Steve Cohen wrote:
> Kenneth Porter wrote:
>
>> --On Monday, February 07, 2005 8:45 PM -0600 Steve Cohen
>> <scohen@javactivity.org> wrote:
>>
>>> Thanks, but I don't think it's a firewall issue. I don't have one. This
>>> is basically a home PC that runs Linux. It is protected from the
>>> outside
>>> world by a broadband router with a firewall, but I think this just
>>> affects eth0, not lo. Or do I have that wrong?? At any rate,
>>> iptables is
>>> running on my system but is set to allow everything.
>>
>>
>>
>> What does "iptables -L" report? That will definitively tell you the
>> contents of the firewall.
>
> # /sbin/iptables -L
> Chain INPUT (policy ACCEPT)
> target prot opt source destination
>
> Chain FORWARD (policy ACCEPT)
> target prot opt source destination
>
> Chain OUTPUT (policy ACCEPT)
> target prot opt source destination
>
>>
>>> No, that isn't it either. It now appears that port 15835 is
>>> specified in
>>> basic_tests.py as the port to listen on. So that makes sense as the
>>> port
>>> to be specified in the spec.
>>>
>>> I think that what the test is trying to tell me is that something is
>>> wrong with my apache setup. Still don't know what, though.
>>
>>
>>
>> I don't think Apache is even involved. It sounds like a standalone
>> Subversion server is set up for the tests to work against. It listens
>> on that port for connections from the test harnesses.
>>
>
> I think you're wrong about that. Here's the test output just before the
> block of tests that failed:
>
> + echo '*** Running regression tests on RA_DAV (HTTP method) layer ***'
> *** Running regression tests on RA_DAV (HTTP method) layer ***
> + killall httpd
> + sleep 1
> ++ pwd
> + sed -e 's;@SVNDIR@;/usr/src/redhat/BUILD/subversion-1.1.3;'
> ++ pwd
> + /usr/sbin/httpd -f /usr/src/redhat/BUILD/subversion-1.1.3/httpd.conf
> httpd: Could not determine the server's fully qualified domain name,
> using 127.0.0.1 for ServerName
> + sleep 1
>
> It looks to me that the test is killing the apache daemon (httpd) and
> then restarting it with the config that comes with the build.
>
> Interestingly, after the build completes, reporting failure, my apache
> daemon is in a bad state:
>
> # /sbin/service httpd status
> httpd dead but subsys locked
>
> It is easily restarted using
> /sbin/service httpd restart
>
> and then everything works fine, including my already
> installed-from-binary-rpm subversion.
>
> So I think that quite definitely apache is involved here, somehow. The
> question is, how?
>
> Looking at the files in /var/log/httpd, I see a few entries that are
> possibly of interest at the time of the tests, but I don't know what
> they signify here:
>
> /var/log/httpd/error_log:
> [Mon Feb 07 22:29:37 2005] [warn] child process 7277 still did not exit,
> sending a SIGTERM
> [Mon Feb 07 22:29:37 2005] [warn] child process 7278 still did not exit,
> sending a SIGTERM
> [Mon Feb 07 22:29:37 2005] [warn] child process 7279 still did not exit,
> sending a SIGTERM
> [Mon Feb 07 22:29:37 2005] [warn] child process 7280 still did not exit,
> sending a SIGTERM
> [Mon Feb 07 22:29:37 2005] [warn] child process 7281 still did not exit,
> sending a SIGTERM
> [Mon Feb 07 22:29:37 2005] [notice] caught SIGTERM, shutting down
>
> Judging from the timestamps, and the test output above, I surmise that
> perhaps the shutdown of httpd was not complete. Is "sleep 1" enough
> time to for the shutdown to complete? If I try to shutdown manually:
>
> /sbin/service httpd stop
>
> there is a noticeable delay before it comes back, although I'm not sure
> it's as much as a second.
>
> Well, can anyone make sense of all these symptoms and tell me what's
> going on?
>
>
I had one more idea - let's eliminate the possibility of not shutting
down apache properly, by turning the daemon off before beginning the
tests. No dice. I get the same errors as before. And the log
timestamps are separated from the point in time where the scripts fail
by enough so that I am doubting the connection.
So I'm stumped. Why are these tests failing? What do they indicate?
The output seems more geared to the developers writing the tests and
subversion than to the user trying to build it. Can someone who knows
these tests please provide me some information on where to go next?
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Feb 8 14:03:41 2005