"allow-stderr" tests that produce stderr output always pass

Steps to reproduce:

1. Write a test that produces stderr output, but then returns exit status 1.
2. Run adt-run against the test.

Expected results: test failure.
Actual results: test pass.

This seems to be due to the following code ("rc" contains the exit status of the test run).

        if stab.st_size != 0:
                l = open(se_read).readline()
                l = l.rstrip('\n \t\r')
                if 'allow-stderr' not in t.restriction_names:
                    if len(l) > 35: l = l[:35] + '...'
                    t.reportfail('status: %d, stderr: %s' % (rc, l))
                else:
                    t.report('PASS')
                t._debug(' - - - - - - - - - - stderr - - - - - - - - - -')
                debug_file(se_read)
        elif rc != 0:
                t.reportfail('non-zero exit status %d' % rc)
        else:
                t.report('PASS')