3:12 p.m.
In the case of the provisioning tools, we don't generate module.xml files.
Modules are provided by the developers/build process. We provide the
mechanism to generate packages from the modules and then the feature-pack.
When a feature-pack is generated, we know which other feature-packs it
depends on. Then when we generate a package from a module we process the
module dependencies and figure out on which package(s) from which
feature-pack(s) the package we generate should depend on.
I am still not sure whether that answers your question since it's
formulated in the context of Swarm and I have a trouble translating it into
the context of the feature-pack-based provisioning. I am interested in
covering your requirements though. So if it makes sense to you let's
continue the discussion or schedule a call to clear things out.
On Wed, Oct 4, 2017 at 7:23 PM, Scott Stark <sstark(a)redhat.com> wrote:
Correct. Starting from a pom that contains the artifact dependencies,
how
to translate this into an appropriate module.xml for the output fraction is
a trial and error process currently. Does the information exist so that one
could go from GAV and/or package names dependencies to the module that
provides said dependencies from amongst a set of wildfly feature pack?
On Tue, Oct 3, 2017 at 8:03 AM, Bob McWhirter <bmcwhirt(a)redhat.com> wrote:
> I think the issue (at least as I’ve found it) is that we either:
>
> a) scrub around in existing feature-packs and modules/ directory to
> determine if there’s a module.xml matching the thing and version we need or
>
> b) hand-author a module.xml based upon pom.xml, adding a chance to mess
> things up.
>
> -Bob
>
> On Tue, Oct 3, 2017 at 10:37 AM, Alexey Loubyansky <
> alexey.loubyansky(a)redhat.com> wrote:
>
>> It's not exactly clear to me what issue you are describing. But I can
>> provide some basic info on how feature-packs are authored for the wildfly
>> releases (core, servlet, full). Perhaps then you could ask a more specific
>> question.
>>
>> A feature-pack represents a specific release. So there will be a
>> feature-pack for the core, another one for the servlet distribution and
>> another one for the full one. In feature-packs, modules are organized into
>> packages (which is an atomic unit of content possibly with dependencies on
>> other packages from the same or another feature-pack).
>> When a feature-pack is generated, the packages are generated from the
>> modules. Each module becomes are package in a feature-pack. And module
>> dependencies become package dependencies. Is it any close to the issue you
>> described?
>>
>> On Tue, Oct 3, 2017 at 1:51 AM, Scott Stark <sstark(a)redhat.com> wrote:
>>
>>> Interesting.
>>>
>>> It brings up a discussion point I have been meaning to raise across the
>>> Wildfly and Wildfly-Swarm teams regarding tools for the step before this
>>> assembly step of feature-packs and fractions into a distributable archive.
>>>
>>> The issue I have seen is that when authoring a feature-pack or
>>> fraction, it is difficult to know how to configure the module dependencies.
>>> One is often starting with GAV dependencies from a spec, and it is
>>> difficult to know how those map onto modules in existing feature-packs for
>>> fractions. Do we have any tooling in this area to make this task not a
>>> trial and error effort?
>>>
>>>
>>>
>>> On Mon, Oct 2, 2017 at 7:34 AM, Emmanuel Hugonnet
<ehugonne(a)redhat.com>
>>> wrote:
>>>
>>>> Hi,
>>>>
>>>> For those who can take 7'26 of their time to look at what we can do
>>>> with the new provisioning tool to build and share custom feature packs.
>>>>
>>>> https://www.dropbox.com/s/84133sgsjef7pqs/feature_pack.mp4?dl=0
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> wildfly-dev mailing list
>>>> wildfly-dev(a)lists.jboss.org
>>>> https://lists.jboss.org/mailman/listinfo/wildfly-dev
>>>>
>>>
>>>
>>> _______________________________________________
>>> wildfly-dev mailing list
>>> wildfly-dev(a)lists.jboss.org
>>> https://lists.jboss.org/mailman/listinfo/wildfly-dev
>>>
>>
>>
>