Peter Donald wrote: > > 1. Manipulating build files to add in "branding"/"versioning" > information or some other sort of post-processing (ie > supporting collections in 1.1 and 1.2
So copy just those files. I'd agree if those files were a significant percentage of the overall project (say 20%), but if they are only few infrequently modified files... > 2. Some people mount the source on a shared drive (ie > samba/nfs) and want to compile on a local drive. The > reason is that compiling on nfs/samba is slooooow. Why can't they copy the entire tree and then work locally? What you propose is that ant copies the source and leave such items as the build instructions behind. This doesn't make much sense to me. > 3. Multiple builds out of same tree. ie you can > simultaneously have 3-4 different builds with diff This does not require a copy of the source, just an adjustment to the destination property. > Of course this still means unadorned win32 without cygnus has > "issues" but most people don't use emacs on windows ;) You may be surprised. Me? I'm typically using vim on unadorned win32. Many of my colleagues use emacs on win32. By the way, I believe that much of the history of this started with Stefano Mazzocchi who needed conditional compilation so he added if checks on targets and selectively copied based on available. Filters were added later and do give a valid justification for copying. I added the ability to selectively include/exclude which was a better solution (IMHO) to Stefano's original problem. The need for filtering is valid, but often can be scoped. I don't see this as a showstopper item, but it might be educational to put this up for a majority vote within the scope of this project. - Sam Ruby
