Changes not in summary

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
16 messages Options
Reply | Threaded
Open this post in threaded view
|

Changes not in summary

daniel.galan
Changes not in summary

Hi there,

even we have changes in the CVS which have been recieved with the following statement (from the console output), there are no entries in the changes summary on the project page.

[workspace] $ cvs -q -z9 update -PdC
P my/files..
U another/file

Some ideas?
Greetings

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Kohsuke Kawaguchi-2
[hidden email] wrote:
> Hi there,
>
> even we have changes in the CVS which have been recieved with the following statement (from the console output), there are no entries in the changes summary on the project page.
>
> [workspace] $ cvs -q -z9 update -PdC
> P my/files..
> U another/file
>
> Some ideas?

I've seen a few bugs like that in the past, but to the best of my
knowledge all of them are fixed by now.

There are a few ways a bug like this happens, and it's bit tricky to
diagnose remotely. Does it happen consistently? Is your project on a
branch? Is the server and the client located in a physically different
timezone?

Maybe I should add a lot of debug logging for this part of the code so
that I can ask you to turn on the debug switch and send me the log.

--
Kohsuke Kawaguchi
Sun Microsystems                   [hidden email]

smime.p7s (4K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Jesse Glick
Kohsuke Kawaguchi wrote:
> I've seen a few bugs like that in the past, but to the best of my
> knowledge all of them are fixed by now.

I don't think so:

http://deadlock.nbextras.org/hudson/job/trunk/1102/changes
http://deadlock.nbextras.org/hudson/job/trunk/1102/console

Note the 2 unlisted file revisions.

> Does it happen consistently?

Rare, I think, but not sure.

> Is your project on a branch?

No.

> Is the server and the client located in a physically different
> timezone?

Not sure; in the above case the server first pulls changes from the CVS
server using cvsup (in response to a commit log mail), then runs Hudson
against a local mirror repo. The timezones could be different.

-J.

--
[hidden email]  x22801  netbeans.org  ant.apache.org
       http://google.com/search?q=e%5E%28pi*i%29%2B1

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Kohsuke Kawaguchi-2
Jesse Glick wrote:

> Kohsuke Kawaguchi wrote:
>> I've seen a few bugs like that in the past, but to the best of my
>> knowledge all of them are fixed by now.
>
> I don't think so:
>
> http://deadlock.nbextras.org/hudson/job/trunk/1102/changes
> http://deadlock.nbextras.org/hudson/job/trunk/1102/console
>
> Note the 2 unlisted file revisions.
Indeed. So it missed

U external/jaxws20.zip
U golden/files-layout.txt

Can you tell me the CVSROOT that this Hudson is configured to? I'd like
to run some tests myself.

>> Does it happen consistently?
>
> Rare, I think, but not sure.
>
>> Is your project on a branch?
>
> No.
>
>> Is the server and the client located in a physically different
>> timezone?
>
> Not sure; in the above case the server first pulls changes from the CVS
> server using cvsup (in response to a commit log mail), then runs Hudson
> against a local mirror repo. The timezones could be different.
Hudson makes a conservative estimate and asks for logs (of those files
whose changes are detected) starting 2 days before the build start time
(previously it was 1 day), and that was the last bug around this. So I
thought it might be related.


--
Kohsuke Kawaguchi
Sun Microsystems                   [hidden email]

smime.p7s (4K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Jesse Glick
Kohsuke Kawaguchi wrote:
> Can you tell me the CVSROOT that this Hudson is configured to?

Well it's a local CVSROOT as I said, using cvsup; the original CVS
server can be accessed e.g. as

   :pserver:[hidden email]:2401/shared/data/ccvs/repository

-J.

--
[hidden email]  x22801  netbeans.org  ant.apache.org
       http://google.com/search?q=e%5E%28pi*i%29%2B1

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Juozas Šalna
In reply to this post by Kohsuke Kawaguchi-2
On Fri, 17 Nov 2006 04:53:40 +0200, Kohsuke Kawaguchi  
<[hidden email]> wrote:

>
> Indeed. So it missed
>
> U external/jaxws20.zip
> U golden/files-layout.txt

>>> Does it happen consistently?
>> Rare, I think, but not sure.

We have the same error but we use hudson 1.57
will update soon.

I'd say one in ten builds leave changelog empty
or maybe one in twenty, but none the less quite often to see this.

