Paul Eggert <[EMAIL PROTECTED]> wrote: > Jim Meyering <[EMAIL PROTECTED]> writes: > >> * tests/du/basic: Make the test larger than 64 bytes, > > The idea of putting small files' contents into the inode is an old > one, and it keeps coming up (a quick Google search reports that the > TFS folks are thinking of doing it: > <http://groups.google.com/groups?selm=b0abr8%242hfc%241%40FreeBSD.csie.NCTU.edu.tw>). > > Though 64 bytes is enough for Netapp, I'd expect other filesystems to > use larger values, if not now then in the future. Would it be easy to > increase that value to 8192 (or better yet, but this is merely icing > on the cake: the maximum of 8192 and st_blksize)?
Sure it'd be easy. Buy why 8k and not the current size of 4k? Here's the test I'm using now:
#!/bin/sh # Compare actual numbers from du, assuming block size matches mine. if test "$VERBOSE" = yes; then set -x du --version fi # DU_BLOCK_SIZE could cause problems . $srcdir/../envvar-check . $srcdir/../lang-default pwd=`pwd` t0=`echo "$0"|sed 's,.*/,,'`.tmp; tmp=$t0/$$ trap 'status=$?; cd $pwd; chmod -R u+rwx $t0; rm -rf $t0 && exit $status' 0 trap '(exit $?); exit $?' 1 2 13 15 framework_failure=0 mkdir -p $tmp || framework_failure=1 cd $tmp || framework_failure=1 mkdir -p a/b d d/sub || framework_failure=1 printf 'make-sure-the-file-is-non-empty\n' > a/b/F || framework_failure=1 # Ensure that these files contain more than 64 bytes, so that we don't # immediately disqualify file systems (e.g., NetApp) on which smaller # files take up zero disk blocks. yes | head --bytes=4k > d/1 yes | head --bytes=4k > d/sub/2 if test $framework_failure = 1; then echo "$0: failure in testing framework" 1>&2 (exit 1); exit 1 fi # Ensure that blocksize and block-count are what we expect for directories. if test "`stat --format=%bx%B a`" != 8x512; then echo "$0: different block count/size, so skipping this test" 1>&2 (exit 77); exit 77 fi # Ensure that blocksize and block-count are what we expect for files. if test "`stat --format=%bx%B d/1`" != 8x512; then echo "$0: different block count/size, so skipping this test" 1>&2 (exit 77); exit 77 fi fail=0 du -a a > out || fail=1 echo === >> out du -a -S a >> out || fail=1 echo === >> out du -s a >> out || fail=1 cat <<\EOF | sed 's/ *#.*//' > exp 4 a/b/F 8 a/b 12 a === 4 a/b/F # size of file, a/b/F 8 a/b # size of dir entry, a/b, + size of file, a/b/F 4 a # size of dir entry, a === 12 a EOF cmp out exp || fail=1 test $fail = 1 && diff -u out exp 2> /dev/null rm -f out exp du -a d | sort -r -k2,2 > out || fail=1 echo === >> out du -S d | sort -r -k2,2 >> out || fail=1 cat <<\EOF | sed 's/ *#.*//' > exp 4 d/sub/2 8 d/sub 4 d/1 16 d === 8 d/sub 8 d # d + d/1; don't count the dir. entry for d/sub EOF cmp out exp || fail=1 test $fail = 1 && diff out exp 2> /dev/null (exit $fail); exit $fail
_______________________________________________ Bug-coreutils mailing list [EMAIL PROTECTED] http://mail.gnu.org/mailman/listinfo/bug-coreutils
