Re: [Xen-devel] [PATCH v2 COLOPre 11/13] tools/libxl: rename remus device to checkpoint device

[Ian Campbell <ian.campbell@xxxxxxxxxx>]

On Mon, 2015-06-15 at 17:24 +0100, Wei Liu wrote:
> On Mon, Jun 15, 2015 at 09:45:54AM +0800, Yang Hongyang wrote:
> > 
> > 
> > On 06/12/2015 10:57 PM, Ian Jackson wrote:
> > >Wei Liu writes ("Re: [Xen-devel] [PATCH v2 COLOPre 11/13] tools/libxl: 
> > >rename remus device to checkpoint device"):
> > >>On Fri, Jun 12, 2015 at 02:30:46PM +0100, Wei Liu wrote:
> > >>>On Mon, Jun 08, 2015 at 11:43:15AM +0800, Yang Hongyang wrote:
> > >>>>-    (-18, "REMUS_DEVOPS_DOES_NOT_MATCH"),
> > >>>>-    (-19, "REMUS_DEVICE_NOT_SUPPORTED"),
> > >>>>+    (-18, "CHECKPOINT_DEVOPS_DOES_NOT_MATCH"),
> > >>>>+    (-19, "CHECKPOINT_DEVICE_NOT_SUPPORTED"),
> > >>>
> > >>>You should add two new error numbers.
> > >>>
> > >>
> > >>And in that case you might also need to go through all places to make
> > >>sure the correct error numbers are return. I.e. old remus code path
> > >>still returns REMUS error code and new CHECKPOINT code path returns new
> > >>error code.
> > >>
> > >>I merely speak from API backward compatibility point of view. If you
> > >>think what I suggest doesn't make sense, please let me know.
> > >
> > >To me this line of reasons prompts me to ask: what would be wrong with
> > >leaving the word REMUS in the error names, and simply updating the
> > >descriptions ?
> > >
> > >After all AFIACT the circumstances are very similar.  I don't think it
> > >makes sense to require libxl to do something like
> > >    rc = were_we_doing_colo_not_remus ? CHECKPOINT_BLAH : REMUS_BLAH;
> > >
> > >Please to contradict me if I have misunderstood...
> > 
> > COLO and REMUS both are checkpoint device. We use checkpoint device layer
> > as a more abstract layer for both COLO and REMUS, come to the error code,
> > these can be used by both COLO and REMUS. So we don't distinguish if we
> > are doing COLO or REMUS, uses are aware of what they're executing(colo
> > or remus).
> > 
> 
> Right. So continue using REMUS_ error code is fine.

Seems like it would also be OK to switch the name and then in libxl,h

#ifdef LIB_API_VERSION < 0xWHENEVER
#define REMUS_BLAH CHECKPOINT_BLAH
#define ...
#endif

_If_ we think the new names make more sense going fwd...



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel