Re: Toolchains proposal

[Shane Isbell <sh...@gmail.com>]

Where are we sitting on the toolchain support? At NMaven, we're going back
to basics to get better alignment and integration with the rest of Maven.
Toolchain support ranks highly.  I see some interfaces and Java support
within the toolchain project, but I don't know how complete this is. If the
interfaces are stable, I can work on getting a dotnet implementation.

Thanks,
Shane

On Nov 10, 2007 4:10 AM, Jörg Schaible <jo...@gmx.de> wrote:

> Milos Kleint wrote:
>
> > On Nov 6, 2007 4:35 PM, Jason van Zyl <ja...@maven.org> wrote:
> >
> >>
> >> On 6 Nov 07, at 4:00 AM 6 Nov 07, Milos Kleint wrote:
> >>
> >> > Hello,
> >> >
> >> > I've got a working prototype of the toolchains proposal. I'm able to
> >> > define
> >> > the jdk toolchains and have them used in a project. Works with
> patched
> >> > compiler, surefire, javadoc plugins.
> >> > details are at http://docs.codehaus.org/display/MAVEN/Toolchains
> >> >
> >> > issues for resolution:
> >> > 1. currently using build-context, I heard stories about it going
> away.
> >>
> >> My only concern with the build-context, and John can counters as he
> >> deems fit, but it's hard to tell where through the core of Maven the
> >> context actually pops out. You still need to inspect the session, but
> >> the session would be the one place that you could look to see what is
> >> changing as it passes through the core. The rub right now is that many
> >> components internally are not setup to use a session. That's my
> >> opinion: that the session passing through the core could just as
> >> easily serve as a build context it's just architecturally the context
> >> is easier to wormhole through the code.
> >
> >
> > Possibly true. In order to move the toolchains code to session, we would
> > need a way to serialize/deserialize Objects. The actual live instances
> > cannot be used due to plugun classloading.
>
> Just a suggestion: You might give XStream with the binary driver a try.
> Not
> as bloated as XML and your classes do not have to implement Serializable.
>
> - Jörg
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> For additional commands, e-mail: dev-help@maven.apache.org
>
>

Re: Toolchains proposal

[Shane Isbell <sh...@gmail.com>]

Cool. If it's not tied to 2.1, then I'll get out an implementation for .NET
soon.

Thanks,
Shane

On Dec 17, 2007 2:26 PM, Milos Kleint <mk...@gmail.com> wrote:

> I've reworked the toolchain stuff to work without the build-context
> component that was removed in the trunk. As a side-effect it became easier
> to backport to 2.0.x. it should work in current 2.1-SNAPSHOT and
> 2.0.9-SNAPSHOT binaries. the shared/toolchains project artifacts with
> components needs to be manually put into the M2-HOME/lib folder.
>
> I consider the current version stable myself, I've rewritten 3 plugins so
> far to use the toolchains (all java/jdk related) and it seems to work fine
> and is reasonably simple.
> What is currently necessary is to get some peer review on the current
> api/implementation. Writing additional toolchain implementation is indeed
> a
> good way to review the code.
> Eventually I'd like to move the shared/toolchains project into components/
> and make it part of the maven's core as it's not going to work otherwise.
>
> Milos
>
>
> On Dec 17, 2007 10:44 PM, Shane Isbell <sh...@gmail.com> wrote:
>
> > Where are we sitting on the toolchain support? At NMaven, we're going
> back
> > to basics to get better alignment and integration with the rest of
> Maven.
> > Toolchain support ranks highly.  I see some interfaces and Java support
> > within the toolchain project, but I don't know how complete this is. If
> > the
> > interfaces are stable, I can work on getting a dotnet implementation.
> >
> > Thanks,
> > Shane
> >
> > On Nov 10, 2007 4:10 AM, Jörg Schaible <jo...@gmx.de> wrote:
> >
> > > Milos Kleint wrote:
> > >
> > > > On Nov 6, 2007 4:35 PM, Jason van Zyl <ja...@maven.org> wrote:
> > > >
> > > >>
> > > >> On 6 Nov 07, at 4:00 AM 6 Nov 07, Milos Kleint wrote:
> > > >>
> > > >> > Hello,
> > > >> >
> > > >> > I've got a working prototype of the toolchains proposal. I'm able
> > to
> > > >> > define
> > > >> > the jdk toolchains and have them used in a project. Works with
> > > patched
> > > >> > compiler, surefire, javadoc plugins.
> > > >> > details are at http://docs.codehaus.org/display/MAVEN/Toolchains
> > > >> >
> > > >> > issues for resolution:
> > > >> > 1. currently using build-context, I heard stories about it going
> > > away.
> > > >>
> > > >> My only concern with the build-context, and John can counters as he
> > > >> deems fit, but it's hard to tell where through the core of Maven
> the
> > > >> context actually pops out. You still need to inspect the session,
> but
> > > >> the session would be the one place that you could look to see what
> is
> > > >> changing as it passes through the core. The rub right now is that
> > many
> > > >> components internally are not setup to use a session. That's my
> > > >> opinion: that the session passing through the core could just as
> > > >> easily serve as a build context it's just architecturally the
> context
> > > >> is easier to wormhole through the code.
> > > >
> > > >
> > > > Possibly true. In order to move the toolchains code to session, we
> > would
> > > > need a way to serialize/deserialize Objects. The actual live
> instances
> > > > cannot be used due to plugun classloading.
> > >
> > > Just a suggestion: You might give XStream with the binary driver a
> try.
> > > Not
> > > as bloated as XML and your classes do not have to implement
> > Serializable.
> > >
> > > - Jörg
> > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> > > For additional commands, e-mail: dev-help@maven.apache.org
> > >
> > >
> >
>

