You are viewing a plain text version of this content. The canonical link for it is here.
Posted to server-dev@james.apache.org by "Noel J. Bergman" <no...@devtech.com> on 2008/08/06 21:34:50 UTC
RE: Server's test failure and CI (Was: JAMES Server Nightly Build Report)
Stefano Bagnara wrote:
> The current process is a simple nighly build [that] is good
> for the nighlies, but I think I don't have to explain the
> difference between a real CI environment and a nighly build script.
That is correct. We have a CI at the ASF. The nightly wasn't intended to
replace it. However ...
> The specific issue I was recording in this message was the inability to
> know what exact test failed (we have only summaries for each test class).
The entire build log is posted to the build site. It seems to me that if
you want more detail, just enhance the build. The nightly process is just a
shell script that runs the build, generates the e-mail summary, and uploads
the build artifacts with full log.
Is there something you want that can't be done via the above?
> > Ubuntu 7.10.
> JVM used? (IIRC it didn't find the Java 1.4 issues, so I thought it was
> 1.5, but if you tell us the specific version it will be helpful)
> JVM parameters used by ant.
I don't recall if it is Java 5 or 6, and freakishly, I don't have access to
it right now. I need to fix the SSH connection back through my firewall
when I get home. Right now it is on autopilot.
> > We've talked about having a JAMES zone for testing. Of course, you were
> > against it the last time, apparently for the sole reason that I
suggested
> > it.
> I've never been against this. You may have misunderstood me.
LOL Actually, you're right -- I went back and cbecked. It is always you who
disagrees, but in this case it was BERND! LOL OK, I apologize. :-)
--- Noel
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org
Re: Server's test failure and CI (Was: JAMES Server Nightly Build
Report)
Posted by Stefano Bagnara <ap...@bago.org>.
Noel J. Bergman ha scritto:
> Stefano Bagnara wrote:
>
>> The current process is a simple nighly build [that] is good
>> for the nighlies, but I think I don't have to explain the
>> difference between a real CI environment and a nighly build script.
>
> That is correct. We have a CI at the ASF. The nightly wasn't intended to
> replace it. However ...
Cool! I'm very happy with Hudson :-) We have an issue because it rarely
stuck during an Hadoop+jSPF build (weird) and I'm working with the zone
mantainers to fix this.
>> The specific issue I was recording in this message was the inability to
>> know what exact test failed (we have only summaries for each test class).
>
> The entire build log is posted to the build site. It seems to me that if
> you want more detail, just enhance the build. The nightly process is just a
> shell script that runs the build, generates the e-mail summary, and uploads
> the build artifacts with full log.
> Is there something you want that can't be done via the above?
I know. But the exact test was not in the log. The ant build does not
give that output. I see it instead in the maven build.
I don't want to deal with improving the ant build now, but I set up both
ant and maven builds on hudson and I've been able to debug the issue.
Also I found it unvaluable to be able to run a "kill -3 <buildpid>" on
the hudson zone to get the stackdump in recent debug for JSPF-65.
I think we have all I need now, thank you for your collaboration!
>>> Ubuntu 7.10.
>
>> JVM used? (IIRC it didn't find the Java 1.4 issues, so I thought it was
>> 1.5, but if you tell us the specific version it will be helpful)
>> JVM parameters used by ant.
>
> I don't recall if it is Java 5 or 6, and freakishly, I don't have access to
> it right now. I need to fix the SSH connection back through my firewall
> when I get home. Right now it is on autopilot.
On hudson we now hava a java 1.4 ant build and a java 5 maven build: a
little diversification should help. If you are on java 6 this would
complete the range.
>>> We've talked about having a JAMES zone for testing. Of course, you were
>>> against it the last time, apparently for the sole reason that I
> suggested
>>> it.
>
>> I've never been against this. You may have misunderstood me.
>
> LOL Actually, you're right -- I went back and cbecked. It is always you who
> disagrees, but in this case it was BERND! LOL OK, I apologize. :-)
>
> --- Noel
;-) It seems we have an agreement then...
Stefano
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org
RemoteDelivery test issue (Was: Server's test failure and CI)
Posted by Stefano Bagnara <ap...@bago.org>.
