Docker: Test failures with Kernel::System::GenericInterface::Webservice

[bschmalhofer]

For example:

scripts/test/DynamicField/Driver/WebService/PossibleValuesGet.t .................................. 
ok 1 - Created test user 18
ok 2 - Added test user test9050002509000002 to group admin
ok 3 - Added test user test9050002509000002 to group users
ok 4 - Set user UserLanguage to en
ok 5 - TicketCreate() Ticket 1 53
ok 6 - TicketCreate() Ticket 2 54
ok 7 - WebserviceAdd WebService 8
ok 8 - Updated Webservice 8 - TestWebservicetest9050002509100000
ok 9 - DynamicFieldAdd() ID 39
ok 10 - State/Queue Ticket 1 Updated Webservice 8 - TestWebservicetest9050002509100000
not ok 11 - State/Queue Ticket 1 PossibleValuesGet() result (is not equal, see below)
Diff:
+--+-------------+--+---------------------+
|  |Actual data  |  |Expected data        |
+--+-------------+--+---------------------+
* 1|$VAR1 = {};  * 1|$VAR1 = {            *
|  |             * 2|  'Queue' => 'Raw',  *
|  |             * 3|  'State' => 'new'   *
|  |             * 4|};                   *
+--+-------------+--+---------------------+

Actual data:
$VAR1 = {};

Expected data:
$VAR1 = {
  'Queue' => 'Raw',
  'State' => 'new'
};
ok 12 - State/Queue Ticket 2 Updated Webservice 8 - TestWebservicetest9050002509100000

[bschmalhofer]

It looks like the Webserver is hit with Requests like http://localhost:5000/otobo//nph-genericinterface.pl. Note the double // Supporting the //fixes propably the immediate problem.

But another issue might be that Proxies discard the douple // . In this case functionality might be lost. More Testing is needed.

[bschmalhofer]

The first guess did not work out. Now let's ponder another idea. On observation is that some of the test cases sometimes work. Maybe it's not a general prolem, but a problem with timing. The test script changes the Configuration by writing small .pm files into Kernel/Config/Files. This causes the Gazelle Webserver to restart and this takes some time.

TODO:

• instead of restarting Gazelle, do something similar to Apache::Reload, even though this is nasty

[bschmalhofer]

The trick was to remove the -R Kernel when calling plackup.
Looks good now.