You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@gump.apache.org by Eric Pugh <ep...@opensourceconnections.com> on 2004/11/29 20:32:05 UTC
Picking up the ball from Niclas (ugh!) on Velocity
So,
I thought I would just submit a patch to Velocity to fix the velocity/log4j
problem and everything would be fine. Wrong.. The issue is the non
backwards compatible API change to log4j. And, I saw Niclas email about it
as well.
So, I dug some more and according to this email:
http://www.archivum.info/log4j-dev@logging.apache.org/2004-11/msg00271.html
there is a binary version. So I thought "hey, use that".. Only to discover
that Gump can check out a tagged branch of code! And that there is a
project called logging-log4j-12.xml already building that most dependees of
log4j use.
So, I made the change, and that should fix Velocity... Of course, it
basically is like dynamically building a package.. Fulcrum Configuration
relies on the older version of commons-configuration. I am going to apply
the same trick there.
Eric
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Stefano Mazzocchi <st...@apache.org>.
Niclas Hedhman wrote:
>>The rule should be to start fixing gump and fix the communication
>>channels rather than fixing the metadata to route around the problems.
>
> Cool, but then you told me "Don't change people" & "we should work around them
> [projects not willing to co-operate]"...
>
> I would call that mixed signals... ;o)
Correct. We don't change people, we try to route around them, but we try
first :-)
Log4j has been, historically, a very collaborative project, but they (or
just Ceki?) seemed to have turned sour. Can't tell why.
I don't think he realizes that the reason why people used log4j was
exactly this collaborative attitude and, without that, it will be a lot
harder to market it.
We tried, he's not listening, so we route around him.
--
Stefano.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Geir Magnusson Jr <ge...@4quarters.com>.
On Nov 30, 2004, at 8:18 AM, Niclas Hedhman wrote:
> On Tuesday 30 November 2004 20:18, Geir Magnusson Jr wrote:
>
>> On Nov 30, 2004, at 6:19 AM, Eric Pugh wrote:
>>> Geir's email highlights a very clear issue with the whole
>>> deprecation/version cycle. He can't switch to log4j until it
>>> releases.
>>> They don't want to keep deprecated code around for forever.
>>
>> I'm not asking for forever. Just one version... even a short lived
>> one...
>>
>> My take away lesson is really more of a reaffirmation - don't depend
>> on
>> outside things if you can help it...
>
> Log4J has otherwise been very generous with having deprecated code
> around, but
> when I asked explicitly if this was a chagne for 1.3 and whether that
> means
> that everyone who uses RFA have to change the config files, I got from
> Ceki
> the answer; "Yes" to those two questions.
> Very strange change of attitude, and I don't know why this is
> happening.
Yes, historically, log4j has bent over backwards to make things easy
for users.
geir
>
>
> Cheers
> Niclas
> --
> +------//-------------------+
> / http://www.dpml.net /
> / http://niclas.hedhman.org /
> +------//-------------------+
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
> For additional commands, e-mail: general-help@gump.apache.org
>
>
--
Geir Magnusson Jr +1-203-665-6437
geir@gluecode.com
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Niclas Hedhman <ni...@hedhman.org>.
On Tuesday 30 November 2004 20:18, Geir Magnusson Jr wrote:
> On Nov 30, 2004, at 6:19 AM, Eric Pugh wrote:
> > Geir's email highlights a very clear issue with the whole
> > deprecation/version cycle. He can't switch to log4j until it releases.
> > They don't want to keep deprecated code around for forever.
>
> I'm not asking for forever. Just one version... even a short lived
> one...
>
> My take away lesson is really more of a reaffirmation - don't depend on
> outside things if you can help it...
Log4J has otherwise been very generous with having deprecated code around, but
when I asked explicitly if this was a chagne for 1.3 and whether that means
that everyone who uses RFA have to change the config files, I got from Ceki
the answer; "Yes" to those two questions.
Very strange change of attitude, and I don't know why this is happening.
