| [Fwd: Re: [cdt-dev] CDT Build (was: CDT Call Tomorrow)] |
|
My recent experience of new tool chain definition shows me that now
there is a big complexity of model expansion and still not possible to
reach even relatively simple goals without hacking existing plugins.
Expandable Java model is incomplete now. Besides that, take in account
that there is a nice jdt debugger, but there are no debugger to locate
the problem in 250K xml file (of course, I don't mean syntax problems).
Anyway, to do this in most cases I traced Java code in order to detect
logic inconsistence in XML file. I'd also not overestimate a
possibility to define new tool chain without any Java knoledge. This
is not the work that people do every day. More important to make it
easier to qualified developer than to make it a tool for anybody who
just knows how to work with XML editor. So I rather agree with Jesper than disagree. PS. To clarify my position: of course I do believe that in spite of written before, there are many good features in MBS that should be saved or just improved. -------- Original Message --------
+0.25 from me...
+1 The original goal for defining toolchains with extension points was so that integrators didn't need to know java to add in their toolchains, and it gave us a nice editor and storage management for the definitions for free. But in the end, we've added so much cool functionality that requires you to write Java classes anyway, that I think it was a mistake. We've even added the capability for you to create a factory to instantiate some of the model yourself. I used it on a previous project and it let me make the model more dynamic. And, face it, you need to be a Java programmer to get anywhere near Eclipse plug-in development, so if you made it this far, allowing you to create the build definitions in Java won't be a hardship on many. So yes, I agree, and will keep this in mind as we take a step back and re-evaluate where we want to go. Cheers, Doug. > -----Original Message----- > From: cdt-dev-bounces@xxxxxxxxxxx > [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Jesper Eskilson > Sent: Thursday, June 12, 2008 4:11 AM > To: CDT General developers list. > Subject: Re: [cdt-dev] CDT Call Tomorrow > > Elena Laskavaia wrote: > > Sorry I have to miss this call. > > About 5.1 > > - MBS: projects > > - wizards fixes (I think we won't apply it in 5.0 at this point?) > > - we need to extend API to achieve better customization > > The possibility of defining a toolchain using a Java API > instead of in XML would be really, REALLY, nice. > > I've complained about this before > (http://thisisnotaprogrammersblog.blogspot.com/2007/10/writing > -toolchain-plugin-for-cdt-or.html), > but it's such a pain to try to write the toolchain definition in XML. > It's clumsy and awkward at best and if you've made > non-trivial mistakes in your XML-code, you're basically on your own. > > I really don't get the point of using XML for the toolchain > definition in the first place. Yes, it might be good to be > able to define extensions without having to write Java code, > but only for simple extensions (which toolchains are not). > After that it's more pain to write XML than Java. > > Unless you're writing a toolchain for a compiler which is > almost identical to GCC you will still need to write Java > code. (macro providers, dependency scanners, etc.) > > -- > /Jesper > > _______________________________________________ > cdt-dev mailing list > cdt-dev@xxxxxxxxxxx > https://dev.eclipse.org/mailman/listinfo/cdt-dev > _______________________________________________ cdt-dev mailing list cdt-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/cdt-dev |
_______________________________________________ cdt-dev mailing list cdt-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/cdt-dev
"Schaefer, Doug"
