Randy McMurchy wrote:
And, unfortunately for me, what's made me feel like even more of an idiot, is that after his two messages to try and explain it, I still don't have a clue.
Not sure if this will help, but the entire purpose of these changes is to document dependencies. Since LFS is linear, for each package, we want to know exactly *why* we're building it *when* we are. If it doesn't matter when it's built, it gets moved to the end and placed in the alphabetical queue.
So the purpose of this isn't to alphabetize, it is to start with a minimal linear build, satisfying dependencies as we go. The rest of the packages are dropped in afterwards to help make what Gerard calls a "well-rounded development model". And those packages are built in alphabetical order only as a default rule - some inter-dependencies are worked out by doing it that way, as well.
So when autotools shows a dependency in its testsuite on something else, the *first* thing we look at doing is moving that dependency up in the build order - but we find that causes other little niggles and that in fact autotools has circular dependencies. So we have to find the best method to achieve our goal, or compare the trouble against the weight of the package and determine if it's worth it.
-- JH -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
