Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Aaron Mulder

Dain,

I fixed GERONIMO-2270 (redeploy when module ID includes version but
not type) but this did not fix GERONIMO-2269 (redeploy with
resource-ref when module ID includes artifact only).  There's a stack
trace in the Jira -- any suggestions on how to pursue it?

http://issues.apache.org/jira/browse/GERONIMO-2269


Matt,

I was assuming the point of branches and RC builds was to test for
obvious problems, and I think the deploy tool is fairly central to the
user experience, and since we appear to have some TCK issues anyway...
Well, whatever, I attached the patch to GERONIMO-2270 and you can
apply it or not as you see fit.

Thanks,
Aaron

On 8/3/06, Matt Hogstrom <[EMAIL PROTECTED]> wrote:

Aaron,

I missed the cut off for 1.1.1 on the three JIRA's I had.  I'll move the 
remaining 1.1.1s to 1.1.2
and then we'll cut a 1.1.2 a few weeks after 1.1.1.   Perhaps this is the start 
of the fast and
furious release cycle that Hiram loves so much :)


Dain Sundstrom wrote:
> On Aug 3, 2006, at 9:18 AM, Aaron Mulder wrote:
>
>> I think this is a blocker.
>>
>> I tried redeploying a web application, both with and without a version
>> in the module ID.  It failed both ways, though differently for each.
>>
>> Matt, do you mind if we work on this in the 1.1.1 branch?
>
> I'd say it should be worked on in 1.1.2.
>
>> Dain or David Jencks, any chance you could look at the "without a
>> version in the module ID" case?  It complained about a dead proxy when
>> Spring accessed a JDBC pool at java:comp/env/jdbc/DataSource during
>> startup of the new version (GERONIMO-2269)
>>
>> Thanks,
>> Aaron
>>
>> http://issues.apache.org/jira/browse/GERONIMO-2269
>
> My guess is this one is a side effect of the next one, so I'd fix 2270
> first and rerun this test case.
>
>> http://issues.apache.org/jira/browse/GERONIMO-2270
>
> Didn't you write the redeploy code?
>
> -dain
>
>
>
>



Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Matt Hogstrom

Aaron,

I missed the cut off for 1.1.1 on the three JIRA's I had.  I'll move the remaining 1.1.1s to 1.1.2 
and then we'll cut a 1.1.2 a few weeks after 1.1.1.   Perhaps this is the start of the fast and 
furious release cycle that Hiram loves so much :)



Dain Sundstrom wrote:

On Aug 3, 2006, at 9:18 AM, Aaron Mulder wrote:


I think this is a blocker.

I tried redeploying a web application, both with and without a version
in the module ID.  It failed both ways, though differently for each.

Matt, do you mind if we work on this in the 1.1.1 branch?


I'd say it should be worked on in 1.1.2.


Dain or David Jencks, any chance you could look at the "without a
version in the module ID" case?  It complained about a dead proxy when
Spring accessed a JDBC pool at java:comp/env/jdbc/DataSource during
startup of the new version (GERONIMO-2269)

Thanks,
Aaron

http://issues.apache.org/jira/browse/GERONIMO-2269


My guess is this one is a side effect of the next one, so I'd fix 2270 
first and rerun this test case.



http://issues.apache.org/jira/browse/GERONIMO-2270


Didn't you write the redeploy code?

-dain






Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Aaron Mulder

OK, well, I'm working on it.  :)

Thanks,
Aaron

On 8/3/06, Dain Sundstrom <[EMAIL PROTECTED]> wrote:

On Aug 3, 2006, at 10:11 AM, Aaron Mulder wrote:

> On 8/3/06, Dain Sundstrom <[EMAIL PROTECTED]> wrote:
>> > http://issues.apache.org/jira/browse/GERONIMO-2269
>>
>> My guess is this one is a side effect of the next one, so I'd fix
>> 2270 first and rerun this test case.
>
> I will, though I disagree with your guess -- 2270 is a problem before
> the call ever gets to the server, while 2269 is a problem after the
> call has gone to the ConfigurationManager and the app has been
> distributed and the new version is being started.

Sorry, I don't understand what you are saying.

My guess is based on this text "UPDATE: a simple redeploy of the
working application causes the error"  If there is a problem during
redeploy, I'd question any JIRA about problems after that.

>> > http://issues.apache.org/jira/browse/GERONIMO-2270
>>
>> Didn't you write the redeploy code?
>
> No, couldn't have been me -- my code never has bugs!  :)  I'm sure I
> should look at 2270.  I definitely can't take full credit for the
> version-less redeploy because that goes through ConfigurationManager,
> and I dropped the ball on trying to refactor that code.

I wrote big chuncks of this, but I think you wrote the part that
matches versionless ids to preexisting installations.

> But the more relevant question probably is -- how can we end up with a
> dead proxy error when looking up a JNDI resource?  The database pool
> wasn't redeployed so a reference to it shouldn't become dead, and the
> pool isn't in the web app module so the version number for the
> database pool shouldn't have changed...  Its a little confusing.

My guess is the pool was actually redeployed, but the only way to
tell is to sick a breakpoint in the code.

