>>>>> "Richard" == Richard Boulton <[EMAIL PROTECTED]> writes:
Richard> Surely one thing we could do is to implement wildcards that Richard> are evaluated at the time that automake is run. Just to be clear, I thought that was the only kind under consideration. I don't think there's any way we could implement wildcards at make time. Richard> It is still true that it would be impossible to ensure that Richard> the Makefile.in is regenerated when a new source file is Richard> added, but I would imagine it is rare to add a new source Richard> file to a project without modifying any of the existing Richard> source files. It happens all the time for test suites. For instance, automake's. Or any of several that appear in gcc. Also this happens a lot in libjava, where we sometimes add a new .java file without touching anything else. I imagine the same is true for many Java libraries. For automake I often make test cases I don't check in. Having these be picked up in Makefile.in would be pretty bad -- it would break the build. So the next step will be someone suggesting that $(awildcard) needs to know how to look at CVS files only. Bleah. (This is a glum prediction, not an actual argument against -- we can always say "no" when that comes along.) That said, I'm still open to understanding why people want this. Are there projects where adding a new line to Makefile.am is a significant burden? What's an example of one? Tom
