>>>>> "Harlan" == Harlan Stenn <[EMAIL PROTECTED]> writes:
Harlan> One of the biggest apparent hurdles is that these folks would Harlan> benefit from having a single monolithic Makefile.am at the top Harlan> level, as that would provide the best parallel makes, provide Harlan> the best library dependency stuff, run the fastest, etc. Harlan> OTOH, it would be Difficult to coordinate changes to this Harlan> Makefile.am (lots of developers) and they still want the Harlan> ability to run a make in a subdir. These goals (monolithic Makefile.am and run make in a subdir) are really not compatible. The `import' plan would let you run `make subdir/all', which is similar, but that plan is still on the drawing boards. Harlan> Other problems include about 160 Makefile.am's that use Harlan> automake conditionals to decide if certain _PROGRAMS need to Harlan> be made or not, and in many cases, which target-specific Harlan> _SOURCES need to be used. Wow. Harlan> I arrived just after the conversion from automake-1.4 to 1.5, Harlan> and based on my current efforts I'm probably going to Harlan> recommend we look at upgrading to 1.6.1 just to quiet some of Harlan> the "automake does not support XXX being defined Harlan> conditionally" messages. I think that's smart. From what I know right now I'd say 1.6.1 is a lot less buggy and difficult to use than 1.5. Harlan> I'm actually considering looking at keeping the Makefile.am's Harlan> in each subdir and looking at AutoXport to combine this info Harlan> into a top-level .xml file and then see if that .xml file can Harlan> be easily converted to a Jakarta/ant build.xml file. That sounds frightening. I haven't looked at Ant too much. I read a bit about it last week. It sounds like it doesn't really track dependencies at all. Perhaps I missed that section of the manual. Tom