
>
> On 09.09.2010 19:56, Justin Edelson wrote:
>> My full-reactor build failure problem is back and it seems to be related
>> to the state of the sling directory in the root of the source tree.
>>
>> $ rm -rf sling
>> $ mvn clean install
>>
>> consistently works
>
> Ok, I generally also remove the "root" sling folder.
>
> But then: can we do something about the launchpad/testing bundle to not
> use the reactor root as the base for sling.home ? IIRC we once had this.
In SLING-845, Bertrand added an ant tasks which deletes [root]/sling.
Not sure why that's not doing the trick.
Justin
>
> Regards
> Felix
>
>>
>> but
>>
>> $ mvn clean install
>> $ mvn clean install
>>
>> consistently fails. This should be reasonably simple to resolve by
>> adding an execution of the clean plugin to the root pom.
>>
>> The node in ContentLoaderMiscPropertyTest is created by the content
>> loader. See
>> launchpad/test-services/src/main/resources/SLING-INF/content/property-types-test.json
>>
>> Which assertion is failing for you?
>>
>> Justin
>>
>> On 9/9/10 10:05 AM, Felix Meschberger wrote:
>>> Hi,
>>>
>>> Addendum: I am running on Linux with either Java 5 or Java 6.
>>>
>>> Also: The ContentLoaderMiscPropertyTest test on a virgin Sling instance
>>> (e.g. by running as a JUnit test from within Eclipse) will always fail.
>>> So I suspect the test to expect some left-over file from another test.
>>>
>>> Regards
>>> Felix
>>>
>>> On 09.09.2010 15:55, Justin Edelson wrote:
>>>> I _was_ seeing this until about a week ago, but it was slightly
>>>> different... a full reactor build would fail, but mvn clean install in
>>>> launchpad/testing would always pass.
>>>>
>>>> Since about a week ago (sorry, don't know the revision), both have
>>>> passed pretty consistently.
>>>>
>>>> Running a build now...
>>>>
>>>> Justin
>>>>
>>>> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>>>>> Hi all,
>>>>>
>>>>> Maybe its just me, but currently the Sling (main) integration tests run
>>>>> extremely unstable. Most of the time I only have one failure. But 1 out
>>>>> of 4 runs, I have a random (but rather high) number of test failures.
>>>>>
>>>>> I run "mvn -o clean install" in the launchpad/testing folder after a
>>>>> full reactor build.
>>>>>
>>>>> The single test that constantly fails is
>>>>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>>>>> cannot find where it is created.
>>>>>
>>>>> Do you observe the same ?
>>>>>
>>>>> Thanks and Regards
>>>>> Felix
>>>>
>>>>
>>
>>
Re: integration tests
Posted by Felix Meschberger <fm...@gmail.com>.
Hi,
On 09.09.2010 19:56, Justin Edelson wrote:
> My full-reactor build failure problem is back and it seems to be related
> to the state of the sling directory in the root of the source tree.
>
> $ rm -rf sling
> $ mvn clean install
>
> consistently works
Ok, I generally also remove the "root" sling folder.
But then: can we do something about the launchpad/testing bundle to not
use the reactor root as the base for sling.home ? IIRC we once had this.
Regards
Felix
>
> but
>
> $ mvn clean install
> $ mvn clean install
>
> consistently fails. This should be reasonably simple to resolve by
> adding an execution of the clean plugin to the root pom.
>
> The node in ContentLoaderMiscPropertyTest is created by the content
> loader. See
> launchpad/test-services/src/main/resources/SLING-INF/content/property-types-test.json
>
> Which assertion is failing for you?
>
> Justin
>
> On 9/9/10 10:05 AM, Felix Meschberger wrote:
>> Hi,
>>
>> Addendum: I am running on Linux with either Java 5 or Java 6.
>>
>> Also: The ContentLoaderMiscPropertyTest test on a virgin Sling instance
>> (e.g. by running as a JUnit test from within Eclipse) will always fail.
>> So I suspect the test to expect some left-over file from another test.
>>
>> Regards
>> Felix
>>
>> On 09.09.2010 15:55, Justin Edelson wrote:
>>> I _was_ seeing this until about a week ago, but it was slightly
>>> different... a full reactor build would fail, but mvn clean install in
>>> launchpad/testing would always pass.
>>>
>>> Since about a week ago (sorry, don't know the revision), both have
>>> passed pretty consistently.
>>>
>>> Running a build now...
>>>
>>> Justin
>>>
>>> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>>>> Hi all,
>>>>
>>>> Maybe its just me, but currently the Sling (main) integration tests run
>>>> extremely unstable. Most of the time I only have one failure. But 1 out
>>>> of 4 runs, I have a random (but rather high) number of test failures.
>>>>
>>>> I run "mvn -o clean install" in the launchpad/testing folder after a
>>>> full reactor build.
>>>>
>>>> The single test that constantly fails is
>>>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>>>> cannot find where it is created.
>>>>
>>>> Do you observe the same ?
>>>>
>>>> Thanks and Regards
>>>> Felix
>>>
>>>
>
>
Re: integration tests
Posted by Justin Edelson <ju...@gmail.com>.
My full-reactor build failure problem is back and it seems to be related
to the state of the sling directory in the root of the source tree.
$ rm -rf sling
$ mvn clean install
consistently works
but
$ mvn clean install
$ mvn clean install
consistently fails. This should be reasonably simple to resolve by
adding an execution of the clean plugin to the root pom.
The node in ContentLoaderMiscPropertyTest is created by the content
loader. See
launchpad/test-services/src/main/resources/SLING-INF/content/property-types-test.json
Which assertion is failing for you?
Justin
On 9/9/10 10:05 AM, Felix Meschberger wrote:
> Hi,
>
> Addendum: I am running on Linux with either Java 5 or Java 6.
