Re: [Qemu-devel] Re: [PATCH 02/11] qerror: expose a function to format an error

[Luiz Capitulino <lcapitulino@redhat.com>]

On Mon, 14 Mar 2011 15:41:49 -0500
Anthony Liguori <aliguori@us.ibm.com> wrote:

> On 03/14/2011 03:22 PM, Luiz Capitulino wrote:
> > On Mon, 14 Mar 2011 14:45:13 -0500
> > Anthony Liguori<aliguori@us.ibm.com>  wrote:
> >
> >> On 03/14/2011 02:37 PM, Luiz Capitulino wrote:
> >>> qerror_abort() only exists for debugging purposes. I won't say its perfect,
> >>> but it's better than nothing and has already saved some time when writing
> >>> new errors.
> >>>
> >>> I'm fine dropping it as long as there's a better replacement, which is
> >>> not the case here. There's even a hunk that replaces qerror_abort() for
> >>> a plain abort().
> >> Yes, that's the replacement.
> > It's not a good one: it makes the current code a bit worse and we don't know
> > how and when the error classes are going to be replaced.
> 
> Yes, we do, before 0.15.0.

Very optimistic :) I don't doubt you can post patches quickly, but we're
likely going to have fun discussions, respins, tests etc. And all the QAPI
stuff in parallel.

> > Maybe a better merge plan would be to work on errors first. Completely drop
> > qerror according to qapi needs, and then put the rest of the stuff on top.
> 
> Can't be done until we introduce new QMP commands to get rid of the old 
> HMP commands because there are HMP commands without equivalents that 
> make use of qerror_report().
> 
> There are just a handful of these left in my QAPI branch so once I can 
> start adding these QMP commands, they'll be gone quickly.  I don't want 
> to introduce a bunch of new QMP commands without this stuff getting 
> merged upstream first though.
> 
> Regards,
> 
> Anthony Liguori
>