[augeas-devel] RFC: Augeas Check function

[David Lutterkort]

On Mon, 2009-01-12 at 13:59 -0500, Bryan Kearney wrote:
> David Lutterkort wrote:
> > On Fri, 2009-01-09 at 13:34 -0500, Bryan Kearney wrote:
> > 
> > I've been meaning to change that so that you can do something like
> >   set /augeas/save overwrite
> >   save
> >   .. more changes to the tree ..
> >   set /augeas/save backup
> >   save
> > so that the first batch of changes just overwrites existing files, and
> > the second also saves .augsave files.
> > 
> > With that model, we'd add another option for /augeas/save, say 'check'
> > or 'none' and do what your aug_check call does. With that, we wouldn't
> > need to add another API call, and the implementation would just be a few
> > additional lines in aug_save and tree_save.
> > 
> > What do you think about approaching it that way ?
> > 
> > David
> 
> 
> David:
> 
> Please take a look at this branch:
> 
> http://github.com/bkearney/augeas/tree/flags
> 
> It implements the feature mostly per our discussion. Couple of things:

This looks pretty good. I'll make a couple small changes and pull that
into my tree tomorrow.

> 1) I made noop into a first class flag, as it seemed odd to have part of 
> the flag->text public and the other half not.

The reason I don't want to add this flag is that I feel all the
AUG_SAVE_* flags were a mistake in the public API, and
modifying /augeas/save should be the way to influence that going
forward.

> 2) in aug_save the logic causes some of the initial flag values to be 
> lost. Since some of the flags effect saving and some do not, the 
> "correct" way would be to toggle on the bits which effect saving or to 
> split out the flags into save versus init. I figured you would not want 
> to do the latter.. and I would hold off on the former to make sure this 
> approach passed muster.

I'll fix that when I pull the patch in. It also needs to check that
(a) /augeas/save is still there and (b) has a non-NULL value; in fact,
if /augeas/save is not there or nor set to one of the four fixed values,
aug_save should fail with an error.

David