Project is built from cvs branch.
and project repository has many branches.




--
juozas šalna

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

daniel.galan
In reply to this post by Kohsuke Kawaguchi-2
Hello again,

> > even we have changes in the CVS which have been recieved with the
> > following statement (from the console output), there are no
> entries in
> > the changes summary on the project page.
> >
> > [workspace] $ cvs -q -z9 update -PdC
> > P my/files..
> > U another/file

> There are a few ways a bug like this happens, and it's bit tricky to
> diagnose remotely. Does it happen consistently?
Yes

> Is your project on a branch?
We use branches as well, but this behaviour occures in every case.
Actually I noticed them on the CVS head/trunk. But it's the same behaviour in the branches.

> Is the server and the client located in a physically different timezone?
physically .. no.
The cvs repository is on the same machine.

> Maybe I should add a lot of debug logging for this part of
> the code so
> that I can ask you to turn on the debug switch and send me the log.

We use CVS update. But also with a full checkout it seems to have this bug.

I just noticed another bug, the changes he displays for a certain build references to another build, which was 2 builds ago.

:(

Greetings
Daniel

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Kohsuke Kawaguchi-2
[hidden email] wrote:

> Hello again,
>
>> > even we have changes in the CVS which have been recieved with the
>> > following statement (from the console output), there are no
>> entries in
>> > the changes summary on the project page.
>> >
>> > [workspace] $ cvs -q -z9 update -PdC
>> > P my/files..
>> > U another/file
>
>> There are a few ways a bug like this happens, and it's bit tricky to
>> diagnose remotely. Does it happen consistently?
> Yes
>
>> Is your project on a branch?
> We use branches as well, but this behaviour occures in every case.
> Actually I noticed them on the CVS head/trunk. But it's the same behaviour in the branches.
>
>> Is the server and the client located in a physically different timezone?
> physically .. no.
> The cvs repository is on the same machine.
>
>> Maybe I should add a lot of debug logging for this part of
>> the code so
>> that I can ask you to turn on the debug switch and send me the log.
>
> We use CVS update. But also with a full checkout it seems to have this bug.
I added the logging code around this. So please try the following steps:

1. download the snapshot build and overwrite your hudson.war with it
    (if your container is not set up for auto deployment, restart the
    container as well.)
2. check the bottom of the top page, to see if the timestamp correctly
    shows a snapshot build.
3. go to http://localhost:8080/hudson/script, and type in:

    hudson.scm.CVSSCM.debugLogging=true;
    print hudson.scm.CVSSCM.debugLogging;

4. hit "Run". You should see 'true' printed out in the result part.

5. perform a build. Now you should see a lot more messages in the
    console output. This should give me enough information.



I realized another possible cause of this bug. Imagine the cvs server
and Hudson running in different machines. A developer commits a change
on 9:00, but say the CVS server is little behind in clock and recorded
it as 8:50.

Then Hudson notices a change, and it now tries to run a build. Hudson
computes a changelog, and when it does so, it filters out the changelog
by the date span (from the timestamp of the previous build.) Let's say
Hudson's clock is accurate and so it only records changelogs from 8:55
(which is the previous build time, on Hudson's clock) to 9:00.

In this scenario, Hudson would miss a change.


> I just noticed another bug, the changes he displays for a certain build references to another build, which was 2 builds ago.
>
> :(

I guess you mean Hudson showed a change that actually happened 2 builds
ago, to the current build, right?


--
Kohsuke Kawaguchi
Sun Microsystems                   [hidden email]

smime.p7s (4K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

AW: Re: Changes not in summary

daniel.galan
Hi Kohsuke,

> I added the logging code around this. So please try the
> following steps:
>
> 1. download the snapshot build and overwrite your hudson.war with it
>     (if your container is not set up for auto deployment, restart the
>     container as well.)
done, but I get a 404 when trying to access /hudson, and no log is written.
This only happens when I use the snapshot :( Any idea?

> I realized another possible cause of this bug. Imagine the cvs server
> and Hudson running in different machines. A developer commits
> a change on 9:00, but say the CVS server is little behind in clock and
> recorded it as 8:50.
In my environment the cvs and hudson are running on the same machine.
Beside of this we build periodically (every half hour).

> In this scenario, Hudson would miss a change.
I don't think so, because of the fact that both run on the same machine.

