If you do this (and you are not an Emacs user who uses PAGER=cat
in your *shell* buffer):

        $ git init
        Initialized empty Git repository in .git/
        $ echo hello world >foo
        $ H=$(git hash-object -w foo)
        $ git tag -a foo-tag -m "Tags $H" $H
        $ echo $H
        3b18e512dba79e4c8300dd08aeb37f8e728b8dad
        $ rm -f .git/objects/3b/18e5*
        $ git show foo-tag
        tag foo-tag
        Tagger: Junio C Hamano <gitster@pobox.com>
        Date:   Sat Feb 16 10:43:23 2008 -0800

        Tags 3b18e512dba79e4c8300dd08aeb37f8e728b8dad

you do not get any indication of error.  If you are careful, you
would notice that no contents from the tagged object is
displayed, but that is about it.  If you run the "show" command
without pager, however, you will see the error:

        $ git --no-pager show foo-tag
        tag foo-tag
        Tagger: Junio C Hamano <gitster@pobox.com>
        Date:   Sat Feb 16 10:43:23 2008 -0800

        Tags 3b18e512dba79e4c8300dd08aeb37f8e728b8dad
        error: Could not read object 3b18e512dba79e4c8300dd08aeb37f8e728b8dad

Because we spawn the pager as the foreground process and feed
its input via pipe from the real command, we cannot affect the
exit status the shell sees from git command when the pager is in
use (I think there is not much gain we can have by working it
around, though).  But at least it may make sense to show the
error message to the user sitting in front of the pager.

[jc: Edgar Toernig suggested a much nicer implementation than
what I originally posted, which I took.]

Signed-off-by: Junio C Hamano <gitster@pobox.com>