Robert Burrell Donkin ha scritto:
> On Thu, Aug 7, 2008 at 10:54 AM, Stefano Bagnara <ap...@bago.org> wrote:
>> Robert Burrell Donkin ha scritto:
>>> On Wed, Aug 6, 2008 at 8:34 PM, Noel J. Bergman <no...@devtech.com> wrote:
>>>> Stefano Bagnara wrote:
>>>>
>>>>> The current process is a simple nighly build [that] is good
>>>>> for the nighlies, but I think I don't have to explain the
>>>>> difference between a real CI environment and a nighly build script.
>>>> That is correct. We have a CI at the ASF. The nightly wasn't intended
>>>> to
>>>> replace it. However ...
>>>>
>>>>> The specific issue I was recording in this message was the inability to
>>>>> know what exact test failed (we have only summaries for each test
>>>>> class).
>>>> The entire build log is posted to the build site. It seems to me that if
>>>> you want more detail, just enhance the build. The nightly process is
>>>> just a
>>>> shell script that runs the build, generates the e-mail summary, and
>>>> uploads
>>>> the build artifacts with full log.
>>>>
>>>> Is there something you want that can't be done via the above?
>>> think this failing test has just been getting everyone down :-/
>>>
>>> it's been particularly irratating (it now fails about 60% of the time
>>> for) and difficult to diagnose. it seems to be some sort of
>>> serialization race condition but it's beyond me. don't know if you
>>> have any ideas.
>> If we agree this probably is an issue in JVM/classpath but we don't have
>> time dealing with SUN for this IMHO we can even add a comment to that test..
>> e.g: catch the assert failure and add a comment about this whole thing.
>
> if this is a JVM issue then couldn't we just narrow down the known bad versions?
>
> - robert
I think we saw it on every platform running very often the tests.
Even if it only happened once in my environment it did. I also saw it on
hudson, and you saw it.
AFAIKT it could even be a non fixed issue existing in every Sun jvm.
Trying IBM/BEA/ASF JVM could give some more answers (I don't know what
we can conclude if they don't give the issue, only what we can guess if
they give the same issue).
Stefano
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org
Re: Server's test failure and CI (Was: JAMES Server Nightly Build Report)
Posted by Robert Burrell Donkin <ro...@gmail.com>.
On Thu, Aug 7, 2008 at 10:54 AM, Stefano Bagnara <ap...@bago.org> wrote:
> Robert Burrell Donkin ha scritto:
>>
>> On Wed, Aug 6, 2008 at 8:34 PM, Noel J. Bergman <no...@devtech.com> wrote:
>>>
>>> Stefano Bagnara wrote:
>>>
>>>> The current process is a simple nighly build [that] is good
>>>> for the nighlies, but I think I don't have to explain the
>>>> difference between a real CI environment and a nighly build script.
>>>
>>> That is correct. We have a CI at the ASF. The nightly wasn't intended
>>> to
>>> replace it. However ...
>>>
>>>> The specific issue I was recording in this message was the inability to
>>>> know what exact test failed (we have only summaries for each test
>>>> class).
>>>
>>> The entire build log is posted to the build site. It seems to me that if
>>> you want more detail, just enhance the build. The nightly process is
>>> just a
>>> shell script that runs the build, generates the e-mail summary, and
>>> uploads
>>> the build artifacts with full log.
>>>
>>> Is there something you want that can't be done via the above?
>>
>> think this failing test has just been getting everyone down :-/
>>
>> it's been particularly irratating (it now fails about 60% of the time
>> for) and difficult to diagnose. it seems to be some sort of
>> serialization race condition but it's beyond me. don't know if you
>> have any ideas.
>
> If we agree this probably is an issue in JVM/classpath but we don't have
> time dealing with SUN for this IMHO we can even add a comment to that test..
> e.g: catch the assert failure and add a comment about this whole thing.
if this is a JVM issue then couldn't we just narrow down the known bad versions?
- robert
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org
Re: Server's test failure and CI (Was: JAMES Server Nightly Build
Report)
Posted by Stefano Bagnara <ap...@bago.org>.
