geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dain Sundstrom <dsundst...@gluecode.com>
Subject Re: [PROPOSAL] Next milestone release (M4?)
Date Tue, 29 Mar 2005 17:04:27 GMT
On Mar 29, 2005, at 8:30 AM, Bruce Snyder wrote:

> On Tue, 29 Mar 2005 07:40:13 -0800, Jeremy Boynes <jboynes@apache.org> 
> wrote:
>> -1
>>
>> Whilst I agree with the intention, we do not have a process defined 
>> that
>>   would allow us to generate a reproducable release. This led to 
>> several
>> of the issues with the last M3 release that ultimately made is 
>> unusable.
>>   We must fix this before we can release another version.
>>
>> Specific things I think we need include in such a process:
>> * an mechanical process for producing the candidate binaries that can 
>> be
>>    executed against any SVN tag. This would reduce the potential for
>>    minor variations by people doing the release that would result in
>>    potentially different binaries
>>
>> * elimination of SNAPSHOT dependencies - these are by nature ephemeral
>>    making it impossible to later regenerate the same distribution
>>
>> * a testing/review period that is at least comprehensive enough to 
>> catch
>>    the blaring defects that plagued M3
>>
>> * verification that the src bundle actually builds and results in the
>>    same binary as we are distibuting
>
> +100 on creating a reproduceable, mechanical process for these tasks.
> I was told by someone that much of this already exists in the form of
> shell/perl scripts somewhere. Is this true or not? I think that
> spending time on this task takes a fairly high priority, no matter
> whether we decide to produce a release at this time or not. That said,
> any idea how long it might take to accomplish this set of tasks? In
> working toward this goal, the first thing I'd like to see is a
> comprehensive list of tasks necessary to complete this work as well as
> a list of artifacts that must be produced. Jeremy's list above is a
> very good start, let's build upon it.

This is a lot harder then it looks.  All of these items and sever 
others (such as tagging cvs/svn) were on the list last time  David and 
I worked on the last release.  Also you need to coordinate with OpenEJB 
to get a tagged set of binaries released.

I guess practice makes perfect.  I suggest you make a dry run branch of 
the server and work out the steps (script) required for a release.  
Once you get a reproducible process worked out, take a current branch 
and cut a release.

-dain


Mime
View raw message