buildbot failure in on ofbiz-trunk

[buildbot]

The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
    https://ci.apache.org/builders/ofbiz-trunk/builds/1861

Buildbot URL: https://ci.apache.org/

Buildslave for this Build: silvanus_ubuntu

Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build
Build Source Stamp: [branch ofbiz/trunk] 1780133
Blamelist: shijh

BUILD FAILED: failed shell_1

Sincerely,
 -The Buildbot

[Shi Jinghai-3]

Hmm, I read the log and found some errors. I don't know how to resolve these problems.

This error is the first one and I guess it causes other errors:
Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
java.net.BindException: Address already in use

I need Jacques's help ... HELP ...

-----邮件原件-----
发件人: [hidden email] [mailto:[hidden email]]
发送时间: 2017年1月25日 11:18
收件人: [hidden email]
主题: buildbot failure in on ofbiz-trunk

The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
    https://ci.apache.org/builders/ofbiz-trunk/builds/1861

Buildbot URL: https://ci.apache.org/

Buildslave for this Build: silvanus_ubuntu

Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
Blamelist: shijh

BUILD FAILED: failed shell_1

Sincerely,
 -The Buildbot

[Jacques Le Roux]

Hi Jinghai,

You are welcome, I'll have a look ASAP (not right now, I must move)

Cheers

Jacques

PS: No worries, it works here (locally on Win7) so it's a Builbot hiccup, I'll check that



Le 25/01/2017 à 04:59, Shi Jinghai a écrit :

> Hmm, I read the log and found some errors. I don't know how to resolve these problems.
>
> This error is the first one and I guess it causes other errors:
> Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
> SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
> java.net.BindException: Address already in use
>
> I need Jacques's help ... HELP ...
>
> -----邮件原件-----
> 发件人: [hidden email] [mailto:[hidden email]]
> 发送时间: 2017年1月25日 11:18
> 收件人: [hidden email]
> 主题: buildbot failure in on ofbiz-trunk
>
> The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
>      https://ci.apache.org/builders/ofbiz-trunk/builds/1861
>
> Buildbot URL: https://ci.apache.org/
>
> Buildslave for this Build: silvanus_ubuntu
>
> Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
> Blamelist: shijh
>
> BUILD FAILED: failed shell_1
>
> Sincerely,
>   -The Buildbot
>
>
>
>
>

[Jacques Le Roux]

Mmm, there is definitely an embarrassment with Buildbot after r1780133 https://ci.apache.org/builders/ofbiz-trunk

I though don't see how it could be related with your changes and as it works perfectly locally (even on Win7!) I guess it's rather on Buildbot.

I'll check and see with infra if I can't find it by myself...

Jacques


Le 25/01/2017 à 07:55, Jacques Le Roux a écrit :

> Hi Jinghai,
>
> You are welcome, I'll have a look ASAP (not right now, I must move)
>
> Cheers
>
> Jacques
>
> PS: No worries, it works here (locally on Win7) so it's a Builbot hiccup, I'll check that
>
>
>
> Le 25/01/2017 à 04:59, Shi Jinghai a écrit :
>> Hmm, I read the log and found some errors. I don't know how to resolve these problems.
>>
>> This error is the first one and I guess it causes other errors:
>> Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
>> SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
>> java.net.BindException: Address already in use
>>
>> I need Jacques's help ... HELP ...
>>
>> -----邮件原件-----
>> 发件人: [hidden email] [mailto:[hidden email]]
>> 发送时间: 2017年1月25日 11:18
>> 收件人: [hidden email]
>> 主题: buildbot failure in on ofbiz-trunk
>>
>> The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
>>      https://ci.apache.org/builders/ofbiz-trunk/builds/1861
>>
>> Buildbot URL: https://ci.apache.org/
>>
>> Buildslave for this Build: silvanus_ubuntu
>>
>> Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
>> Blamelist: shijh
>>
>> BUILD FAILED: failed shell_1
>>
>> Sincerely,
>>   -The Buildbot
>>
>>
>>
>>
>>
>
>

[Jacques Le Roux]

Tests pass on Ubuntu locally. Buildbot runs also on Ubuntu (though newer version).

So I'm now pretty sure it's a setting in Buildbot which has changed. I'll see with infra. Please devs until it's fixed refer to your local builds

Jacques


Le 25/01/2017 à 09:06, Jacques Le Roux a écrit :

> Mmm, there is definitely an embarrassment with Buildbot after r1780133 https://ci.apache.org/builders/ofbiz-trunk
>
> I though don't see how it could be related with your changes and as it works perfectly locally (even on Win7!) I guess it's rather on Buildbot.
>
> I'll check and see with infra if I can't find it by myself...
>
> Jacques
>
>
> Le 25/01/2017 à 07:55, Jacques Le Roux a écrit :
>> Hi Jinghai,
>>
>> You are welcome, I'll have a look ASAP (not right now, I must move)
>>
>> Cheers
>>
>> Jacques
>>
>> PS: No worries, it works here (locally on Win7) so it's a Builbot hiccup, I'll check that
>>
>>
>>
>> Le 25/01/2017 à 04:59, Shi Jinghai a écrit :
>>> Hmm, I read the log and found some errors. I don't know how to resolve these problems.
>>>
>>> This error is the first one and I guess it causes other errors:
>>> Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
>>> SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
>>> java.net.BindException: Address already in use
>>>
>>> I need Jacques's help ... HELP ...
>>>
>>> -----邮件原件-----
>>> 发件人: [hidden email] [mailto:[hidden email]]
>>> 发送时间: 2017年1月25日 11:18
>>> 收件人: [hidden email]
>>> 主题: buildbot failure in on ofbiz-trunk
>>>
>>> The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
>>>      https://ci.apache.org/builders/ofbiz-trunk/builds/1861
>>>
>>> Buildbot URL: https://ci.apache.org/
>>>
>>> Buildslave for this Build: silvanus_ubuntu
>>>
>>> Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
>>> Blamelist: shijh
>>>
>>> BUILD FAILED: failed shell_1
>>>
>>> Sincerely,
>>>   -The Buildbot
>>>
>>>
>>>
>>>
>>>
>>
>>
>
>

[Jacques Le Roux]

https://issues.apache.org/jira/browse/INFRA-13402

Jacques


Le 25/01/2017 à 10:13, Jacques Le Roux a écrit :

