On Sun, 2004-10-17 at 19:36 -0700, matt whiteley wrote: > On Sat, 16 Oct 2004 15:27:46 -0400, Scot Mc Pherson > <scot@xxxxxxxxxxxxxxxxxxxx> wrote: > > On Saturday 16 October 2004 05:58 am, matt whiteley wrote: > > > > > > > As an introduction, I have been building a few rpms for about two > > > years now for personal use. Now that I am having to create them for > > > work, I am trying to learn more about them. I just spent a day > > > figuring out that the default strip post command was causing the > > > binary I had to be invalid. That was a fun one to figure out. > > > > > > My question is this: Why is the rpm produced from a .spec file differ > > > in multiple build attemps? > > > > > > What I mean is that I repeatedly built the same file from a .spec and > > > each time it had a varrying size and md5sum. I am curious as to what > > > might cause this. (hopefully I am not missing something terribly > > > obvious). > > > > > > thanks, > > > > Generally this means you have an unstable toolchain. When all things are equal > > the toolchain should produce the same output every time. When I build glibc > > several time, my files are always the same from one build to the next unless > > I change something. Try rebuilding binutils, gcc and glibc a few times until > > they settle into producing consistent code. > > > > -- > > Scot Mc Pherson > > scot@xxxxxxxxxxxxxxxxxxxx > > http://linuxfromscratch.org/~scot > > http://beewiki.linuxfromscratch.org > > AIM: ScotLFS ICQ: 342949 MSN:behomet@xxxxxxxxxxxxx > > > So in trying to do some more testing I downloaded the FC2 Bison SRPM > and wrote a script to build it 100 times and name the rpms with a > timestamp. Well all 100 had different md5sums, and a nice bell curve > distribution on the file sizes. What is interesting is that I > extracted the RPMS and every file in the RPM save one has the same > md5sum. For the Bison rpm this is 'liby.a'. I'd say this is to be expected because (ar) archives contain time stamps which change with every build. That is also why I would expect every RPM to have different checksums, because it again contains a (cpio) archive which contains timestamps. Nils -- Nils Philippsen / Red Hat / nphilipp@xxxxxxxxxx "They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety." -- B. Franklin, 1759 PGP fingerprint: C4A8 9474 5C4C ADE3 2B8F 656D 47D8 9B65 6951 3011 _______________________________________________ Rpm-list mailing list Rpm-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/rpm-list
