[Pkg-giraffe-discuss] Getting autopkgtest for kopanocore working (finally)

[Carsten Schoenert]

Am 17.12.18 um 09:33 schrieb Carsten Schoenert:
> Hi Guido,
> 
> Am 16.12.18 um 13:03 schrieb Guido Günther:
> 
>> I was wondering if the previous 'help' test would trigger it but *thought*
>> each test is supposed to start out from a clean test bed. If this is not
>> the case there's a bug somewhere else since I'm pretty sure this used to
>> be the case. This might depend on the used runner. When you use e.g.
>> the nullrunner in pbuilder tests are not properly isolated. I thought
>> the ssh runner would handle deinstalling packages but I was wrong there.
> 
> mhh, interesting.
> My view was always from other other side, I thought always it's one
> environment and all tests are running within this.

Seems your memory isn't that bad :) , the debci environment seems to
work exactly that way you remembering.

After the incredible fast tracking of my yesterdays upload a first debci
run did happen and the package did fail here. The log shows that the
package ucf is updated within the testbed, as far I remember ucf came in
yesterday as an update into unstable.

While running yesterday the autopkgtest locally before the final upload
I just have seen the update of ucf only once here while the autopkgtest
did running! So I need to add kopano-utils too to the help test run to
get this "fixed".

But piuparts is also not happy as the logrotate configuration for
kopano-server did not have a 'missingok' and so logrotate is stopping
with an error because there is simply no log file existing at this state.

https://piuparts.debian.org/sid/fail/kopano-server_8.6.90-1.log

> cat /etc/logrotate.d/kopano-server
> /var/log/kopano/server.log {
>   rotate 5
>   weekly
>   postrotate
>     /usr/bin/killall -HUP kopano-server
>   endscript
> }

I'm wondering if this extra file is really needed as there is also the
file /etc/logrotate.d/kopano (which comes from kopano-common) also holds
a configuration for the server log. If I remove the dedicated file for
kopano-server logrotate is working normally. If I add the missingok
entry logrotate is complaining about double entries for
/var/log/kopano/server.log of course.

> Dez 18 12:58:01 debian systemd[1]: Starting Rotate log files...
> Dez 18 12:58:01 debian logrotate[5548]: error: kopano-server:1 duplicate log entry for /var/log/kopano/server.log
> Dez 18 12:58:01 debian logrotate[5548]: error: found error in file kopano-server, skipping
> Dez 18 12:58:01 debian systemd[1]: logrotate.service: Main process exited, code=exited, status=1/FAILURE
> Dez 18 12:58:01 debian systemd[1]: logrotate.service: Failed with result 'exit-code'.
> Dez 18 12:58:01 debian systemd[1]: Failed to start Rotate log files.

I vote for dropping the additional logrotate config file. To me it makes
not really sense to have an extra file for kopano-server which also
looks to me as written a bit on a rush compared to /etc/logrotate.d/kopano.

The log rotation for server.log and also audit.log from/e/l.d/kopano
have a bit more options.

> /var/log/kopano/server.log /var/log/kopano/audit.log {
>         weekly
>         missingok
>         rotate 52
>         compress
>         delaycompress
>         notifempty
>         postrotate
> 			killall -HUP kopano-server 2> /dev/null || true
>         endscript
> }
-- 
Regards
Carsten Schoenert