> Tests pass on Ubuntu locally. Buildbot runs also on Ubuntu (though newer version).
>
> So I'm now pretty sure it's a setting in Buildbot which has changed. I'll see with infra. Please devs until it's fixed refer to your local builds
>
> Jacques
>
>
> Le 25/01/2017 à 09:06, Jacques Le Roux a écrit :
>> Mmm, there is definitely an embarrassment with Buildbot after r1780133 https://ci.apache.org/builders/ofbiz-trunk
>>
>> I though don't see how it could be related with your changes and as it works perfectly locally (even on Win7!) I guess it's rather on Buildbot.
>>
>> I'll check and see with infra if I can't find it by myself...
>>
>> Jacques
>>
>>
>> Le 25/01/2017 à 07:55, Jacques Le Roux a écrit :
>>> Hi Jinghai,
>>>
>>> You are welcome, I'll have a look ASAP (not right now, I must move)
>>>
>>> Cheers
>>>
>>> Jacques
>>>
>>> PS: No worries, it works here (locally on Win7) so it's a Builbot hiccup, I'll check that
>>>
>>>
>>>
>>> Le 25/01/2017 à 04:59, Shi Jinghai a écrit :
>>>> Hmm, I read the log and found some errors. I don't know how to resolve these problems.
>>>>
>>>> This error is the first one and I guess it causes other errors:
>>>> Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
>>>> SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
>>>> java.net.BindException: Address already in use
>>>>
>>>> I need Jacques's help ... HELP ...
>>>>
>>>> -----邮件原件-----
>>>> 发件人: [hidden email] [mailto:[hidden email]]
>>>> 发送时间: 2017年1月25日 11:18
>>>> 收件人: [hidden email]
>>>> 主题: buildbot failure in on ofbiz-trunk
>>>>
>>>> The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
>>>>      https://ci.apache.org/builders/ofbiz-trunk/builds/1861
>>>>
>>>> Buildbot URL: https://ci.apache.org/
>>>>
>>>> Buildslave for this Build: silvanus_ubuntu
>>>>
>>>> Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
>>>> Blamelist: shijh
>>>>
>>>> BUILD FAILED: failed shell_1
>>>>
>>>> Sincerely,
>>>>   -The Buildbot
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>
>

[Jacques Le Roux]

I don't know why my last message did not reach the ML.

Anyway, I have fixed the issue, and made some cleaning, in the Buildbot config script.

But we have an issue "with Gradle" when running test

     java -jar build/libs/ofbiz.jar -t --portoffset 1

while loading data is OK

     java -jar build/libs/ofbiz.jar -l --portoffset 1

These is one of the 2 regressions I experienced so far. The other being unable to run OFBiz inside Eclipse (in Debug mode or not).

We already spoke about Eclipse. And if I regret a bit to no longer being able to run a build in Eclipse with 1 click and then, also with 1 click, run
OFBiz in Debug mode or not, other good aspects largely overweight this small drawbacks (sorry I could not resist to this rant)

But in the case of running integration tests with a portoffset we have an issue which is blocking BuildBot now for trunk and

I'll try to have a look today, but of course I'd be very glad if someone (hint Taher ;)) beats me on it

Thanks

Jacques


Le 26/01/2017 à 06:58, Jacques Le Roux a écrit :

> https://issues.apache.org/jira/browse/INFRA-13402
>
> Jacques
>
>
> Le 25/01/2017 à 10:13, Jacques Le Roux a écrit :
>> Tests pass on Ubuntu locally. Buildbot runs also on Ubuntu (though newer version).
>>
>> So I'm now pretty sure it's a setting in Buildbot which has changed. I'll see with infra. Please devs until it's fixed refer to your local builds
>>
>> Jacques
>>
>>
>> Le 25/01/2017 à 09:06, Jacques Le Roux a écrit :
>>> Mmm, there is definitely an embarrassment with Buildbot after r1780133 https://ci.apache.org/builders/ofbiz-trunk
>>>
>>> I though don't see how it could be related with your changes and as it works perfectly locally (even on Win7!) I guess it's rather on Buildbot.
>>>
>>> I'll check and see with infra if I can't find it by myself...
>>>
>>> Jacques
>>>
>>>
>>> Le 25/01/2017 à 07:55, Jacques Le Roux a écrit :
>>>> Hi Jinghai,
>>>>
>>>> You are welcome, I'll have a look ASAP (not right now, I must move)
>>>>
>>>> Cheers
>>>>
>>>> Jacques
>>>>
>>>> PS: No worries, it works here (locally on Win7) so it's a Builbot hiccup, I'll check that
>>>>
>>>>
>>>>
>>>> Le 25/01/2017 à 04:59, Shi Jinghai a écrit :
>>>>> Hmm, I read the log and found some errors. I don't know how to resolve these problems.
>>>>>
>>>>> This error is the first one and I guess it causes other errors:
>>>>> Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
>>>>> SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
>>>>> java.net.BindException: Address already in use
>>>>>
>>>>> I need Jacques's help ... HELP ...
>>>>>
>>>>> -----邮件原件-----
>>>>> 发件人: [hidden email] [mailto:[hidden email]]
>>>>> 发送时间: 2017年1月25日 11:18
>>>>> 收件人: [hidden email]
>>>>> 主题: buildbot failure in on ofbiz-trunk
>>>>>
>>>>> The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
>>>>> https://ci.apache.org/builders/ofbiz-trunk/builds/1861
>>>>>
>>>>> Buildbot URL: https://ci.apache.org/
>>>>>
>>>>> Buildslave for this Build: silvanus_ubuntu
>>>>>
>>>>> Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
>>>>> Blamelist: shijh
>>>>>
>>>>> BUILD FAILED: failed shell_1
>>>>>
>>>>> Sincerely,
>>>>>   -The Buildbot
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>>
>

[Jacques Le Roux]

Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
> But in the case of running integration tests with a portoffset we have an issue which is blocking BuildBot now for trunk and
R16

Jacques

[Jacques Le Roux]

Ah, I forgot to say, that I have dropped BuildBot support for R12 and R13.

Also that the pre-Gradle versions don't need portoffset


Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>> But in the case of running integration tests with a portoffset we have an issue which is blocking BuildBot now for trunk and
> R16
>
> Jacques
>
>

[taher]

Why do we need to run the tests on portoffset?

On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
[hidden email]> wrote:

> Ah, I forgot to say, that I have dropped BuildBot support for R12 and R13.
>
> Also that the pre-Gradle versions don't need portoffset
>
>
>
> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>
>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>
>>> But in the case of running integration tests with a portoffset we have
>>> an issue which is blocking BuildBot now for trunk and
>>>
>> R16
>>
>> Jacques
>>
>>
>>
>

[Jacques Le Roux]

I explained in this thread at http://markmail.org/message/63ik3ofjuztdjrcf

Jacques


Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :

> Why do we need to run the tests on portoffset?
>
> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
> [hidden email]> wrote:
>
>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and R13.
>>
>> Also that the pre-Gradle versions don't need portoffset
>>
>>
>>
>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>
>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>
>>>> But in the case of running integration tests with a portoffset we have
>>>> an issue which is blocking BuildBot now for trunk and
>>>>
>>> R16
>>>
>>> Jacques
>>>
>>>
>>>

[Jacques Le Roux]

OK I thought about this. It seems we have 2 issues:

1) when using portoffset with testIntegration (on Builbot or locally), on error OFBiz does not report to runtime\logs\test-results\html eg:
https://ci.apache.org/projects/ofbiz/logs/trunk/

2) We need a patch or another mechanism to test with portoffset. Because of XMLRPC and SOAP tests, see port 8080 harcoded in XmlRpcTests class,
serviceengine.xml and testRemoteSoap service definition

To be continued later...

