Bug 1053985

yum reinstall kernel-3.13.0-0.rc8.git0.1.fc21.x86_64.rpm
...
Running transaction
Warning: RPMDB altered outside of yum.
  Installing : kernel-3.13.0-0.rc8.git0.1.fc21.x86_64                                                                                   1/1 
warning: %post(kernel-3.13.0-0.rc8.git0.1.fc21.x86_64) scriptlet failed, exit status 127
Non-fatal POSTIN scriptlet failure in rpm package kernel-3.13.0-0.rc8.git0.1.fc21.x86_64
warning: %posttrans(kernel-3.13.0-0.rc8.git0.1.fc21.x86_64) scriptlet failed, exit status 127
Non-fatal POSTTRANS scriptlet failure in rpm package kernel-3.13.0-0.rc8.git0.1.fc21.x86_64
  Verifying  : kernel-3.13.0-0.rc8.git0.1.fc21.x86_64                                                                                   1/1 

Installed:
  kernel.x86_64 0:3.13.0-0.rc8.git0.1.fc21   

I previously yum installed rc7 and rc6, those worked. Since then, other packages have been updated so it's possible this is a problem with one of those rather than the kernel package.

Created attachment 850863 [details]
rpm -vv output

We haven't changed these scriptlets at all.  Did you update systemd, dracut, grubby, grub, bash, etc?  Is your box actually stock F20, or are you on rawhide or pulling in updates?

There's one other person reporting similar issues but they also sent an email to test list basically showing that their machine is broken.  Maybe an RPM or shell update is something you two have in common?  Their's is bug 1053948

It's a current F20 with u-t enabled, plus 3.13 kernels which is the only F21 component.

systemd-208-9.fc20.x86_64
grub2-efi-2.00-25.fc20.x86_64
grubby-8.28-1.fc20.x86_64
dracut-034-64.git20131205.fc20.1.x86_64
bash-4.2.45-4.fc20.x86_64

Those are all the same as on F20 media, they haven't been updated.

In between a successful rc7 and the failing rc8, dnf became irritable in the middle of downloading updates, claiming it had checked all mirrors and couldn't find any available. I waited a day, and it persisted until I whacked it with a hammer 'dnf clean all' and it applied a bunch of updates at that point. But no errors.

Now, when I "dnf update" all it wants to do is remove 3.13.0-0.rc5 but that fails with the same scriplet errors. I'll attach strace.

Created attachment 851137 [details]
strace dnf update

This is strace output after saying 'y' to the single task listed by dnf, which is to remove kernel 3.13.0-0.rc5. This doesn't tell me anything about what's broken. This is on Btrfs with snapshots so I can easily roll back but some ordinary operation has caused breakage.


pread(6, "\0\0\0\0\1\0\0\0P\t\0\0Q\t\0\0O\t\0\0\1\0\346\17\0\7ax\0- B"..., 4096, 9764864) = 4096
pread(6, "\0\0\0\0\1\0\0\0O\t\0\0P\t\0\0N\t\0\0\1\0\346\17\0\7\0\0\0\0\0\0"..., 4096, 9760768) = 4096
pread(6, "\0\0\0\0\1\0\0\0N\t\0\0O\t\0\0\0\0\0\0\1\0V\f\0\7_MESSA"..., 4096, 9756672) = 4096
rt_sigprocmask(SIG_BLOCK, ~[RTMIN RT_1], [], 8) = 0
rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
stat("/var", {st_mode=S_IFDIR|0755, st_size=184, ...}) = 0
stat("/var/tmp", {st_mode=S_IFDIR|S_ISVTX|0777, st_size=128, ...}) = 0
umask(077)                              = 022
open("/var/tmp/rpm-tmp.itNNIQ", O_RDWR|O_CREAT|O_EXCL, 0600) = 37
umask(022)                              = 077
dup(37)                                 = 38
close(37)                               = 0
write(38, "/bin/kernel-install remove 3.13."..., 115) = 115
close(38)                               = 0
dup(23)                                 = 37
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7f592ce45a10) = 1081
wait4(1081, [{WIFEXITED(s) && WEXITSTATUS(s) == 127}], 0, NULL) = 1081
--- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=1081, si_status=127, si_utime=0, si_stime=0} ---
write(21, "error: %preun(kernel-3.13.0-0.rc"..., 88) = 88
close(37)                               = 0
unlink("/var/tmp/rpm-tmp.itNNIQ")       = 0
fstat(20, {st_mode=S_IFREG|0600, st_size=88, ...}) = 0
lseek(20, 0, SEEK_CUR)                  = 0
read(20, "error: %preun(kernel-3.13.0-0.rc"..., 4096) = 88
read(20, "", 4096)                      = 0
write(1, "error: %preun(kernel-3.13.0-0.rc"..., 88error: %preun(kernel-3.13.0-0.rc5.git0.1.fc21.x86_64) scriptlet failed, exit status 127

It would be helpful if you could roll back to a working state and then install each update from updates-testing and see if it breaks then.  Based on some of the other emails on the test list thread, it seems like something other than the usual suspects is causing RPM scriptlets to fail in multiple packages.  selinux, pcre, libpcap, initscripts, and file are all pending there.

I'm going to mark this as a dupe of the selinux bug.  Seems very related.

*** This bug has been marked as a duplicate of bug 1054350 ***