Hi, I have seen several discussions about the new build method proposed by Jeremy.;-) For myself, I have no idea, although I like this method because I tried to implement it myself (and not succeeded because I took a wrong path). One crucial point always with a novelty is "does it work?" Of course in our case, it builds, but a further test is: "does it build as before?" I haven't really tried to ensure the full compatibility with before, but just performed ICA on the new build method. Actually, since chapter 6 is not modified in that method, I would expect that mods in chapter 5 show up with differences between the first and second build of chapter 6. So here are the results (without diffs coming from timestamps and/or some use of random bits, as in groff doc, and a different ordering of info/dir): ----------------------- diff -ur iteration-1/usr/include/etip.h iteration-2/usr/include/etip.h --- iteration-1/usr/include/etip.h 2012-02-29 17:02:33.000000000 +0100 +++ iteration-2/usr/include/etip.h 2012-02-29 19:10:39.000000000 +0100 @@ -62,7 +62,7 @@ #endif
#ifndef ETIP_NEEDS_MATH_H -#define ETIP_NEEDS_MATH_H 0 +#define ETIP_NEEDS_MATH_H 1 #endif #ifndef ETIP_NEEDS_MATH_EXCEPTION --------------------------- (same issue as with current method, comes from ncurses bug) --------------------------- diff -ur iteration-1/usr/lib/libgmpxx.la iteration-2/usr/lib/libgmpxx.la --- iteration-1/usr/lib/libgmpxx.la 2012-02-29 15:54:39.000000000 +0100 +++ iteration-2/usr/lib/libgmpxx.la 2012-02-29 17:59:46.000000000 +0100 @@ -17,7 +17,7 @@ inherited_linker_flags='' # Libraries that this one depends upon. -dependency_libs=' /usr/lib/libgmp.la /usr/lib64/../lib64/libstdc++.la' +dependency_libs=' /usr/lib/libgmp.la /usr/lib/../lib64/libstdc++.la' # Names of additional weak libraries provided by this library weak_library_names='' ---------------------------- (there is a similar issue with the current method, but interestingly, it manifests differently. With the current method, the last argument on the '-' line is /usr/lib/libstdc++.la. The lib64 controversy may be not over:-) ) ---------------------------- There are also differing binary files. For the majority of them, those are executable, and the most likely reason is that they differ because of timestamps. So I content myself with checking the outputs of 'readelf -a' do not differ. The only differing result is: libgmpxx.so.4.2.4, where the library rpath's differ (and so the length of the string pool): iteration 1 : Library rpath : [/usr/lib64/../lib64] iteration 2 : Library rpath : [/usr/lib/../lib64] This is just the same as what is in libgmpxx.la. The current build has more differences, because there is no rpath at all in the first iteration library. Another differing file is ld.so.cache, because ldconfig is not run at the end of the builds... The last differing file is locale-archive, but its format is a complicated hastable, and locale -a gives the same results on each iteration. Those last 2 are the same with the current build. Hope it helps Pierre -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
