Re: Another atf_check question

[Paul Goyette <paul%whooppee.com@localhost>]

On Sun, 9 Jan 2011, Julio Merino wrote:

> On Sat, Jan 8, 2011 at 9:35 PM, Paul Goyette <paul%whooppee.com@localhost> 
> wrote:
> > I've got a helper program that I want to invoke multiple times for a single
> > test case:
> >
> >        atf_test_case empty
> >        empty_head()
> >        {
> >                atf_set "descr" "Testing empty name"
> >        }
> >        empty_body()
> >        {
> >                atf_check 'empty_tests' 0 ${files}empty.exp null
> >        }
> >        empty_tests()
> >        {
> >                $TEST ::1 ''
> >                $TEST 127.0.0.1 ''
> >                $TEST localhost ''
> >                $TEST '' ''
> >        }
> >
> > So, the "command" that I want to pass to atf-check(1) is actually a shell
> > function.
> >
> > As expected, this does not work.  I thought of using atf-check directly, and
> > specifying -x, but that would start another shell to execute the command.
> >
> > Is there a sane way to do this?
>
> What I have done sometimes is to write the script out to a temporary
> file and then run it.  It is ugly, specially because with
> here-documents it is not nice to type shell scripts, but does the job.
> I have also just put such auxiliary scripts in their own files.
>
> Could you maybe rework your auxiliary function to do the atf_check
> calls by itself?  (The sample code above does not really provide
> enough detail on what you are trying to accomplish.)  I've also
> sometimes generated test cases on-the-fly by means of eval; see the
> ipf tests.

What I ended up doing was to call the shell function with output 
redirection, then using cmp(1) to compare with the golden output.

        xxx_body() {
                xxx_test > out 2>&1
                atf-check -s 0 -o ignore -e ignore \
                        "cmp -s out ${files}xxx.exp"
        }

This was for a rework of the old regress-getaddrinfo test.  I've since 
decided to abandon this for now, since the correct results are not 
deterministic (they depend heavily on host configuration, particularly 
network interfaces and /etc/hosts file).


> And lastly, any reason why the above didn't go to atf-devel? ;-)

My bad - sorry.



-------------------------------------------------------------------------
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:       |
| Customer Service | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com    |
| Network Engineer | 0786 F758 55DE 53BA 7731 | pgoyette at juniper.net |
| Kernel Developer |                          | pgoyette at netbsd.org  |
-------------------------------------------------------------------------