On Thu, Mar 10, 2005 at 12:13:45AM -0800, Kin H. Au wrote: > I am using FC3 and notice very wired behavior. The rpm -q --dump output > is inconsistent with actual file attribute. > > # rpm -q --dump glibc | grep lib/tls/libc-2.3.4.so > /lib/tls/libc-2.3.4.so *1451615* 1103537616 > *3f269651d6f0953ce3f753760efa5026* 0100755 root root 0 0 0 X > # ls -l /lib/tls/libc-2.3.4.so > -rwxr-xr-x 1 root root *1454835* Dec 20 02:13 /lib/tls/libc-2.3.4.so > # md5sum /lib/tls/libc-2.3.4.so > *d6abef51fef1c5df35398f5fff34b714 * /lib/tls/libc-2.3.4.so > > The funny thing is that rpm -qV glibc doesn't report any md5sum > inconsistent. It takes me 2 days and finally find out the different > file size and different md5sum is caused by prelink daily cornjob ( > /etc/cron.daily/prelink ). I wonder why md5sum and file size are > different from rpm -q --dump output? Can someone explain to me? rpm is prelink aware and during -V for prelinked binaries and shared libraries verifies that their md5sum and filesize after unprelinking is the same as the recorded one and that prelinking the object again results in bitwise identical binary/library to the one present on the filesystem. For more info, see /usr/share/doc/prelink-*/prelink.pdf Jakub
