When glibc-common is installed in chroot (when building a livecd) it emits a warning, as shown by this example:

# rm -rf root
# rpm --root `pwd`/root -ihv \
    glibc-common-2.14.90-4.1.x86_64.rpm \
    glibc-2.14.90-4.1.x86_64.rpm \
    bash-4.2.10-4.fc16.x86_64.rpm \
    tzdata-2011h-1.fc16.noarch.rpm \
    libgcc-4.6.1-7.fc16.x86_64.rpm \
    basesystem-10.0-5.fc16.noarch.rpm \
    setup-2.8.35-1.fc16.noarch.rpm \
    filesystem-2.4.44-1.fc16.x86_64.rpm \
    libselinux-2.0.102-6.fc16.x86_64.rpm \
    nss-softokn-freebl-3.12.10-4.fc16.x86_64.rpm \
    libsepol-2.1.1-1.fc16.x86_64.rpm \
    ncurses-libs-5.9-2.20110716.fc16.x86_64.rpm \
    ncurses-base-5.9-2.20110716.fc16.x86_64.rpm
warning: glibc-common-2.14.90-4.1.x86_64.rpm: Header V3 RSA/SHA256 Signature, key ID a82ba4b7: NOKEY
Preparing...                ########################################### [100%]
   1:setup                  ########################################### [  8%]
   2:libgcc                 ########################################### [ 15%]
   3:filesystem             ########################################### [ 23%]
   4:basesystem             ########################################### [ 31%]
   5:ncurses-base           ########################################### [ 38%]
   6:tzdata                 ########################################### [ 46%]
   7:nss-softokn-freebl     ########################################### [ 54%]
   8:glibc-common           ########################################### [ 62%]
warning: (null) scriptlet failed, exit status 127
   9:glibc                  ########################################### [ 69%]
  10:ncurses-libs           ########################################### [ 77%]
  11:bash                   ########################################### [ 85%]
  12:libsepol               ########################################### [ 92%]
  13:libselinux             ########################################### [100%]

Why do glibc-common installation cause a warning, and what is it that is failing?


glibc.spec has:

%triggerin common -p <lua> -- glibc
if posix.stat("%{_prefix}/lib/locale/locale-archive.tmpl", "size") > 0 then
  pid = posix.fork()
  if pid == 0 then
    posix.exec("%{_prefix}/sbin/build-locale-archive")
  elseif pid > 0 then
    posix.wait(pid)
  end
end

%post common -p <lua>
if posix.access("/etc/ld.so.cache") then
  if posix.stat("%{_prefix}/lib/locale/locale-archive.tmpl", "size") > 0 then
    pid = posix.fork()
    if pid == 0 then
      posix.exec("%{_prefix}/sbin/build-locale-archive")
    elseif pid > 0 then
      posix.wait(pid)
    end
  end
end

%triggerin common -p /usr/sbin/tzdata-update -- tzdata

strace shows that %post correctly is skipped after 
525   access("/etc/ld.so.cache", F_OK)  = -1 ENOENT (No such file or directory)
but tzdata has already been installed and is thus run before glibc is available and we get:
526   execve("/usr/sbin/tzdata-update", ["/usr/sbin/tzdata-update", "/var/tmp/rpm-tmp.8o2qZw", "1", "1"], [/* 28 vars */]) = -1 ENOENT (No such file or directory)
526   exit_group(127)                   = ?
... 18 lines ...
525   write(2, "warning: ", 9)          = 9
525   write(2, "(null) scriptlet failed, exit st"..., 41) = 41

AFAICS rpm does the right thing so I guess the tzdata scriptlet should be fixed somehow.

Thinking out loud: it seems like the simplest way to break the strong circular glibc/glibc-common dependency would be to merge the packages ... but I guess there is a reason they are separate ... Forcing or tricking rpm to install glibc first would perhaps also solve the problem - perhaps with some tricky explicit Requires(post).


glibc-common-2.14.90-4.1.x86_64
rpm-4.9.1-3.fc16.x86_64