> In a sense, yes. But having nightly builds for which you can provide no
> guarantee - for example, ones that might include viruses - sounds a whole
> lot worse to me that having no nightly builds.

I don't know that you can guarantee the absence of viruses on an ASF
server anymore than you can on an ASF server.  Also, I thought the
Gump server was an ASF server but I guess I was mistaken about that. 
Anyways, I don't really have a preference as to what server hosts the
builds so I will defer to you on that one.

> OK, here's a suggestion: Why don't we (you and I) approach the Gump team
> for (potentially temporary) access to brutus. 

OK.  FYI I sent an email to [EMAIL PROTECTED] (the only email
address I could find) with a few questions about how their procedures
works.  I haven't heard back from them yet.  Why don't you give it a
try yourself?  I will try my contact with the Ant group and see if we
can't make headway with their contacts.

> Once we have that, we can
> work out the kinks in the process of getting a nightly build up and
> running. Simultaneously, we (meaning I) can approach infra@ and ask for
> access to an appropriate box for running the real nightlies on, once the
> process is ironed out. This isn't going to be a surprise to them - they
> know we want to get all of the nightlies running on ASF hardware, and 'm
> pretty sure they're eager to have that happen.

Sounds good.
 
> At this point, I'm talking only about MyFaces nightlies, to appease your
> fear of this process dragging out. Either after the MyFaces nightlies are
> sorted out, or while that's in process, depending on my time and energy, I
> can also work on adding Struts, Commons and Taglibs builds into the mix
> (with the first of these having a fairly high priority, due to the newly
> restructured build process).
> 
> How does that sound?

That sounds great!  If infra@ people get back to us quickly with a
server (lets say within 4 weeks) then there would be no big deal in
waiting.  I don't want to give the impression that we must have the
nightlies at all costs.  My only objection is waiting around for
something that might take a while to happen when there are other
perfectly acceptable alternatives available now.

> Not quite. The nightly script also needs to be able to push up the builds
> to the official nightly build location. That would have been the advantage
> of running on minotaur - it's the same box, so 'cp' works, and 'scp' is
> not necessary. ;-)

I'm assuming your proposed ASF box would not need scp either right?

> I think we should strive to have the nightly build process and the release
> build process be the same, or at least as similar as possible. That way,
> you can have greater confidence that a release build will retain the
> solidity of its most recent nightly cousin, rather than having a whole new
> enchilada to test.

Agreed.  Tomorrow (Friday) I will spend some time familiarizing myself
with the current scripts and see what changes I think need to be made.
 I will submit the changes to the dev list for review.
 
> I see the nightly process (and the release process) as being comprised of
> three different pieces: checkout, build, deploy. The first and last of
> these will be slightly different between nightly and release, but the
> build itself should be the same.

Agreed.  I don't think there is really a formal release script right
now though.  Yes there is a target for release but it doesn't do steps
one and three (checkout and publish).  So we have step 2 more or less.

I will start working on a "bootstrap" build file that can be used to
checkout from CVS, run the build.xml contained within the project, and
publish to specified directory (I'll assume regular copy to another
directory on the same machine for now.)
 
Let me know what other help you need from me.

> Martin Cooper

sean

Reply via email to