> I guess you mean Hudson showed a change that actually
> happened 2 builds ago, to the current build, right?
erm, right :)

Greetings
Daniel

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: AW: Re: Changes not in summary

Kohsuke Kawaguchi-2
[hidden email] wrote:
> done, but I get a 404 when trying to access /hudson, and no log is written.
> This only happens when I use the snapshot :( Any idea?

Hmm. The only difference that Ican think of is that snapshot versions
might reuire Tiger. Could that be it?

I can release 1.64, if that's easier.


>> I realized another possible cause of this bug. Imagine the cvs server
>> and Hudson running in different machines. A developer commits
>> a change on 9:00, but say the CVS server is little behind in clock and
>> recorded it as 8:50.
> In my environment the cvs and hudson are running on the same machine.
> Beside of this we build periodically (every half hour).

OK. Hopefully the added logging would reveal more information. I also
noticed that Jesse made one change in this area, could that be related
to this, Jesse?


--
Kohsuke Kawaguchi
Sun Microsystems                   [hidden email]

smime.p7s (4K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: AW: Re: Changes not in summary

Jesse Glick
Kohsuke Kawaguchi wrote:
> I also noticed that Jesse made one change in this area, could that be
> related to this, Jesse?

I'm not sure either which change you are referring to or what exactly
the problem is that is being discussed...?

-J.

--
[hidden email]  x22801  netbeans.org  ant.apache.org
       http://google.com/search?q=e%5E%28pi*i%29%2B1

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: AW: Re: Changes not in summary

Kohsuke Kawaguchi-2
Jesse Glick wrote:
> Kohsuke Kawaguchi wrote:
>> I also noticed that Jesse made one change in this area, could that be
>> related to this, Jesse?
>
> I'm not sure either which change you are referring to or what exactly
> the problem is that is being discussed...?

Well, I saw you made one commit in CVSSCM this morning, which I thought
might be related to the general issue of Hudson missing changes in CVS
occasionally.

I looked at your commit message now, and it doesn't seem like it's
related. Sorry about the confusion.

--
Kohsuke Kawaguchi
Sun Microsystems                   [hidden email]

smime.p7s (4K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

AW: Re: AW: Re: Changes not in summary

daniel.galan
In reply to this post by Kohsuke Kawaguchi-2
> > done, but I get a 404 when trying to access /hudson, and no log is
> > written. This only happens when I use the snapshot :( Any idea?
> Hmm. The only difference that Ican think of is that snapshot versions
> might reuire Tiger. Could that be it?
I don't think so, we use linux, not Mac OS X. Strange behaviour, maybe the tomcat is wrong configured.

> I can release 1.64, if that's easier.
I would be pleased :)

Thanks
Daniel

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

daniel.galan
In reply to this post by Kohsuke Kawaguchi-2
> > done, but I get a 404 when trying to access /hudson, and no log is
> > written. This only happens when I use the snapshot :( Any idea?
> I can release 1.64, if that's easier.

Ok, I installed 1.64 and executed:
    hudson.scm.CVSSCM.debugLogging=true;
    print hudson.scm.CVSSCM.debugLogging;

(with result true). After that i checked in a dummy file and started a build.
All output related to hudson I got on the console was (catalina.out):

Nov 22, 2006 9:41:14 AM org.apache.catalina.startup.Catalina start
INFO: Server startup in 5999 ms
Nov 22, 2006 9:44:39 AM hudson.model.Run run
INFO: immonet #253 main build action completed: SUCCESS

thats it, nothing more. do I have to configure anything else?

I thought that the loglevel was not debug, so I added a log4j.properties file in tomcat/common/classes
(doesn't changed the output to catalina.out)

log4j.rootLogger=debug, R
log4j.appender.R=org.apache.log4j.RollingFileAppender
log4j.appender.R.File=${catalina.home}/logs/tomcat.log
log4j.appender.R.MaxFileSize=10MB
log4j.appender.R.MaxBackupIndex=10
log4j.appender.R.layout=org.apache.log4j.PatternLayout
log4j.appender.R.layout.ConversionPattern=%p %t %c - %m%n
log4j.logger.org.apache.catalina=DEBUG, R

(taken from: http://tomcat.apache.org/tomcat-5.5-doc/logging.html)

Greetings
Daniel

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: Changes not in summary

Kohsuke Kawaguchi
Administrator
The output should have gone to the console of the build, not to the
system log screen. Sorry for confusing property name.

