--- Peter Reilly <[EMAIL PROTECTED]> wrote: > I think that this is more fundamental that modifing > a number > of tasks. > > The fact that a fileset has a directory base is I > think its basic design. > > It is used *a lot* for mappers, in-fact without the > base-dir > for the filesets, most uses of mappers by tasks make > no sense.
I like mappers, don't get me wrong, but I have the general feeling that different tasks may not always use them the same way. I usually find myself having to play a little to create a mapper to do exactly what I want. This would be the same thing. Your filenames are no longer relative; so what? Especially with the container and chained mappers... if I have a fileset rooted in ten different directories, even on different roots (where available) I can set up a container mapper to map each of them appropriately. -Matt > > Peter > > Matt Benson wrote: > > >Time for controversy! There is an interesting > thread > >at > >http://issues.apache.org/bugzilla/show_bug.cgi?id=5035 > >that touches on this issue. The key issue was that > >some tasks (including 3rd party ones) would break > if > >AFS.getDir() were to return null. This is indeed > >true. I have implemented the subject line, and the > >following tasks/types had to be touched: > > > >M src/main/org/apache/tools/ant/taskdefs/Copy.java > >(made copying abs. paths imply flattening) > > > >M > src/main/org/apache/tools/ant/taskdefs/Delete.java > >(log message accessed dir) > > > >M > >src/main/org/apache/tools/ant/taskdefs/DependSet.java > >(depend stuff needs a basedir for package > resolution) > > > >M > src/main/org/apache/tools/ant/taskdefs/Javadoc.java > >(requires dir w/ packagesets b/c of package > >resolution) > > > >M src/main/org/apache/tools/ant/taskdefs/ > > optional/ide/VAJImport.java > >(easier to assume with untestables) > >M src/main/org/apache/tools/ant/taskdefs/ > > optional/metamata/AbstractMetamataTask.java > >(easier to assume with untestables) > >M src/main/org/apache/tools/ant/taskdefs/ > > optional/ssh/Scp.java > >(too complex to deal with yet) > >M src/main/org/apache/tools/ant/types/ > > optional/depend/ClassfileSet.java > >(depend stuff needs a basedir for package > resolution) > > > > > >However, as I stated on the referenced bug entry, > the > >API has never AFAICT promised that getDir would > return > >a non-null result. The overwhelming majority of > tasks > >would be unaffected by this as many tasks would > simply > >use the directory as the first parameter of new > >File(File, String). No harm done. This has been > an > >outstanding request for a long time. I feel that > it > >represents little risk; fileset's documentation can > be > >liberally sprinkled with warnings that errors might > be > >encountered using dir-less filesets with some > >third-party tasks, and we can encourage third-party > >providers to make sure they are compatible. If we > >scheduled this for 1.7 we could put ample warnings > >into 1.6.3 that this is coming. > > > >So what say you all? > > > >-Matt > > > > > > > > > >__________________________________ > >Celebrate Yahoo!'s 10th Birthday! > >Yahoo! Netrospective: 100 Moments of the Web > >http://birthday.yahoo.com/netrospective/ > > > >--------------------------------------------------------------------- > >To unsubscribe, e-mail: > [EMAIL PROTECTED] > >For additional commands, e-mail: > [EMAIL PROTECTED] > > > > > > > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: > [EMAIL PROTECTED] > For additional commands, e-mail: > [EMAIL PROTECTED] > > __________________________________ Celebrate Yahoo!'s 10th Birthday! Yahoo! Netrospective: 100 Moments of the Web http://birthday.yahoo.com/netrospective/ --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
