Re: Proposal for using std::unique_ptr to signal transfer of ownership

2016-03-12 Thread Thorsten Behrens
Markus Mohrhard wrote:
> I hope that these are not really controversial [...]
>
Not at all - to state the obvious (and not let the mail linger
unanswered).

Cheers,

-- Thorsten


signature.asc
Description: Digital signature
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/libreoffice


Proposal for using std::unique_ptr to signal transfer of ownership

2016-03-11 Thread Markus Mohrhard
Hey,

so while looking through some of the current calc memory leaks I noticed
that at least some of them are related to our poor API and I noticed that
quite a few of the problems can easily be avoided with better API design.

For now I would like to propose that we try to use std::unique_ptr in two
places to show ownership of data. A method that returns an object for which
ownership transfers to the caller we should return the object wrapped in a
std::unique_ptr. The same should be done for a method taking ownership of a
pointer.

I hope that these are not really controversial and I'm aware that they are
not yet useful for all the old code but we need to start somewhere. Also if
there is some agreement on this there are some simple easy hacks that can
be done to move at least a bit of our code into this direction (e.g. the
horrible SfxItemPool::Clone method that results in quite a few of my leaks.)

Regards,
Markus
___
LibreOffice mailing list
LibreOffice@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/libreoffice