2006/11/22, [hidden email] <[hidden email]>:

> > > done, but I get a 404 when trying to access /hudson, and no log is
> > > written. This only happens when I use the snapshot :( Any idea?
> > I can release 1.64, if that's easier.
>
> Ok, I installed 1.64 and executed:
>     hudson.scm.CVSSCM.debugLogging=true;
>     print hudson.scm.CVSSCM.debugLogging;
>
> (with result true). After that i checked in a dummy file and started a build.
> All output related to hudson I got on the console was (catalina.out):
>
> Nov 22, 2006 9:41:14 AM org.apache.catalina.startup.Catalina start
> INFO: Server startup in 5999 ms
> Nov 22, 2006 9:44:39 AM hudson.model.Run run
> INFO: immonet #253 main build action completed: SUCCESS
>
> thats it, nothing more. do I have to configure anything else?
>
> I thought that the loglevel was not debug, so I added a log4j.properties file in tomcat/common/classes
> (doesn't changed the output to catalina.out)
>
> log4j.rootLogger=debug, R
> log4j.appender.R=org.apache.log4j.RollingFileAppender
> log4j.appender.R.File=${catalina.home}/logs/tomcat.log
> log4j.appender.R.MaxFileSize=10MB
> log4j.appender.R.MaxBackupIndex=10
> log4j.appender.R.layout=org.apache.log4j.PatternLayout
> log4j.appender.R.layout.ConversionPattern=%p %t %c - %m%n
> log4j.logger.org.apache.catalina=DEBUG, R
>
> (taken from: http://tomcat.apache.org/tomcat-5.5-doc/logging.html)
>
> Greetings
> Daniel
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [hidden email]
> For additional commands, e-mail: [hidden email]
>
>


--
--
Kohsuke Kawaguchi

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]

Reply | Threaded
Open this post in threaded view
|

AW: Re: Changes not in summary

daniel.galan
> The output should have gone to the console of the build, not
> to the system log screen. Sorry for confusing property name.
No problem, I filtered the log and found following parts useful
(I cloaked some parts (like path and package infos)):

<start>

[workspace] $ cvs -q -z9 update -PdC
P project/src/de/packages/MyClass.java
$ computing changelog
Running log at /userhome/.hudson/jobs/project/workspace
Using cvs passfile: /userhome/.cvspass
Executing 'cvs' with arguments:
'-d:pserver:user@host:/path/to/cvs/repository'
'log'
'-d'
'>=2006-11-21'
'project/src/de/packages/MyClass.java'

[..snip..]

keyword substitution: kv
total revisions: 13; selected revisions: 2
description:
----------------------------
revision 1.11
Recorded a change: 2006-11-22 17:31:29,dgalan,1.11(branch=null),CVS Message goes here
date: 2006-11-22 17:31:29 +0100;  author: dgalan;  state: Exp;  lines: +3 -4;  commitid: 3bb945647b614567;
CVS Message goes here
----------------------------
revision 1.9.4.2
date: 2006-11-22 17:32:33 +0100;  author: dgalan;  state: Exp;  lines: +3 -4;  commitid: 3eb345647ba14567;
CVS Message goes here
Recorded a change: 2006-11-22 17:32:33,dgalan,1.9.4.2(branch=mybranch),CVS Message goes here
=============================================================================
retCode=0
Filtering entries
Filtering out [dgalan,Wed Nov 22 18:31:29 CET 2006,[hudson.org.apache.tools.ant.taskdefs.cvslib.RCSFile@4065c4],CVS Message goes here] because it's too late compare to Wed Nov 22 17:31:27 CET 2006
Filtering out [dgalan,Wed Nov 22 18:32:33 CET 2006,[hudson.org.apache.tools.ant.taskdefs.cvslib.RCSFile@1d92803],CVS Message goes here] because it's too late compare to Wed Nov 22 17:31:27 CET 2006

<end>

As we can see, hudson is filtering the changes, because the time is somehow not in sync.
A date on the server gives me "Wed Nov 22 17:58:21 CET 2006"
So the problems come from "18:31:29" and "18:32:33". Where does they come from?

Is it possible, that the CVS works internally only in GMT?

Greetings
Daniel

(ps.: And happy Vacation to you Kohsuke :)

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]