Paolo Bonzini <[email protected]> writes:

> On 12/03/2015 16:10, Peter Maydell wrote:
>> So is the non-deprecated approach to have the board call
>>     object_property_set_bool(OBJECT(dev), true, "realized", &err);
>> directly itself (handling errors however it feels like) rather
>> than calling qdev_init_nofail() ?
>
> qdev_init_nofail seems a good enough name for a shortcut...  The problem
> with qdev_init was that it was a symptom of bad errors, but I would
> happily un-deprecate qdev_init_nofail, especially until we figure out
> whether we really want recursive realization.

I agree.

I'm getting rid of qdev_init() because it's fundamentally tied to
qerror_report().  And I'm getting rid of qerror_report() because it
doesn't play nicely with error_set().

I can't see anything wrong with qdev_init_nofail().  Except perhaps the
name, but improving it doesn't seem to be worth the churn.

Reply via email to