PolicyKit releases and !AWOL

Doug Klima cardoe at gentoo.org
Thu Dec 6 08:33:35 PST 2007


David Zeuthen wrote:
> On Thu, 2007-12-06 at 17:13 +0100, Michael Biebl wrote:
>   
>> 2007/12/6, David Zeuthen <david at fubar.dk>:
>>     
>>> Hi,
>>>       
>> Hi
>>
>>     
>>> First sorry for not being more active on the list. Been busy with lots
>>> and lots of stuff including PolicyKit and trying to put my thoughts down
>>> about the next major version of HAL (basically a rewrite, more on that
>>> soon!).
>>>       
>> TBH, sentences like that scare me a little bit.
>>     
>
> That's fine, that's the reaction I get from most people which is
> expected. The interface (apart form syntax) won't change a lot; a lot of
> the internals will. But more on that later.
>
>   
>> PK-gnome failes to compile with -Wl,--as-needed, a -lpolkit-grant is missing.
>> Please find the attached patch.
>>     
>
> I know that in PolicyKit we set these automatically. How about a patch
> to do the same for PK-gnome? Thanks.
>
>   
>>> Also, packagers should be careful, there's a few new setgid helpers; the
>>> blurb at the end of ./configure should be useful. And here's the Fedora
>>> spec files for reference
>>>       
>> "polkit-set-default-helper" seems to be missing in the blurb at the
>> end of ./configure.
>>     
>
> I'll fix that.
>
>   
>> The PK changelog mentions as highlight, that the glib dependency was removed.
>> Instead you implemented your on sortof glib, called libkit.
>> I don't see the real benefit here, smells a bit of NIH. glib is a
>> rather well tested lib and already used in the utopia stack. Could you
>> elaborate a bit on that decision?
>>     
>
> Sure. Two things.
>
> First, in order for libpolkit to be useful in projects like dbus
> (specifically the bus) and X.org it needs to handle OOM. And glib can't
> do that; the glib API is not designed to do that. And that's fine.
>
> Seriously, libkit is 4700 lines of C including 100% doc coverage and 91%
> test coverage. It's hardly reinventing glib.
>
>       David
>   
It's still providing some functionality that's provided by glib and as
such still re-implementing a sub-set of glib and to what benefit?


More information about the hal mailing list