Conor MacNeill wrote:
>
>> Accurate description.  See http://jakarta.apache.org/gump/why.html#2000 for
>> my rendition.
>
> "But it was the only way I could regain control."
>
> :-)

I do firmly believe that while the person writing the build script should
get the first say as to what dependencies ought to be used, the person
running the build should get the final say.

One thing worth discussing: what would it take for this information to be
inferred from build.xml files?  What I would most like to know is what
pathelements represent directories inside of the tree, and are referenced
by classpaths which are referenced by javac tasks.

With today's ant, that would sometimes actually require running through the
motions of executing Ant to see what the values of various variables are.

If you accept that sometimes people want to use Ant's build.xml files as
building blocks (look at taglibs, for example), then it would be helpful if
some of this information was factored out into declarative sections.  I
require the following inputs.  I can optionally take advantage of these
other inputs.  I require the following work directories to be available and
in my classpath.  I produce the following outputs (jars/wars/ears,
javadocs, zip/tar/whatever).

I have plenty of raw data available here to analyze:
http://cvs.apache.org/viewcvs/jakarta-alexandria/proposal/gump/project/

- Sam Ruby


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to