Core rebuild in mock - BuildRequires - rpmdiffs normal ?

[David Timms]

Hi, while learning a bit about doing this, I find similar to others eg: 
Alexander in https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=191647#c4
"all files of the rpm are built, but there are still some MD5 
differences shown with rpmdiff"

For some combinations,
Eg ncpfs: rpmdiff on i386
..5....T   /usr/bin/nwpurge
..5....T   /usr/bin/nwrevoke
..5....T   /usr/bin/nwrights
* most common case where the binary is different contents, but of 
identical size to the published rpm.

S.5....T   /usr/bin/nwsfind
* rarer case detected where the original rpm's binary and the mock 
build's size differ, and are not the same contents.

.......T   /usr/bin/pqrm
* A rare case where a mock built binary is identical to the original 
packages binary, except for time.

.......T   /usr/include
.......T   /usr/include/ncp
.......T   /usr/include/ncp/eas.h
.......T   /usr/include/ncp/ext
.......T   /usr/include/ncp/ext/socket.h
* diff due to T=modifiedTime. Mock just extracted/made the dirs/files 
this moment, so this should be ignored.

So, if a BuildRequires patch:
1. allows the rpms to get built *and*
2. there are no missing files shown via rpmdiff compared to devel i386 rpm,
should the patch be considered complete for BuildRequires purposes ?

I did read that one of the points of rpm packaging was to have 
repeatable package builds. While this doesn't seem to fit this scenario, 
perhaps it is normal when re-compiling an i386 package on an AMD Athlon 
with possibly updated libraries / compilers etc (from mock devel as 
well) that the built binaries would change (eg, even if it was just text 
fields like:
# abcdtool
Built on blah machine on Fri blah) ?

Would the fact that cpp/gcc/gcc-c++ has been updated 2006-05, but the 
ncpfs package not since 2006-04 explicitly cause this difference ?

DaveT.