Jacques

Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :

> I don't know why my last message did not reach the ML.
>
> Anyway, I have fixed the issue, and made some cleaning, in the Buildbot config script.
>
> But we have an issue "with Gradle" when running test
>
>     java -jar build/libs/ofbiz.jar -t --portoffset 1
>
> while loading data is OK
>
>     java -jar build/libs/ofbiz.jar -l --portoffset 1
>
> These is one of the 2 regressions I experienced so far. The other being unable to run OFBiz inside Eclipse (in Debug mode or not).
>
> We already spoke about Eclipse. And if I regret a bit to no longer being able to run a build in Eclipse with 1 click and then, also with 1 click,
> run OFBiz in Debug mode or not, other good aspects largely overweight this small drawbacks (sorry I could not resist to this rant)
>
> But in the case of running integration tests with a portoffset we have an issue which is blocking BuildBot now for trunk and
>
> I'll try to have a look today, but of course I'd be very glad if someone (hint Taher ;)) beats me on it
>
> Thanks
>
> Jacques
>
>
> Le 26/01/2017 à 06:58, Jacques Le Roux a écrit :
>> https://issues.apache.org/jira/browse/INFRA-13402
>>
>> Jacques
>>
>>
>> Le 25/01/2017 à 10:13, Jacques Le Roux a écrit :
>>> Tests pass on Ubuntu locally. Buildbot runs also on Ubuntu (though newer version).
>>>
>>> So I'm now pretty sure it's a setting in Buildbot which has changed. I'll see with infra. Please devs until it's fixed refer to your local builds
>>>
>>> Jacques
>>>
>>>
>>> Le 25/01/2017 à 09:06, Jacques Le Roux a écrit :
>>>> Mmm, there is definitely an embarrassment with Buildbot after r1780133 https://ci.apache.org/builders/ofbiz-trunk
>>>>
>>>> I though don't see how it could be related with your changes and as it works perfectly locally (even on Win7!) I guess it's rather on Buildbot.
>>>>
>>>> I'll check and see with infra if I can't find it by myself...
>>>>
>>>> Jacques
>>>>
>>>>
>>>> Le 25/01/2017 à 07:55, Jacques Le Roux a écrit :
>>>>> Hi Jinghai,
>>>>>
>>>>> You are welcome, I'll have a look ASAP (not right now, I must move)
>>>>>
>>>>> Cheers
>>>>>
>>>>> Jacques
>>>>>
>>>>> PS: No worries, it works here (locally on Win7) so it's a Builbot hiccup, I'll check that
>>>>>
>>>>>
>>>>>
>>>>> Le 25/01/2017 à 04:59, Shi Jinghai a écrit :
>>>>>> Hmm, I read the log and found some errors. I don't know how to resolve these problems.
>>>>>>
>>>>>> This error is the first one and I guess it causes other errors:
>>>>>> Jan 25, 2017 3:14:06 AM org.apache.coyote.AbstractProtocol init
>>>>>> SEVERE: Failed to initialize end point associated with ProtocolHandler ["http-nio-8080"]
>>>>>> java.net.BindException: Address already in use
>>>>>>
>>>>>> I need Jacques's help ... HELP ...
>>>>>>
>>>>>> -----邮件原件-----
>>>>>> 发件人: [hidden email] [mailto:[hidden email]]
>>>>>> 发送时间: 2017年1月25日 11:18
>>>>>> 收件人: [hidden email]
>>>>>> 主题: buildbot failure in on ofbiz-trunk
>>>>>>
>>>>>> The Buildbot has detected a new failure on builder ofbiz-trunk while building . Full details are available at:
>>>>>> https://ci.apache.org/builders/ofbiz-trunk/builds/1861
>>>>>>
>>>>>> Buildbot URL: https://ci.apache.org/
>>>>>>
>>>>>> Buildslave for this Build: silvanus_ubuntu
>>>>>>
>>>>>> Build Reason: The AnyBranchScheduler scheduler named 'on-ofbiz-commit' triggered this build Build Source Stamp: [branch ofbiz/trunk] 1780133
>>>>>> Blamelist: shijh
>>>>>>
>>>>>> BUILD FAILED: failed shell_1
>>>>>>
>>>>>> Sincerely,
>>>>>>   -The Buildbot
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>
>
>

[Jacques Le Roux]

Hi Taher,

Prior to Gradle migration we were not using the 8080 port during tests. For instance we could have an OFBiz instance running, playing with its UI and
then run tests on another instance or vice-versa.

This is no longer possible because now the port 8080 (and others) are kept open while running tests. So far it was only a "small" annoyance, but it's
now blocking BuildBot because infra puppetized the BuildBot slaves https://s.apache.org/pWQD.

And, I don't know why and this could be a question to infra, for a reason there is always a Tomcat instance running on port 8080 these hosts
https://s.apache.org/iSaN

I thought we could bypass this issue by using portoffset with testIntegration task since it's now required. But I just tested locally and it seems to
not work. I'm not quite sure and will try a last thing directly "on" Buildbot. If it does not work we have to think about removing this need of open
ports while running tests, IOW to remove this regression introduced with Gradle.

Before anything else, I will though ask infra why Tomcat is always running on these hosts...

Jacques


Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :

> Why do we need to run the tests on portoffset?
>
> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
> [hidden email]> wrote:
>
>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and R13.
>>
>> Also that the pre-Gradle versions don't need portoffset
>>
>>
>>
>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>
>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>
>>>> But in the case of running integration tests with a portoffset we have
>>>> an issue which is blocking BuildBot now for trunk and
>>>>
>>> R16
>>>
>>> Jacques
>>>
>>>
>>>

[taher]

Okay, thank you for the efforts Jacques.

The only thing that I suspect is related might be in Config.java where the
portoffset is evaluated. I don't think I've changed the logic, and so I
suspect that it is not related to Gradle. If the tests do not run on a
portoffset, then I suspect perhaps they don't do the same on ant either.

On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
[hidden email]> wrote:

> Hi Taher,
>
> Prior to Gradle migration we were not using the 8080 port during tests.
> For instance we could have an OFBiz instance running, playing with its UI
> and then run tests on another instance or vice-versa.
>
> This is no longer possible because now the port 8080 (and others) are kept
> open while running tests. So far it was only a "small" annoyance, but it's
> now blocking BuildBot because infra puppetized the BuildBot slaves
> https://s.apache.org/pWQD.
>
> And, I don't know why and this could be a question to infra, for a reason
> there is always a Tomcat instance running on port 8080 these hosts
> https://s.apache.org/iSaN
>
> I thought we could bypass this issue by using portoffset with
> testIntegration task since it's now required. But I just tested locally and
> it seems to not work. I'm not quite sure and will try a last thing directly
> "on" Buildbot. If it does not work we have to think about removing this
> need of open ports while running tests, IOW to remove this regression
> introduced with Gradle.
>
> Before anything else, I will though ask infra why Tomcat is always running
> on these hosts...
>
> Jacques
>
>
> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>
>> Why do we need to run the tests on portoffset?
>>
>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>> [hidden email]> wrote:
>>
>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and R13.
>>>
>>> Also that the pre-Gradle versions don't need portoffset
>>>
>>>
>>>
>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>
>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>
>>>> But in the case of running integration tests with a portoffset we have
>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>
>>>>> R16
>>>>
>>>> Jacques
>>>>
>>>>
>>>>
>>>>
>

[Jacques Le Roux]

Taher,

I did not say that "ant run-tests" can use portoffset. Indeed it's not, and I was overzealous when, long ago and yesterday in r1780660, I changed the
Java code of tests for that (related with portoffset). Because I did not change how the tests are running in pre Gradle versions. Actually I was then
unsure and preferred to cover all cases. So, I repeat: we never use/block ports (at least 8080 and I guess 8443) when running tests in pre Gradle
versions.

