Re: [libvirt] [PATCH 4/8] tests: don't alter state in $HOME

On 03/27/2011 02:59 AM, Laine Stump wrote: > On 03/26/2011 08:12 AM, Eric Blake wrote: >> Diego reported a bug where virsh tries to initialize a readline >> history directory during 'make check' run as root, but fails >> because /root was read-only. >> >> It turns out that I could reproduce this as non-root, by using: >> >> mv ~/.virsh{,.bak} >> chmod a-w ~ >> make check -C tests TESTS=int-overflow >> chmod u+w ~ >> mv ~/.virsh{.bak,} >> >> * tests/int-overflow: Don't trigger interactive mode. >> Reported by Diego Elio Pettenò. >> --- >> >> >> echo "error: failed to get domain '4294967298'"> exp || fail=1 >> -echo domname 4294967298 | $abs_top_builddir/tools/virsh --quiet \ >> +$abs_top_builddir/tools/virsh --quiet \ >> --connect test://$abs_top_srcdir/examples/xml/test/testnode.xml \ >> -> /dev/null 2> err || fail=1 >> + 'domname 4294967298; quit'> /dev/null 2> err || fail=1 >> diff -u err exp || fail=1 > > Why the "; quit" in the command? a quit is implicit anyway. A quit may be implicit, but the problem at hand is what $? gets set to in the shell after virsh is done. In batch mode, virsh exits with the status of the last command executed, and we expect the domname command to fail (invalid argument), whereas quit (trivially) succeeds. That is, since the shell script expects $? to be zero, the quit command is necessary to avoid polluting the exit status with the failure from domname while still proving that virsh didn't abort command processing merely because one command failed. We didn't need the quit in interactive mode, because EOF is treated the same as an implicit quit command. > > ACK, with or without the quit - it works either way. Pushed with the quit.