Re: Toolchains proposal

[Milos Kleint <mk...@gmail.com>]

I've reworked the toolchain stuff to work without the build-context
component that was removed in the trunk. As a side-effect it became easier
to backport to 2.0.x. it should work in current 2.1-SNAPSHOT and
2.0.9-SNAPSHOT binaries. the shared/toolchains project artifacts with
components needs to be manually put into the M2-HOME/lib folder.

I consider the current version stable myself, I've rewritten 3 plugins so
far to use the toolchains (all java/jdk related) and it seems to work fine
and is reasonably simple.
What is currently necessary is to get some peer review on the current
api/implementation. Writing additional toolchain implementation is indeed a
good way to review the code.
Eventually I'd like to move the shared/toolchains project into components/
and make it part of the maven's core as it's not going to work otherwise.

Milos


On Dec 17, 2007 10:44 PM, Shane Isbell <sh...@gmail.com> wrote:

> Where are we sitting on the toolchain support? At NMaven, we're going back
> to basics to get better alignment and integration with the rest of Maven.
> Toolchain support ranks highly.  I see some interfaces and Java support
> within the toolchain project, but I don't know how complete this is. If
> the
> interfaces are stable, I can work on getting a dotnet implementation.
>
> Thanks,
> Shane
>
> On Nov 10, 2007 4:10 AM, Jörg Schaible <jo...@gmx.de> wrote:
>
> > Milos Kleint wrote:
> >
> > > On Nov 6, 2007 4:35 PM, Jason van Zyl <ja...@maven.org> wrote:
> > >
> > >>
> > >> On 6 Nov 07, at 4:00 AM 6 Nov 07, Milos Kleint wrote:
> > >>
> > >> > Hello,
> > >> >
> > >> > I've got a working prototype of the toolchains proposal. I'm able
> to
> > >> > define
> > >> > the jdk toolchains and have them used in a project. Works with
> > patched
> > >> > compiler, surefire, javadoc plugins.
> > >> > details are at http://docs.codehaus.org/display/MAVEN/Toolchains
> > >> >
> > >> > issues for resolution:
> > >> > 1. currently using build-context, I heard stories about it going
> > away.
> > >>
> > >> My only concern with the build-context, and John can counters as he
> > >> deems fit, but it's hard to tell where through the core of Maven the
> > >> context actually pops out. You still need to inspect the session, but
> > >> the session would be the one place that you could look to see what is
> > >> changing as it passes through the core. The rub right now is that
> many
> > >> components internally are not setup to use a session. That's my
> > >> opinion: that the session passing through the core could just as
> > >> easily serve as a build context it's just architecturally the context
> > >> is easier to wormhole through the code.
> > >
> > >
> > > Possibly true. In order to move the toolchains code to session, we
> would
> > > need a way to serialize/deserialize Objects. The actual live instances
> > > cannot be used due to plugun classloading.
> >
> > Just a suggestion: You might give XStream with the binary driver a try.
> > Not
> > as bloated as XML and your classes do not have to implement
> Serializable.
> >
> > - Jörg
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@maven.apache.org
> > For additional commands, e-mail: dev-help@maven.apache.org
> >
> >
>