Robert Burrell Donkin ha scritto:
> On Wed, Aug 6, 2008 at 8:34 PM, Noel J. Bergman <no...@devtech.com> wrote:
>> Stefano Bagnara wrote:
>>
>>> The current process is a simple nighly build [that] is good
>>> for the nighlies, but I think I don't have to explain the
>>> difference between a real CI environment and a nighly build script.
>> That is correct. We have a CI at the ASF. The nightly wasn't intended to
>> replace it. However ...
>>
>>> The specific issue I was recording in this message was the inability to
>>> know what exact test failed (we have only summaries for each test class).
>> The entire build log is posted to the build site. It seems to me that if
>> you want more detail, just enhance the build. The nightly process is just a
>> shell script that runs the build, generates the e-mail summary, and uploads
>> the build artifacts with full log.
>>
>> Is there something you want that can't be done via the above?
>
> think this failing test has just been getting everyone down :-/
>
> it's been particularly irratating (it now fails about 60% of the time
> for) and difficult to diagnose. it seems to be some sort of
> serialization race condition but it's beyond me. don't know if you
> have any ideas.
If we agree this probably is an issue in JVM/classpath but we don't have
time dealing with SUN for this IMHO we can even add a comment to that
test.. e.g: catch the assert failure and add a comment about this whole
thing.
Stefano
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org
Re: Server's test failure and CI (Was: JAMES Server Nightly Build Report)
Posted by Robert Burrell Donkin <ro...@gmail.com>.
On Wed, Aug 6, 2008 at 8:34 PM, Noel J. Bergman <no...@devtech.com> wrote:
> Stefano Bagnara wrote:
>
>> The current process is a simple nighly build [that] is good
>> for the nighlies, but I think I don't have to explain the
>> difference between a real CI environment and a nighly build script.
>
> That is correct. We have a CI at the ASF. The nightly wasn't intended to
> replace it. However ...
>
>> The specific issue I was recording in this message was the inability to
>> know what exact test failed (we have only summaries for each test class).
>
> The entire build log is posted to the build site. It seems to me that if
> you want more detail, just enhance the build. The nightly process is just a
> shell script that runs the build, generates the e-mail summary, and uploads
> the build artifacts with full log.
>
> Is there something you want that can't be done via the above?
think this failing test has just been getting everyone down :-/
it's been particularly irratating (it now fails about 60% of the time
for) and difficult to diagnose. it seems to be some sort of
serialization race condition but it's beyond me. don't know if you
have any ideas.
- robert
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org
Re: Server's test failure and CI (Was: JAMES Server Nightly Build Report)
Posted by Bernd Fondermann <be...@googlemail.com>.
On Wed, Aug 6, 2008 at 21:34, Noel J. Bergman <no...@devtech.com> wrote:
> Stefano Bagnara wrote:
>
>> The current process is a simple nighly build [that] is good
>> for the nighlies, but I think I don't have to explain the
>> difference between a real CI environment and a nighly build script.
>
> That is correct. We have a CI at the ASF. The nightly wasn't intended to
> replace it. However ...
>
>> The specific issue I was recording in this message was the inability to
>> know what exact test failed (we have only summaries for each test class).
>
> The entire build log is posted to the build site. It seems to me that if
> you want more detail, just enhance the build. The nightly process is just a
> shell script that runs the build, generates the e-mail summary, and uploads
> the build artifacts with full log.
>
> Is there something you want that can't be done via the above?
>
>> > Ubuntu 7.10.
>
>> JVM used? (IIRC it didn't find the Java 1.4 issues, so I thought it was
>> 1.5, but if you tell us the specific version it will be helpful)
>> JVM parameters used by ant.
>
> I don't recall if it is Java 5 or 6, and freakishly, I don't have access to
> it right now. I need to fix the SSH connection back through my firewall
> when I get home. Right now it is on autopilot.
>
>> > We've talked about having a JAMES zone for testing. Of course, you were
>> > against it the last time, apparently for the sole reason that I
> suggested
>> > it.
>
>> I've never been against this. You may have misunderstood me.
>
> LOL Actually, you're right -- I went back and cbecked. It is always you who
> disagrees, but in this case it was BERND! LOL OK, I apologize. :-)
I supported this, too. I volontered to aquire, admin and donate a domain for it.
But the proposal never kicked off, so I didn't do it. Should I do so, now?
Bernd
---------------------------------------------------------------------
To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org