For instance if you look for

f_ofb_15.addStep(Compile(command=["ant" , "load-demo"],
and
f_ofb_15.addStep(Compile(command=["ant" , "run-tests"],

in https://svn.apache.org/repos/infra/infrastructure/buildbot/aegis/buildmaster/master1/projects/ofbiz.conf (for committers only)

You will see that we don't use portoffset in Buildbot and it works well:

https://ci.apache.org/projects/ofbiz/logs/15.12/html/ (ran yesterday)

But since Gradle migration something has surely changed. As Buildbot shows (it's a side effect here, but it's also Buildbot role to track
regressions), the ports are used(?)/blocked when running tests in post Gradle versions.
I believe the best for us is to remove that, if it's possible. I'll have a look in Config.java. I'll not feel offended if someone beats me on it...

Thanks

Jacques


Le 29/01/2017 à 11:12, Taher Alkhateeb a écrit :

> Okay, thank you for the efforts Jacques.
>
> The only thing that I suspect is related might be in Config.java where the
> portoffset is evaluated. I don't think I've changed the logic, and so I
> suspect that it is not related to Gradle. If the tests do not run on a
> portoffset, then I suspect perhaps they don't do the same on ant either.
>
> On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
> [hidden email]> wrote:
>
>> Hi Taher,
>>
>> Prior to Gradle migration we were not using the 8080 port during tests.
>> For instance we could have an OFBiz instance running, playing with its UI
>> and then run tests on another instance or vice-versa.
>>
>> This is no longer possible because now the port 8080 (and others) are kept
>> open while running tests. So far it was only a "small" annoyance, but it's
>> now blocking BuildBot because infra puppetized the BuildBot slaves
>> https://s.apache.org/pWQD.
>>
>> And, I don't know why and this could be a question to infra, for a reason
>> there is always a Tomcat instance running on port 8080 these hosts
>> https://s.apache.org/iSaN
>>
>> I thought we could bypass this issue by using portoffset with
>> testIntegration task since it's now required. But I just tested locally and
>> it seems to not work. I'm not quite sure and will try a last thing directly
>> "on" Buildbot. If it does not work we have to think about removing this
>> need of open ports while running tests, IOW to remove this regression
>> introduced with Gradle.
>>
>> Before anything else, I will though ask infra why Tomcat is always running
>> on these hosts...
>>
>> Jacques
>>
>>
>> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>>
>>> Why do we need to run the tests on portoffset?
>>>
>>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>>> [hidden email]> wrote:
>>>
>>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and R13.
>>>> Also that the pre-Gradle versions don't need portoffset
>>>>
>>>>
>>>>
>>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>>
>>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>> But in the case of running integration tests with a portoffset we have
>>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>>
>>>>>> R16
>>>>> Jacques
>>>>>
>>>>>
>>>>>
>>>>>
>

[taher]

What does gradle have to do with blocking these ports? This is stuff
related to Tomcat. There is nothing in both ant and gradle that has
anything to do with these ports afaik.

On Jan 29, 2017 4:16 PM, "Jacques Le Roux" <[hidden email]>
wrote:

> Taher,
>
> I did not say that "ant run-tests" can use portoffset. Indeed it's not,
> and I was overzealous when, long ago and yesterday in r1780660, I changed
> the Java code of tests for that (related with portoffset). Because I did
> not change how the tests are running in pre Gradle versions. Actually I was
> then unsure and preferred to cover all cases. So, I repeat: we never
> use/block ports (at least 8080 and I guess 8443) when running tests in pre
> Gradle versions.
>
> For instance if you look for
>
> f_ofb_15.addStep(Compile(command=["ant" , "load-demo"],
> and
> f_ofb_15.addStep(Compile(command=["ant" , "run-tests"],
>
> in https://svn.apache.org/repos/infra/infrastructure/buildbot/a
> egis/buildmaster/master1/projects/ofbiz.conf (for committers only)
>
> You will see that we don't use portoffset in Buildbot and it works well:
>
> https://ci.apache.org/projects/ofbiz/logs/15.12/html/ (ran yesterday)
>
> But since Gradle migration something has surely changed. As Buildbot shows
> (it's a side effect here, but it's also Buildbot role to track
> regressions), the ports are used(?)/blocked when running tests in post
> Gradle versions.
> I believe the best for us is to remove that, if it's possible. I'll have a
> look in Config.java. I'll not feel offended if someone beats me on it...
>
> Thanks
>
> Jacques
>
>
> Le 29/01/2017 à 11:12, Taher Alkhateeb a écrit :
>
>> Okay, thank you for the efforts Jacques.
>>
>> The only thing that I suspect is related might be in Config.java where the
>> portoffset is evaluated. I don't think I've changed the logic, and so I
>> suspect that it is not related to Gradle. If the tests do not run on a
>> portoffset, then I suspect perhaps they don't do the same on ant either.
>>
>> On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
>> [hidden email]> wrote:
>>
>> Hi Taher,
>>>
>>> Prior to Gradle migration we were not using the 8080 port during tests.
>>> For instance we could have an OFBiz instance running, playing with its UI
>>> and then run tests on another instance or vice-versa.
>>>
>>> This is no longer possible because now the port 8080 (and others) are
>>> kept
>>> open while running tests. So far it was only a "small" annoyance, but
>>> it's
>>> now blocking BuildBot because infra puppetized the BuildBot slaves
>>> https://s.apache.org/pWQD.
>>>
>>> And, I don't know why and this could be a question to infra, for a reason
>>> there is always a Tomcat instance running on port 8080 these hosts
>>> https://s.apache.org/iSaN
>>>
>>> I thought we could bypass this issue by using portoffset with
>>> testIntegration task since it's now required. But I just tested locally
>>> and
>>> it seems to not work. I'm not quite sure and will try a last thing
>>> directly
>>> "on" Buildbot. If it does not work we have to think about removing this
>>> need of open ports while running tests, IOW to remove this regression
>>> introduced with Gradle.
>>>
>>> Before anything else, I will though ask infra why Tomcat is always
>>> running
>>> on these hosts...
>>>
>>> Jacques
>>>
>>>
>>> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>>>
>>> Why do we need to run the tests on portoffset?
>>>>
>>>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>>>> [hidden email]> wrote:
>>>>
>>>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and
>>>> R13.
>>>>
>>>>> Also that the pre-Gradle versions don't need portoffset
>>>>>
>>>>>
>>>>>
>>>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>>>
>>>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>>
>>>>>> But in the case of running integration tests with a portoffset we have
>>>>>>
>>>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>>>
>>>>>>> R16
>>>>>>>
>>>>>> Jacques
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>
>

[Jacques Le Roux]

Come on Taher,

As I said below (my last message) BuildBot is only revealing a symptom.

