ant elder wrote:
On Fri, Feb 11, 2011 at 9:24 AM, Simon Nash <n...@apache.org> wrote:
ant elder wrote:
On Thu, Feb 10, 2011 at 10:36 AM, Florian Moga <moga....@gmail.com> wrote:
I'm not keen into having an ant build file in every sample. If someone
wants
to use ant, shouldn't they download all dependencies manually? Won't they
use maven for that task anyway? What about ivy?
I've been thinking the same thing lately, especially now that you
pointed out that we can do "mvn ant:ant" to have Ant scripts
automatically generated. Instead of Ant builds for each sample there
could just be some doc in the top level samples folder or README that
describes how to use "mvn ant:ant".
The build scripts generated with mvn ant:ant download all the
dependencies from the Maven repositories, that does mean that not much
would be actually using all the jars from the Tuscany binary
distribution so we might want to consider what all thats really for.
...ant
Actually I would wonder what is the point of using maven to generate
an ant script that does exactly the same as the maven build.
In 1.x the ant scripts were provided as an alternative to maven that
use local artifacts from the binary distro instead of depending on
remote repositories. Another purpose was to make it very clear and
explicit what steps are involved "under the covers" to build and run
Tuscany applications. This information is useful to people who want
to develop solutions that embed Tuscany.
Simon
In 1.x at least some of the Ant scripts are/were just generated from
the maven build but using the Tuscany plugin instead of the Ant plugin
aren't they, so its not so different?
Many of these are generated, but they use local jars rather than
downloading them from maven repositories, and the Ant scripts are
included in the binary distro so there's no need for users to have
maven to create them. In some cases only the list of dependencies
is generated as build-dependency.xml, and this is included by a
hand-written build.xml file.
I think we need to agree what the purpose of them is, what it is that
they're demonstrating, and then decide if the best way to show that is
a separate sample(s), or to include both Maven and Ant in every
sample, I'm wondering if separate might be better.
I think that's a good suggestion. For a group of samples where the
building/running logistics are similar, we could just provide one
Ant script. For samples where the logistics are substantially different,
a separate Ant script could be provided. This reduces the considerable
burden of providing a working tested Ant script for every sample.
One think is that the 2.x distribution is a bit unusual with the way
it stores the jars in both top and subdirectories, and there are now
all the generated manifest jars, and the various aggregated jars. I
can't imagine anyone would have their own jars in this sort of mix so
i wonder how useful the hand written Ant builds which use them really
would be.
...ant
Isn't that quite a problem for the embedded scenario? Developing an
application that embeds Tuscany isn't particularly easy even with the
simpler 1.x structure. What would we recommend to someone who wants
to develop a Tuscany application using 2.x that embeds/redistributes
the Tuscany runtime?
Simon