Cheers
Niclas
--
+------//-------------------+
/ http://www.dpml.net /
/ http://niclas.hedhman.org /
+------//-------------------+
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Geir Magnusson Jr <ge...@4quarters.com>.
On Nov 30, 2004, at 6:19 AM, Eric Pugh wrote:
> I think what he means is that we can't expect people to make changes
> just to
> make Gump happy. But, we can *attempt* to influence them to help.
> And I
> think that is where we are going wrong. For instance, Fulcrum
> components
> didn't build very well until I got involved (and got lots of help!)..
> You
> need a committer from the project being gumped to keep things running
> well.
> As far as I know, there are no log4j committers actively involved in
> keeping
> gump working. Hence, they made a backwards incompatible change, and
> the
> fact that gump keeled over doesn't bother then.
>
> Now, partly that may be a communication thing.. If Log4j fails, they
> get
> emailed. If log4j breaks every body else, they don't... Without
> active
> involvement by a group, the prospect of keeping things working becomes
> a
> thankless task (witness Niclas's frustration). I thought "hey, I'll
> try
> and help" and ran into, in an hour, the same frustration Niclas sees..
> I am
> not a committer (or even involved beyond the occasional email) with
> Log4j or
> Velocity, so who do I got to prod for action?
>
> Geir's email highlights a very clear issue with the whole
> deprecation/version cycle. He can't switch to log4j until it releases.
> They don't want to keep deprecated code around for forever.
I'm not asking for forever. Just one version... even a short lived
one...
My take away lesson is really more of a reaffirmation - don't depend on
outside things if you can help it...
--
Geir Magnusson Jr +1-203-665-6437
geir@gluecode.com
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
RE: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Eric Pugh <ep...@opensourceconnections.com>.
I think what he means is that we can't expect people to make changes just to
make Gump happy. But, we can *attempt* to influence them to help. And I
think that is where we are going wrong. For instance, Fulcrum components
didn't build very well until I got involved (and got lots of help!).. You
need a committer from the project being gumped to keep things running well.
As far as I know, there are no log4j committers actively involved in keeping
gump working. Hence, they made a backwards incompatible change, and the
fact that gump keeled over doesn't bother then.
Now, partly that may be a communication thing.. If Log4j fails, they get
emailed. If log4j breaks every body else, they don't... Without active
involvement by a group, the prospect of keeping things working becomes a
thankless task (witness Niclas's frustration). I thought "hey, I'll try
and help" and ran into, in an hour, the same frustration Niclas sees.. I am
not a committer (or even involved beyond the occasional email) with Log4j or
Velocity, so who do I got to prod for action?
Geir's email highlights a very clear issue with the whole
deprecation/version cycle. He can't switch to log4j until it releases.
They don't want to keep deprecated code around for forever. I'll argue that
neither party is at fault. It's just an icky sore spot that will be worked
out at somepoint, and gump is just bringing up the incompatiblity sooner.
I'd like Gump to now just say "Yes, log4j HEAD fails on Velocity, lets step
back to the last successful build with Velocity and use that", or, we work
around it by building in Gump the older version (logging_log4j_12) and use
that.
I don't specifically think it is a tooling issue.. This stuff is hard, and
while some rote stuff could be made simpler with better tooling, we are
always going to have odd situations to deal with.
So, I'll again argue about the benefit of packages/custom branches.. We
need to build up mindshare the gump is this great build tool/continous
integration tool. But we certainly can't wait a couple years, and I think
that certain components that break frequently need to be packaged to provide
some stability, as well as pruning some leaves that never build. Anything
that hasn't built in the last 300 cycles for instance should be canned! At
least, until we get some stability in gump...
Eric
> -----Original Message-----
> From: Niclas Hedhman [mailto:niclas@hedhman.org]
> Sent: Tuesday, November 30, 2004 6:36 AM
> To: Gump code and data
> Subject: Re: Picking up the ball from Niclas (ugh!) on Velocity
>
>
> On Tuesday 30 November 2004 07:45, Stefano Mazzocchi wrote:
>
> > Gump is about establishing communication channels between development
> > communities.