-dain



Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Dain Sundstrom

On Aug 3, 2006, at 10:11 AM, Aaron Mulder wrote:


On 8/3/06, Dain Sundstrom <[EMAIL PROTECTED]> wrote:

> http://issues.apache.org/jira/browse/GERONIMO-2269

My guess is this one is a side effect of the next one, so I'd fix
2270 first and rerun this test case.


I will, though I disagree with your guess -- 2270 is a problem before
the call ever gets to the server, while 2269 is a problem after the
call has gone to the ConfigurationManager and the app has been
distributed and the new version is being started.


Sorry, I don't understand what you are saying.

My guess is based on this text "UPDATE: a simple redeploy of the  
working application causes the error"  If there is a problem during  
redeploy, I'd question any JIRA about problems after that.



> http://issues.apache.org/jira/browse/GERONIMO-2270

Didn't you write the redeploy code?


No, couldn't have been me -- my code never has bugs!  :)  I'm sure I
should look at 2270.  I definitely can't take full credit for the
version-less redeploy because that goes through ConfigurationManager,
and I dropped the ball on trying to refactor that code.


I wrote big chuncks of this, but I think you wrote the part that  
matches versionless ids to preexisting installations.



But the more relevant question probably is -- how can we end up with a
dead proxy error when looking up a JNDI resource?  The database pool
wasn't redeployed so a reference to it shouldn't become dead, and the
pool isn't in the web app module so the version number for the
database pool shouldn't have changed...  Its a little confusing.


My guess is the pool was actually redeployed, but the only way to  
tell is to sick a breakpoint in the code.


-dain


Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Aaron Mulder

On 8/3/06, Dain Sundstrom <[EMAIL PROTECTED]> wrote:

> http://issues.apache.org/jira/browse/GERONIMO-2269

My guess is this one is a side effect of the next one, so I'd fix
2270 first and rerun this test case.


I will, though I disagree with your guess -- 2270 is a problem before
the call ever gets to the server, while 2269 is a problem after the
call has gone to the ConfigurationManager and the app has been
distributed and the new version is being started.


> http://issues.apache.org/jira/browse/GERONIMO-2270

Didn't you write the redeploy code?


No, couldn't have been me -- my code never has bugs!  :)  I'm sure I
should look at 2270.  I definitely can't take full credit for the
version-less redeploy because that goes through ConfigurationManager,
and I dropped the ball on trying to refactor that code.

But the more relevant question probably is -- how can we end up with a
dead proxy error when looking up a JNDI resource?  The database pool
wasn't redeployed so a reference to it shouldn't become dead, and the
pool isn't in the web app module so the version number for the
database pool shouldn't have changed...  Its a little confusing.

Thanks,
Aaron


Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Dain Sundstrom

On Aug 3, 2006, at 9:18 AM, Aaron Mulder wrote:


I think this is a blocker.

I tried redeploying a web application, both with and without a version
in the module ID.  It failed both ways, though differently for each.

Matt, do you mind if we work on this in the 1.1.1 branch?


I'd say it should be worked on in 1.1.2.


Dain or David Jencks, any chance you could look at the "without a
version in the module ID" case?  It complained about a dead proxy when
Spring accessed a JDBC pool at java:comp/env/jdbc/DataSource during
startup of the new version (GERONIMO-2269)

Thanks,
Aaron

http://issues.apache.org/jira/browse/GERONIMO-2269


My guess is this one is a side effect of the next one, so I'd fix  
2270 first and rerun this test case.



http://issues.apache.org/jira/browse/GERONIMO-2270


Didn't you write the redeploy code?

-dain



Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Aaron Mulder

On 8/3/06, Matt Hogstrom <[EMAIL PROTECTED]> wrote:

Is this a regression or is the first test?


It looks like 1.1 has the same problems for the same test application.
I haven't tried this particular application before -- it is a Spring
app I pulled off the shelf to make sure 1.1.1 didn't have problems
with Spring/Hibernate web apps.

Still, I'd really rather not ship 1.1.1 with known errors in the deploy tool.

Thanks,
Aaron


Aaron Mulder wrote:
> I think this is a blocker.
>
> I tried redeploying a web application, both with and without a version
> in the module ID.  It failed both ways, though differently for each.
>
> Matt, do you mind if we work on this in the 1.1.1 branch?
>
> Dain or David Jencks, any chance you could look at the "without a
> version in the module ID" case?  It complained about a dead proxy when
> Spring accessed a JDBC pool at java:comp/env/jdbc/DataSource during
> startup of the new version (GERONIMO-2269)
>
> Thanks,
> Aaron
>
> http://issues.apache.org/jira/browse/GERONIMO-2269
> http://issues.apache.org/jira/browse/GERONIMO-2270
>
>
>



Re: Redeploy problems in branches/1.1.1

2006-08-03 Thread Matt Hogstrom

Is this a regression or is the first test?

Aaron Mulder wrote:

I think this is a blocker.

I tried redeploying a web application, both with and without a version
in the module ID.  It failed both ways, though differently for each.

Matt, do you mind if we work on this in the 1.1.1 branch?

Dain or David Jencks, any chance you could look at the "without a
version in the module ID" case?  It complained about a dead proxy when
Spring accessed a JDBC pool at java:comp/env/jdbc/DataSource during
startup of the new version (GERONIMO-2269)

Thanks,
Aaron

http://issues.apache.org/jira/browse/GERONIMO-2269
http://issues.apache.org/jira/browse/GERONIMO-2270