Hopefully, these are my last ICA results before looking at a toolchain
which doesn't initially link to /tools. The book is from 20060110 with
mktemp back where it used to be, a sed for gccbug, and Greg's change to
bison. Built three times, without locales or testsuites, all except the
first build didn't have /tools on the path.
I was also keeping the perl source around each time (omitted from these
results), but the only thing that showed up was a randomness in the RCS
version info in the perl/x2p/{M,m}akefiles.
The total number of files is probably slightly more than from the book,
I had backup copies of a couple of files to assist in my rerun process.
The copies were extracted with --one-file-system, so only a minimal
/dev.
Item / builds being compared: 0-1 1-2
Files in only one build 3 0
Files compared 18809 18812
unreadable 0 0
identical 17658 18100
expected to differ 7 8
allowable 1140 704
different 4 0
The files in question:
only in one build
libhistory.so.5.1.old seems to be from updating in-place
libreadline.so.5.1.old ditto
/root/.viminfo (perhaps from running 'view')
expected to differ
/root/.bash_history
the two c++ precompiled headers
libstdc++.a
libstdc++.so.6.0.6
libsupc++.a
/usr/share/info/dir
and, in second comparison, /root/.viminfo
different (between first two builds)
libhistory.so.5 (symlink, points to .old after in-place rebuild)
libreadline.so.5 (ditto)
/usr/lib/libstdc++.la
/usr/lib/libsupc++.la
Current issues:
(i.) the libtool libraries from glibc.
(ii.) for ELF files, after two builds most were accepted after
stripping and/or tokenizing. After three builds, most were
identical and only a few were accepted after stripping and/or
tokenizing.
Ken
--
das eine Mal als Trag?die, das andere Mal als Farce--
http://linuxfromscratch.org/mailman/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
