Randy McMurchy wrote:

Right now,
I think the build order is because it was developed through years
of experience, trial and error and testing.

experience + trial and error + testing != proof.

What Jeremy wants is the exact knowledge of package dependencies that gives such proof. As a side effect, the freedom to re-arrange the build order is gained.

BUT, I disagree with the proposed method of obtaining this knowledge. As far as I can understand from the discussion, the method is to build all possible combinations of Current and Alphabetical chapter 6 from each other, and do some IPA. This will give us the knowledge whether those two build orders produce identical results. This knowledge is of zero value if the current build order misses some optional dependency. The only known true (but very expensive) method of determining package dependencies absolutely correctly without reading the whole source code is to run the whole build process under strace.

--
Alexander E. Patrakov
--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to