Testing considerations for an SI
Couple of oddities in testing an SI....
One has to do with mounts; once you mount /proc inside
the SI chroot, the mounts from the underlying system are
visible. To do a reasonable job of passing tests, you
must make sure that those are accounted for: if the "df"
command returns errors, lots of tests will fail. df will
return errors if a path appears to be mounted that does
not exist, so those may need to be created, manually.
I've also taken to adding most of the information from the
underlying fstab to the one inside the chroot, but frankly,
I'm not sure if this is absolutely required.
On RH and Mdk systems, /dev/shm and /dev/pts are separate
filesystems, their mount points need to exist, and they
need to be mounted.
Given this, there's no way to make a generic LSB-si that
is absolutely correct for the system under test, maybe we
could create a script that is run after the LSB-si is
installed to take care of any final oddities like this.
The other issue I wanted to mention is that a syslogs is
needed in order to run a large number of tests - 124 will
be uninitated if a suitable syslogd, which is running, is
not found. It appears that the tests expect to find a
running syslog, stop it, and restart it with a custom
syslog.conf for the duration of the syslog testing, then
restore the old one. I thought it was sufficient to have
syslogd present, so the test could start it, but apparently
not finding one running is grounds for failure, and the
tests don't proceed.
So to make this work, you must:
(a) put a syslog and a working syslog.conf into the si
(b) stop the host system's syslogd
(c) start the chroot
(d) deal with other stuff (the mounts)
(e) start the syslogd
(f) run login, and log in as vsx0 (NOTE: *still*
not sure if this is sufficient to handle all
the logged-in-as tests)
(e) run the tests
(f) clean up
(g) quit the chroot
Kind of a pain, but I guess ordinary folks won't be
doing this sort of stuff.
On the topic of syslogd, I've compiled an LSB-conforming
syslogd, but with that one, five syslog tests fail.
I'm not sure if that's a side effect of having "ported"
the syslogd to LSB (which isn't *strictly* necessary,
I guess), or if it's a side effect of running an unpatched
syslogd from the upstream. I started with this package:
sysklogd-1.4.1.tar.gz
I can add building of this package to our application
battery scheme so that it can be available as an
installable rpm for use inside the LSB-si, but before
that happens, I'd like to know if it's worthwhile.
Does anyone have any insights into the syslogd issue -
are there known problems with passing LSB testing, for
example? Since these are all format problems, is this
a glibc issue?
FYI, the failing tests are:
syslogd 89, 98, 99, 117, 118
The journal excerpts are appended.
Mats
200|5 89 17:11:36|TP Start
520|5 89 00005037 1 1|syslog() failed to format argument correctly
520|5 89 00005037 1 2| Received: "Jul 10 17:11:37 mdwichma-lnx
test-syslog-89: 1074081849 75535 0 7654 ca9 CA9
520|5 89 00005037 1 3|"
520|5 89 00005037 1 4| Expected: "1074081849 1074931471 0 7654 ca9 CA9"
220|5 89 1 17:11:38|FAIL
200|5 98 17:11:58|TP Start
520|5 98 00005037 1 1|syslog() failed to format argument correctly
520|5 98 00005037 1 2| Received: "Jul 10 17:11:59 mdwichma-lnx
test-syslog-98a: 12345 1234 11145401322 322 3e+02
5
520|5 98 00005037 1 3|"
520|5 98 00005037 1 4| Expected: " 12345 1074922706 11145401322
322 3e+02 5 "
520|5 98 00005037 1 5|syslog() failed to format argument correctly
520|5 98 00005037 1 6| Received: "Jul 10 17:12:01 mdwichma-lnx
test-syslog-98b: 12345 1234 11145401322 322 3e+02
5
520|5 98 00005037 1 7|"
520|5 98 00005037 1 8| Expected: "12345 1074922706 11145401322 322
3e+02 5 "
220|5 98 1 17:12:02|FAIL
200|5 99 17:12:02|TP Start
520|5 99 00005037 1 1|syslog() failed to format argument correctly
520|5 99 00005037 1 2| Received: "Jul 10 17:12:04 mdwichma-lnx
test-syslog-99a: 12345 134612178 11145401322 321.765
3.21765432e+02 5 test-string
520|5 99 00005037 1 3|"
520|5 99 00005037 1 4| Expected: " 12345 -1073806126 11145401322
321.765 3.21765432e+02 5 test-string"
520|5 99 00005037 1 5|syslog() failed to format argument correctly
520|5 99 00005037 1 6| Received: "Jul 10 17:12:06 mdwichma-lnx
test-syslog-99b: 12345 134612178 11145401322 321.765432
3.217654e+02 5 test-string
520|5 99 00005037 1 7|"
520|5 99 00005037 1 8| Expected: "12345 -1073806126 11145401322
321.765432 3.217654e+02 5 test-string "
220|5 99 1 17:12:07|FAIL
200|5 117 17:12:41|TP Start
520|5 117 00005037 1 1|syslog() failed to format argument correctly
520|5 117 00005037 1 2| Received: "Jul 10 17:12:42 mdwichma-lnx
test-syslog-117: 011145401322 +321.765432 +3.217654e+02 +5.00000
+12345 +134612178 a test-string
520|5 117 00005037 1 3|"
520|5 117 00005037 1 4| Expected: "011145401322 +321.765432
+3.217654e+02 +5.00000 +12345 +1074136274 a test-string "
220|5 117 1 17:12:43|FAIL
200|5 118 17:12:43|TP Start
520|5 118 00005037 1 1|syslog() failed to format argument correctly
520|5 118 00005037 1 2| Received: "Jul 10 17:12:44 mdwichma-lnx
test-syslog-118: 011145401322 321.765 3.21765432e+02 5.000
$9$#-*12$.5d 134612178 test-strin
520|5 118 00005037 1 3|"
520|5 118 00005037 1 4| Expected: "011145401322 321.765
3.21765432e+02 5.000 $9$#-*12$.5d 1074922706 test-strin "
220|5 118 1 17:12:45|FAIL
--
To UNSUBSCRIBE, email to lsb-test-request@lists.linuxbase.org
with subject of "unsubscribe". Trouble? Email listmaster@lists.linuxbase.org
Reply to: