Re: [PATCH] tools: remove unused wrappers for python

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

On Tue, 2015-10-06 at 17:31 +0100, Wei Liu wrote:
> On Tue, Oct 06, 2015 at 04:38:08PM +0100, Andrew Cooper wrote:
> > On 06/10/15 16:26, Ian Campbell wrote:
> > > On Tue, 2015-10-06 at 17:21 +0200, Juergen Gross wrote:
> > > > On 10/06/2015 05:11 PM, Ian Campbell wrote:
> > > > > On Tue, 2015-10-06 at 16:51 +0200, Juergen Gross wrote:
> > > > > > On 10/06/2015 03:40 PM, Ian Campbell wrote:
> > > > > > > On Tue, 2015-10-06 at 12:39 +0100, Wei Liu wrote:
> > > > > > > 
> > > > > > > > And for the record, if my google-fu doesn't fail me, it's
> > > > > > > > possible to
> > > > > > > > load shared library into python interpreter using "dl"
> > > > > > > > module in
> > > > > > > > 2.7
> > > > > > > > and
> > > > > > > > "ctypes" module in 3.x.
> > > > > > > Possible, but not especially convenient since you need to
> > > > > > > convert
> > > > > > > the C
> > > > > > > prototype manually, plus the result is not necessarily very
> > > > > > > "pythonic".
> > > > > > > 
> > > > > > > I could totally see why people would prefer these bindings
> > > > > > > (or an
> > > > > > > argument
> > > > > > > for us providing a ctypes based wrapper).
> > > > > > How often is such a debugging interface being used? Please
> > > > > > consider
> > > > > > the amount of code (my patch removed nearly 3000 lines of
> > > > > > code!) and
> > > > > > the availability of the xl wrapper.
> > > > > My understanding was that this was used by the "xen-bugtool"
> > > > > stuff in
> > > > > XenServer, so for actual functionality (gathering debug info) and
> > > > > not
> > > > > debugging (I supposed that the reference to being used for
> > > > > debugging was
> > > > > due to the name of the tool).
> > > > And this functionality isn't available via the xl bindings?
> > > I don't know, we'll have to wait for those who are using it to chime
> > > in.
> > 
> > The python xl bindings? They don't even compile.
> 
> Urgh. It does compile for me.

Are you talking about tools/python/xen/lowlevel/xl/? Because that seems
unlikely.

In any case they certain don't work, they have a model for updating the C
versions of the data structures in sync with the python code, as opposed to
marshalling in and out around the libxl calls like the other language
bindings do, which is IIRC broken wrt at least nested structures and keyed
unions IIRC (and probably other stuff now).

Ian.