>
> Also: The ContentLoaderMiscPropertyTest test on a virgin Sling instance
> (e.g. by running as a JUnit test from within Eclipse) will always fail.
> So I suspect the test to expect some left-over file from another test.
>
> Regards
> Felix
>
> On 09.09.2010 15:55, Justin Edelson wrote:
>> I _was_ seeing this until about a week ago, but it was slightly
>> different... a full reactor build would fail, but mvn clean install in
>> launchpad/testing would always pass.
>>
>> Since about a week ago (sorry, don't know the revision), both have
>> passed pretty consistently.
>>
>> Running a build now...
>>
>> Justin
>>
>> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>>> Hi all,
>>>
>>> Maybe its just me, but currently the Sling (main) integration tests run
>>> extremely unstable. Most of the time I only have one failure. But 1 out
>>> of 4 runs, I have a random (but rather high) number of test failures.
>>>
>>> I run "mvn -o clean install" in the launchpad/testing folder after a
>>> full reactor build.
>>>
>>> The single test that constantly fails is
>>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>>> cannot find where it is created.
>>>
>>> Do you observe the same ?
>>>
>>> Thanks and Regards
>>> Felix
>>
>>
Re: integration tests
Posted by Pontus Amberg <po...@comhem.se>.
I'm also running Linux with Java 6 and I also encounter random test
failures. The tests that I think fails most frequently for me is
testCanAddUser(org.apache.sling.launchpad.webapp.integrationtest.userManager.PrivilegesInfoTest)
testCanAddGroup(org.apache.sling.launchpad.webapp.integrationtest.userManager.PrivilegesInfoTest)
testCanUpdateUserProperties(org.apache.sling.launchpad.webapp.integrationtest.userManager.PrivilegesInfoTest)
testCanUpdateGroupProperties(org.apache.sling.launchpad.webapp.integrationtest.userManager.PrivilegesInfoTest)
testCanRemoveUser(org.apache.sling.launchpad.webapp.integrationtest.userManager.PrivilegesInfoTest)
testCanRemoveGroup(org.apache.sling.launchpad.webapp.integrationtest.userManager.PrivilegesInfoTest)
/Pontus
On 2010-09-09 16:05, Felix Meschberger wrote:
> Hi,
>
> Addendum: I am running on Linux with either Java 5 or Java 6.
>
> Also: The ContentLoaderMiscPropertyTest test on a virgin Sling instance
> (e.g. by running as a JUnit test from within Eclipse) will always fail.
> So I suspect the test to expect some left-over file from another test.
>
> Regards
> Felix
>
> On 09.09.2010 15:55, Justin Edelson wrote:
>
>> I _was_ seeing this until about a week ago, but it was slightly
>> different... a full reactor build would fail, but mvn clean install in
>> launchpad/testing would always pass.
>>
>> Since about a week ago (sorry, don't know the revision), both have
>> passed pretty consistently.
>>
>> Running a build now...
>>
>> Justin
>>
>> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>>
>>> Hi all,
>>>
>>> Maybe its just me, but currently the Sling (main) integration tests run
>>> extremely unstable. Most of the time I only have one failure. But 1 out
>>> of 4 runs, I have a random (but rather high) number of test failures.
>>>
>>> I run "mvn -o clean install" in the launchpad/testing folder after a
>>> full reactor build.
>>>
>>> The single test that constantly fails is
>>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>>> cannot find where it is created.
>>>
>>> Do you observe the same ?
>>>
>>> Thanks and Regards
>>> Felix
>>>
>>
>>
>
Re: integration tests
Posted by Felix Meschberger <fm...@gmail.com>.
Hi,
Addendum: I am running on Linux with either Java 5 or Java 6.
Also: The ContentLoaderMiscPropertyTest test on a virgin Sling instance
(e.g. by running as a JUnit test from within Eclipse) will always fail.
So I suspect the test to expect some left-over file from another test.
Regards
Felix
On 09.09.2010 15:55, Justin Edelson wrote:
> I _was_ seeing this until about a week ago, but it was slightly
> different... a full reactor build would fail, but mvn clean install in
> launchpad/testing would always pass.
>
> Since about a week ago (sorry, don't know the revision), both have
> passed pretty consistently.
>
> Running a build now...
>
> Justin
>
> On 9/9/10 8:53 AM, Felix Meschberger wrote:
>> Hi all,
>>
>> Maybe its just me, but currently the Sling (main) integration tests run
>> extremely unstable. Most of the time I only have one failure. But 1 out
>> of 4 runs, I have a random (but rather high) number of test failures.
>>
>> I run "mvn -o clean install" in the launchpad/testing folder after a
>> full reactor build.
>>
>> The single test that constantly fails is
>> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
>> cannot find where it is created.
>>
>> Do you observe the same ?
>>
>> Thanks and Regards
>> Felix
>
>
Re: integration tests
Posted by Justin Edelson <ju...@gmail.com>.
I _was_ seeing this until about a week ago, but it was slightly
different... a full reactor build would fail, but mvn clean install in
launchpad/testing would always pass.
Since about a week ago (sorry, don't know the revision), both have
passed pretty consistently.
Running a build now...
Justin
On 9/9/10 8:53 AM, Felix Meschberger wrote:
> Hi all,
>
> Maybe its just me, but currently the Sling (main) integration tests run
> extremely unstable. Most of the time I only have one failure. But 1 out
> of 4 runs, I have a random (but rather high) number of test failures.
>
> I run "mvn -o clean install" in the launchpad/testing folder after a
> full reactor build.
>
> The single test that constantly fails is
> ContentLoaderMiscPropertyTest.testLoaded. This expects a node, which I
> cannot find where it is created.
>
> Do you observe the same ?
>
> Thanks and Regards
> Felix