Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [b3-dev] try not to mirror antlr

On 2011-04-12 07:59, joerg@xxxxxxxxxxxxx wrote:

Can you please explain why only legacy sites are affected?

Legacy sites are transformed on the fly into p2 sites. It's p2 and not the aggregator that does this but obviously there is something in the correlation between the aggregator and p2 that causes this failure when that happens. That's all I know without investigating this further.


Then, maybe its better to get in contact with the ANTLR people, telling them they should rebuild their update site (http://antlreclipse.sourceforge.net/updates/).

Absolutely. It's always better to provide a proper p2 site with meta-data. When p2 encounters a legacy site, it must download all feature jars in order to build the p2 meta-data. That is a slow process.


The b3 aggregator deals solely with p2 or Maven artifacts. Not sure I
follow your line of reasoning when you talk about the b3 in concjunction
with SVN. Can you please elaborate?

When a Eclipse developer has to configure his Eclipse to get ready for developing, (1) he first has to install a bunch of upate sites required by the project, and after this (2) he often checks out a bunch of SVN projects where the code resides, maybe from different SVN repositories.

With b3, (1) can be automated very well, however (2) is not automated by now. It would be nice, if not only p2 artifacts can be added and grouped with b3, but also SVN projects within a category. So you could put SVN projects (actually only the SVN URLs) into a category, and maybe the category also contains all required p2 dependencies required for the SVN projects to compile.

This is out of scope for the b3 aggregator. Partly because what you find in SVN is a moving target. It doesn't have fixed versions yet since it hasn't been built and therefore it can not participate in a p2 resolution.

We provide another tool, Buckminster, that will perform a resolution of both SVN repositories and p2 repositories and then divide the downloaded material into workspace projects and target platform binaries. Very special requirements, and a very special logic must be used when transitively resolve features and bundles in source form. Buckminster is also capable of building a p2 update site based on your workspace content. Might be worth looking at.

- thomas




Back to the top