Re: [Boost-users] [program_options] copy option_description from one options_description to another

Subject: Re: [Boost-users] [program_options] copy option_description from one options_description to another
From: Juraj Ivančić (juraj.ivancic_at_[hidden])
Date: 2013-06-30 17:57:26

On 28.6.2013. 20:20, Ghyslain Leclerc wrote:

> First post here.  Not a programmer by trade, but a medical physicist, so my apologies
> if my question seems too much of a newbie question.  And hopefully, my programing
> jargon won’t be too bad.

Your post is clearer than most.

> In either case (pointer or object), my actual problem comes when I want to add options
> to a options_description.  I can easily use the add function :
>    myOptionsDescriptionObject.add( theOtherOptionsDescriptionObject );
>
> but that adds empty lines in the output format and creates artificial grouping of options,
> which I would like to avoid if possible.  That being said, given the current state of our code,
> I will implement this version now as this is the "best" solution that works.

There is an 'add' overload which accepts shared_ptr. This one does not 
add grouping. If you make your base class contain shared pointers I 
think you could avoid this issue. Alternatively - you can create a new 
shared_ptr on demand:

shared_ptr< option_description > tmp( new option_description(
     theOtherOptionsDescriptionObject ) );
myOptionsDescriptionObject.add( tmp );

> The way I am trying to get around that is to extract for each option_description
> contained in the options_description object the name, the description and the
> value_semantic and then use these three extracted values with the add_options()
> function like so:
>    muOptionsDescriptionObject.add_options()( other.options()->name(),
>                                              other.options()->semantic().get(),
>                                              other.options()->description()    );
>
> Works perfectly for name and description, but fails for value_semantic.  It compiles, but
> generates a "The program has unexpectedly finished." message.  My hypothesis for what is
> happening is that I am using the .get() to extract the raw pointer from the shared_ptr
> returned, but the shared_ptr somehow deletes the memory and my pointer becomes invalid.

I don't think that is what happens. As far as I can see - description 
object tries to take ownership of the value semantic object. If you pass 
value which is owned by a shared_ptr you get double delete crash.

Please note that I am not familiar with program_options, most of what I 
wrote is based on a quick program_options code review, so I could be 
wrong. Anyway, I hope this helps.