Re: __attribute__ ((cleanup) patch

[Colin Walters <walters verbum org>]

On Thu, 2012-10-18 at 11:28 -0500, Dan Williams wrote:

> I don't consider this an experiment at this point.

Well, I do think Pavel has a point; it's true that NetworkManager is
an order of magnitude more code than any other project I've used it in
before, and it's larger in terms of people too.

>  Yes,
> there is a bit of education required, but perhaps we could also work on
> making some things compile-time errors instead of runtime gotchas.

Would require a GCC plugin.  But the way I see it is these patches don't
*regress* type safety - the compiler isn't going to warn you right now
if you call g_free (some_object) either.

Anyways, bottom line: given the benefits, my goal here is to figure out
where the bar is in the opinion of the NetworkManager developers, and do
what I need to reach it.