And btw, your commit now makes at least two automated build definitions fail at the Bamboo server, e.g.
http://bamboo.ci.codehaus.org/browse/CASTOR-DEF/latest Can you please let me know how to go about this ? There's lots of things in place to assure that we have high quality (unit tests, maven, bamboo, ...), so I am sure you must have run the Maven test suite before committing your changes, right ? Werner Werner Guttmann wrote: > Ralf, > > I am sure you have read my email where I spent some 10 minutes to > explain that in length that the problem is not about adding some 10 > lines of code to the Ant scripts, and everything works. It that were an > option, I would have done it anyhow (despite me not liking to maintain > Ant scripts any more). > > Problem is that there is a major design flaw with the way we link in > source generation with our Ant scripts (I have tried to explain things > below). As such, there's no easy fix. And as such, there will be hours > needed to fix that. > > If you are willing to discuss these things in detail with me, feel free. > Then I will be more than willing to invest some more of my time to > describe the problem. > > If your only reaction is > > <quote> > build scripts for both build scripts we support > </quote> > > I freely admit that I am willing to spend my personal time improving > Castor, but not fixing (again and again) a build system that has major > design flaws (per se, i.e. Ant) and build scripts (again, with major > problems the way they have been built historically). > > Again, for me it's the code base that matters, and not the (flawed) > build scripts I have been fixing repeatedly when I have been extracting > all the modules in the past 1.5 years (to give Castor the modularity it > requires from a user perspective). > > As mentioned in my original reply, we have a source that compiles > without problems, and test suites (junit, functional) that complete just > fine. > > For me (personally), I am not willing to maintain two build systems any > more, especially when one has to code around issues to do useful things > (and that time could be made available to progress the tool, not the > build scripts). > > a) delete generated classes from the SVN repo (i.e. have them generated > during the build). > b) profile code during test (quality assurance) > c) define dependencies in a clear and concise way (through Maven, that is). > > Regards > Werner > > Ralf Joachim wrote: >> Hi Werner, >> >> it is fine if you do not check in generated sources. But you should at >> least provide: >> >> 1. a note that one have to generate sources >> 2. build scripts for both build scripts we support >> >> Regards >> Ralf >> >> >> Werner Guttmann schrieb: >>> Joachim, >>> >>> the project is perfectrly compilable and all tests run (to my knowledge, >>> incl. XML and JDO), but only if you use Maven. In other words, with >>> whatever I committed, I paid well attention that the project compiles >>> and all tests run. >>> >>> The problem is with Ant, I am aware of this, and there is NO easy >>> solution. There's a work-around though: >>> >>> With Maven installed, position yourself in the schema module, and run >>> >>>> mvn castor:generate >>> This will run the schema generator and generate the sources in >>> schema/target/generated-sources/castor >>> >>> Once you have done that, the build script works completely again. >>> >>> Werner >>> >>> PS I have spent hours yesterday and today trying to amend the Ant build >>> scripts to get this working, but there's so many bugs in these scripts, >>> and sequence of compilation matters so much that I am growing really >>> tired to modify those messy scripts (again and again), especially when >>> all it takes on the Maven side is one plugin configuration, and >>> everything works seamlessly. >>> >>> The problem is that you have to run the source generator *before* >>> building your sources, and this is not easy with Ant (especially when >>> the Castor Ant scripts are configured to rely on building e.g. the Ant >>> task definitions before being able to use them. This - imho - is >>> completely wrong, as Maven relies on a deployed version of e.g. the >>> ANttask JARs in order to run source code compilation (which is how it >>> should work). >>> >>> As the work-around is easy and absolute minimal in terms of steps to >>> follow, I am not willing to spend any more time on this for the time >>> being, as I'd rather spend those 6+ hours on committing absolutely >>> brilliant code that enhances Castor (on the functionality side of things). >>> >>> Just to make one thing clear: I even deployed a snapshot release after >>> making those commits, and a few other people have picked those up and >>> are using them. In other words: the code compiles with the right tool(s) >>> and tests run successfully (XML: 440, JDO: 70 on mySQL). >>> >>> Ralf Joachim wrote: >>>> Hi Werner, >>>> >>>> currently SVN trunk has lots of compile failures as some classes of >>>> schema.annotation package are missing. Can you please fix this soon and >>>> try to leave SVN in a compilable state after your commit sessions next >>>> time. >>>> >>>> Regards >>>> Ralf >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe from this list, please visit: >>>> >>>> http://xircles.codehaus.org/manage_email >>>> >>>> >>> --------------------------------------------------------------------- >>> To unsubscribe from this list, please visit: >>> >>> http://xircles.codehaus.org/manage_email >>> >> --------------------------------------------------------------------- >> To unsubscribe from this list, please visit: >> >> http://xircles.codehaus.org/manage_email >> >> > > --------------------------------------------------------------------- > To unsubscribe from this list, please visit: > > http://xircles.codehaus.org/manage_email > > --------------------------------------------------------------------- To unsubscribe from this list, please visit: http://xircles.codehaus.org/manage_email