I already explained in this thread http://markmail.org/message/h4i7cn4i2btugdwn:

<<Prior to Gradle migration we were not using the 8080 port during tests. For instance we could have an OFBiz instance running, playing with its UI
and then run tests on another instance or vice-versa.
This is no longer possible because now the port 8080 (and others) are kept open while running tests. So far it was only a "small" annoyance, but it's
now blocking BuildBot because infra puppetized the BuildBot slaves https://s.apache.org/pWQD.>>

I suggest that you test by yourself to be convinced.

I also wrote in this same email <<Before anything else, I will though ask infra why Tomcat is always running on these hosts...>>

I did, now waiting their answer... and focusing on other matters in the meantime...

Jacques

Le 29/01/2017 à 14:27, Taher Alkhateeb a écrit :

> What does gradle have to do with blocking these ports? This is stuff
> related to Tomcat. There is nothing in both ant and gradle that has
> anything to do with these ports afaik.
>
> On Jan 29, 2017 4:16 PM, "Jacques Le Roux" <[hidden email]>
> wrote:
>
>> Taher,
>>
>> I did not say that "ant run-tests" can use portoffset. Indeed it's not,
>> and I was overzealous when, long ago and yesterday in r1780660, I changed
>> the Java code of tests for that (related with portoffset). Because I did
>> not change how the tests are running in pre Gradle versions. Actually I was
>> then unsure and preferred to cover all cases. So, I repeat: we never
>> use/block ports (at least 8080 and I guess 8443) when running tests in pre
>> Gradle versions.
>>
>> For instance if you look for
>>
>> f_ofb_15.addStep(Compile(command=["ant" , "load-demo"],
>> and
>> f_ofb_15.addStep(Compile(command=["ant" , "run-tests"],
>>
>> in https://svn.apache.org/repos/infra/infrastructure/buildbot/a
>> egis/buildmaster/master1/projects/ofbiz.conf (for committers only)
>>
>> You will see that we don't use portoffset in Buildbot and it works well:
>>
>> https://ci.apache.org/projects/ofbiz/logs/15.12/html/ (ran yesterday)
>>
>> But since Gradle migration something has surely changed. As Buildbot shows
>> (it's a side effect here, but it's also Buildbot role to track
>> regressions), the ports are used(?)/blocked when running tests in post
>> Gradle versions.
>> I believe the best for us is to remove that, if it's possible. I'll have a
>> look in Config.java. I'll not feel offended if someone beats me on it...
>>
>> Thanks
>>
>> Jacques
>>
>>
>> Le 29/01/2017 à 11:12, Taher Alkhateeb a écrit :
>>
>>> Okay, thank you for the efforts Jacques.
>>>
>>> The only thing that I suspect is related might be in Config.java where the
>>> portoffset is evaluated. I don't think I've changed the logic, and so I
>>> suspect that it is not related to Gradle. If the tests do not run on a
>>> portoffset, then I suspect perhaps they don't do the same on ant either.
>>>
>>> On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
>>> [hidden email]> wrote:
>>>
>>> Hi Taher,
>>>> Prior to Gradle migration we were not using the 8080 port during tests.
>>>> For instance we could have an OFBiz instance running, playing with its UI
>>>> and then run tests on another instance or vice-versa.
>>>>
>>>> This is no longer possible because now the port 8080 (and others) are
>>>> kept
>>>> open while running tests. So far it was only a "small" annoyance, but
>>>> it's
>>>> now blocking BuildBot because infra puppetized the BuildBot slaves
>>>> https://s.apache.org/pWQD.
>>>>
>>>> And, I don't know why and this could be a question to infra, for a reason
>>>> there is always a Tomcat instance running on port 8080 these hosts
>>>> https://s.apache.org/iSaN
>>>>
>>>> I thought we could bypass this issue by using portoffset with
>>>> testIntegration task since it's now required. But I just tested locally
>>>> and
>>>> it seems to not work. I'm not quite sure and will try a last thing
>>>> directly
>>>> "on" Buildbot. If it does not work we have to think about removing this
>>>> need of open ports while running tests, IOW to remove this regression
>>>> introduced with Gradle.
>>>>
>>>> Before anything else, I will though ask infra why Tomcat is always
>>>> running
>>>> on these hosts...
>>>>
>>>> Jacques
>>>>
>>>>
>>>> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>>>>
>>>> Why do we need to run the tests on portoffset?
>>>>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>>>>> [hidden email]> wrote:
>>>>>
>>>>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and
>>>>> R13.
>>>>>
>>>>>> Also that the pre-Gradle versions don't need portoffset
>>>>>>
>>>>>>
>>>>>>
>>>>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>>>>
>>>>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>>>
>>>>>>> But in the case of running integration tests with a portoffset we have
>>>>>>>
>>>>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>>>>
>>>>>>>> R16
>>>>>>>>
>>>>>>> Jacques
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>

[Scott Gray-3]

Try and be patient Jacques, I'm also having trouble understanding what you
are saying.

Here's what I've been able to understand so far:
- Pre-gradle OFBiz tests run fine on buildbot
- Gradle OFBiz fails on buildbot with a complaint that the default ports
are in use
- The test framework in OFBiz has never supported port offset
- Jacques thinks that maybe using a port offset for tests will fix the
problem but he's unsure why

Is that mostly correct Jacques?
Is there a possibility that buildbot isn't correctly shutting down OFBiz
from the previous test run and the ports are being left open?  Just a
guess, I don't know much about buildbot but that's the usual cause of the
default ports being unavailable in my experience.

Regards
Scott

On 30 January 2017 at 21:08, Jacques Le Roux <[hidden email]>
wrote:

