Re: [PATCH for-4.5 1/2] libxl: continue when encounter ERROR_JSON_CONFIG_EMPTY

[Ian Campbell <Ian.Campbell@citrix.com>]

On Fri, 2014-11-14 at 11:10 +0000, Wei Liu wrote:
> On Fri, Nov 14, 2014 at 10:55:35AM +0000, Ian Campbell wrote:
> > On Fri, 2014-11-14 at 10:52 +0000, Ian Campbell wrote:
> > > On Wed, 2014-11-12 at 17:04 +0000, Wei Liu wrote:
> > > > Continue when libxl_retrieve_domain_configuration encounters
> > > > ERROR_JSON_CONFIG_EMPTY, as caller might be interested in the partial
> > > > configuration pulled from xenstore.  In this case
> > > > ERROR_JSON_CONFIG_EMPTY is used as return value as before, if no other
> > > > error happens along the way.
> > > > 
> > > > Signed-off-by: Wei Liu <wei.liu2@citrix.com>
> > > > Cc: Zhigang Wang <zhigang.x.wang@oracle.com>
> > > 
> > > Acked-by: Ian Campbell <ian.campbell@citrix.com>
> > 
> > On second thoughts, I think this really needs an update to libxl.h to
> > describe the semantics of this function, i.e. to what extent the output
> > is valid for various error codes, especially ERROR_JSON_CONFIG_EMPTY.
> > 
> 
> Any non-zero return code means the output is invalid (as in "Is this
> output valid to rebuild a domain?").

But your second patch prints it as if it is at least somewhat
meaningful, if not entirely valid. According to what you just said it
shouldn't do so.

You effectively have three return states now: Fully valid, domain exists
but it's configuration is unsure or somehow incomplete (~= JSON EMPTY),
some sort of error occurred.

It might even be a good idea to have some new externally visible error
code for the middle state, since JSON_EMPTY may not be the only reason
for being in that state (at least in theory).

Anyway, since this is all more subtle than the existing 0 is good,
non-zero is completely bad it should be written down.

Ian.