>
> > On the other hand, if you are interested in creating a social
> > engineering support tool and you are willing to get your hands dirty in
> > python and prepare to have a huge ton of patience to convert a few
> > hundreds people to a very difficult concept,
>
> > The rule should be to start fixing gump and fix the communication
> > channels rather than fixing the metadata to route around the problems.
>
> Cool, but then you told me "Don't change people" & "we should
> work around them
> [projects not willing to co-operate]"...
>
> I would call that mixed signals... ;o)
>
> Cheers
> Niclas
>
> P.S. Let me know when you figured out that Java is a better tool
> after all :o)
> so I can help more actively.
>
>
> --
> +------//-------------------+
> / http://www.dpml.net /
> / http://niclas.hedhman.org /
> +------//-------------------+
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
> For additional commands, e-mail: general-help@gump.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Niclas Hedhman <ni...@hedhman.org>.
On Tuesday 30 November 2004 07:45, Stefano Mazzocchi wrote:
> Gump is about establishing communication channels between development
> communities.
> On the other hand, if you are interested in creating a social
> engineering support tool and you are willing to get your hands dirty in
> python and prepare to have a huge ton of patience to convert a few
> hundreds people to a very difficult concept,
> The rule should be to start fixing gump and fix the communication
> channels rather than fixing the metadata to route around the problems.
Cool, but then you told me "Don't change people" & "we should work around them
[projects not willing to co-operate]"...
I would call that mixed signals... ;o)
Cheers
Niclas
P.S. Let me know when you figured out that Java is a better tool after all :o)
so I can help more actively.
--
+------//-------------------+
/ http://www.dpml.net /
/ http://niclas.hedhman.org /
+------//-------------------+
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
RE: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Stephen McConnell <mc...@apache.org>.
> -----Original Message-----
> From: Stefano Mazzocchi [mailto:stefano@apache.org]
> Gump is about establishing communication channels between development
> communities.
ROTFLOL
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Picking up the ball from Niclas (ugh!) on Velocity
Posted by Stefano Mazzocchi <st...@apache.org>.
Eric Pugh wrote:
> So,
>
> I thought I would just submit a patch to Velocity to fix the velocity/log4j
> problem and everything would be fine. Wrong.. The issue is the non
> backwards compatible API change to log4j. And, I saw Niclas email about it
> as well.
>
> So, I dug some more and according to this email:
> http://www.archivum.info/log4j-dev@logging.apache.org/2004-11/msg00271.html
>
> there is a binary version. So I thought "hey, use that".. Only to discover
> that Gump can check out a tagged branch of code! And that there is a
> project called logging-log4j-12.xml already building that most dependees of
> log4j use.
>
> So, I made the change, and that should fix Velocity... Of course, it
> basically is like dynamically building a package.. Fulcrum Configuration
> relies on the older version of commons-configuration. I am going to apply
> the same trick there.
I have spoken to Geir Magnuson, who was not aware that velocity was
being a problem and I indicated the problem and he said he would solve
it in a few minutes.
People, Gump is *NOT* about 100% success.
Gump is about establishing communication channels between development
communities.
If you are interested in the first and not in the second, run a cron job
every night on your build files and send your list an email if that
build goes wrong. it's a two seconds job and I'll be happy to give
access to brutus if you need that.
On the other hand, if you are interested in creating a social
engineering support tool and you are willing to get your hands dirty in
python and prepare to have a huge ton of patience to convert a few
hundreds people to a very difficult concept, then stick around for a few
more years, because that's how much it will take before gump can be
considered a success.
The recent 'velocity stall' indicates that our way to signal problems is
too noisy to convey information and Eric identified a few issues that I
also have.
The rule should be to start fixing gump and fix the communication
channels rather than fixing the metadata to route around the problems.
--
Stefano.
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org