beehive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rich Feit <richf...@gmail.com>
Subject Re: Beehive/tooling issue
Date Fri, 09 Sep 2005 16:22:44 GMT
Sounds good -- will do.
Rich

Eddie O'Neil wrote:

>  Yeah -- that would work.  I always run that target in a clean sync,
>so I don't tend to see SVN version numbers that reflect local changes.
> Feel free to just add that to the block that determines the version.
>
>Eddie
>
>
>
>On 9/9/05, Rich Feit <richfeit@gmail.com> wrote:
>  
>
>>OK, great.  Thanks for taking a look at it so quickly.
>>
>>Does anyone have any opposition to this fix getting checked in?
>>
>>Also, one quick thing:  the nightly-test script failed on my Windows box
>>because it tried to create a directory with ':' in it:
>>     build\dist\apache-beehive-20050909-svn279336:279683M
>>
>>I think all we need to do is filter the output of 'svnversion' in
>>nightly.xml, to replace the ':' with something else.  Does that make sense?
>>
>>Rich
>>
>>Eddie O'Neil wrote:
>>
>>    
>>
>>>Rich--
>>>
>>> The diff looks good -- making this change definitely makes the
>>>webapp build process simpler as it removes another tmp directory to
>>>clean / manage.
>>>
>>> Nice...
>>>
>>>Eddie
>>>
>>>
>>>
>>>On 9/9/05, Rich Feit <richfeit@gmail.com> wrote:
>>>
>>>
>>>      
>>>
>>>>Ah, that's great.  I'd been doing a build.dist.full, then running a
>>>>script to expand the dists and run tests against them.  This is nicer.  :)
>>>>
>>>>Rich
>>>>
>>>>Eddie O'Neil wrote:
>>>>
>>>>
>>>>
>>>>        
>>>>
>>>>>You can run a full-on distribution test run by:
>>>>>
>>>>>cd trunk/ant
>>>>>ant -f nightly.xml run
>>>>>
>>>>>It should run end-to-end without any trouble, though there are
>>>>>occasionally intermittant failures during controls testing on Linux.
>>>>>
>>>>>Eddie
>>>>>
>>>>>
>>>>>
>>>>>On 9/9/05, Rich Feit <richfeit@gmail.com> wrote:
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>          
>>>>>
>>>>>>That's great, thanks -- hopefully there'll be agreement on this. 
FYI,
>>>>>>this passes tests in the tree (will be important to test against the
>>>>>>distributions though).
>>>>>>
>>>>>>Anyone else have comments about this?
>>>>>>
>>>>>>Rich
>>>>>>
>>>>>>Eddie O'Neil wrote:
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>            
>>>>>>
>>>>>>>I'd fix it too.  :)  It makes sense to have the generated
>>>>>>>struts-config files end up in WEB-INF/classes because they're
really
>>>>>>>not meant to be read / write configuration files (like web.xml
or
>>>>>>>beehive-netui-config.xml).
>>>>>>>
>>>>>>>Also, better to ship 1.0 without having to change the behavior
later.
>>>>>>>
>>>>>>>Will take a look over the patch...
>>>>>>>
>>>>>>>Eddie
>>>>>>>
>>>>>>>
>>>>>>>On 9/9/05, Rich Feit <richfeit@gmail.com> wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>              
>>>>>>>
>>>>>>>>OK, I've added a patch for this in the bug.  Feel free to
give it a good
>>>>>>>>once-over.
>>>>>>>>
>>>>>>>>I'm running tests now -- looks OK so far.  The only ill effect
of this
>>>>>>>>change is one which would only get worse over time (the longer
we wait
>>>>>>>>to do it): for *legacy* apps, where the root Struts-config
file path is
>>>>>>>>specified in web.xml, we can no longer honor that path when
we generate
>>>>>>>>the file.  In these apps, you will get an error logged at
Servlet
>>>>>>>>startup about Struts not being able to load the file (though
everything
>>>>>>>>does work fine).
>>>>>>>>
>>>>>>>>Rich
>>>>>>>>
>>>>>>>>Rich Feit wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>                
>>>>>>>>
>>>>>>>>>Hi all,
>>>>>>>>>
>>>>>>>>>I just entered http://issues.apache.org/jira/browse/BEEHIVE-915
: Page
>>>>>>>>>Flow annotation processors generate files in an IDE-unfriendly
way.  It
>>>>>>>>>turns out that the way we generate files when running
under apt works
>>>>>>>>>fine on the command line, but makes it hard for an IDE
(implementing the
>>>>>>>>>Mirror interfaces) to know when/where our generated files
are created.
>>>>>>>>>
>>>>>>>>>This from the bug:
>>>>>>>>>
>>>>>>>>>"Unfortunately, [apt's] Filer offers only two choices
of places to
>>>>>>>>>create files: the source directory and the build directory.
This means
>>>>>>>>>that to fix this bug, our generated files need to move
out of WEB-INF,
>>>>>>>>>into WEB-INF/classes, i.e., they will not only be generated
into a
>>>>>>>>>different place, but they will be read through a different
mechanism in
>>>>>>>>>the runtime."
>>>>>>>>>
>>>>>>>>>So, this is definitely not a trivial change.  I am working
on the fix
>>>>>>>>>for this, and I'd like to have the discussion about whether
to get it
>>>>>>>>>into v1.0.  If we don't, v1.0 won't be toolable in an
IDE.
>>>>>>>>>Additionally, it could cause back-compat problems between
this and the
>>>>>>>>>next version, if people begin to depend on our current
location for
>>>>>>>>>generated files.  On the other hand, it's a risky change
in that we
>>>>>>>>>don't have a lot of time to have people hammer on this.
 The one saving
>>>>>>>>>grace is that the file location is so fundamental that
presumably, any
>>>>>>>>>bug would cause blatantly bad behavior across the framework.
>>>>>>>>>
>>>>>>>>>Let me know what you think.
>>>>>>>>>
>>>>>>>>>Thanks,
>>>>>>>>>Rich
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>                  
>>>>>>>>>
>>>>>>>
>>>>>>>              
>>>>>>>
>>>>>
>>>>>          
>>>>>
>>>
>>>      
>>>
>
>  
>

Mime
View raw message