In the orginal form of the blocks-build.xsl, we had separate targets for the patch files. But it was incredible slow. Then I merge these targets to one target, and rewrote to the XConf task to a MatchingTask, which allow to execute more than one patches. But it doesn't preserves the dependencies, then Carsten cuts the target in to several target again, to solve this problem. Now, with latest change it works again.
I tend to agree with you Joerg, separate targets are much more elegant. But in the real world I have real problems, like a build time von 4min 25sec on a 2.4GHz Intel system. Which is, by the way, unacceptable, IMHO.
So, should I revert the change to have a more elegant build file with bigger build time?! .... ehrmm ... I think not.
To be honest, such statements enrage me at least a bit. You talk about time, but you forget the time to maintain this additional dependency resolving. Starting with the missing .xweb patches you have now to go on searching for bugs - things that already have been working. For having a look on this issue I removed ojb, database and hsqldb block from the excluded ones. A simple build (Cocoon was previously built with only cforms and xsp enabled) - and many patches of those blocks were not applied. Only a clean build made it working - partly, see above. If I need to do every time a clean build to get this thing correctly working, I don't see how you can gain time. This might be only a simple bug somewhere, maybe only a typo - but I talk about the principle - which, I know, often ends in obstinacy.
IMO, yes, we should revert it. I prefer the elegancy much more about the speed. And to add Carsten's argument: Additionally it forces us "keeping the dependencies correct".
Excuse me, if I have forgotten to add 'rant' around it ...
Joerg
