Re: [Xen-devel] [PATCH v2 07/15] libxl: convert libxl_domain_destroy to an async op

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Roger Pau Monne writes ("Re: [PATCH v2 07/15] libxl: convert 
libxl_domain_destroy to an async op"):
> Ian Jackson wrote:
> > Just a suggestion: rather than this business with *last it might be
> > easier to assign a magic error code, or use +1, for "not all done
> > yet".  (+1 would work since all actual libxl error codes are -ve).
> > By my reading of the current interface the return value is not
> > meaningful if *last==0 on output, which is the opposite of the usual
> > way round.
> 
> This function currently returns two different things, the return value 
> returns the rc code of any operation that has failed, and the parameter 
> last marks if all device events are finished. Without the last 
> parameter, we won't be able to return the failed error code if all 
> devices had finished, since we had to return 0, so the error code would 
> be lost.

I was suggesting that it should return:
   rc==0                if all devices completed successfully
   rc==ERROR_SOMETHING  if all devices completed with at least one error
   rc==+1               if something is still going

Or if you prefer
   rc==ERROR_OPERATION_NOT_YET_COMPLETE    if something is still going

> Since I think this is not really clear, I will split this function into 
> two different ones, one that checks if all devices are finished, and 
> another one that checks if some devices have reported errors.

Please don't do that - you'll have to turn it into two loops and it'll
be a lot more fiddly.

Ian.

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