On an FC6t3 x86_64 system installed 2 days ago, and updated daily since, I just got this : [...] Installing: kernel-xen ####################### [15/54] xargs: nm: No such file or directory Updating : firstboot-tui ####################### [16/54] Installing: kernel-devel ####################### [17/54] Updating : pirut ####################### [18/54] Installing: kernel ####################### [19/54] xargs: nm: No such file or directory Updating : python-xeninst ####################### [20/54] [...] The scriplets for the kernels don't seem to contain "nm" directly, so I dug a little and found out that it's from /sbin/weak-modules : if [ -e /lib/modules/$module_krel/extra ]; then find /lib/modules/$module_krel/extra -name '*.ko' \ | xargs nm \ | sed -nre 's:^([0-9a-f]+) A __crc_(.*):0x\1 \2:p' fi > $tmpdir/extra-symvers-$module_krel So if this xargs nm is important, module-init-tools should depend on it, so that we are sure to have it installed. Otherwise, adding a test for nm's presence and running the find only if it is found?
Today's new redhat-lsb-3.1-10.3 has pulled in binutils now (and m4 and gettext which I didn't have installed either). It's apparently related to bug #202548 (from the %changelog), but I'm not authorized to see that bug... Since redhat-lsb is in the set of default packages for a default install, I guess the issue is a lot less important, but still, you might want to also have module-init-tools require binutils if having "nm" is really important.
The issue still persists as of FC6 final. binutils is a large package, so a solution which wouldn't require a dependency on it in module-init-tools would be nice. +1 to something like described in the initial comment (testing if /usr/bin/nm exists and only proceeding if it does).
Is this still an issue in F7? Jon.
I just looked at /sbin/weak-modules on my F7 system and the section that calls "nm" doesn't seem to have changed, and it's still pretty easy to not have binutils installed (although it's now pulled in by redhat-lsb apparently). So I'd think the bug is still present, yes.
I've updated weak-modules.