Hudson setup status

[Andrus Adamchik <an...@objectstyle.org>]

Looks like I was able to fix configurations for Cayenne-trunk and  
Cayenne-30. So now we have two fully working builds, and can improve  
upon the existing baseline setup. A few comments explaining what I had  
to fix:

1. I was wrong about the "Ubuntu" group including Hadoop* machines.  
Ubuntu = minerva + vesta. So both build configurations are explicitly  
clustered across the group.

2. Both minerva and vesta contain identical ~hudson/.cayenne/ 
connection.properties files. We need to manually keep those in sync.  
At the same time we don't really need one on hudson.zones.apache.org.

3. Cayenne-30 does not support cayenneTestConnection property and I am  
not comfortable patching it from trunk at this point. And as Ari  
noticed, using cayenne.test.connection notation doesn't play well with  
Hudson (I can't imagine why, but I can confirm this finding). As a  
result we can't run 3.0 on Hudson with Derby or H2. So I removed those  
options. Now it runs with Java 5 & 6 on HSQLDB only.

4. Cayenne-trunk on the other hand can support specific DB's (see #2  
above), and so it does. Derby runs without test failures BTW. However  
I've seen occasional test failures that can be attributed to race  
conditions in the tests (those happen on local machines as well). One  
day we are going to track them down.

5. Builds themselves are pretty fast. Usually the slowest step is SVN  
checkout.


Andrus

Re: Hudson setup status

[Aristedes Maniatis <ar...@maniatis.org>]

On 5/01/10 7:28 AM, Andrus Adamchik wrote:
>
> 1. I was wrong about the "Ubuntu" group including Hadoop* machines.
> Ubuntu = minerva + vesta. So both build configurations are explicitly
> clustered across the group.
>
> 2. Both minerva and vesta contain identical
> ~hudson/.cayenne/connection.properties files. We need to manually keep
> those in sync. At the same time we don't really need one on
> hudson.zones.apache.org.

I was thinking that in the longer term we could get rid of that file for Hudson purposes and just hardcode some defaults right into a file within svn. After all, HSQLDB is hardcoded in.

> 3. Cayenne-30 does not support cayenneTestConnection property and I am
> not comfortable patching it from trunk at this point. And as Ari
> noticed, using cayenne.test.connection notation doesn't play well with
> Hudson (I can't imagine why, but I can confirm this finding).

It is because Hudson uses Groovy for scripting. And Groovy gets confused by variable names with dots.

>  As a
> result we can't run 3.0 on Hudson with Derby or H2. So I removed those
> options. Now it runs with Java 5 & 6 on HSQLDB only.

I think that is a good option for now.


> 4. Cayenne-trunk on the other hand can support specific DB's (see #2
> above), and so it does. Derby runs without test failures BTW. However
> I've seen occasional test failures that can be attributed to race
> conditions in the tests (those happen on local machines as well). One
> day we are going to track them down.
>
> 5. Builds themselves are pretty fast. Usually the slowest step is SVN
> checkout.


Hey, maybe if we are game we could try building against Harmony. I wonder if that would work...

Ari


-- 

-------------------------->
Aristedes Maniatis
GPG fingerprint CBFB 84B4 738D 4E87 5E5C  5EFA EF6A 7D2E 3E49 102A