Description of Problem: Using rpm-build-4.0.4-7x.5.i386.rpm, ELF binaries aren't being stripped anymore. (I may be wrong and the fault may be binutils-2.11.92.0.12-10.i386.rpm.) It's not "file"'s mistake, as I checked some packaged binaries with different versions of "file", e.g. on Solaris. Version-Release number of selected component (if applicable): rpm-build-4.0.4-7x.5.i386.rpm How Reproducible: always so far, seemingly also with RedHat's RawHide build system Steps to Reproduce: 1. rpm -bb <your spec file> 2. rpm -Uvh <your resulting binary RPM> 3. file /usr/bin/<your binaries> Actual Results: Your binaries are not classifed as stripped ("file" says: "not stripped"). Expected Results: Your binaries should be stripped. Additional Information: I noticed by chance while trying rpmlint for the first time and getting "non-stripped-binary" errors, both with packages of my own and e.g RawHide's ImageMagick-5.4.3.5-1.i386.rpm. The "file" commands confirms rpmlint's hints. ./usr/lib/rpm/brp-strip calls "strip -g" on detected ELF binaries. This seems insufficient in terms of what "file" believes is "stripped". There may be a design decision for this "debug symbol stripping only", but I don't know, so I'm reporting. ;-) (I couldn't find anything in CHANGES.) I've put the prio to "high" because the whole distribution is affected by your build system. :) Thanks for listening, Moritz :-)
Yes, only debugging symbols are now stripped. The intent is to provide better info on segfaults.