[PATCHES] Merge loadparm further

[Garming Sam]

Regarding any work further from this point, it has become increasingly 
apparent that the
approach of passing around pointers and the use of the s3 helper 
functions isn't really the
right approach.

Fundamentally, the issue lies with the life of the temporary contexts. 
Their purpose was to be
be temporary, but they point to global variables which are longer lived 
and change. I first noticed
the issue when attempting to merge 'add a service', and realized how the 
service pointers could be
realloc'd beneath it.

The work up to this point should not be affected, but I don't think the 
problem can continue
to be sidestepped. At least, not if there is ever to be a single 
well-integrated loadparm
system. It really needs an overhaul in how the two interact, which can't 
be achieved bit by bit.
This would probably involve lifting the loadparm context up and making 
it global in lib/param.
At one point, a little while back, I took a brief crack at this and ran 
into a number of initialization issues
and implicit assumptions, with fields being set to zero for instance. 
These kinds of issues would need
to be dealt with.


It would be nice to hear any thoughts on this.


Cheers,

Garming Sam


On 20/06/14 23:55, Andrew Bartlett wrote:
> Jeremy,
>
> Sorry for the odd message flow, git-send-email got away from me :-)
>
> What I was going to say in the cover message is that following is the
> next set of patches to merge parts of loadparm, with a focus on the
> parsing of parameters and the handling of special functions.
>
> Some of these are quite old, but that's because Garming needed to spend
> quite some time ensuring that every patch passes the key tests on their
> own.
>
> Anyway, now that I've had time to review these, would you please be able
> to review and push them to master?
>
> Thanks,
>
> Andrew Bartlett
>