> Come on Taher,
>
> As I said below (my last message) BuildBot is only revealing a symptom.
>
> I already explained in this thread http://markmail.org/message/h4
> i7cn4i2btugdwn:
>
> <<Prior to Gradle migration we were not using the 8080 port during tests.
> For instance we could have an OFBiz instance running, playing with its UI
> and then run tests on another instance or vice-versa.
> This is no longer possible because now the port 8080 (and others) are kept
> open while running tests. So far it was only a "small" annoyance, but it's
> now blocking BuildBot because infra puppetized the BuildBot slaves
> https://s.apache.org/pWQD.>>
>
> I suggest that you test by yourself to be convinced.
>
> I also wrote in this same email <<Before anything else, I will though ask
> infra why Tomcat is always running on these hosts...>>
>
> I did, now waiting their answer... and focusing on other matters in the
> meantime...
>
> Jacques
>
>
> Le 29/01/2017 à 14:27, Taher Alkhateeb a écrit :
>
>> What does gradle have to do with blocking these ports? This is stuff
>> related to Tomcat. There is nothing in both ant and gradle that has
>> anything to do with these ports afaik.
>>
>> On Jan 29, 2017 4:16 PM, "Jacques Le Roux" <[hidden email]>
>> wrote:
>>
>> Taher,
>>>
>>> I did not say that "ant run-tests" can use portoffset. Indeed it's not,
>>> and I was overzealous when, long ago and yesterday in r1780660, I changed
>>> the Java code of tests for that (related with portoffset). Because I did
>>> not change how the tests are running in pre Gradle versions. Actually I
>>> was
>>> then unsure and preferred to cover all cases. So, I repeat: we never
>>> use/block ports (at least 8080 and I guess 8443) when running tests in
>>> pre
>>> Gradle versions.
>>>
>>> For instance if you look for
>>>
>>> f_ofb_15.addStep(Compile(command=["ant" , "load-demo"],
>>> and
>>> f_ofb_15.addStep(Compile(command=["ant" , "run-tests"],
>>>
>>> in https://svn.apache.org/repos/infra/infrastructure/buildbot/a
>>> egis/buildmaster/master1/projects/ofbiz.conf (for committers only)
>>>
>>> You will see that we don't use portoffset in Buildbot and it works well:
>>>
>>> https://ci.apache.org/projects/ofbiz/logs/15.12/html/ (ran yesterday)
>>>
>>> But since Gradle migration something has surely changed. As Buildbot
>>> shows
>>> (it's a side effect here, but it's also Buildbot role to track
>>> regressions), the ports are used(?)/blocked when running tests in post
>>> Gradle versions.
>>> I believe the best for us is to remove that, if it's possible. I'll have
>>> a
>>> look in Config.java. I'll not feel offended if someone beats me on it...
>>>
>>> Thanks
>>>
>>> Jacques
>>>
>>>
>>> Le 29/01/2017 à 11:12, Taher Alkhateeb a écrit :
>>>
>>> Okay, thank you for the efforts Jacques.
>>>>
>>>> The only thing that I suspect is related might be in Config.java where
>>>> the
>>>> portoffset is evaluated. I don't think I've changed the logic, and so I
>>>> suspect that it is not related to Gradle. If the tests do not run on a
>>>> portoffset, then I suspect perhaps they don't do the same on ant either.
>>>>
>>>> On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
>>>> [hidden email]> wrote:
>>>>
>>>> Hi Taher,
>>>>
>>>>> Prior to Gradle migration we were not using the 8080 port during tests.
>>>>> For instance we could have an OFBiz instance running, playing with its
>>>>> UI
>>>>> and then run tests on another instance or vice-versa.
>>>>>
>>>>> This is no longer possible because now the port 8080 (and others) are
>>>>> kept
>>>>> open while running tests. So far it was only a "small" annoyance, but
>>>>> it's
>>>>> now blocking BuildBot because infra puppetized the BuildBot slaves
>>>>> https://s.apache.org/pWQD.
>>>>>
>>>>> And, I don't know why and this could be a question to infra, for a
>>>>> reason
>>>>> there is always a Tomcat instance running on port 8080 these hosts
>>>>> https://s.apache.org/iSaN
>>>>>
>>>>> I thought we could bypass this issue by using portoffset with
>>>>> testIntegration task since it's now required. But I just tested locally
>>>>> and
>>>>> it seems to not work. I'm not quite sure and will try a last thing
>>>>> directly
>>>>> "on" Buildbot. If it does not work we have to think about removing this
>>>>> need of open ports while running tests, IOW to remove this regression
>>>>> introduced with Gradle.
>>>>>
>>>>> Before anything else, I will though ask infra why Tomcat is always
>>>>> running
>>>>> on these hosts...
>>>>>
>>>>> Jacques
>>>>>
>>>>>
>>>>> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>>>>>
>>>>> Why do we need to run the tests on portoffset?
>>>>>
>>>>>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>>>>>> [hidden email]> wrote:
>>>>>>
>>>>>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and
>>>>>> R13.
>>>>>>
>>>>>> Also that the pre-Gradle versions don't need portoffset
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>>>>>
>>>>>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>>>>
>>>>>>> But in the case of running integration tests with a portoffset we
>>>>>>>> have
>>>>>>>>
>>>>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>>>>>
>>>>>>>>> R16
>>>>>>>>>
>>>>>>>>> Jacques
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>

[Jacques Le Roux]

Thanks Scott,

Inline...


Le 30/01/2017 à 10:03, Scott Gray a écrit :

> Try and be patient Jacques, I'm also having trouble understanding what you
> are saying.
>
> Here's what I've been able to understand so far:
> - Pre-gradle OFBiz tests run fine on buildbot
> - Gradle OFBiz fails on buildbot with a complaint that the default ports
> are in use
> - The test framework in OFBiz has never supported port offset
> - Jacques thinks that maybe using a port offset for tests will fix the
> problem but he's unsure why
>
> Is that mostly correct Jacques?

Almost, It's not that I'm unsure, actually I tried this weekend to use portoffset in Gradle versions. From the documentation, it's theoretically
possible but does not work OOTB (I agree it's a complicated matter)
So I  committed r1780660  and locally hardcoded 8080 port changes to 8081 in XML files serviceengine.xml, services_test.xml and axis2.xml w/o success
when running tests.

Then I thought more at the problem and I see now 2 options:
1) Infra does not need to have always Tomcat using the 8080 port on Buildbot hosts, problem resolved. I asked a question in the related infra:
https://issues.apache.org/jira/browse/INFRA-13402
2) Infra needs to have always Tomcat using the 8080 port on Buildbot hosts. Then, it's maybe possible, as Taher suggested, to get back to the previous
behaviour were tests were not using/blocking the 8080 port. Taher suggested to have a look into config.xml...

I first wait an answer to 1), then I'll consider 2)

Note though that it would be nice to not use/block ports during tests, as it was before Gradle. Then you can use another OFBiz instance while running
tests, and vice-versa. I don't know if it's me, but it seems that I'm rehearsing the same things :/

> Is there a possibility that buildbot isn't correctly shutting down OFBiz
> from the previous test run and the ports are being left open?  Just a
> guess, I don't know much about buildbot but that's the usual cause of the
> default ports being unavailable in my experience.

