RFC: mass rebuild results

[Karsten Hopp]

Hello,

I've done a mass rebuilt of Rawhide in an (almost) full install chroot with x86_64 packages only.
Then I've run rpmdiff of the resulting packages against the rawhide packages and put together a rather large list
of differences. You can grab the list from http://people.redhat.com/karsten/RPMDIFF . This list doesn't even contain the failed
packages which suddenly have unpackaged files not catched by some wildcard in the %files section.

- additional requirements usually mean either a missing configure flag to disable a certain feature or a missing buildrequirement
   p.e.
   RPM in Rawhide: vim-enhanced-7.1.266-1.fc9.x86_64.rpm Rebuilt RPM: vim-enhanced-7.1.266-1.fc9.x86_64.rpm
   added      REQUIRES libruby.so.1.8()(64bit)
   This mean I either should have added --without-rubyinterp to vim's configure flags if I don't want ruby support or I need to add
   a build requirement on ruby-devel

- almost every *-javadoc package gets a huge amount of additional documentation in a full install rebuild

- some noarch packages put files in either /usr/lib64 or in /usr/lib, depending on which arch they have been built (example: gcstar)

- some packages, p.e. evolution switch from using libgnutls to libgio ?



In my opinion rebuilding packages outside of mock should give me packages with the same features than the original ones.
In addition to that I shouldn't have to cope with build failures because of unpackaged files just because some other installed
random package changes the filelist of my package via configure detection.
The packaging guidelines mention only that the buildrequirements for the used configure flags need to be added to the spec file
but I think that we should add to the guidelines that all other (unwanted) features need to be disabled via configure flags.

Should I even bother to start filing the most obvious bugs or shall we leave it 'as is' ?

    Karsten