[eq-dev] Server::useConfig segmentation fault in 0.6 RC1

[Stefan Eilemann]

Hello Juan,

On 4. Dec 2008, at 16:40, Juan Hernando wrote:

> Dear Equalizer developers,
>
> Recently I've started porting and MPK based application to Equalizer.

Ola! You're the first one I am aware of being on the same path as me. :)

What kind of application is this? Where we in contact during MPK  
'times'?

> So instead of doing that I'd like to use a local server inside the  
> application node process and use Server::useConfig to give the  
> configuration as a string. I know the function is undocumented and  
> may disappear but in my case, it's useful, unless there is an easier  
> way of starting the server with a user given configuration.

Right now this is the easiest way. The API will be replaced by another  
API eventually, but a similar possibility will be available.

> I've written a small test, which now is working. However it was  
> crashing in the beginning and after some debugging I've come across  
> something that, to me, seems to be a bug in Server::useConfig.  
> Analyzing the code I've found that there is a difference between  
> chooseConfig and useConfig in the sequence of operations performed  
> in the server, while the first one calls Server::addConfig for the  
> default config at some moment, the latter doesn't.
> After the following change in Server::_cmdUseConfig:
>    EQINFO << "Using config: " << endl << Global::instance() <<  
> config << endl;
>    config->setApplicationNetNode( node );
>    +addConfig( config );
>    mapConfig( config );
> the test application works.

It is intentional that there is no addConfig(). Active configs are in  
_appConfigs, and it's added to this. The bug was something else, but  
it's fixed in the trunk now:

http://equalizer.svn.sourceforge.net/equalizer/?rev=2405&view=rev


Cheers,

Stefan.