Re: failing regression tests (please fix them!)

[christos%astron.com@localhost (Christos Zoulas)]

In article <20100602120543.GA1993%cs.hut.fi@localhost>,
Antti Kantee  <pooka%cs.hut.fi@localhost> wrote:
>Hi,
>
>We still have 24 failing regression tests in the standard "anita test" run:
>
>Failed test cases:
>    ipf/t_ipf:bpf1, ipf/t_ipf:bpf_f1, ipf/t_ipf:n1, ipf/t_ipf:n11,
>    ipf/t_ipf:n2, ipf/t_ipf:n4, ipf/t_ipf:n5, ipf/t_ipf:n6, ipf/t_ipf:ni10,
>    ipf/t_ipf:ni11, ipf/t_ipf:ni12, ipf/t_ipf:ni19, ipf/t_ipf:ni20,
>    ipf/t_ipf:ni5, kernel/t_umount:umount, lib/libevent/t_event:kqueue,
>    lib/libevent/t_event:poll, lib/libevent/t_event:select,
>    net/sys/t_connect:low_port, util/df/t_df:normal,
>    util/grep/t_grep:basic, util/grep/t_grep:file_exp,
>    util/sh/t_expand:strip, util/sh/t_set_e:all
>
>Most of them look like broken / unupdated tests.  (There's also a bunch
>of tests which are skipped for incorrect reasons.)

All the failed sh(1) tests are because of the extra level of quoting/eval
the atf introduces:

    eval '(set -e; false; echo ERR$?); echo -n OK$?'

vs.

    (set -e; false; echo ERR$?); echo -n OK$?

Looks like eval is busted :-)

christos