Infra said in INFRA-13402 that it's due to Tomcat being puppetized, please read INFRA-13402

Jacques

>
> Regards
> Scott
>
> On 30 January 2017 at 21:08, Jacques Le Roux <[hidden email]>
> wrote:
>
>> Come on Taher,
>>
>> As I said below (my last message) BuildBot is only revealing a symptom.
>>
>> I already explained in this thread http://markmail.org/message/h4
>> i7cn4i2btugdwn:
>>
>> <<Prior to Gradle migration we were not using the 8080 port during tests.
>> For instance we could have an OFBiz instance running, playing with its UI
>> and then run tests on another instance or vice-versa.
>> This is no longer possible because now the port 8080 (and others) are kept
>> open while running tests. So far it was only a "small" annoyance, but it's
>> now blocking BuildBot because infra puppetized the BuildBot slaves
>> https://s.apache.org/pWQD.>>
>>
>> I suggest that you test by yourself to be convinced.
>>
>> I also wrote in this same email <<Before anything else, I will though ask
>> infra why Tomcat is always running on these hosts...>>
>>
>> I did, now waiting their answer... and focusing on other matters in the
>> meantime...
>>
>> Jacques
>>
>>
>> Le 29/01/2017 à 14:27, Taher Alkhateeb a écrit :
>>
>>> What does gradle have to do with blocking these ports? This is stuff
>>> related to Tomcat. There is nothing in both ant and gradle that has
>>> anything to do with these ports afaik.
>>>
>>> On Jan 29, 2017 4:16 PM, "Jacques Le Roux" <[hidden email]>
>>> wrote:
>>>
>>> Taher,
>>>> I did not say that "ant run-tests" can use portoffset. Indeed it's not,
>>>> and I was overzealous when, long ago and yesterday in r1780660, I changed
>>>> the Java code of tests for that (related with portoffset). Because I did
>>>> not change how the tests are running in pre Gradle versions. Actually I
>>>> was
>>>> then unsure and preferred to cover all cases. So, I repeat: we never
>>>> use/block ports (at least 8080 and I guess 8443) when running tests in
>>>> pre
>>>> Gradle versions.
>>>>
>>>> For instance if you look for
>>>>
>>>> f_ofb_15.addStep(Compile(command=["ant" , "load-demo"],
>>>> and
>>>> f_ofb_15.addStep(Compile(command=["ant" , "run-tests"],
>>>>
>>>> in https://svn.apache.org/repos/infra/infrastructure/buildbot/a
>>>> egis/buildmaster/master1/projects/ofbiz.conf (for committers only)
>>>>
>>>> You will see that we don't use portoffset in Buildbot and it works well:
>>>>
>>>> https://ci.apache.org/projects/ofbiz/logs/15.12/html/ (ran yesterday)
>>>>
>>>> But since Gradle migration something has surely changed. As Buildbot
>>>> shows
>>>> (it's a side effect here, but it's also Buildbot role to track
>>>> regressions), the ports are used(?)/blocked when running tests in post
>>>> Gradle versions.
>>>> I believe the best for us is to remove that, if it's possible. I'll have
>>>> a
>>>> look in Config.java. I'll not feel offended if someone beats me on it...
>>>>
>>>> Thanks
>>>>
>>>> Jacques
>>>>
>>>>
>>>> Le 29/01/2017 à 11:12, Taher Alkhateeb a écrit :
>>>>
>>>> Okay, thank you for the efforts Jacques.
>>>>> The only thing that I suspect is related might be in Config.java where
>>>>> the
>>>>> portoffset is evaluated. I don't think I've changed the logic, and so I
>>>>> suspect that it is not related to Gradle. If the tests do not run on a
>>>>> portoffset, then I suspect perhaps they don't do the same on ant either.
>>>>>
>>>>> On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
>>>>> [hidden email]> wrote:
>>>>>
>>>>> Hi Taher,
>>>>>
>>>>>> Prior to Gradle migration we were not using the 8080 port during tests.
>>>>>> For instance we could have an OFBiz instance running, playing with its
>>>>>> UI
>>>>>> and then run tests on another instance or vice-versa.
>>>>>>
>>>>>> This is no longer possible because now the port 8080 (and others) are
>>>>>> kept
>>>>>> open while running tests. So far it was only a "small" annoyance, but
>>>>>> it's
>>>>>> now blocking BuildBot because infra puppetized the BuildBot slaves
>>>>>> https://s.apache.org/pWQD.
>>>>>>
>>>>>> And, I don't know why and this could be a question to infra, for a
>>>>>> reason
>>>>>> there is always a Tomcat instance running on port 8080 these hosts
>>>>>> https://s.apache.org/iSaN
>>>>>>
>>>>>> I thought we could bypass this issue by using portoffset with
>>>>>> testIntegration task since it's now required. But I just tested locally
>>>>>> and
>>>>>> it seems to not work. I'm not quite sure and will try a last thing
>>>>>> directly
>>>>>> "on" Buildbot. If it does not work we have to think about removing this
>>>>>> need of open ports while running tests, IOW to remove this regression
>>>>>> introduced with Gradle.
>>>>>>
>>>>>> Before anything else, I will though ask infra why Tomcat is always
>>>>>> running
>>>>>> on these hosts...
>>>>>>
>>>>>> Jacques
>>>>>>
>>>>>>
>>>>>> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>>>>>>
>>>>>> Why do we need to run the tests on portoffset?
>>>>>>
>>>>>>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>>>>>>> [hidden email]> wrote:
>>>>>>>
>>>>>>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and
>>>>>>> R13.
>>>>>>>
>>>>>>> Also that the pre-Gradle versions don't need portoffset
>>>>>>>>
>>>>>>>>
>>>>>>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>>>>>>
>>>>>>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>>>>>
>>>>>>>> But in the case of running integration tests with a portoffset we
>>>>>>>>> have
>>>>>>>>>
>>>>>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>>>>>> R16
>>>>>>>>>>
>>>>>>>>>> Jacques
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>

[Jacques Le Roux]

I created https://issues.apache.org/jira/browse/OFBIZ-9196 for that ("Regression: the testIntegration Gradle taks shoud not use/block the ports")

Jacques


Le 30/01/2017 à 11:15, Jacques Le Roux a écrit :

