[Ctk-developers] ctkPlugins::install hides the original cause of ctkPluginExceptions

[Miklos Espak]

Thank you, Sascha!
The output is now even more detailed, and it contains the error that I
was looking for.

Miklos


On Tue, Dec 4, 2012 at 6:32 PM, Sascha Zelzer
<s.zelzer at dkfz-heidelberg.de> wrote:
> Hi Miklos,
>
> thanks for reporting this. In my opinion, code in ctkPluginStorageSQL.cpp
> shouldn't throw ctkPluginException objects but rather ctkRuntimeException or
> similar. This code misuses ctkPluginException in a semantic sense.
>
> But you are right that the error messages from wrapped exceptions could be
> suppressed in the console output if not handled properly. For MITK, the
> following commit fixed this:
>
> https://github.com/MITK/MITK/commit/298de77d3de67c228e51e6c4b9959902b3840372
>
> Thanks,
> Sascha
>
>
> On 12/04/2012 06:09 PM, Miklos Espak wrote:
>>
>> Hi,
>>
>> CTK failed to install one of our MITK plugins. The error turned to be
>> in our plugin, but I found a little mistake in the CTK code while
>> debugging, that made it a bit difficult to find the error in the
>> plugin.
>>
>> In ctkPluginStorageSQL.cpp:436 a ctkPluginException is thrown when a
>> plugin could not be loaded.
>>
>> This is fine, but in ctkPlugins.cpp:125-139 it is caught as
>> ctkException, and re-wrapped to another ctkPluginException.
>>
>> Because of this, the original error does not appear in the log
>> (BlueBerry.consoleLog), what makes it a bit troublesome to find out
>> what is wrong with the plugin.
>>
>> Inserting these lines at line 125 has solved the problem for me:
>>
>>      catch (const ctkPluginException& e)
>>      {
>>        if (!pa.isNull())
>>        {
>>          pa->purge();
>>        }
>>        throw e;
>>      }
>>
>>
>> Cheers,
>> Miklos
>> _______________________________________________
>> Ctk-developers mailing list
>> Ctk-developers at commontk.org
>> http://public.kitware.com/cgi-bin/mailman/listinfo/ctk-developers
>
>