On 30/03/12 09:22, Simone Tripodi wrote:
> Hi Lewis,
>
> there are just few minor additions I'd like to discuss before to start
> cutting an RC:
>
> 1 update the plugin page according to the ServiceLoader pattern we
adopted;
> 2 improving the CLI code - we can do a little better maintainable
> code, switching to JCommander[1] pattern
> 3 add a maven archetype to simplify plugins development
This isn't the only release - indeed, some would say that that being too
perfect discourages others getting involved as the barrier to
contribution is higher.
So I suggest setting a timescale with hard date for the code freeze.
If something gets done, great, if it doesn't, shrug. There will be
another release coming along. First release may go through easily, it
may cause comment - focus on process this time. Once that is bedded
down in the project, it's easy to release but first release is about the
process.
Sometimes, I think a podling should do the release thing, then throw it
away. At least that way, it's clear the release is much more about
Apache process than a normal release.
High ceremony ... with people watching.
> 4 understand how to concurrently deploy mvn artifacts on nexus and
> binaries bundles (included asc signs and md5/sha1 checksums) on the
> builds server, in one shot (I would avoid the RC manager has to
> perform manual operations as much as I can)
I don't know how much you've discussed this already - I haven't been
able to follow everything recently ...
In Jena, we have got this down to three steps:
mvn release:clean release:prepare -Papache-release
mvn release:perform -Papache-release
then run
<shell script that builds the dist/ area>
mvn puts everything into staging (yukness => you get .asc.md5, .asc.sha1
-- everyone finds this and eventually just lives with it as a fact of life).
This also creates the "source-release.zip" item which is the thing that
is *the* release. Strictly, that's what the vote is about. All the
other stuff is not "the release". It just the stuff people use!
The shell script picks bytes out of the local machine maven repository.
That does not include .md5/.sha1.
In fact, we now recalculate those rather than go pull them back from
Nexus staging. maven creates .md5 with just the md5 has in the file.
md5sum(1) puts the file name in and then "md5sum -c *.md5" works.
Some other projects have different processes - some have good write ups.
See "sling" and "rave" for examples. "rave" has a 130 line shell
script that does everything - but for this first release, it might be
easier to be a bit more manual because it means you get the knowledge to
capture in a script.
> while 3 and 4 are minor, I see 1 and 2 more important - it will define
> the *Apache* Any23 design and the thousands (hopefully :P) of users
> that start develop their tools on top of Any23, will do in a way that
> will be backward/forward compatible.
>
> WDYT?
1, 2 and 3 are technical.
4 is major.
I went checking ...
The build is in a good state and most of the N&L files look OK (I didn't
see if the N&L files inside a jar were right - I didn't see a combined
redistribution N&L in the with-deps jar though).
The RAT report is quite good - most unlicensed files are test materials
(yet another area for "debate" on g@i).
Andy
((
My experience: the first Jena release took a while to set up the N&L
files, but votes on -dev and g@i were straight forward. The first TDB
release got a somewhat detailed analysis on g@i and provoked some
discussion about what actually is correct and required practice. It's
all-meant but wearing a flame retardant jacket is advised.
And the Nexus buttons in staging are very close together. "Release" and
"Drop" are rather different actions but right next to each other :-)
))
> I'll dedicate the coming WE to Any23 to start arranging things.
>
> Have a nice day, all the best!
> -Simo
>
> [1] http://jcommander.org/
>
> http://people.apache.org/~simonetripodi/
> http://simonetripodi.livejournal.com/
> http://twitter.com/simonetripodi
> http://www.99soft.org/
>
>
>
> On Thu, Mar 29, 2012 at 6:22 PM, Lewis John Mcgibbney
> <[email protected]> wrote:
>> Hi Everyone,
>>
>> The final remaining 0.7.0 issue which I created is certainly not a
blocker,
>> and can wait until a later date to be addressed as I don't have time
this
>> week.
>>
>> Are we ready to blow a 0.7.0-incubating RC?
>>
>> Here's to hoping :)
>>
>> Thanks
>>
>> Lewis
>>
>> --
>> *Lewis*