> Thanks Scott,
>
> Inline...
>
>
> Le 30/01/2017 à 10:03, Scott Gray a écrit :
>> Try and be patient Jacques, I'm also having trouble understanding what you
>> are saying.
>>
>> Here's what I've been able to understand so far:
>> - Pre-gradle OFBiz tests run fine on buildbot
>> - Gradle OFBiz fails on buildbot with a complaint that the default ports
>> are in use
>> - The test framework in OFBiz has never supported port offset
>> - Jacques thinks that maybe using a port offset for tests will fix the
>> problem but he's unsure why
>>
>> Is that mostly correct Jacques?
> Almost, It's not that I'm unsure, actually I tried this weekend to use portoffset in Gradle versions. From the documentation, it's theoretically
> possible but does not work OOTB (I agree it's a complicated matter)
> So I  committed r1780660  and locally hardcoded 8080 port changes to 8081 in XML files serviceengine.xml, services_test.xml and axis2.xml w/o
> success when running tests.
>
> Then I thought more at the problem and I see now 2 options:
> 1) Infra does not need to have always Tomcat using the 8080 port on Buildbot hosts, problem resolved. I asked a question in the related infra:
> https://issues.apache.org/jira/browse/INFRA-13402
> 2) Infra needs to have always Tomcat using the 8080 port on Buildbot hosts. Then, it's maybe possible, as Taher suggested, to get back to the
> previous behaviour were tests were not using/blocking the 8080 port. Taher suggested to have a look into config.xml...
>
> I first wait an answer to 1), then I'll consider 2)
>
> Note though that it would be nice to not use/block ports during tests, as it was before Gradle. Then you can use another OFBiz instance while
> running tests, and vice-versa. I don't know if it's me, but it seems that I'm rehearsing the same things :/
>
>> Is there a possibility that buildbot isn't correctly shutting down OFBiz
>> from the previous test run and the ports are being left open? Just a
>> guess, I don't know much about buildbot but that's the usual cause of the
>> default ports being unavailable in my experience.
>
> Infra said in INFRA-13402 that it's due to Tomcat being puppetized, please read INFRA-13402
>
> Jacques
>>
>> Regards
>> Scott
>>
>> On 30 January 2017 at 21:08, Jacques Le Roux <[hidden email]>
>> wrote:
>>
>>> Come on Taher,
>>>
>>> As I said below (my last message) BuildBot is only revealing a symptom.
>>>
>>> I already explained in this thread http://markmail.org/message/h4
>>> i7cn4i2btugdwn:
>>>
>>> <<Prior to Gradle migration we were not using the 8080 port during tests.
>>> For instance we could have an OFBiz instance running, playing with its UI
>>> and then run tests on another instance or vice-versa.
>>> This is no longer possible because now the port 8080 (and others) are kept
>>> open while running tests. So far it was only a "small" annoyance, but it's
>>> now blocking BuildBot because infra puppetized the BuildBot slaves
>>> https://s.apache.org/pWQD.>>
>>>
>>> I suggest that you test by yourself to be convinced.
>>>
>>> I also wrote in this same email <<Before anything else, I will though ask
>>> infra why Tomcat is always running on these hosts...>>
>>>
>>> I did, now waiting their answer... and focusing on other matters in the
>>> meantime...
>>>
>>> Jacques
>>>
>>>
>>> Le 29/01/2017 à 14:27, Taher Alkhateeb a écrit :
>>>
>>>> What does gradle have to do with blocking these ports? This is stuff
>>>> related to Tomcat. There is nothing in both ant and gradle that has
>>>> anything to do with these ports afaik.
>>>>
>>>> On Jan 29, 2017 4:16 PM, "Jacques Le Roux" <[hidden email]>
>>>> wrote:
>>>>
>>>> Taher,
>>>>> I did not say that "ant run-tests" can use portoffset. Indeed it's not,
>>>>> and I was overzealous when, long ago and yesterday in r1780660, I changed
>>>>> the Java code of tests for that (related with portoffset). Because I did
>>>>> not change how the tests are running in pre Gradle versions. Actually I
>>>>> was
>>>>> then unsure and preferred to cover all cases. So, I repeat: we never
>>>>> use/block ports (at least 8080 and I guess 8443) when running tests in
>>>>> pre
>>>>> Gradle versions.
>>>>>
>>>>> For instance if you look for
>>>>>
>>>>> f_ofb_15.addStep(Compile(command=["ant" , "load-demo"],
>>>>> and
>>>>> f_ofb_15.addStep(Compile(command=["ant" , "run-tests"],
>>>>>
>>>>> in https://svn.apache.org/repos/infra/infrastructure/buildbot/a
>>>>> egis/buildmaster/master1/projects/ofbiz.conf (for committers only)
>>>>>
>>>>> You will see that we don't use portoffset in Buildbot and it works well:
>>>>>
>>>>> https://ci.apache.org/projects/ofbiz/logs/15.12/html/ (ran yesterday)
>>>>>
>>>>> But since Gradle migration something has surely changed. As Buildbot
>>>>> shows
>>>>> (it's a side effect here, but it's also Buildbot role to track
>>>>> regressions), the ports are used(?)/blocked when running tests in post
>>>>> Gradle versions.
>>>>> I believe the best for us is to remove that, if it's possible. I'll have
>>>>> a
>>>>> look in Config.java. I'll not feel offended if someone beats me on it...
>>>>>
>>>>> Thanks
>>>>>
>>>>> Jacques
>>>>>
>>>>>
>>>>> Le 29/01/2017 à 11:12, Taher Alkhateeb a écrit :
>>>>>
>>>>> Okay, thank you for the efforts Jacques.
>>>>>> The only thing that I suspect is related might be in Config.java where
>>>>>> the
>>>>>> portoffset is evaluated. I don't think I've changed the logic, and so I
>>>>>> suspect that it is not related to Gradle. If the tests do not run on a
>>>>>> portoffset, then I suspect perhaps they don't do the same on ant either.
>>>>>>
>>>>>> On Sun, Jan 29, 2017 at 12:57 PM, Jacques Le Roux <
>>>>>> [hidden email]> wrote:
>>>>>>
>>>>>> Hi Taher,
>>>>>>
>>>>>>> Prior to Gradle migration we were not using the 8080 port during tests.
>>>>>>> For instance we could have an OFBiz instance running, playing with its
>>>>>>> UI
>>>>>>> and then run tests on another instance or vice-versa.
>>>>>>>
>>>>>>> This is no longer possible because now the port 8080 (and others) are
>>>>>>> kept
>>>>>>> open while running tests. So far it was only a "small" annoyance, but
>>>>>>> it's
>>>>>>> now blocking BuildBot because infra puppetized the BuildBot slaves
>>>>>>> https://s.apache.org/pWQD.
>>>>>>>
>>>>>>> And, I don't know why and this could be a question to infra, for a
>>>>>>> reason
>>>>>>> there is always a Tomcat instance running on port 8080 these hosts
>>>>>>> https://s.apache.org/iSaN
>>>>>>>
>>>>>>> I thought we could bypass this issue by using portoffset with
>>>>>>> testIntegration task since it's now required. But I just tested locally
>>>>>>> and
>>>>>>> it seems to not work. I'm not quite sure and will try a last thing
>>>>>>> directly
>>>>>>> "on" Buildbot. If it does not work we have to think about removing this
>>>>>>> need of open ports while running tests, IOW to remove this regression
>>>>>>> introduced with Gradle.
>>>>>>>
>>>>>>> Before anything else, I will though ask infra why Tomcat is always
>>>>>>> running
>>>>>>> on these hosts...
>>>>>>>
>>>>>>> Jacques
>>>>>>>
>>>>>>>
>>>>>>> Le 27/01/2017 à 18:15, Taher Alkhateeb a écrit :
>>>>>>>
>>>>>>> Why do we need to run the tests on portoffset?
>>>>>>>
>>>>>>>> On Fri, Jan 27, 2017 at 7:48 AM, Jacques Le Roux <
>>>>>>>> [hidden email]> wrote:
>>>>>>>>
>>>>>>>> Ah, I forgot to say, that I have dropped BuildBot support for R12 and
>>>>>>>> R13.
>>>>>>>>
>>>>>>>> Also that the pre-Gradle versions don't need portoffset
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Le 27/01/2017 à 05:29, Jacques Le Roux a écrit :
>>>>>>>>>
>>>>>>>>> Le 27/01/2017 à 05:14, Jacques Le Roux a écrit :
>>>>>>>>>
>>>>>>>>> But in the case of running integration tests with a portoffset we
>>>>>>>>>> have
>>>>>>>>>>
>>>>>>>>>> an issue which is blocking BuildBot now for trunk and
>>>>>>>>>>> R16
>>>>>>>>>>>
>>>>>>>>>>> Jacques
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>
>