continuum-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Emmanuel Venisse <emman...@venisse.net>
Subject Re: long delay between maven completion and build completion
Date Sat, 20 Oct 2007 12:39:14 GMT


Emmanuel Venisse a écrit :
> 
> 
> Brett Porter a écrit :
>> Sorry, missed this reply:
>>
>> On 26/09/2007, at 4:55 PM, Trygve Laugstøl wrote:
>>
>>> Brett Porter wrote:
>>>> I'm seeing this on the zone:
>>>> INFO   | jvm 1    | 2007/09/24 03:29:17 | 2007-09-24 03:29:17,421 
>>>> [pool-1-thread-1] INFO  ContinuumBuildExecutor:maven2  - Exit code: 0
>>>> then for a long interval, on the web interface, the build is still 
>>>> "in progress". It doesn't happen all the time, though.
>>>> The log is already saying:
>>>> [INFO] 
>>>> ------------------------------------------------------------------------

>>>>
>>>> [INFO] BUILD SUCCESSFUL
>>>> [INFO] 
>>>> ------------------------------------------------------------------------

>>>>
>>>> [INFO] Total time: 6 minutes 4 seconds
>>>> [INFO] Finished at: Mon Sep 24 03:29:17 GMT+00:00 2007
>>>> [INFO] Final Memory: 14M/255M
>>>> [INFO] 
>>>> ------------------------------------------------------------------------

>>>>
>>>> The build finally completed after 10 minutes:
>>>> INFO   | jvm 1    | 2007/09/24 03:33:58 | 2007-09-24 03:33:58,430 
>>>> [pool-1-thread-1] INFO  BuildController:default        - Performing 
>>>> action deploy-artifact
>>>> Note that the build result had said it was going for 8, 9, 10 
>>>> minutes as it kept going, but then after finishing completely it 
>>>> reported the duration as "6 min 8 sec"
>>>> Any ideas?
>>>
>>> Initial hunch says signaling between the external process and the JVM.
>>>
>>> Do you see this on all platforms? In particular a UNIX platform vs 
>>> Windows
>>>
>>> What about with running non-jvm builds (IWO shell scripts)?
>>
>> I've not tried others - I've just seen it on Solaris so far with the 
>> Maven 2 builds. Anyone seen this elsewhere?
> 
> I remove a FileUtils.fileRead(ouput) in AbstractBuildExecutor because we 
> don't need the output there.
> 
> others seconds are consumed by some jdo requests and the copy/analyze of 
> tests result files. This task is done in ExecuteBuilderContinuumAction 
> (line 163)

Test results parsing can be (must be) removed (CONTINUUM-1532) because we don't need/use them

Emmanuel

> 
> Emmanuel
> 
>>
>>>
>>> I have no time to debug this right now, but perhaps in a couple of days.
>>>
>>> -- 
>>> Trygve
>>
>> -- 
>> Brett Porter - brett@apache.org
>> Blog: http://www.devzuz.org/blogs/bporter/
>>
>>
> 
> 
> 


Mime
View raw message